Bearing surface defect detection system based on statistical projection training and detection method

Abstract

The invention discloses a bearing surface defect detection system based on statistical projection training and a detection method. The system comprises a front-end information acquisition device, a computer, and an execution system. The front-end information acquisition device comprises a camera and a light source, and is placed on a bearing line to be detected. The computer comprises a client, animage processing module, and a decision output module. The client controls and displays the system and is connected to the image processing module and the decision output module. The computer makes the execution system carry out screening and overturning of a bearing. The system adopts ordered peak value energy to calculate and construct a characteristic vector of a projection curve, and trains asupport vector machine SVM to obtain a defect classifier and realizes binary classification detection of normal and defective bearings. The image processing module carries out preprocessing, polar coordinate transformation, image segmentation, characteristic extraction, and SVM training. The decision output module detects whether the bearing is defective through the trained classifier. The structure of the system is simple, robustness is high, and a double-sided defect of a bearing dust cover can be detected in online and real-time modes and can be removed.

Description

technical field [0001] The invention relates to the field of part defect detection and machine vision image processing, mainly relates to a bearing surface defect detection system and detection method based on statistical projection training, which is applicable to bearing defect detection scenarios in most industrial detection fields, and can realize bearing dustproof Real-time online detection of defects on both sides of the cover. Background technique [0002] As the cornerstone of the machinery industry and equipment manufacturing industry, the quality of bearings directly affects the smooth progress of machinery production activities. Therefore, the inspection of bearing balls, peripherals, inner rings, dust covers, etc. is a problem that manufacturers attach great importance to. Among them, the surface of the dust cover is prone to various defects such as pits, scratches, rust spots, and oil stains during the production process, causing greater losses to this large-sca...

AI Technical Summary

Problems solved by technology

[0005] (2) The method introduced in the prior art only involves the single-side defect detection algorithm of the bearing. There is no follow-up judgment execution system, and the defective bearing still needs to be removed manually. Defects are automatically detected and removed online;

[0006] (3) The method based on machine learning usually needs to obtain a large number of manually labeled samples to train the model before detection, which is more difficult for the bearing production process with a low defective rate

Images