What is AOI?

Automatic optical inspection (AOI for short in English) is an efficient optical image inspection system and mechanism. The mechanism can be divided into four parts, namely camera, lens, light source and materials required for its installation. The other is the algorithm of the software. Traditional algorithms can solve most problems, while AI can process a certain number of images through the software to become AI learning models. These models can easily enable the software to distinguish between good and defective products. distinguish. At present, the most commonly used methods of Chernger Technology are Image segmentation, Object detection, Anomaly detection and Image classification.

Segmentation will cut the image into small pieces for detection, which takes longer than Object detection. Anomaly detection will divide the image into good products and defective products. As long as foreign objects are detected, they will be judged as defective products. Finally, Image classification will It is an algorithm for classifying samples and classifying data into specified categories.

The defective samples faced by Chernger in various industrial applications are extremely different, ranging from flat samples, roll materials, printed matter, plastic/rubber parts, metal parts, medical equipment, glass samples, contact lenses… and so on. Common defects include “black spots/stains”, “scratches”, “bubbles”, “burrs”, “deformation”, “warping”, “chips”… and so on. After the introduction of AI, technology that cannot be achieved by AOI algorithms has more complete solutions.

In AI artificial intelligence, the existence of the Anomaly detection algorithm provides a more effective and complete detection model for complex backgrounds and samples with multiple defects. This algorithm only needs to collect the correct images of OK products and multiple NG products as data to quickly establish a detection model, reduce the error rate, effectively increase the speed of verification, and significantly reduce manpower. The client Being able to provide only fewer defective samples, Our engineers use a small amount of sample amplification data to obtain a small number of photos of defective products, take images of various angles, colors and other defects, and copy them into more different samples. Good product information provides more data for AI to learn.

Chernger has self-built light field technology that can match the appearance of the product to create the most suitable light field, allowing the product’s OK/NG resolution to reach over 98%.