Design And Implementation Of Fruit Recognition And Sorting System Based On Convolutional Neural Network

Our country is a big fruit producing country,the output of each kind of fruit is tens of millions tons every year,therefore,the fruit has become one of the focal points of our country agricultural development.But at present,China's fruit industry,especially in the stage of fruit recognition and sorting,because the automatic recognition and classification technology for fruits is not mature,often still through manual way to complete the recognition and sorting work,this way not only increases the human cost but also the work efficiency is not high.But the use of convolutional neural network(CNN)in computer vision in agriculture makes it possible to solve these problems.In order to solve the problem of low sorting efficiency in fruit recognition and sorting,a fruit recognition and classification method based on convolutional neural network is proposed in this paper.The proposed method is based on the classical LENET-5 model with some improvements,and combined with the Census transformation theory to solve the problem of image local region information loss in existing convolutional neural network.The experiments show that this paper can recognize the fruit image more accurately and effectively,and improve the accuracy of fruit recognition by using the model of Census transformation and convolutional neural network,the fruit image recognition method has been greatly improved,and can reflect the necessity and importance of intelligent recognition to modern agriculture.Specifically,the main contents of this article are as follows.(1)Two kinds of fruit images,4200 sample data,were collected and transformed into a fruit image database for research training set and test set.(2)In this paper,QT software is used to design a mechanical arm to simulate the automatic recognition and sorting of fruit.Specifically,the trained CNN model was first introduced into the robotic arm,giving it the ability to recognize the fruit,and then the images of the fruit were transmitted through a conveyor belt,with each image passing through the robotic arm,the robot will automatically sort the fruits according to the requirements to achieve the goal of automatic sorting.After testing,on the basis of deep learning theory,the application of CNN to fruit recognition and classification can significantly improve the performance of fruit recognition and sorting.(3)Combining Census transformation theory with convolutional neural network theory,a disease recognition model based on convolutional neural network algorithm was designed.After a lot of training on the data set,the disease recognition model based on convolutional neural network algorithm was designed,the capability of feature representation at each level be enhanced gradually,and higher level features can be obtained at a deeper level,with better image representation and generalization capability.