Rapid high-precision defect detection method based on visual image

Abstract

The invention discloses a rapid high-precision defect detection method based on a visual image, and the method comprises the following steps: positioning a reference standard object, collecting an image, creating a matching template, determining a detection region, positioning a to-be-detected object, collecting an image, comparing a standard region image with a to-be-detected region image, obtaining a defect detection result.Extracting abnormal areas, and extracting and screening abnormal area features. According to the rapid high-precision defect detection method based on the visual image, the detection process is simple and rapid, a high-precision calibration plate or an auxiliary tool does not need to be specially customized, the requirement for the assembly precision of the alignmentmodule is low, calculation of the translation position and the rotation angle can be completed through one-time photographing of multiple cameras, and the alignment process is high in efficiency and precision, therefore, rapid and high-precision defect detection of the target object can be realized.

Description

technical field [0001] The invention relates to the technical field of defect detection of industrial products, in particular to a fast and high-precision defect detection method based on visual images. Background technique [0002] In industries such as chip semiconductors, display panels, mobile phone panels, and PCB manufacturing, the manufacturing process requires extremely high precision, and the tolerance for defects generated in the manufacturing process is low, so high-precision defect detection methods are usually required. The production and processing process of display panels and mobile phone panels generally includes: material cutting, slotting, chamfering, fine carving, flat grinding, cleaning, electroplating / screen printing, lens cleaning, and packaging. [0003] When using visual images for detection, in order to ensure the detection accuracy, it is usually necessary to align the product first, and then compare it with the standard template to locate the defe...

AI Technical Summary

Problems solved by technology

[0006] First, the calibration process is complicated. Many methods need to rely on customized high-precision calibration boards. Not only is the production cost high, but the calibration algorithm logic is complex. When there are errors in the parallelism and perpendicularity of the alignment module, it will also lead to alignment accuracy. decline; there are still many calibration methods that need to calculate the alignment error multiple times, and then iteratively calculate the linear compensation parameters, and it cannot be done automatically

[0007] Second, the alignment process is complicated. Many alignment methods do not calibrate the center of rotation during calibration, or do not establish a suitable alignment model, resulting in the need for multiple photographs and multiple platform movements to complete the alignment process, resulting in less efficient

[0008] Third, the alignment accuracy is low

Many alignment systems do not consider all factors or ignore certain parameters when establishing the positional relationship model between multiple cameras, resulting in low alignment accuracy

[0009] There is another visual image defect detection method in the prior art that uses manual visual inspection during final inspection. This method cannot prevent batch failures due to equipment and production processes in advance. A cover plate usually needs to pass through from the transparent substrate to the finished product. As many as twenty printing and baking, cumulative time-consuming hours

At present, the single-pass printing capacity of conventional automated screen printing machines is 600-1400 pieces per hour. From the first occurrence of defective products to the final inspection, thousands of defective products may have occurred, resulting in increased production costs and decreased effective production capacity. work efficiency is very low

Images