Search

It is important to judge the density of pear flower accurately for automatic thinning. In order to judge the density of pear flowers precisely and achieve the purpose of automatic flower thinning, a density classification of pear flower images method based on improved density peak clustering algorithm was proposed in this study. The pear flower image samples used in the experiment were taken from the research base of Hebei Agricultural University, Niugang Village, Yi County, Baoding City, Hebei Province, and the variety was three-year-old Qiuyue. In this study, the images of pear flowers in natural environment were collected on April 18, 2021, when the pear flowers were in full bloom. The pear flower recognition images with high detection accuracy were obtained by using the depth learning model to detect the pear flower. The central coordinates of the target detection frame were extracted from the recognition images to obtain the data points that needed to be clustered. Secondly, in order to achieve the density classification of pear flower images, the method of selecting the cluster center was improved according to the shortcomings of the original density peak clustering algorithm in the density classification of pear flower and the requirements of pear flower density classification. The decision graph was divided into four parts by four groups of local density and center offset distance segmentation thresholds to select the cluster centers. These four parts corresponded to four grades of high, medium and low density and no thinning treatment. The accurate classification of pear flower images with reasonable density was realized. Finally, to solve the problem of inaccurate clustering classification of pear flower distribution with only cluster distribution, sparse distribution and large-scale close-up, the calculation method of dij parameter of distance between two points was improved. The pear flower scale and density grading standard were unified, and reasonable density grading could be achieved for all distribution types of pear flower images. The experimental results showed that the proposed algorithm could adapt to pear flower images of different scales. The accuracy of scale prediction was 94.89%, and the accuracy of density classification was 94.29%. Compared with the existing methods, the proposed method could achieve the density classification of local flower clusters in natural environment. the algorithm in this paper was compared with K-means algorithm and DBSCAN algorithm, and the accuracy of density classification was 94.29% and 68.57% respectively. Although the density classification accuracy of K-Means clustering algorithm was high, it lacked the algorithm steps to automatically select the clustering centers of different density classification, and was too sensitive to noise points. The density calculation method of DBSCAN algorithm lacked continuity, and the division of density grade was not precise enough. Based on the comprehensive analysis of density classification accuracy and cluster center selection method, the improved density peak clustering algorithm in this paper had the best effect. The clustering centers of different density grade could be automatically selected and the error of random initialization of clustering centers could be reduced. The proposed algorithm was robust to noise points and could eliminate noise points effectively. The proposed algorithm adopted soft statistics method for density calculation, which had continuity and was more accurate for density classification. It provides technical support for analyzing the density of pear flower and machine intelligent flower thinning. [ABSTRACT FROM AUTHOR]

Copyright of Transactions of the Chinese Society of Agricultural Engineering is the property of Chinese Society of Agricultural Engineering and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Transactions of the Chinese Society of Agricultural Engineering is the property of Chinese Society of Agricultural Engineering and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Transactions of the Chinese Society of Agricultural Engineering is the property of Chinese Society of Agricultural Engineering and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Transactions of the Chinese Society of Agricultural Engineering is the property of Chinese Society of Agricultural Engineering and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Transactions of the Chinese Society of Agricultural Engineering is the property of Chinese Society of Agricultural Engineering and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Transactions of the Chinese Society of Agricultural Engineering is the property of Chinese Society of Agricultural Engineering and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Deep learning has been widely applied to the recognition and segmentation of plant leaf disease. However, traditional recognition models of plant leaf disease usually lack transparency, since these end-to-end deep classifiers are susceptible to shadow, occlusion, and light intensity. To address this drawback, in this paper, we proposed a new recognition and segmentation model of plant leaf disease based on deconvolution-guided, called Deconvolution-Guided VGGNet (DGVGGNet). An encoder-decoder architecture with symmetric convolutional-deconvolutional layers was applied to DGVGGNet so that the plant leaf disease recognition and segmentation can be carried simultaneously. Our model consists of three main phases: recognition, inversion, and deconvolution. In the recognition phase, we first fed the VGGNet with plenty of plant leaf disease images, and then Categorical Cross-Entropy Loss was utilized to train the recognition model. VGGNet was made up of the convolution layers of VGG16 and 2 fully connected layers, and the weights of convolution layers were pre-trained on ImageNet. Ten kinds of tomato leaf disease images in the PlantVillage dataset were used in this paper, concretely, 30% of the pictures were used for training, and the rest 70% were used for the test. Besides, on-the-fly data augmentation was also exploited during the training stage, such as flipping the images and corrupting the origin images by brightness, saturation, salt noise, and Gaussian noise. In the inversion phase, two fully connected layers were used to reverse the category prediction vector. Simultaneously, two skip connections were used to reinforce the decoder by adding the vector of the VGGNet’s fully connected layers. Then, the feature vector was reshaped to feature maps to input into the deconvolution module. In the deconvolution phase, the feature maps were fed into the deconvolution module to acquire the segmentation result of the disease area, where each pixel was trained with the Binary Cross-Entropy Loss. The deconvolution module consists of the upsample and convolution operation. Meanwhile, five skip connections were used to fuse the multiple features, which can refine the segmentation results. Besides, only a few samples of the training set were given pixel-level labels of plant disease to supervise the output of the deconvolution module. At the end of the deconvolution module, the reconstruction layer was used to smooth the segmentation edge. To explore the influence of the number of pixel-level labels used in the model, 9 and 45 pixel-level labels were used to supervise the segmentation results, respectively. To simulate the natural conditions, different kinds of interference were added to the test data, such as translation, fruit occlusion, soil occlusion, leaf occlusion, and brightness reduction with different percentages. Experimentally, we evaluate our recognition module by exploring the performance of VGGNet, DGVGGNet-9, and DGVGGNet-45 on different interference datasets, respectively. We also evaluate our deconvolution module by exploring 4 different evaluation metrics, i.e., PA, MPA, MIoU, and FWIoU, compared with 3 popular semantic segmentation models, i.e., FCN-8s, U-Net, and SegNet. The results show that DGVGGNet-45 has the highest recognition accuracy as well as with the highest PA, MIoU, and FWIoU among the four segmentation evaluation metrics, which are 94.66%, 75.36%, and 90.46%, respectively. Compared with VGGNet, the deconvolution module of DGVGGNet-45 can guide the recognition module to pay more attention to the actual area of disease, which is effective in improving the segmentation accuracy. The recognition results demonstrate that DGVGGNet had strong robustness in tough conditions. Furthermore, DGVGGNet only took 12 ms to identify a single picture on the GPU, which can meet the real-time requirements. [ABSTRACT FROM AUTHOR]

Copyright of Transactions of the Chinese Society of Agricultural Engineering is the property of Chinese Society of Agricultural Engineering and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Transactions of the Chinese Society of Agricultural Engineering is the property of Chinese Society of Agricultural Engineering and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Rice seed surface morphology is an important aspect of seed purity identification and recognition. Considering that artificial recognition and identification methods have some faults, which including low efficiency, high labor costs, and poor accuracy. So scientifically selecting quality rice seeds by using computer vision methods is important. Different models and methods have been established in the field of crop seed identification. Studies on rice seed speciation analysis methods indicate that the current detection methods in computer vision mainly analyze 2D information and that the use of 3D models is lacking. This paper proposes a 3D rice seed reconstruction system which can be used to measure the morphology of rice seed, with more accurate shape measure parameters and more comprehensive appearance characteristics and defect expression. In this paper, a new crop seed reconstruction system that supports fast and accurate recognition was designed to build a 3D surface morphology. The depth-from-focus (DFF) method was applied in the analysis of crop surface morphology. Image sequences were acquired by using a specific vision device through setting different distances between the camera lens and the rice seed. High-pass filtering was used to extract pixels and analyze strength value changes in the frequency domain. The second-order differential was employed to strengthen the value in the frequency domain by using the improved Laplacian operator. The threshold statistical analysis was conducted in pixel windows, by which each pixel generated a value which showed the focusing condition. The focusing measure of the image sequence effectively determined the estimated depth value of a pixel, and a focusing pixel stack could be defined based on these values. Using the characteristics of the Gaussian distribution of the focal depth estimation value, the Gaussian interpolation was calculated to obtain a more precise surface morphology depth value. As a result, a depth image collected based on the estimated depth value of the pixel was developed. Finally, through depth image smoothing and edge pixel processing, a 3D point cloud could be produced. Thus, a rice seed reconstruction system which can be used in rice seed identification and recognition was designed. This novel system supports three main patterns, namely, shape, texture, and 3D recognition. Through further calculations, the surface morphology characteristics of seed are obtained. The new 3D surface morphology reconstruction system can effectively overcome the deficiencies of traditional seed speciation analysis methods and can be served as an important reference for researchers. Finally, the BP neural network model was constructed to support the variety identification. Suitable neural network algorithm was selected for five different sorts of rice seed, and the final identification rate is 90%. The research can provide a reference for study of three-dimension shape and texture in automation crops variety identification field. [ABSTRACT FROM AUTHOR]

In order to achieve automatic recognition of hair impurities in the canned mushroom production process, an image recognition method based on the hair centerline feature is proposed in this paper. The proposed approach recognizes the hair centerline based on the Hessian matrices with Gaussian filtering. Under ideal lighting conditions and general lighting conditions, Hessian matrices are obtained after the original image is filtered 6 times using one-dimensional Gaussian derivative filters. The centerline pixels of the hairs and the shadows are obtained by calculating the eigenvalues of every pixel's Hessian matrix. After employing non-maximum suppression, 8-neighbour linkage, and parallel edge analysis, the hairs' centerlines are recognized, and the confusing shadow centerlines are eliminated at the same time. Lighting conditions have a great impact on recognizing hairs in canned mushrooms because (a) the canned mushroom contains water and (b) its surface is not flat. Ideal lighting can eliminate the reflection caused by water and reduce the shadow caused by the shape of the mushroom. Under general lighting conditions, reflection and shadow interference are severe. Therefore it is very difficult to recognize hairs on canned mushroom under general lighting conditions. Unfortunately, general lighting conditions are most widely used in industrial applications. In our experiment, the ideal lighting condition is realized using a HDL-160W-type diffuse light source. The general lighting source consists of two 36W/840 fluorescent strip lamps, which form low-angle, fluorescent strip-light illumination. The hardware of the experimental system mainly consists of a computer, a digital monochrome CCD camera, and a 1394 interface card. The monochrome CCD camera is HD-SV2000FM with a resolution of 2 million pixels (1628 pixelx1236 pixel). The lens is a 12-36 mm, 1:2.8, 2/3'' Computar lens. The vertical distance between the lens and the subject is 12 cm and the FOV is 5.50 cmx4.15 cm. Under 2 different lighting conditions, we extract 4 images with representative hair shapes to be the original images in this paper, viz. common hair (black, 70 um in diameter) under ideal lighting conditions, common hair under general lighting conditions, special hair (yellow, 50 um in diameter) under ideal lighting conditions, and special hair under general lighting conditions. There are hairs of simple shape and complex shape in each original image. Simple shapes consist of a straight line and a circular curve. Complex shapes consist of two cross laps and two cross hyperbolic characteristics. Under the ideal lighting condition and the general lighting condition, this paper compares the hair centerline feature extraction effects of 5 methods. The recognition results of the 5 methods are also compared. The 5 methods are parallel edges, steerable quadrature filter pair, optimized 4-order steerable filter, linear combination of shifted Gaussian kernels, and the Hessian matrices with Gaussian filtering proposed in this paper. From the feature extraction effects and the recognition results, it can be concluded that under general lighting conditions, the former 4 methods are infeasible and only the proposed method is feasible. Under general lighting conditions, using different characteristic thresholds, we obtain the receiver operating characteristic (ROC) curves for each of the five methods' hair-recognition results. By comparing the five methods' ROC curves, it can be easily seen that: 1) our method far outperforms the other four methods; 2) using the same method, common hairs are easier to recognize, and special hairs are difficult to recognize. Under general lighting conditions, the accuracy rates of our recognition results for complex-shaped common hairs and complex-shaped special hairs are 0.98673 and 0.97007 respectively. These results show that the proposed method also performs well in recognizing complex special hairs under poor lighting conditions. The proposed approach is able to recognize accurately hairs of various types and various shapes in canned mushrooms in either ideal or general lighting conditions. This suggests that the proposed approach can be applied in automatic impurity image recognition for industrial canned mushroom production. [ABSTRACT FROM AUTHOR]

Body shape parts acquisition and recognition from moving cattle are difficult to realize automatically especially with the similar color and texture frames or images. Usually, the cattle’s abnormal behavior is hardly detected by human because abnormal behavior detection needs continuous observation, but we can solve this problem with our method. In this paper, we use the Microsoft Company's smart vision sensor (named Kinect) to collect the information of 2 modals from cattle movement (the depth information and RGB information, DRGB). The depth information can be acquired by the infrared sensor and the depth value means the distance between the object and the sensor. The RGB information can be acquired by the normal camera. Based on the color modal information, we propose a randomized nearest neighbor pixel comparison (RNNPC) method to snatch at continuous action frames without static frames. By comparing the distinction of pixels between 2 adjacent frames, we can judge whether there is obvious or micro actions appearing in these 2 neighboring images. And if the distinction value is more than the threshold, the camera would record the cattle’s continuous movements or actions automatically. Based on the depth modal information, we can calculate the mean value of depth which is obtained in the dynamic region by the RNNPC method. And with the mean value of depth, we can filter and transform the invalid pixels into dark pixels in the continuous frames. Meanwhile, the intact shape of the cattle in the original picture is preserved by our method (save the pixels with no change). From the results of filtration, we can see that the original image background area is reduced, and the method (we use SelectiveSearch in this paper) also directly reduces the number of candidate regions compared with the conventional recognition algorithm. According to the training samples scale, the network performance and the characteristic of the key parts of cattle from continuous frames, we adjust the parameters of the deep convolutional neural network, and we set the iterating upper limit and finish the network training when the number of iterating reaches this limit. Finally, with the trained network, we can realize the recognition of key parts from moving cattle in the processed continuous frames. The experimental results show that RNNPC method can save 72% of storage space, and the ratio of valid data for the other 38% of continuous frames can reach 94%. By filtering the invalid pixels, the invalid background information or objects of the original images can be almost removed, and the number of candidate objects generated by the conventional object recognition algorithm can be reduced by an order of magnitude. Compared to the original image, 90% candidate regions are reduced with SelectiveSearch algorithm on DRGB image. By adjusting network parameters, we can improve the convergence speed of deep convolutional neural network in training the samples with similar color and texture, and the single training iteration time does not significantly increase. The average classification accuracy of the net can reach 75.88%, and the image processing rate is 4.32 FPS, and under the same condition, the effect is better than that by the original Fast RCNN. By using the method mentioned above, we can realize body shape parts acquisition and recognition from moving cattle. [ABSTRACT FROM AUTHOR]

Copyright of Transactions of the Chinese Society of Agricultural Engineering is the property of Chinese Society of Agricultural Engineering and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Transactions of the Chinese Society of Agricultural Engineering is the property of Chinese Society of Agricultural Engineering and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Transactions of the Chinese Society of Agricultural Engineering is the property of Chinese Society of Agricultural Engineering and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Transactions of the Chinese Society of Agricultural Engineering is the property of Chinese Society of Agricultural Engineering and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Transactions of the Chinese Society of Agricultural Engineering is the property of Chinese Society of Agricultural Engineering and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Transactions of the Chinese Society of Agricultural Engineering is the property of Chinese Society of Agricultural Engineering and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Automatic classification of insect species in field crops such as corn, soybean, wheat, and canola is more difficult than the generic object classification because of complex background in filed and high appearance similarity among insect species. In this paper, we propose an insect recognition system on the basis of advanced sparse coding and spatial pyramid model. We firstly learn features from a large amount of unlabeled insect image patches to construct an over-complete dictionary. The sparse coding of insect image patches is obtained by encoding over the dictionary. To enhance discriminative ability of the sparse coding, we then apply multiple scales of filters coupled with different spaces. Finally, the multiple space features of sparse coding are seamlessly embedded into a multi-kernel framework for robust classification. Traditionally, insect recognition has mainly relied on manual identification by expert entomologists. However, for laymen without a thorough understanding of the terminology of insect taxonomy and morphological characteristics, it is hard to discriminate insect categories at the species level. Therefore, effective identification of insects is a key issue that needs to be well addressed. To improve the recognition accuracy, we develop an insect recognition system using advanced sparse coding, spatial pyramid model and multiple-kernel learning techniques. Different from traditional feature representation, a novel feature representation that is multiple-space sparse coding of insect objects is proposed by this work. The work flow of our method can be decomposed into 2 stages. The first stage focuses on image or insect object representation. At this stage, the features of insect images are extracted using advanced sparse coding and spatial pyramid model. The second stage, which deals with effective fusion of multiple insect-categorization features, constructs a kernel-level fusion classifier using all the sparse coding features. At the first stage, for an insect image given, the features of insect images are extracted firstly. The features are then represented as a linear combination of the corresponding training feature dictionary. Then, a multiple-space sparse coding with spatial pyramid model is used to represent insect image in a joint sparse way over all the features. In this process, for the object image, an over-complete dictionary with unlabeled insect images is learned first. Then, the local image patches of the insect object are represented by their sparse codes with the training dictionary. Despite the fact that appearance is modeled using local patches, the global structure information is necessary for accurate insect identification. Consequently, insect appearance is represented by concatenating the location and orientation sparse-coding of all image patches. To obtain the more compact representations of insect images, we use spatial pyramid model at multi-scale levels, which achieves better robustness to noise and clutter, and thus better copes with severe variations in the pose, scale or rotation. In this paper, we use the 3-scale level pyramid to represent insect image. At the top level of the pyramid, there are 4 image patches which represent the whole image. Each image block size is 50×50. The middle level contains 16 equal size non-overlapping image patches, for which each image block size is 25×25; and the bottom level has 64 image patches for which each image block size is 12×12. Then, the local image features of the 3-level pyramid are combined to represent the insect appearance. The larger scale level provides better geometric features when the classifying insects undergo large appearance variations, while smaller scale level obtains finer features. Finally，the features from fine to coarse levels across different scales are concatenated together to generate the final feature representation of insect images for insect classification. At the second stage, the multiple-kernel learning approach is adopted to combine multiple-space sparse coding. As different features of insect images contribute differently to the classification of insect species, the multiple-space sparse coding technique can combine multiple features of insect species to enhance the recognition performance. Given positive and negative insect samples, the features are extracted. Local image patches of the samples are then represented by multiple-space sparse coding using the corresponding training dictionary. Finally, the multiple-kernel learning classifier is constructed by learning the multiple-space sparse coding of the negative and positive samples for insect categorization and recognition. To meet the need of practical insect image identification, we collected insect images covering various species across several common field crops including corn, soybean, wheat, and canola. Samples of 35 common pest species found in field crops were collected, such as Pieris rapae (Linnaeus), and Leptocorisa acuta (Thunberg). Experimental results showed that our proposed method performed well on the classification of insect species, and outperformed the state-of-the-art methods of the generic insect recognition. Our method improved the recognition rate by more than 9% compared to other methods for the same data sets. In addition, the proposed method had also a good performance and enhanced the average recognition accuracy by 14.1% for the different data sets. [ABSTRACT FROM AUTHOR]

Accurate identification of diseases and insect pests has been the key link to the yield, quality and safety of crops in modern agriculture. However, the same category of diseases and insect pests showed obvious differences in intra-class representation and slight similarities in inter-class representation of various diseases, due to the influence of environmental conditions, disease cycle and damaged tissues. At present, the traditional deep transfer learning methods have been difficult to cope with large-scale, multi-class fine-grained identification of pests and diseases, particularly unsuitable to the practice in complicated scenes. This paper aims to apply multiple source cameras in agricultural Internet of Things (IoT) and different intelligent precision equipment, including picking robots and smart phones, to capture high resolution 122,000 images of insect pests and diseases, covering a total of 181 fine-grained categories, including 49 types pests and 77 types diseases on different plant parts of different crops. A fine-grained recognition model was then proposed for pests and diseases based on the Multi-Stream Gaussian Probability Fusion Network (MPFN). In detail, a data-augmented method was first employed to enlarge the dataset, and then to pretrain the basic VGG19 and ResNet networks on high-quality images, in order to learn common and domain knowledge, as well fine-tuning with professional skill. Next, the refined multiple deep learning networks, including Fast-MPN and NTS-Net with transfer learning, were applied to design a multi-stream feature extractor, utilizing the mixture-granularity information to exploit high-dimensionality features, thereby to distinguish interclass discrepancy and tolerate intra-class variances. Finally, an integrated optimization was developed combining the NetVLAD feature aggregation layer with the gaussian probability fusion layer, in order to fuse various components model with Gaussian distribution as a unified probability representation for the ultimate fine-grained recognition. The input of this module was the various features of multi-models, whereas, the output was the fused classification probability. The end-to-end implementation of framework included an inner loop about the expectation maximization algorithm within an outer loop with the gradient back-propagation optimization of the whole network, indicating multi-model fusion information complementation and confidence for the overall model. The experimental results demonstrated that the MPFN model presented the excellent performance in the average recognition accuracy rate of 93.18% for a total of 181 classes of pests and diseases, indicating 5.6 percentage points better than that of the coarse-grained and fine-grained deep learning methods. In terms of test time, the average processing time of the MPFN was 61ms, indicating the basic needs of fine-grained image recognition of pests and diseases at the terminals of the IoT and intelligent equipment. In the contradistinctive analysis, training loss curves and various sub-categories identification showed that the feature aggregation fusion and Gaussian probability fusion can greatly enhance the efficiency of model with the training speed, recognition accuracy and generalized robustness for fine-grained identification of crop pests and diseases in practical scenarios. Therefore, this work can provide a technical application reference for the intelligent recognition of pests and diseases in agricultural production. In the follow-up study, more images will be taken under natural conditions to develop a more robust MPFN model deployed on actual IoTs or equipment for the pre-warning, prevention, and control of crop pests and diseases in modern agriculture. [ABSTRACT FROM AUTHOR]

Copyright of Transactions of the Chinese Society of Agricultural Engineering is the property of Chinese Society of Agricultural Engineering and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Transactions of the Chinese Society of Agricultural Engineering is the property of Chinese Society of Agricultural Engineering and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Crop disease is one of the main disasters for Chinese agriculture and it seriously affects the yield and quality of crops, and causes economic losses to farmers. Early detection and prevention of crop diseases is critical to control the diseases, improve crop yields, reduce the economic losses and control pesticide pollution. Therefore, the research of recognition methods for crop diseases is necessary. In this study, a disease recognition method of cucumber disease, based on leaf image and environmental information, is proposed. In this method, the cucumber disease features, including environmental classifying features and disease leaf classifying features, were extracted by image processing and statistical analysis methods. The classifying features were then selected by SAS discriminant analysis, and the cucumber diseases were identified by using the rule of maximum membership degree. The cucumber leaves and their environmental information of Downy mildew, Brown Speck, and Anthracnose were collected for disease recognition. The diseased cucumber leaf images were processed by using a series of image pre-processing methods, such as image transforming, smoothing, and segmentation. White was chosen as the background of the diseased leaf images, a median filter was applied to effectively remove the disturbance of noise, and the color image segmentation method, based on statistical pattern recognition, was applied to separate the disease spot images from the diseased leaf images. The five features of environmental information were extracted, and the 35 statistical eigen vectors of color, shape, and texture of the diseased leaf images could be extracted by statistical analysis. Then, 40 disease union classifying features were obtained. Ten strong classifying features were then selected by the SAS discriminant analysis method. The feature vectors of the clustering center were then computed. Finally, three kinds of cucumber diseases were recognized by the maximum membership degree. The disease recognition method proposed in this study differs essentially from the traditional ones. Traditional methods only take into account features extracted from diseased leaf images, which makes the recognition rate of traditional methods low because the diseased leaf image is quite complex. However, the proposed disease recognition method contains not only the leaf image information, but also the environmental information, which improves the robustness and recognition rate of the proposed method. The recognition results of three kinds of cucumber diseases by the proposed method were more than 93 percent. The experiment results show that cucumber diseased leaf images obtained under different conditions were effectively recognized by the integrated application of image processing technology, analysis of image texture, color and figure characteristics, and analysis of the disease environmental information, etc. It has provided a technical basis and support for the automatic recognition of crop diseases with diseased leaf images and environmental information obtained in the field. The analysis and experimental results in this paper demonstrated that the crop disease recognition method is feasible by computer vision, statistics, and comprehensive utilization technology of spot color, texture, shape information, and crop environment information. As there are many factors affecting crop diseases, various diseases in different periods will show different symptoms. Therefore, how to use the computer vision technology and the disease environmental information to build a powerful and practical crop disease recognition method still needs further study. [ABSTRACT FROM AUTHOR]

Image detection of crops' disease spots can be helpful for real-time monitoring and diagnosis of diseases and pests. Plant Disease Images were diverse and complex. Many methods have been studied to detect and recognize disease spots. In order to realize real-time and accurate detection, the detection algorithm of disease spots must be both steady and simplified. This paper proposed a rapid detection method of disease spots based on edge detection and improved Hough transform. Edge detection needed a gray image. Disease images were calculated into feature images by R, G, B, or difference components, according to disease spots in the images. The edges were first detected by a Canny operator. A mended template was designed to repair the broken edge into the close edge. The close edge was picked by morphological operation. The close edges contained disease spots, also other background edges. Farther filtration was needed to obtain the close edge of disease spots. Edge detection and repair was the base for the Hough transform. As the disease spots grew and developed approximately circular, round Hough transform was improved to detect the edge of the disease spot. Direct application of Hough transform to the disease spots edges would result in great calculation and inaccuracy fitting, as there might be a large gap between the radiuses of the disease spots. The edges of disease spots were coded and divided into quasi-circular disease spots and irregular disease spots, according to the ratio between the long radius and the short radius of each spot. The quasi-circular disease spots were fitted by a single circle, the irregular disease spots were fitted by multiple circles. The transform parameters of each edge were determined according to the disease spots edge. Hough transform were carried out on each edge in independent space. So that error accumulation could be avoided. The largest intersected Hough circle would be kept. The edge of the disease spot was fitted by the circular and the center of the disease spot so it could be located. 90 crop diseases images were used to detect disease spots. These images which included leaf spots and fruit spots were randomly pictured in the field. The detection algorithm was programmed to carry out the detection. The results showed that the detected circles covered most of the disease spots; centers of the disease spots were also well located. Statistics results showed that the circle fitting accuracy was 87.01%, the central error was 4.44%. The running times of the Hough transform for each image were less than 5 s. The results demonstrated that this detection algorithm could detect disease spots both quickly and accurately. Meanwhile, the edges of the disease spots could be effectively identified and better recognized. [ABSTRACT FROM AUTHOR]

Cotton (Gossypium hirsutum) is one of the most important cash crops in China, The timely and effective removal of weeds in cotton seedling stage is an important measure to ensure high and stable yield of cotton. Nowadays, weed recognition based on machine vision is widely used. The fast and effective recognition of crop and weed in the field under natural illumination is one of the key technologies for the development of intelligent mechanization weeding pattern. In the one hand, cotton and weeds have similar color feature in the field. Feature presentation of the natural property of target is difficult to be obtained by the hand-engineered feature extractor. The spatial consistency of the obtained features is not good, and the real-time performance of recognition system is reduced for the complex feature extraction algorithm. On the other hand, the effect of image preprocessing has important influence on recognition results. In order to solve the main problems in the current research, we explored the way to improve the recognition accuracy, stability and real-time performance, and a recognition method of crop and weed based on Faster R-CNN．In this paper, cotton seedling at 2-5 leaves stages and weeding during the same stage were used as research objects under natural illumination. Weed identification from digital images taken under natural illumination at field level is still challenging in agricultural image processing applications, though a lot of research has been conducted related to this topic. To address this problem, images including cottons and weeds were taken vertically from top to bottom. A method based on Faster R-CNN convolutional neural network was proposed to identify weeds from cotton plants more accurately and quickly. The residual network was used to extract image features, with ReLU as the activation function and Max-pooling as the down-sampling method. In the region of proposal network, feature pyramid network was introduced to generate target candidate frame, and Softmax regression classifier was utilized to optimize the CNN network. The proposed methodology was implemented on 200 digital images taken under natural illumination. The experimental results demonstrated that, the average accuracy of weed identification reached 95.5%, and the average time for individual weed plant identification was 1.51 s, which was reduced to 0.09 s by using GPU. To test the efficiency of the proposed methodology, YOLOv3 method was also carried out on the same training and test datasets. The weed identification results were assessed by mean average precision and average precision. The experimental results showed that better performance was achieved by using our proposed methodology, and better identification accuracy was reached as well. This indicated that the proposed method had a good effect on weed detection under natural illumination, and it will greatly promote the development of precise weed control. [ABSTRACT FROM AUTHOR]

The real-time and accurate recognition for the organs in tomato is crucial to achieve automated agricultural production, such as automatic harvesting and targeted drug application. Due to the difference of fruit maturity and flower age, and the color of tomato organs varying frequently during the growth period, it is very difficult for the simultaneous detection of different tomato organs with the traditional image segmentation method based on color space. In the meanwhile, because of the non-real-time nature of filters such as SIFT (scale-invariant feature transform) and Haar-like, they cannot be directly applied to the real-time detection of plant organs. The convolution neural networks (CNNs) can automatically extract the low-level image feature and high-level semantics of image, and they have real-time performance on GPU (graphics processing unit) devices. Therefore, inspired by Faster R-CNN, especially YOLOv2, in this paper, we proposed a real-time recognition method of main organs in tomato based on CNNs, and designed a corresponding recognition network model which can predict the object boundary and type only using feature map. In greenhouse, the images of various forms of tomato flower, fruit and stem organs were collected, and the image data set of tomato organs was constructed according to the influence of illumination of the image which was considered during the acquisition process. For the sake of screening the underlying network structure for recognition network, the performance and applicability of several typical CNN-based classification networks were analyzed based on the criterion of model size, statistical separability, classification performance, and computation speed. Inspired by the advantages of these typical networks, a channel wise group convolutional (CWGC) block and a corresponding classification network (CWGCNet) were designed. A sample extension training method was presented to further improve the feature extraction ability of these classification networks. The CWGCNet, Darknet-19 and Inception v2 were selected as candidate infrastructure for recognition network. Subsequently, a CWGC block with dropout layer and 3 full convolution layers were respectively attached to the infrastructure to form the overall recognition architecture. Based on the Microsoft Cognitive Toolkit (CNTK), all CNN-based classification networks and recognition networks were implemented by using of Python, and the relevant experiment was performed on a computer equipped with a Tesla K40c GPU. The results show that, compared with the typical CNN-based classification networks, CWGCNet combines high feature statistical separability and real-time performance. On tomato organ image dataset, using Caltech256 to perform sample extension training can significantly improve the feature extraction ability of the classification networks. Compared with the exponential function, the nonlinear scaling factor in the Sigmoid form makes the recognition networks easier to train. In contrast to the 3 full convolution layers, using CWGC block with dropout as an additional convolution layer to the recognition network CNN infrastructure can dramatically reduce the size of the model, while significantly improve the network recall rate, recognition speed and average precision (AP). The convolution part of CWGCNet and the CWGC block with dropout are used as the final structure of the recognition network. The final recognition network can identify the different maturity and different forms of tomato organs, which gets the AP of 96.52%, 97.85% and 82.62% respectively for flower, fruit and stem. The growth stage and maturity of tomato organs have a certain influence on the recognition accuracy, and especially the flowering flower, full maturity fruit and lower stem have higher recognition accuracy. The final network can recall different forms of tomato organs, and the recall rates of flower, fruit and stem can reach 77.39%, 69.33% and 64.23% separately. And the recognition speed of the final network is 62 fps. Compared with YOLOv2, the recall rate can be improved by 14.03 percentage points, and AP can be improved by 2.51 percentage points. [ABSTRACT FROM AUTHOR]

Copyright of Transactions of the Chinese Society of Agricultural Engineering is the property of Chinese Society of Agricultural Engineering and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Vision system is one of the most important parts of apple picking robot, which, to some extent, determines the quality and the speed of picking task implemented by apple picking robot. In this review, we enumerated the existing apple picking robots. Some information, such as type of visual sensors, hardware of vision system, success rate of harvesting and run time, was illustrated in details. Meanwhile, on the basis of discussing the vision system of apple picking robot, we focused on summarizing hardware structure of existing vision systems and apple image segmentation methods as well as apple recognition and localization methods applied in vision systems. The vision system of apple picking robot mainly includes machine vision system, laser vision system, three-dimensional vision system and vision system formed by machine vision and other vision system. And the machine vision system can be classified into 3 types according to the number of image sensors used, that is, monocular vision system, binocular vision system and multi vision system. The recognition and localization of apple target is the first step of the implementation of the picking task for picking robot. The currently used apple segmentation and recognition methods include threshold segmentation algorithm, chromatic aberration based algorithm, K-mean clustering algorithm, region growing algorithm, segmentation method combining 2 or more algorithms, and so on. There are 4 methods that are commonly used in the localization of apple target. They are the methods based on centroid, fitting circle, symmetry axes, and three-dimensional coordinates, respectively. In natural scene, the recognition and localization of apple target may be affected by many factors. Hence, recognition and localization of apple target under different conditions, such as color nonuniformity, different illumination, shadow on the surface, oscillation, overlapping and occlusion, was reviewed and analyzed. Among all these conditions, occlusion can be regarded as the most serious factor. The condition of occlusion can be roughly divided into 4 kinds, i.e. apple target blocked by other apple, by branches, by leaves and by branches, leaves and other apple simultaneously. As for apple targets blocked by branches, one apple may be separated by branches, thus causing that an apple may be recognized as several apples. For the apple targets blocked by leaves, the symmetry of apple can be utilized to localize apple targets. The apple targets blocked by other apple can be considered as overlapping. There are overlapped apples with series connection, parallel connection, and blend connection. Because of the complexity of overlapping, the recognition and localization of apple targets blocked by other apple target is a little more difficult. In addition, the detection of obstacles like tree trunk and branches in apple orchard is important for apple picking robots to avoid obstacles, and thus obstacles detection methods were summarized in this review. In the process of target recognition and localization, binocular vision technology was commonly used in vision system. The key point of binocular vision technology is stereo matching. Therefore, stereo matching was then reviewed, and the image matching methods in existence can generally be divided into 2 categories, i.e. region-based image matching method and feature-based image matching method. What's more, the problems exist in recognition and localization methods used in the vision system of apple picking robot, including accuracy, effectiveness, character of real-time and universal applicability, were analyzed. Further study will concentrate on optimizing the structure of vision system, optimizing the intelligent algorithms used in the vision system, improving the real-time capability, recognizing and locating apple targets when the apple targets and vision systems are influenced by oscillation, and improving cost performance. The paper has summarized and analyzed vision system of apple picking robot comprehensively, which can provide reference for future research. [ABSTRACT FROM AUTHOR]

Weeds is one of the main harmful factors to the yield and quality of wheat and other main crops during seedling stage. Image processing technology is often used in weed recognition, but the method mainly cares about the weeds between different rows, which is always inefficient and wasteful for unmanned aerial vehicle (UAV) and machine spraying ways. In order to overcome the limitations above, this paper proposes a regional weed identification method, which takes advantage of properties of shearlets. Shearlets have attracted much attention in the field of image recognition because of its good sensitivity and fast computation in texture recognition. Meanwhile, it is a multi-scale analysis method with the characteristic of direction independence. Through the comparison of the regional images of the wheat and weed, it shows that the texture of the weed leaves is more complex while the wheat leaves are relatively regular. So we first choose 8 images including 4 wheat images and 4 weed images. Then we obtain shearlet transform coefficient (STC) at diverse scales and directions according to the different texture characteristics of wheat and weeds. In the STC images of different scales, the brightness from black to white represents different coefficient value. Moreover, the complexity of bright regional distribution represents the textural complexity, which can be used to distinguish wheat and weeds. Shearlets have self-adaptability because of different directions on these scales, so that obvious textural features in images taken from different angles can be detected. In our research, we take the self-adaptability of shearlets and the differences of STC images into account, and we choose the STC in the second scale of vertical cone to distinguish wheat weeds as experimental object. The result shows that the STC mean of wheat in the second scale is lower than that of weeds. Additionally, the fluctuation of STC mean of wheat is smaller than that of weeds. This study chooses 16 wheat images and 16 weeds images, aiming to distinguish weed and wheat more intuitively; we take a further statistical analysis on the mean and variance of coefficient matrixes of shearlets in the second scale of vertical cone. After normalization treatment, the distinction mean values and mean square error between wheat seedling and weeds are about 0.07 and 0.08 respectively. We randomly select 13 pictures of weeds and wheat seedling, and the recognition accuracy is 69.2%. The experimental results of contrast experiment show that the shearlet-transform method performs better than gray level co-occurrence matrix (GLCM) method to distinguish wheat seedling and weeds. We can get an explanation for the experimental results from the different theory of the shearlet-transform and GLCM. The theory of shearlet-transform shows that it can get different directions information adaptively. On the contrary, GLCM can only get the directions assigned, so the number of directions for image processing can't be changed. In addition, the method of splitting blocks of larger image gathered by UAV is used to realize the effective identification of non-wheat region. From the experimental results, we can see that the difference between wheat and weeds is based on effective shearlet-transform, and we can generalize our method to other image classification based on textural features. Furthermore, this method performs with high flexibility and stability and it has the potential for herbicide spraying in the field. [ABSTRACT FROM AUTHOR]

In order to overcome the problems of low and non-adjustable resolution existing in wireless images sensor nodes applied in agricultural images acquisition at present, a wireless image sensor node with real-time adjustable resolution was designed and realized in this paper. The node was composed of an image acquisition module, a processor module, a wireless communication module and a power module. The node needed to not only capture, compress, and transmit image data, but also perform multiple task schedules and network protocols, so a powerful ARM (advanced RISC machines) processor S3C6410 was chosen as the processor module of the node. Considering the cost and power consumption, a CMOS (complementary metal oxide semiconductor) type image sensor chip was chosen to design the image acquisition module of the node. The design of image acquisition module included image sensor PCB (printed circuit board) design, chip pin interface design, appropriate lens selection, and development of the sensor chip driver. In order to monitor a larger area of the crop and remotely transmit crop images, the node integrated a WiFi (wireless fidelity) or 4G (the 4th generation) module. A solar power supply system was designed to make the node work stably in the field for a long time. To ensure the stability and reliability of the node, the powerful embedded Linux operating system was employed as the software development platform, and a modular designing method was adopted to program the software system of the node in C/C++ language based on this platform. In order to realize real-time adjustment of the resolution, an algorithm of resolution real-time adjustment based on driver layer and application layer collaboration and multi-thread concurrence was proposed, and all the functions of resolution real-time adjustment of image acquisition, image compression and image transmission were realized in the application layer of software system. In order to verify the performance of the node designed, a series of tests were conducted in Cencun experimental base of South China Agricultural University and Guangdong Dongsheng Farm (Panyu) from April to September in 2016. In the tests, 7 nodes were deployed in the farmland to form an acquisition and transmission network based on WiFi and 4G technologies. To ensure the nodes work stably in the field, where the climate was changeable and the infrastructure was absent, the nodes were encapsulated in a waterproof spherical shield, and a solar panel plus rechargeable batteries was used to supply power for them. The nodes were tested from the aspects of multi-resolution capability, real-time adjustment capability, image acquisition and transmission performance (transmission time and packet loss rate), and the availability of node energy. The test results indicated that the node had 7 different resolutions and its highest resolution was up to a pixel of 5 M, and more importantly, it could adjust its resolution in real time under the control of a remote user, then capture images with different resolution, and finally transmit them to the remote server. The time consumed to capture, compress and transmit 4 images with different resolution of 640×480, 1280×1024, 2048×1536, and 2592×1944 was 4.67, 8.77, 15.38 and 22.74 s respectively, and the average transmission packet loss ratio of 4 images was less than 1%. The tests validate the node designed in this work can capture crop images with different resolution in real time and transmit them remotely, and satisfy the requirement of different users for different crop image precision. [ABSTRACT FROM AUTHOR]

Copyright of Transactions of the Chinese Society of Agricultural Engineering is the property of Chinese Society of Agricultural Engineering and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Crop diseases often seriously affect both the quality and quantity of agricultural products and cause economic losses to farmers. How to accurately and quickly recognize the crop disease information is an important problem in preventing and controlling crop diseases. Crop disease recognition by crop leaf symptoms is a basic method of attempting to address this problem. Studies show that relying on pure naked-eye observing of the leaf symptoms by experts to detect the crop diseases can be prohibitively expensive, especially in developing countries. Automatic detection of crop diseases is an essential research topic, as it may prove benefits in monitoring large fields of crops, and thus automatically detect the symptoms of diseases as soon as they appear on crop leaves. In a research study of identifying and diagnosing crop diseases, the pattern of the disease is important part. Leaf spots are considered the important units indicating the existence of disease and regarded as indicators of crop diseases. A technique to detect the disease spot is needed. It is important to select a threshold of gray level for extracting the disease spot from the crop leaf. In order to classify disease leaf sample categories, a set of spot features for the classification and detection of the different disease leaves are required. The disease leaf images of the crop would be processed by using a series of image pre-processing methods, such as image transforming, image smoothing, and image segmentation. In this paper, crop disease leaf spots were segmented by the seeded region growing based region algorithm. Because the crop leaves look differ in many ways, most of classical pattern recognition methods are not effective to extract the disease features and reduce the dimensionality of diseased leaf images. A novel manifold learning algorithm called local discriminant projects (LDP) was proposed and was applied to crop disease recognition. After being projected into a low-dimensional subspace, the data points in the same class were close to each other, whereas the gaps between the data points from different classes became wider than before. In LDP, the class action was introduced to construct the objection function. There was no need to calculate the inverse matrix, so the small sample size problem occurring in traditional linear discriminant analysis was naturedly avoided, and much computational time would be saved by using LDP for dimensionality reduction. After each spot image was reorganized as one-dimensionality vector and its dimensionality was reduced by LDP, the nearest neighbor classifier was adopted to recognize crop disease. The extensive experiments were performed on a real maize disease leaf image database and compared with the traditional disease recognition methods and the supervised subspace learning algorithms in recognition performance. The mean correct classification rate of the proposed method was 94.4%. The proposed method was compared with the classical crop disease recognition methods (ANN, PCA+PNN, and Bayesian) and supervised subspace algorithms (LDE, DNE). The experiment results showed that the proposed method was effective and feasible for crop disease recognition. The preliminary study showed that there is a potential to establish an online field application in crop leaf disease detection based on leaf image processing techniques. [ABSTRACT FROM AUTHOR]

Copyright of Transactions of the Chinese Society of Agricultural Engineering is the property of Chinese Society of Agricultural Engineering and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Transactions of the Chinese Society of Agricultural Engineering is the property of Chinese Society of Agricultural Engineering and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Transactions of the Chinese Society of Agricultural Engineering is the property of Chinese Society of Agricultural Engineering and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Transactions of the Chinese Society of Agricultural Engineering is the property of Chinese Society of Agricultural Engineering and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Transactions of the Chinese Society of Agricultural Engineering is the property of Chinese Society of Agricultural Engineering and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)