On-line Identification Technology And System Of Wheel Hub Model Based On Machine Vision

The automatic identification of the wheel model plays a vital role in the production,inspection and transportation of the hub.Especially in the field of defect detection,the automatic identification of the wheel type is a necessary prerequisite for the automation of defect detection.However,the traditional manual identification method cannot meet the current mass production due to inefficiency.Therefore,it is urgent to develop an efficient and accurate wheel type identification system to improve the automation level of wheel hub production and inspection.Based on this,this paper develops an online identification system for wheel hub models based on industrial cameras and grating sensors to identify and classify different types of roughing wheels on the production line.The main work of this paper is as follows:(1)According to the technical requirements of the wheel hub manufacturer,combined with the working characteristics of machine vision,a hardware system based on industrial camera and grating sensor is built to obtain the appearance image and height information of the hub.Subsequently,in order to obtain true and effective feature parameters,the camera is calibrated to correct the resulting hub image.(2)The parameters of gray stretching are selected by comparing the histogram of hub image and background image.Then gray-scale stretching is applied to the hub image to improve the contrast.Because of the complex structure of the rough hub,the traditional segmentation method cannot be applied to the segmentation of the hub image,so the minimum intra-class cross-entropy method is selected to segment the hub image.Finally,the burr in the binary image of the hub is removed by morphological operation.(3)By analyzing the appearance characteristics and naming rules of the hub,six parameters such as height,diameter,number of spokes,type of spokes,circularity of spokewindows and area ratio of spoke windows are extracted as the characteristic parameters of hub models identification.Then improve the KNN algorithm: reduce the sample dimension;optimize the similarity distance formula.Finally,the collected hub data are classified and tested.The experimental results show that compared with the traditional KNN algorithm,the improved KNN algorithm has higher recognition accuracy and speed.When K is 5,the accuracy of classification recognition is the highest,which is 98.08%.And the recognition time of the recognition system for a single hub is less than 2s.