Research On Control Method Of Apple Picking Robot Based On Object Recognition

China is the largest apple producer in the world.As a new generation of agricultural intelligent equipment,apple picking robots have become the focus of attention in recent years.With the support of Modern Agriculture Project of Jiangsu Province(BE2020406),this paper studies the detection algorithm of image recognition,servo control of robot arm and path planning method,taking robot as the carrier and apple as the object of target recognition.It enhances the diversity of the apple picking robot system to obtain in-depth information and improves the system’s ability to perceive the external environment.The main contents are as follows:First,for the complex orchard environment,in order to improve the robot environmental cognition and adaptability,the overall design of the apple picking robot system.The system mainly includes track movement module,picking arm module,end gripper module,upper machine module and a variety of sensors.According to the requirements of apple picking,the hardware circuit design and software system development of each module are carried out,thus laying the foundation for the realization of the visual recognition,servo control and path planning of the Apple picking robot.Secondly,construct the vision system model of apple picking robot,and study the fast picking visual detection algorithm.In order to realize rapid target detection of apple picking robot in complex environment and overcome the shortcomings of traditional YOLO v5 network,such as complex structure and weak computing performance,a target detection method of apple picking robot based on depth separable convolution YOLO v5 was proposed.After the apple sample image is collected and the experimental data set is generated,the model is trained and tested.The deep separable convolution YOLO v5 network was used to extract features from apple images,which solved the problem of parameter redundancy in the network and improved the recognition speed of the picking robot.The CIo U-Loss function and DIo U-NMS non-maximum suppression method were used to optimize the loss function and improve the positioning accuracy of the robot vision system.The experimental results show that the performance of the improved algorithm is improved,and it can realize the rapid and accurate detection of the picking robot in the complex environment.Thirdly,modeling and analysis of the apple picking manipulator,research on the adaptive sliding mode control method of the picking manipulator based on RBF neural network.Compared with traditional sliding mode controller,the proposed controller can achieve accurate trajectory tracking and better precision.After analyzing the characteristics and problems of RRT algorithm and RT-Connect algorithm,a Fuzzy-RRT algorithm based on the picking manipulator planning method is designed.Simulation results show that the algorithm improves the efficiency and quality of path planning for the arm of the apple picking robot.Finally,an apple picking robot experiment platform was built in the orchard environment through computer simulation and laboratory simulation,and the robot system was completed.Results The feasibility and efficiency of the proposed method were verified by simulation image and data analysis,and the ideal effect level was achieved.