Research On Peper Classification And Location Technology Based On Deep Learning

At present,people ’s demand for healthy diet is getting higher and higher.Color pepper,as a healthy low-calorie food,has 18 kilocalories per 100 grams of calories.Because of its sweet taste and rich vitamin content,it is favored by more and more people.However,at present,the harvest of color pepper mostly adopts the method of manual picking,which is not only timeconsuming and labor-intensive,but also has high production cost,which seriously restricts the development of color pepper industry.Therefore,it is of great significance to study the intelligent picking equipment of color pepper to improve the harvesting efficiency and reduce the production cost.Aiming at the problems of low recognition efficiency and difficult picking point positioning in pepper picking,a pepper recognition and picking point positioning system based on deep learning is established.The related research contents are as follows :(1)Research on neural network modelIn order to determine the most suitable model for pepper training,the current mainstream neural network is trained and compared : In this study,three mainstream neural network models,Faster-RCNN,Mask-RCNN and YOLOv5,are selected for training and the recognition effects of different neural network models are compared.(2)Improved YOLOv5 network modelIn order to increase the ability of the network to extract the information of the pepper fruit and improve the recognition rate of the network to the pepper fruit,the YOLOv5 network model is improved,and the attention mechanism is added : SE attention mechanism,SA attention mechanism and ECA attention mechanism are added;multi-scale target feature fusion :improve the small target detection head,add the CBAM attention module,improve the Bi FPN network,and compare the recognition effect of the improved model on the color pepper.(3)Prediction and location of color pepper fruitUNet network and corner detection algorithm are used to locate the picking points of pepper by binocular camera.The improved clustering algorithm is used to cluster the corners of the pepper,obtain the clustering center of the corners,and predict the position of the pepper picking points.Using binocular camera calibration,the parameters of the device are obtained,and the real coordinates of the picking point are finally calculated.(4)Design of pepper identification and positioning systemUsing Pycharm software and Pyqt5 design tool,the color pepper recognition and positioning system is designed,and the feasibility of the system is tested and analyzed to ensure that the system can operate normally and provide technical support for subsequent color pepper picking and related equipment development.