Ingot scratch detection system and method based on depth learning and image processing technology

Abstract

The invention relates to an ingot scratch detection system and method based on depth learning and image processing technology. The invention solves the problems of low accuracy, low efficiency and hysteresis in detecting the defects of the wire spindle by manual method in the prior art, and the problems of serious error detection and low classification accuracy in the general automatic detection system. The system includes a tray for loading spindles, a conveyor belt is provided with a dark box on the conveyor belt, a sorting unit is arranged on the drive belt behind the dark box, an image acquisition unit is arranged in the dark box, the image acquisition unit sends the acquired information to the processing unit for scratch target detection, adopts depth learning CNN network detection toextract the scratch target, calculate the scratch target evaluation score, and judge the spindle grade according to the multi-grade division interval. The spindle is sorted according to the scratch grade sorting unit. Detection accuracy is higher, with using multi-level classification, classification accuracy is high.

Description

technical field [0001] The present invention relates to the technical field of wire spindle defect detection, in particular to a wire spindle scratch detection system and method based on deep learning and image processing technology. Background technique [0002] Poor molding is a kind of defect of chemical fiber spindles, which is mainly caused by mechanical or human causes in the production process, and scratches are a common form of poor molding. The main mechanical causes are as follows: [0003] 1. Due to mechanical abnormalities in the winding process, grooves will appear, the top and bottom parts will appear in a concentric shape, and the side will appear in the shape of strangulation; [0004] 2. Due to the displacement deviation during the transfer process between the devices, it will cause collisions, including the collision between the silk surface, the collision between the silk surface and the device, and the serious collision between the silk surface and the d...

AI Technical Summary

Problems solved by technology

[0013] The present invention mainly solves the problems of low accuracy, low efficiency, and hysteresis in detection of wire spindle defects in the prior art, as well as the problems of serious misdetection and low classification accuracy in general automatic detection systems, and provides A wire spindle scratch detection system and method based on deep learning and image processing technology

Images