Information Recognition Of Aluminum Hub Back Cavity Based On Machine Vision

The rapid development of science and technology in the 21 st century,followed by the informationization of people's lives and the automation of industrial production.The aluminum wheel industry has an important impact on the country's economic development and needs more innovation.Using advanced science and technology and equipment to replace the artificial labor force can not only reduce labor costs and industrial accidents,but also improve product quality,increase industrial technology content,and make aluminum wheel production more systematic and intelligent.At present,there are relatively few researches on the automatic production of aluminum wheels,which makes the aluminum hub production line unable to achieve full automation,and still cannot achieve systematic management in terms of product sorting and product traceability.In order to make up for this inadequacy and advance the automation process of the aluminum wheel hub production industry,this paper deals with the proper handling of the aluminum hub based on the special shape of the aluminum hub and the character of the back cavity character,and focuses on the identification of the aluminum hub back cavity information.problem.The main research content includes:(1)Using an industrial camera to collect the image of the aluminum hub back cavity,select appropriate graying,binarization,and filtering methods to preprocess the original image;combine the bi-directional projection method and the Hough circle transformation method to determine the annular region where the aluminum hub back cavity information is located,and extracting the toroidal region by pixel traversal.(2)Using a 360-degree rotating projection segmentation method to segment spokes with aluminum hub back cavity and display the ring region's effective region gray distribution information in a projection histogram through the rotation of the aluminum hub image from the center to the edge of the projection,the projection histogram is processed using the double threshold method and the position angle information of each spoke is finally obtained.(3)Using multiple template matching method to separate single characters,select the best normalization parameter by experiment and introduce the center distance to calibrate the position of the string,use template evaluation and local exact matching to further optimize the segmentation method,and finally complete Segmentation of characters.(4)The scheme of classifying and recognizing the character of the aluminum hub back cavity of the collected convolution neural network based on the inception structure is designed.The appropriate training parameters are selected and the character recognition model is trained.The relationship between training time and number of trainings and the recognition status of different categories of characters were analyzed.At the same time,classification experiments were performed on normal characters,characters with different color depths,skewed characters and incomplete characters,which verified the efficiency of the method,robustness and effectiveness.Compared with the results of traditional template matching recognition,the universality of the proposed method is proved.