Application of visual inspection in laser coded character detection

In visual inspection, OCR recognition is a typical application, and the application scenarios are also rich. Basically, in manufacturing, every factory or printing plant will use visual OCR recognition detection. The principle of OCR character recognition is very simple. After image segmentation, characters are learned and named. After the image is taken by an industrial camera, the software processes it to read the results.

With the development of technology, laser marking technology is more widely used in the production process. Laser marking is used instead of ink for character printing, which is faster and better. However, the ensuing impact is on visual inspection. Most of the character detection equipment currently on the market are software developed by integrators. When detecting ink printing, the characters on the image have no breakpoints and are very consistent The required image segmentation is relatively simple, but the same application is not feasible for characters on laser marking equipment, because the characters marked by laser are composed of N points. The original visual inspection equipment will The small dots are separated separately and cannot be identified.

Vision Bank software can solve this problem perfectly. Through the image pre-processing function, the image is expanded and the characters consisting of N points are connected to achieve image processing. The entire processing time is very short and the fastest can be reached. A few milliseconds, the following is explained by Wang Laoji's case.



 1571296299271770.png

Figure one

1571296316225952.png

Figure II

It can be seen from Figure 1 that each character is composed of dots, and this situation occurs when Figure 2 is directly studied. In this case in Figure 2, there is no way to learn characters, so we need to do image preprocessing.

1571296395297455.png

Figure three

Figure 3 is the processed image. Through the minimum value filtering, the black points are connected together, so that we can do the overall segmentation of the characters.

1571296417898606.png

Figure four

 1571296426516985.png

Figure five

It can be seen in Figure 4 that the entire number is divided as a whole. The software in Figure 5 is bound to come out. The entire recognition process takes 8.6ms.