Chopstick image classification method based on adaptive convolutional neural network

Abstract

The invention discloses a chopstick image classification method based on an adaptive convolutional neural network, which solves the problem of adaptive precise classification of different materials and different types of chopsticks, and comprises the following steps of: normalizing chopstick images; establishing a twelve-layer network model; training the network and calculating the accuracy of thenetwork; randomly selecting one convolutional layer as an extended convolutional layer, expanding the convolutional neural network, respectively adding seven network structures at the extended convolutional layer to obtain seven neural networks, training and testing the seven neural networks, selecting one neural network with the highest test accuracy, and continuing network expansion if the accuracy of the selected network is greater than that of the current network; and training and testing the optimal self-adaptive convolutional neural network. According to the method, the self-adaptive convolutional neural network is constructed, so that the structure of the convolutional neural network can be adjusted along with the training process, the depth and width of the network are increased,and the finally constructed self-adaptive convolutional neural network can accurately classify the chopsticks.

Description

technical field [0001] The invention belongs to the technical field of image processing, and further relates to chopstick image classification, in particular to a chopstick image classification method based on an adaptive convolutional neural network. The invention is mainly used for classifying the images of chopsticks, and can be used for the processing and production of chopsticks. Background technique [0002] Chopsticks refer to chopsticks processed from raw materials such as bamboo and wood in industrial production and processing. In the production process of chopsticks, if the raw materials used are polluted or damaged, the final chopsticks produced will be defective. And even if the raw materials used are not defective, due to the possibility of machine failures, manual operation errors and other problems during the production process, the finished chopsticks will also have various defects. Among them, the defect types of chopsticks mainly include dead spots, green...

AI Technical Summary

Problems solved by technology

[0005]The shortcomings of existing methods are that the existing methods directly classify chopsticks through traditional machine vision technology, without using neural network technology, The traditional machine vision technology will summarize the characteristics of unqualified chopsticks for specific chopstick detection targets in the assembly line, and then set fixed parameters such as color detection threshold, crack width and length detection threshold, and then according to the fixed parameters. Therefore, traditional machine vision technology can only detect unqualified chopsticks that reach the set parameters, but cannot detect unqualified chopsticks that are smaller than the set parameters. The neural network technology extracts features such as whether the chopstick image contains defects or not. Judging whether the chopsticks to be detected are qualified, so the neural network technology can classify the chopsticks more accurately

Images