Search

Crab aquaculture is an important component of the freshwater aquaculture industry in China, encompassing an expansive farming area of over 6000 km2 nationwide. Currently, crab farmers rely on manually monitored feeding platforms to count the number and assess the distribution of crabs in the pond. However, this method is inefficient and lacks automation. To address the problem of efficient and rapid detection of crabs via automated systems based on machine vision in low-brightness underwater environments, a two-step color correction and improved dark channel prior underwater image processing approach for crab detection is proposed in this paper. Firstly, the parameters of the dark channel prior are optimized with guided filtering and quadtrees to solve the problems of blurred underwater images and artificial lighting. Then, the gray world assumption, the perfect reflection assumption, and a strong channel to compensate for the weak channel are applied to improve the pixels of red and blue channels, correct the color of the defogged image, optimize the visual effect of the image, and enrich the image information. Finally, ShuffleNetV2 is applied to optimize the target detection model to improve the model detection speed and real-time performance. The experimental results show that the proposed method has a detection rate of 90.78% and an average confidence level of 0.75. Compared with the improved YOLOv5s detection results of the original image, the detection rate of the proposed method is increased by 21.41%, and the average confidence level is increased by 47.06%, which meets a good standard. This approach could effectively build an underwater crab distribution map and provide scientific guidance for crab farming.

The cucumber fruits have the same color with leaves and their shapes are all long and narrow, which is different from other common fruits, such as apples, tomatoes, and strawberries, etc. Therefore, cucumber fruits are more difficult to be detected by machine vision in greenhouses for special color and shape. A pixel-wise instance segmentation method, mask region-based convolutional neural network (Mask RCNN) of an improved version, is proposed to detect cucumber fruits. Resnet-101 is selected as the backbone of Mask RCNN with feature pyramid network (FPN). To improve the detection precision, region proposal network (RPN) in original Mask RCNN is improved. Logical green (LG) operator is designed to filter non-green background and limit the range of anchor boxes. Besides, the scales and aspect ratios of anchor boxes are also adjusted to fit the size and shape of fruits. Improved Mask RCNN has a better performance on test images. The test results are compared with that of original Mask RCNN, Faster RCNN, you only look once (YOLO) V2 and YOLO V3. The F1 score of improved Mask RCNN in test results reaches 89.47%, which is higher than the other methods. The average elapsed time of improved Mask RCNN is 0.3461 s, which is only lower than the original Mask RCNN. Meanwhile, the mean value and standard deviation of location deviation in improved Mask RCNN are 2.10 pixels and 1.73 pixels respectively, which are lower than the other methods.

The skins of most mature apple fruits are incompletely red and also include green and pale yellow color, which increases the difficulty of fruit detection by machine vision. A detection method based on color and shape features is proposed for this kind of apple fruits. Simple linear iterative clustering (SLIC) is adapted to segment images taken in orchards into super-pixel blocks. The color feature extracted from blocks is used to determine candidate regions, which can filter a large proportion of non-fruit blocks and improve detection precision. Next, the histogram of oriented gradient (HOG) is adopted to describe the shape of fruits, which is applied to detect fruits in candidate regions and locate the position of fruits further. The proposed method was tested by images taken under different illuminations. The average values of recall, precision, and F1 reach 89.80%, 95.12%, and 92.38% respectively. The performance of detecting fruits covered at different levels is also tested. The values of the recall are all more than 85%, which indicates that proposed method can detect a great part of covered fruits. Compared with pedestrian detection method and faster region-based convolutional neural network (RCNN), the proposed method has the best performance and higher than faster RCNN slightly. However, the proposed method is not robust to noise and its elapsed time of one image is 1.94 s and less than faster RCNN.

The bearingless induction motor (BIM) system has the characteristics of non-linear, strongly coupled, and highly accurate. To achieve the stable and accurate control of BIM, the methods of direct torque control (DTC) and direct suspension force control (DSFC) based on space vector modulation (SVM) are introduced first. However, the dynamic performance and uncertain anti-jamming robustness of the system need to be further researched. Based on the BIM control system, the active disturbance rejection control (ADRC) technology is selected in this paper. The total disturbance of the system is estimated in real time by using the extended state observer (ESO). According to the estimated value of the ESO, the disturbance can be compensated by the ADRC in time to eliminate the steady-state tracking error to achieve good anti-interference performance. The ADRC-based direct torque and suspension force control system for the BIM is designed and simulated by the MATLAB software. The simulation and experiment results show that the ADRC technology can effectively suppress the impact of load disturbance, reduce the overshoot, and have strong robustness.

In the green pepper harvesting robot, the color of green pepper is similar to that of leaves, which makes it difficult to recognize the green pepper target. In order to solve this problem, a green pepper recognition method based on least-squares support vector machine optimized by the improved particle swarm optimization (IPSO-LSSVM) is proposed in this paper. Firstly, the green pepper images are segmented by K-Means method under the Lab color space, and the segmentation images of the target and background are obtained. The processed green pepper image was divided into training and testing samples. Then, the shape and texture features of green pepper targets are extracted separately from the training sample using the hu invariant moment and Tamura texture feature. Meanwhile, in order to reduce the complexity of data calculations and improve the efficiency, the extracted feature vectors are normalized. The feature vector is used as the input eigenvector of the least-squares support vector machine (LSSVM). The particle swarm optimization algorithm is used to obtain the optimal regularization parameter and the kernel function width. In order to maintain the particle activity, the mutation strategy is introduced to improve the particle swarm optimization algorithm. The experimental results show that the recognition rate of IPSO-LSSVM is higher than that of other methods, and the recognition accuracy is 89.04%. It could meet the requirements of green pepper identification.

Under the complex agricultural operation environment, reliable navigation system is the basic guarantee to realize the agricultural robot automated operation. This study focuses on improving navigation accuracy and control accuracy and conducts related research on autonomous navigation control of agricultural robots. This article discusses the advantages of using strict convergence criteria and combining Sage–Husa adaptive filtering with strong tracking Kalman filtering and then proposes an improved adaptive Kalman filter algorithm. The new algorithm can effectively suppress the filter divergence, improve the dynamic performance of the filter, and ensure its better filtering accuracy and strong adaptive ability to improve navigation accuracy of GPS. Further variable structure switching method is used to prevent proportional integral differential (PID) controller integral saturation phenomenon, which effectively solves the controller over-saturation problem. And combining this method with an improved adaptive filtering algorithm not only can effectively inhibit control interference but also achieve the anti-saturation effect, thereby enhancing the stability and accuracy of the control system. Finally, the simulation and experiment of the new method show that the proposed method greatly improves the ability of the filter to suppress divergence and control precision.

It has been more than 30 years since the French pioneered the research of the apple harvesting robot, and with the joint efforts of scholars all around the world, a variety of prototypes have been developed. However, the existing apple harvesting robot prototype is still in the experimental research stage because of its low harvesting efficiency. With the help of information technology, the related research has ushered in a milestone development, and it is full of opportunities and challenges for apple harvesting robotic researchers. In this article, it briefly introduced the development history, structure, and composition of apple harvesting robots and the operation process, which makes readers have a clear understanding of apple harvesting robot and its harvesting principle. Then systematically summarizing the research results of apple harvesting robots both at domestic and at foreign, we carried out in following three aspects: rapid and accurate recognition and positioning of target fruit, all-weather operation mode, and application of intelligent computing theory in apple harvesting robots, and it analyzes the research progress of apple harvesting robot in detail. The results show that improving the harvesting efficiency is the key and hot spot for the research on apple harvesting robots. Under the impetus of information technology, how to achieve fast and accurate recognition of the fruits of multienvironment and multiple information, obtain reasonable path planning, and further optimization of control strategies are all important research directions.

In order to improve the harvesting efficiency of apple harvesting robot, this article presents an apple recognition method based on pulse coupled neural network and genetic Elman neural network (GA-Elman). Firstly, we use pulse coupled neural network to segment the captured 150 images, respectively, and extract six color features of R, G, B, H, S, and I and 10 shape features of circular variance, density, the ratio of perimeter square to area, and Hu invariant moments of segmented images, and these 16 features are considered as the inputs of Elman neural network. In order to overcome some defects of Elman neural network, such as, trapping local minimum easily and determining the number of hidden neurons difficultly; in this article, genetic algorithm is introduced to optimize it, and new optimization way is designed, that is, the connection weights and number of hidden neurons separate encoding and evolving simultaneously, in the process of structural evolution at the same time the learning of connection weights is completed, and then the operating efficiency and recognition precision of Elman model are improved. In order to get more precision neural network, and avoid the influence of fruit recognition caused by branches or leaves shadow, apple along with branches and leaves is allowed to train. The results of experiments show that compared with the traditional back-propagation, Elman neural network, and other two recognition algorithms of obscured fruit. the genetic Elman neural network algorithm is the optimal method which successful training rate can reach to 100%, recognition rate of overlapping fruit and obscured fruit can reach to 88.67% and 93.64%, respectively, and the total recognition rate reaches to 94.88%.

Chinese river crabs are important aquatic products in China, and the accurate operation of aquatic plants cleaning workboat is an urgent need for solving various problems in the aquaculture process. In order to achieve the accurate navigation positioning, this article introduces the visual-aided navigation system and combines the advantages of particle filter in nonlinear and non-Gaussian systems. Meanwhile, the generalized regression neural network is used to adjust the particle weights so that the samples are closer to the posterior density, thus avoiding the phenomenon of particle degradation and keeping the diversity of particles. In order to improve the network performance, the fruit fly optimization algorithm is introduced to adjust the smoothing factor of transfer function for the generalized regression neural network model layer. On this basis, the location filtering navigation method based on fruit fly optimization algorithm-generalized regression neural network-particle filter is proposed. According to the simulation results, the mean R of root-mean-square error of the proposed fruit fly optimization algorithm-generalized regression neural network- particle filter method decreases by 12.39% and 6.87%, respectively, compared with those of particle filter and generalized regression neural network methods, and the mean T of running time decreases by 16.04% and 9.14%, respectively. From the repeated experiments on the aquatic plants cleaning workboat in crab ponds, the latitude error of the proposed method decreases by 23.45% and 12.68%, respectively, and that the longitude error decreases by 29.11% and 17.65%, respectively, compared with those of particle filter and generalized regression neural network methods. It is proved that our proposed method can effectively improve the navigation positioning accuracy of aquatic plants cleaning workboat.

Aiming at the problem of apple branch obstacle localization in fruit picking process of harvesting robot manipulator, in order to obtain three-dimensional information of the apple branch obstacle, the binocular stereo vision localization method of apple branch obstacle is proposed. Firstly, branch skeleton is extracted by morphological thinning method and then the feature skeleton is obtained after removing the false branch and recovering the occluded branch. After that, the endpoints and bifurcation points regarded as match feature points are extracted from skeleton, and the stereo matching algorithm based on features is adopted. Then, the depth information of branch obstacle is obtained on the basis of triangulation theory. Finally, the experiment results for apple tree branches localization show that the error lies in ±6.2 mm. Moreover, the error is merely ±1.5 mm when the distance between the object and the binocular camera is 1000 mm, which meets with localization accuracy requirements of apple harvesting robot visual system.

In order to resolve the poor real-time performance problem of the normalized cut (Ncut) method in apple vision recognition of picking robots, a fast segmentation method of colour apple images based on the adaptive mean-shift and Ncut methods is proposed in this paper. Firstly, the traditional Ncut method based on pixels is changed into the Ncut method based on regions by the adaptive mean-shift initial segmenting. In this way, the number of peaks and edges in the image is dramatically reduced and the computation speed is improved. Secondly, the image is divided into regional maps by extracting the R-B colour feature, which not only reduces the quantity of regions, but also to some extent overcomes the effect on illumination. On this basis, every region map is expressed by a region point, so the undirected graph of the R-B colour grey-level feature is attained. Finally, regarding the undirected graph as the input of Ncut, we construct the weight matrix W by region points and determine the number of clusters based on the decision-theoretic rough set. The adaptive clustering segmentation can be implemented by an Ncut algorithm. Experimental results show that the maximum segmentation error is 3% and the average recognition time is less than 0.7s, which can meet the requirements of a real-time picking robot.

In this paper, the de-noising problem of night vision images is studied for apple harvesting robots working at night. The wavelet threshold method is applied to the de-noising of night vision images. Due to the fact that the choice of wavelet threshold function restricts the effect of the wavelet threshold method, the fuzzy theory is introduced to construct the fuzzy threshold function. We then propose the de-noising algorithm based on the wavelet fuzzy threshold. This new method can reduce image noise interferences, which is conducive to further image segmentation and recognition. To demonstrate the performance of the proposed method, we conducted simulation experiments and compared the median filtering and the wavelet soft threshold de-noising methods. It is shown that this new method can achieve the highest relative PSNR. Compared with the original images, the median filtering de-noising method and the classical wavelet threshold de-noising method, the relative PSNR increases 24.86%, 13.95%, and 11.38% respectively. We carry out comparisons from various aspects, such as intuitive visual evaluation, objective data evaluation, edge evaluation and artificial light evaluation. The experimental results show that the proposed method has unique advantages for the de-noising of night vision images, which lay the foundation for apple harvesting robots working at night.

In order to improve the performance of the closed-loop Buck converter control system, a compound control scheme based on nonlinear disturbance observer (DO) and nonsingular terminal sliding mode (TSM) was developed to control the Buck converter. The control design includes two steps. First of all, without considering the dynamic and steady-state performances, a baseline terminal sliding mode controller was designed based on the average model of the Buck converter, such that the desired value of output voltage could be tracked. Secondly, a nonlinear DO was designed, which yields an estimated value as the feedforward term to compensate the lumped disturbance. The compound controller was composed of the terminal sliding mode controller as the state feedback and the estimated value as the feedforward term. Simulation analysis and experimental verifications showed that compared with the traditional proportional integral derivative (PID) and terminal sliding mode state feedback control, the proposed compound control method can provide faster convergence performance and higher voltage output quality for the closed-loop system of the Buck converter.

In order to broaden the limited utilization range of wind power and improve the charging and discharging control performance of the storage battery in traditional small wind power generation systems, a wind power system based on a vertical-axis off-grid induction generator is proposed in this paper. The induction generator not only can run in a wide wind speed range but can also assist the vertical-axis wind turbine to realize self-starting at low wind speed. Combined with the maximum power point tracking method, the slip frequency control strategy is employed to regulate the pulse width modulation (PWM) converter to control the output power of the proposed system when the wind speed and load change. The charge and discharge of the storage battery is realized by the segmented current-limiting control strategy by means of an electric power unloader device connected to the DC bus. All these implement a balanced and stable operation of the proposed power generation system. The experimental research on the 5.5 kW prototype system is developed, and the corresponding results verify the correctness and feasibility of the system design and control strategy. Some comparison experiments with a magnetic suspension permanent magnet synchronous generator (PMSG) demonstrate the application prospect of the proposed vertical-axis off-grid induction generator wind power system.