Research On On-line Detection Method Of Cylinder Defects Of Commutator Based On Machine Vision

The commutator is the core component of the motor,and its quality directly affects the normal use of the motor.At present,the quality inspection of the commutator mainly relies on manual detection,and the detection accuracy and efficiency cannot meet the requirements of the enterprise.With the development of technology,machine vision-based quality inspection technology is widely used in various fields.In this paper,a set of visual inspection schemes for burr defects and copper shell surface defects on the cylinder surface of the commutator is presented.The main work includes:The overall design of the commutator cylinder defect detection system was carried out,including camera and lens selection,illumination mode design,test bench structure design and defect detection algorithm design.In the design of the illumination mode,according to the influence of the groove width and the curvature of the copper shell on the sufficiency and uniformity of the image,the design of the light source and the illumination mode is carried out to realize the illumination supplementation in the inner region of the groove and the uniform illumination of the surface area of the copper shell.A defect detection algorithm based on energy function is studied for the groove edge burr defect with low contrast and large background noise.Based on the Markov random field model,the method constructs an energy function for each pixel based on the neighborhood information of the pixel.By iteratively optimizing the energy function,pixel point classification is realized,and the color image clustering method is combined to realize the groove edge.Complete extraction of burr defects.A defect detection algorithm based on background fitting is studied for the in-groove burr defect with low contrast and uneven background illumination due to increased backlight.The method obtains the non-uniform illumination component by polynomial background fitting on the image,improves the dynamic threshold segmentation method based on the background of the fitting,and combines the region growing method to realize the automatic extraction of the burr defects in the groove,effectively reducing the illumination.Misdetection of burrs in the tank caused by unevenness.A defect detection algorithm based on combined classifier is studied for the surface defects of copper shells with similar features and difficult classification.Based on image grayscale features,the algorithm extracts multiple types of defects,selects the optimal classification feature set based on RelifF and PCA,and constructs a combined classifier based on BP neural network,which improves the accuracy and diversity of basic classifiers.The setting method of the number of hidden layer neurons and the number of training samples are set to achieve accurate classification of defects and reduce false detection and missed detection.Finally,the defect detection algorithm is tested for robustness,accuracy and efficiency.The experiment proves that the defect detection algorithm studied in this paper is suitable for the quality inspection of workpieces with different defect morphology and different surface quality.In terms of accuracy,the zero defect detection of burr defects is basically realized.The classification accuracy of copper shell surface defects reaches 96.6%,and the efficiency meets the current stage.Business requirements.