Research On Detection And Picking Positioning Of Prickly Pears Based On Machine Vision

Rosa roxburghii Tratt has the important functions of enhancing immunity,delaying aging,strengthening stomach,anti-tumor and anti-cancer,and reducing the heavy metal load in the body.At present,the picking of Rosa roxburghii Tratt almost depends on manual work.There is no automatic and intelligent picking device at home and abroad.The cost of picking Rosa roxburghii Tratt is high,the labor intensity is high,and the picking efficiency is low.With the rapid development of national economy,the existing manual picking methods can not meet the needs of the rapid development of Rosa roxburghii Tratt industry.In this paper,image segmentation and machine vision are used to detect and pick Rosa roxburghii fruit in natural environment.This study is of great significance to reduce the labor intensity and improve the efficiency of Rosa roxburghii Tratt.The methods and devices studied in this paper are suitable for the detection,location and recognition of various fruits in natural environment(1)Looking up and understanding the growth characteristics of Rosa roxburghii Tratt,understanding the application of machine vision algorithm in target detection,and designing a research scheme suitable for the detection and location of Rosa roxburghii Tratt fruit in natural environment.(2)The study of Rosa roxburghii fruit recognition based on traditional image segmentation algorithm.By analyzing the characteristics of color and category of Rosa roxburghii fruit image,the methods of color difference component,maximum variance between classes and watershed segmentation are used to detect and locate the fruit under the condition of independence,occlusion and overlap.(3)Research on fruit recognition algorithm of Rosa roxburghii Tratt based on machine vision.By comparing and analyzing the common convolution neural network algorithm,the fast RCNN and Yolo V3 network model which are most suitable for training Rosa roxburghii Tratt fruit data set are selected,and the corresponding algorithm is improved.Through the comparative analysis of the loss function of different iterations in the training process,and the verification analysis of the verification set that does not participate in the training of data set,the best recognition algorithm and recognition model are finally obtained.(4)Vision system platform construction and information acquisition of Rosa roxburghii fruit detection and positioning device.In order to detect the fruit of Rosa roxburghii Tratt in the natural environment,it is necessary to configure the deep learning environment,select the appropriate operating system,binocular camera and bracket,and transform the coordinate system.