Gear defect visual detection method and system based on improved YOLOv5 network

Abstract

The invention discloses a gear defect visual detection method and system based on an improved YOLOv5 network. The system comprises a controller with a built-in improved YOLOv5 network model, a conveying system composed of a first conveying belt and a second conveying belt, and an image acquisition module and a rejection mechanism which are built on the conveying system. According to the improvement mode, an unimproved YOLOv5 network is trained through a sample data set to obtain weight parameters, a convolution attention mechanism module and a repeated weighted bidirectional feature pyramid network are added to a YOLOv5 network model, and the weight parameters are migrated to the improved YOLOv5 network model; and training an improved YOLOv5 network model through the data set, completing the construction of a gear defect detection model, collecting an image through an image collection module, inputting the image into the gear defect detection model for recognition, and rejecting a corresponding defective gear according to a recognition result. According to the invention, accurate identification of gear characteristic defects and automatic detection and sorting of gear multi-surface defects are realized, and the detection efficiency is improved.

Description

technical field [0001] The invention belongs to the field of product defect detection, and relates to a product defect visual detection technology, in particular to a gear defect visual detection method and system based on an improved YOLOv5 network. Background technique [0002] Gears are widely used in various mechanical products. However, gears are prone to wear, chipping, cracking and other defects during use, which will affect the service life and motion accuracy of moving parts and mechanisms. Therefore, it is necessary to detect gear defects. Due to the geometrical characteristics of the gear itself, and at the same time, defects such as chipping are mainly distributed on the tooth surface, and defects such as cracking are mainly distributed in the end face, so it is necessary to detect multiple surfaces of the gear. The traditional manual inspection has a large workload, which is easy to cause visual fatigue of inspectors, resulting in missed inspections and wrong in...

AI Technical Summary

Problems solved by technology

[0008] The invention patent application with the application number CN202110000785.2 discloses a spur gear defect detection and sorting device, which uses an electromagnetic rod to absorb the gear and drives the gear to rotate through the electric rotating seat. At this time, the eddy current sensor and the gear are on the same horizontal line , the relevant parameters of the gear can be measured through the eddy current sensor, but the device is not based on machine vision for detection, and only detects the defects on the side of the gear

Images