Research On The Target Recognition Of The Vision System Of Apple Picking Robot

China is a big apple-growing country in the world,and the picking activities carried out by Apple have high-intensity and intensive basic characteristics.With the continuous adjustment of China's agricultural institutions and the emergence of labor shortages,in order to avoid economic losses caused by failure to pick them in time,it is necessary to research and develop automated picking techniques.At present,because the recognition efficiency in the actual picking environment is not ideal,the working efficiency of the robot is low,and the automatic picking technology is still stagnant in the initial laboratory development stage.The subject analyzes the above-mentioned problems,using natural-growing apples as the analysis object,and using the visual system to conduct research on its identification.The main work is as follows:1.Apple picking site identification technology and experimental platform.The picking environment and identification technology are the main factors affecting the accuracy of picking.Through the analysis of apple growth environment and recognition technology,the appropriate research program is selected to carry out the experiment of this subject.A picking robot experiment platform was built and the software and hardware facilities of the platform were introduced.2.Research and improvement of Apple recognition model algorithm.The identification of the target fruit is the basis of post-image processing.The high efficiency and high precision recognition model can quickly and accurately identify the image containing the target fruit from the video stream,laying a solid foundation for subsequent research.The experiment uses the Decomposable Convolution MobileNet network to replace the VGG16 network in the traditional SSD object recognition algorithm,and builds the SSD_MobileNet algorithm model.Experiments show that the algorithm improves the recognition rate of the model.3.Research on Apple image recognition and segmentation methods.The subject usesdifferent algorithms for segmentation and recognition of apple images in three different forms.The algorithm is verified by experiments,and the proposed HS merge segmentation algorithm has high success rate for image segmentation recognition.4.Target identification and calibration.In this paper,the principle of binocular target recognition and binocular vision are calibrated.The characteristics of the apple's centroid and number in the image are extracted,which lays a foundation for the subsequent research work.According to the algorithm proposed in this paper,the programming design is carried out under the environment of MATLAB,and the algorithm is verified and this paper analyzed the experimental results.The experimental results demonstrate that the algorithm in this paper has better performance in the recognition efficiency and accuracy synthesis.The technical research of the subject has reference significance for other agricultural harvesting technology research.