Glass is a common item in our daily life. With the increasing demand for new energy vehicles, the demand for automotive displays is also increasing. With the upgrading of consumption, more and more enterprises pay more attention to the quality of glass. Affected by the production environment or process, the surface of the glass is prone to appearance defects such as crystal points, scratches, dirt, impurities, etc. during the production process, which seriously affects the quality of the glass, increases the customer complaint rate, and brings losses to the enterprise.

Therefore, the realization of automatic and intelligent detection of various point and line appearance defects of automotive glass can solve the problems of high labor cost, difficult detection, and unstable yield in the glass production process. Effectively improve product quality and production capacity, help customers increase shipments, save costs, and use multi-dimensional analysis of defect data to find out the problems exposed by each process, help customers optimize processes, and gradually improve product yield.

Case study:

Online all-round appearance inspection of automotive display electronic glass. The detection surface includes the front, back and edges. The maximum size of the glass is 1000mm, and the minimum defect is 30um. The defects cover more than ten types of defects such as scratches, dirt, pits, inks, and defects.

Difficulties in detection

1. Some defects are small in size, have many types, and are randomly distributed on both sides;

2. The size of the product is large, and the detection accuracy is high;

3. The blurred edge features of some defects are not easy to accurately quantify;

solution

Using high-resolution industrial cameras and lenses, multi-channel bright and dark field customized light source solutions for defect image acquisition.

The front and back sides of the glass are imaged respectively, and the image is generated by the AI training terminal to generate a deep learning network model. The model is integrated with traditional machine vision algorithms to identify defects, and to achieve surface defect detection of different types of glass, greatly improving the yield of subsequent processes. .