| The most important form of the pork production and processing companies that delivered to dealers is pork carcass,the quality of the carcasses is directly related to the diet of the general public.Traditional pig carcass grading methods generally have problems with inaccurate or high hardware prices.In recent years,with the developing of deep learning theory and the tremendous development of equipment computing capabilities,deep learning has been splendid in many different fields.The convolutional neural network is one of the most effective method of deep learning in the field of image recognition,which has a very wide range of applications in image classification.This paper builds a pig carcass image dataset to train the convolutional neural networks.Through research and design,a system for automatically classifying pig carcasses based on the convolutional neural networks is implemented.Because of the lack of pig carcass images in existed datasets.According to the actual situation of Tieling Jiuxing Group's pig processing production line,pig carcass pictures in the production line situation were collected.Through the data augmentation,SMOTE algorithm and data pretreatment for CNN network characteristics,a binary format pig carcass image data set was constructed.An AlexNet-based convolutional neural network model was designed and applied to the training and recognition experiment of pig carcass images.Through the analysis of the classification results in the experiment,the network model was continuously optimized,and the use of smaller was The number of convolution kernels and more convolutional nuclei,the increase in the number of convolutional layers,and the narrowing of the pooling range can lead to the conclusion that the classification effect can be improved.The CNN-P,as a model of pig carcass image grading,was eventually formed.Based on the model CNN-P,training and testing were carried out using the constructed pig carcass image data sets,and the recognition results were compared with other pig carcass grading methods based on computer vision methods.The results of the study show that by adjusting the size and number of convolution kernels and adjusting the number of convolutional layers.A convolutional neural network model was used to classify pig carcass images.At present,the recognition effect on the image classification of pig carcass can reach 92.7%.As the first attempt to apply theconvolutional neural network to the non-destructive grading of pig carcass at home and abroad,the system has solved the complicated preprocessing steps of traditional grading technology and compared with the low generalization ability and other issues,the price of automated pig carcass grading equipment is greatly reduced,which has certain research significance and practical value. |
