Bearing roller defect detection method based on Fast-RCNN

Abstract

A bearing roller defect detection method based on a Faster-RCNN comprises the following steps: 1) acquiring data, and shooting a bearing roller picture through experimental device equipment; 2) dividing the data set into a training set and a prediction set, and making a single defect label for the data in the training set; 3) training the Faster-RCNN network by using the training set; 4) detectingpictures of the prediction set by using the trained Faster-RCNN model to obtain a detection result of each picture; 5) deleting the original label, making 15 types of defect labels for the data in the training set, and not processing the data in the prediction set; 6) training a Faster-RCNN network by using the new training set, and 7) detecting the pictures in the prediction set by using the newFaster-RCNN model to obtain a detection result of each picture. The Faster-RCNN model is constructed, so that the accuracy is relatively high and the detection speed is relatively high.

Description

technical field [0001] The invention relates to deep learning, computer vision, and target detection, and is a method for detecting defects of bearing rollers based on Faster-RCNN. Background technique [0002] As the core component of the manufacturing industry, bearings are one of the strategic basic industries that our country focuses on. As a key component in the bearing structure, the bearing roller plays a role in supporting the shaft and reducing the rotational friction of the bearing. Its quality determines the performance and life of the bearing. During the manufacturing process of bearing rollers, various surface defects may appear due to the interference of mechanical, environmental and human factors, resulting in a decline in their quality, thereby affecting the performance and service life of the bearing. Once defective bearing roller products enter the market, it will not only have a negative impact on the reputation of the company, but also pose a major safet...

AI Technical Summary

Problems solved by technology

[0003] The traditional detection method for surface defects of bearing rollers is manual detection. Manual detection has many shortcomings, such as the limited spatial and temporal resolution of the human eye, which is prone to false detection and missed detection.

In addition, manual inspection will take up more human resources and increase the production cost of the enterprise.

Images