Search

Tomato spotted wilt virus is a wide-spread plant disease in the world. It can threaten thousands of plants with a persistent and propagative manner. Early disease detection is expected to be able to control the disease spread, to facilitate management practice, and further to guarantee accompanying economic benefits. Hyperspectral imaging, a powerful remote sensing tool, has been widely applied in different science fields, especially in plant science domain. Rich spectral information makes disease detection possible before visible disease symptoms showing up. In the paper, a new hyperspectral analysis proximal sensing method based on generative adversarial nets (GAN) is proposed, named as outlier removal auxiliary classifier generative adversarial nets (OR-AC-GAN). It is an all-in-one method, which integrates the tasks of plant segmentation, spectrum classification and image classification. The model focuses on image pixels, which can effectively visualize potential plant disease positions, and keep experts’ attention on these diseased pixels. Meanwhile, this new model can improve the performances of classic spectrum band selection methods, including the maximum variance principle component analysis (MVPCA), fast density-peak-based clustering, and similarity-based unsupervised band selection. Selecting spectrum wavebands reasonably is an important preprocessing step in spectroscopy/hyperspectral analysis applications, which can reduce the computation time for potential in-field applications, affect the prediction results and make the hyperspectral analysis results explainable. In the experiment, the hyperspectral reflectance imaging system covers the spectral range from 395 nm to 1005 nm. The proprosed model makes use of 83 bands to do the analysis. The plant level classification accuracy gets 96.25% before visible symptoms shows up. The pixel prediction false positive rate in healthy plants gets as low as 1.47%. Combining the OR-AC-GAN with three existing band selection algorithms, the performance of these band selection models can be significantly improved. Among them, MVPCA can leverage only 8 spectrum bands to get the same plant level classification accuracy as OR-AC-GAN, and the pixel prediction false positive rate in healthy plants is 1.57%, which is also comparable to OR-AC-GAN. This new model can be potentially transferred to other plant diseases detection applications. Its property to boost the performance of existing band selection methods can also accelerate the in-field applications of hyperspectral imaging technology.

Maize inspection is an important and time-consuming task in the domain of food engineering. The human-based inspection strategy needs to be brought up to date with the rapid developments in the maize industry. In this paper, an automatic maize-inspection machine is proposed. Our proposed machine integrates several new designs in terms of both hardware and software components. First, a gravity-based dual-side camera design expands the machine's field-of-view to evaluate maize kernels more thoroughly. Second, touching kernels are pre-processed using a new k-means clustering guided-curvature method, which can improve the robustness of our machine. Next, a deep convolutional neural network, which has shown promise for application in image processing, is embedded into the system to evaluate maize kernels. In this work, the ResNet, which is a deep convolutional neural network architecture, was trained by fine-tuning with 1632 images. It achieved a 98.2% prediction accuracy for 408 test images, which outperforms existing approaches.

An automated high-speed strawberry calyx removal machine was designed, built, tested, and evaluated. The three main components of the machine consisted of 1) a strawberry loading and orientation conveyor, 2) a color-based machine vision section for strawberry feature identification, and 3) a synchronized multi-waterjet knife calyx removal system. An overview of the machine as well as a description of the working principles of each component is included. A full-scale model of this machine was evaluated through a 12-week pilot study, during which the machine processed over 70 metric tons of strawberries. Results indicated that the machine could produce an average calyx-free strawberry weight yield of 49.6 percent at a rate of 2270 kg/hr. Furthermore, it was seen that strawberry size had a significant effect on machine efficacy.

University of Maryland, Bio-Imaging and Machine Vision Laboratory, Fischell Department of Bioengineering, College Park, MD, 20740 Corresponding Author: ytao@umd.edu Abstract. The Atlantic Blue Crab is amongst the highest-valued seafood found in the American Eastern Seaboard. Currently, the crab processing industry is highly dependent on manual labor. There is a great potential market for vision guided intelligent machines to automate the meat picking process. Understanding crab morphologies in digital crab images is the first and key step for the intelligent machine. To achieve the goal, a deep learning architecture is integrated into the system, and the fully automatic model can segment crab images into five region of interests in single step with high accuracy and efficiency. Compared to the pervious knuckle detection algorithm proposed by our group, the updated model can accurately locate not only knuckle positions, but also crab legs and crab cores. The average pixel accuracy can get up to 0.9843. From another angle, the computation time of the updated model decreases 50 folds compared to the pervious method. It can further improve the processing speed of the crab machine. The image segmentation results can be used for generating crab cutlines in XY plane, determining starting cutting points in Z plane, and guiding end effectors to harvest crab meat. This work promises a bright further in not only crab industry but also other natural resources meat picking area.

Hyperspectral imaging is a powerful technique in the agriculture field. The subtle changes in spectral reflectance of plants could reflect the invisible symptoms of plant diseases, which promises to be applied for detecting plant disease in the early stage. In practice, compared to the image-level classification, pixel-level (spectrum-level) classification could show tiny defects of plant, and keep experts‘ attentions on the diseased pixels. However, in most cases, because there is no obvious visible difference between healthy spectrum and diseased spectrum, it is still an open research topic to conduct accurate pixel-level classification. Meanwhile, the outliers in the training dataset, the uncertainty of illumination conditions and the imbalance number of healthy pixels and diseased pixels in the training dataset, dramatically increase the difficulties of the problem. Targeting to these issues, based on a well-known deep leaning architecture, generative adversarial nets (GAN), a new deep learning architecture, outlier removal auxiliary classifier generative adversarial nets (OR-AC-GAN) is proposed. It not only integrates the classification task in the model, but also can weaken the side-effects of data outliers and find the intrinsic data features. In the experiment, a wide-spread disease Tomato Spotted Wilt Virus (TSWV) is used for validating the model. For the pixel-level classification, the average false positive rate of plant pixels in the healthy plants is 1.47%. For the plant-level classification, the corresponding sensitivity and specificity values are 92.59% and 100%. Compared to one dimensional convolutional neural network and auxiliary classifier GAN architecture, the proposed OR-AC-GAN model achieves the best results. In theory, the proposed model can be used for hyperspectral image analysis and early detection of any plant diseases.

The Atlantic blue crab is among the highest-valued seafood found in the American Eastern Seaboard. Currently, the crab processing industry is highly dependent on manual labor. However, there is great potential for vision-guided intelligent machines to automate the meat picking process. Studies show that the back-fin knuckles are robust features containing information about a crab’s size, orientation, and the position of the crab’s meat compartments. Our studies also make it clear that detecting the knuckles reliably in images is challenging due to the knuckle’s small size, anomalous shape, and similarity to joints in the legs and claws. An accurate and reliable computer vision algorithm was proposed to detect the crab’s back-fin knuckles in digital images. Convolutional neural networks (CNNs) can localize rough knuckle positions with 97.67% accuracy, transforming a global detection problem into a local detection problem. Compared to the rough localization based on human experience or other machine learning classification methods, the CNN shows the best localization results. In the rough knuckle position, a k-means clustering method is able to further extract the exact knuckle positions based on the back-fin knuckle color features. The exact knuckle position can help us to generate a crab cutline in XY plane using a template matching method. This is a pioneering research project in crab image analysis and offers advanced machine intelligence for automated crab processing.