Research On Intelligent Identification And Quality Inspection Of Wheel Hub Based On Machine Vision

The increase in car ownership has driven the development of wheel production.The processes such as finishing and quality defect detection of the wheel hub after casting are based on the actual model,so the actual wheel model needs to be identified.Because the wheels are susceptible to quality defects during casting and finishing,they need to be inspected to determine whether they are qualified.However,the current model identification and quality defect detection methods of wheels are mostly manual visual inspection,which cannot complete the identification and detection tasks with quality and quantity.Nowadays,machine vision systems can overcome the problems caused by manual identification and improve the flexibility and automation of production systems.Therefore,this paper studies the intelligent identification and quality detection system of the wheel hub based on machine vision,which is used to realize the intelligent identification of the wheel hub and the quality defect detection.First,based on the application of machine vision,research and develop a complete set of machine hub intelligent recognition and quality inspection systems based on machine vision.Design and research the software part and inspection process based on the hardware of intelligent identification of wheel model and quality defect detection.Secondly,the realization basis of the wheel type intelligent identification system is studied.The wheel image collected by model recognition hardware system is pre-processed by graying,filtering and denoising,binary morphology,and edge detection.From the preprocessed image,six characteristic data of wheel radius,number of spokes,type of spokes,distance from the center of the circle to the lower edge of the spoke window,window area to hub area ratio,and width were extracted and normalized and principal component processed.KNN-SVM algorithm combining KNN and SVM and GWO-SVM algorithm using Grey Wolf algorithm Optimizer?GWO?to improve SVM penalty coefficient c and kernel function radius?are studied on the basis of conventional K-Nearest Neighbor?KNN?and Support Vector Machine?SVM?recognition algorithms.Through the application of four algorithms to the wheel model recognition experiment,results show that the GWO-SVM algorithm has the best recognition effect,and the correct recognition rate can reach 95.83%,which is about 6.48%higher than traditional SVM to meet model recognition requirements.Thirdly,the realization basis of the wheel quality defect detection system is studied.Based on the study of wheel hub porosity,shrinkage,shrinkage,cracks,and inclusion defects,multi-frame superimposed denoising and anti-sharp mask processing are performed on the defect images collected by the hardware system.The background image is obtained based on the defect contrast template,and Cuckoo Search algorithm?CS?combined with the Otsu's CS-Otsu algorithm is used to find the optimal threshold T for segmentation display of the defect.After the defect areas to be detected and pseudo-removal operations were selected through ROI,seven defect feature data including defect area,grayscale mean,length diameter,sharpness,perimeter and area ratio,length-width ratio and rectangularity were extracted and normalized and principal component processing was carried out.Based on the BP?Back Propagation?neural network,Glowworm Swarm Optimization?GSO?,a GSO-BP algorithm that improves the weights w_{j i},v_kjand thresholds?_j,?_kof the BP hidden layer to the output layer and the input layer to the hidden layer,is used to detect wheel defects.Results the defect detection recognition rate of the table name using the GSO-BP algorithm reached 95%.Compared with the conventional BP algorithm,the detection rate was improved by about 6.74%.It can meet the defect detection requirements,and the defects are graded and qualified according to ASTM standards and actual conditions.Finally,based on the Visual Studio compilation environment,an Open CV open source computer vision library and a Qt visual program interface are built to implement the software module functions of intelligent model identification and quality defect detection system.The experimental results verify that the system can meet the requirements of intelligent identification and quality defect detection of wheel models.