Machine vision systems are very reliable in high-speed, subtle and repetitive manufacturing processes, so they are widely used in processing and manufacturing companies to complete repetitive inspection tasks in high-volume production processes. The application of machine vision in quality inspection accounts for nearly 80% of the entire industrial application, among which the larger application industries are: automotive, pharmaceutical, electronics and electrical, manufacturing, packaging, food, beverage, etc. Machine vision inspection is non-contact non-destructive inspection. Compared with traditional inspection methods, it has irreplaceable advantages, so it has been widely used. Use linear array CCD to cooperate with the one-dimensional movement of the packaging box to obtain the target image, and then process the image by computer, which can detect the omission and correctness of information such as date number; As a detection device for the outline and dimension of rebar, it can realize online measurement of geometric parameters of hot-rolled rebar; it is also widely used in surface defect detection of various products.

The development trend of machine vision system

In recent years, the development of computer vision (ie machine vision) is generally manifested in the following three aspects:

1. The theoretical system of computer vision computing based on geometric methods has reached a complete computer vision. One of the research goals is to enable machines to perceive the geometric information of objects in a three-dimensional environment, including its shape, position, attitude, motion, etc. Since the mid-1990s, the computer vision community has systematically introduced the descriptions of projective geometry, affine geometry, and Euclidean geometry into visual computing methods, which more accurately correspond to the objects in the visual system from coarse to fine. The description of , reduces the requirement to know the parameters of the camera system in the computer vision system, and improves the robustness of the system to noise.

2. Machine learning methods have received more and more attention, and there are always two branches based on structure and based on statistics in all fields of pattern recognition. If it is said that geometry-based computer vision mainly describes the three-dimensional structure of objects and their motion through geometry, it is a structural method and has been studied systematically; while statistical methods in computer vision are better used in the underlying processing of images. , has always seemed imperfect, let alone systematic.

3. The application research for many specific fields continues to deepen, and the larger-scale application systems are gradually commercialized. With the rapid improvement of the current computer performance-price ratio, the commercialization of many specific fields of computer real-time application systems has become possible. For example, the use of fingerprint, iris, face, voice and other recognition technologies, behavior recognition technology and motion tracking technology, and multi-camera fusion technology constitute a visual monitoring system for information security, intelligent transportation, anti-terrorism and anti-theft, and identity authentication.