Research And Implementation Of Automobile Wheel Classification Alogorithm And Recognition System

In recent years,with the continuous growth of automobile production,there is an urgent need to advance the automation level of the whole production line.There have been more than one hundred types of automobile wheels on the production line,and new wheel types are also designed and produced sometimes.Since the classification accuracy of human is not high enough and long-term observation is harmful to the human eye,the traditional method of identifying the type of automobile wheel by manual classification can no longer meet the needs of the factory.This paper uses technology related with computer vision to design a wheel classification algorithm and a steel sleeve detection algorithm.Based on these algorithms,a wheel production management system is developed in this paper.The main research work is as follows:Firstly,based on the actual requirements of the wheel classification algorithm,the image preprocessing method for separating the wheel from the original image captured on the production line is designed and implemented.Then,combined with template matching method and SURF points matching method,these two classification algorithms are implemented together for the purpose of double verification.The final classification accuracy rate of the proposed wheel classification algorithm can reach 99.9% on the whole production line.Secondly,a dataset of wheel images for different appearances captured during the process of production is built,after which a convolutional neural network is designed to automatically extract the feature maps and classify the wheel images.After training,the network proposed in this paper finally achieved 95.6% classification accuracy in this dataset,which proves the broad application prospect of deep learning on wheel classification of multiple appearances.Thirdly,based on the requirements of the steel sleeve detection,this paper designs a steel sleeve detection algorithm through circular detection to locate the steel sleeve region and gray level analysis to confirm whether there is a steel sleeve in this region.At the same time,in order to improve the accuracy and stability of the algorithm,image contrast enhancement is obtained through gray stretch technique,and an improved K-Means++ algorithm to cluster the detected circles is proposed,which can effectively prevent the Round miss and false detection.The final design and implementation of the steel sleeve detection algorithm can achieve 99.9%recognition accuracy in the actual production line.Finally,based on the proposed wheel classification algorithm and steel sleeve detection algorithm,an automobile wheel production management system is designed.This paper not only gives the specific design for each module,but also shows the application examples of each module in actual production.What’s more,the optimization acceleration methods adopted by the system in algorithm design and software implementation are also listed,which are important guarantees for the real-time and stability of the system.