Research On Application Of Edible Fruit Image Recognition Based On Deep Learning

The wholesale market of edible agricultural products involves a large number of edible fruit merchants.These merchants have a wide variety of fruits and huge transaction volumes.However,at present,merchants still use traditional manual operations in the process of edible fruit transactions.Especially when wholesale sales need to identify the type of fruit by manual judgment,they can enter the corresponding price data to calculate the total amount of the transaction.This method of operation not only does not take full advantage of the advantages of advanced Internet+and current artificial intelligence technology,but also consumes manpower,increases labor intensity,and affects operating costs.Therefore,it is meaningful and valuable to develop a set of edible fruit image recognition system to assist fruit business operators to automatically and efficiently identify the types of fruit images and then complete the cost calculations involved in fruit transactions.The content studied in this article aims to build an edible fruit image recognition model based on advanced deep learning technology,and develop an edible fruit image recognition system for edible fruit merchant transactions in the wholesale market based on this model.The main research contents and work of this article are as follows:(1)Research and analysis of fruit image recognition related technologies based on traditional machine learning and deep learning.Research and analysis of the feature extraction algorithms and classification algorithms in traditional machine learning,focusing on the analysis of convolutional neural network technology,especially the in-depth analysis of the network architecture characteristics and main applications of AlexNet,MobileNet and MobileNetV2.(2)Collect and organize the edible fruit image data sets required for model training.This article is based on the real edible fruit image data set collected in a real wholesale market,and then collected through the network to obtain as many data sets as possible.In addition,in order to train the model,the data set is organized into two parts:the training set and the test set.(3)Study the establishment of edible fruit image recognition models based on traditional machine learning and deep learning networks,and test the effectiveness of the models.The important research content of this article is how to choose an effective and suitable edible fruit image recognition model for the development of application systems.In order to choose an effective model,this paper separately recognizes edible fruit image recognition based on traditional machine learning algorithms such as original LBP,rotation invariant LBP,equivalent LBP,and equivalent rotation LBP and classifiers such as SVM and KNN The model and edible fruit image recognition model based on AlexNet and MobileNetV2 deep learning methods were tested and analyzed.After finally comparing the experiments and analyzing the results,this article chooses an edible fruit image recognition model based on the MobileNetV2 network for system development.(4)Develop an edible fruit image recognition system based on MobileNetV2.The software engineering method is used to carry out feasibility analysis,requirements analysis,overall design,database design and detailed design of the fruit image recognition system,and elaborates the overall architecture design and specific module design of the system.The system is mainly divided into two parts:one is the client,which includes the functions of fruit image upload,fruit image recognition,viewing the retrieval log,and viewing and modifying personal information;the second is the management terminal,which includes user management,model augmentation and other functions.Finally,write the code and complete the development of the system and perform software testing on the client and management side of the system.The test results are in line with expectations and have achieved practical effects.