Machine vision is a branch of artificial intelligence that is developing rapidly. Simply put, machine vision is to use machines instead of human eyes to make measurements and judgments. The machine vision system converts the captured target into an image signal through a machine vision product (ie, image capture device, divided into CMOS and CCD), and transmits it to a dedicated image processing system to obtain the morphological information of the captured target. According to the pixel distribution and Information such as brightness and color is converted into digital signals; the image system performs various operations on these signals to extract the characteristics of the target, and then controls the actions of the equipment on the spot according to the results of the discrimination.
Classified by function:
1. Alignment system: also called positioning.
2. ORC: also called character detection.
3. Measurement: Mainly size measurement, testing speed also counts.
4. Defect detection: automatic detection.
5. Code reading system
Visual system composition:
1. Industrial computer, such as 4U industrial computer, wall-mounted industrial computer, BOX, tablet computer, etc.
The brands of industrial computers currently used in industrial cameras include: Advantech, EVOC, Corrard, Axiomtek, Chenyao, IEI, AAEON, etc.
There are three general characteristics of machine vision industrial computers: Intel network port, high configuration, relatively high requirements, and strong scalability.
2. Lens (too many manufacturers)
3. Light source (too many manufacturers)
4. Cameras, such as JIA, The Imaging Source, Gray Dot, Daheng, Hikvision, Medvision, Keyence, Cognex, etc.
5. Capture card, such as 1394 capture card, Camaralink capture card, NI capture card
6. Motion control card, such as Reese, Gogo, Step, etc.
Camera classification:
According to the overall situation, it is divided into CMOS and CCD;
In general, it is subdivided into:
1. Black and white camera, only 2 colors, black and white image
2. Color camera, RGB three colors
3. Line scan camera
The industrial computer occupies a relatively important position in the vision system and is also the core terminal of computing. How to choose?
1. The size of the camera's pixel. The larger the pixel, the larger the decoding operation is required for shooting, such as 300,000, 1.3 million, 2 million, 2.3 million, 5 million, 10 million, 14 million, 11 million, 20 million, 250 million, 29 million pixels, etc., there are also color and black and white points.
2. Software algorithm, software algorithm depends on the habits and choices of visual engineers. At present, many basic visual softwares have operators, such as halcon, opencv, NI Labliview, etc. There are also companies like Innovation Technology, OPT, Keyence Standard Software and a complete set of standard vision systems.
3. How many cameras are there, single camera or multi-lens camera, synchronous or asynchronous? If it is synchronous, the image is processed at the same time, and the speed is required to be faster. If it is asynchronous, the processing of the image is divided into batches. Multi-lens cameras generally require better multi-thread processing, so the CPU supports multi-threads.
4. Camera shooting speed and processing speed, frame rate and processing speed have requirements for CPU and GPU.
5. Are there any special requirements for GPU? At present, 2D and 3D vision have increasingly adopted deep learning methods and AI intelligence, so some professional capture cards or graphics cards will be used.
6. To expand the number and type of interfaces, multi-lens cameras require multiple USB or multiple network ports, or use a capture card.
Machine vision industry applications:
Dispenser/coating machine, solder paste printer, placement machine, laser welding/cutting, LED bonding, AOI, SPI, taping machine, visual code reader, palletizing, inkjet printer, spectroscope, 4 Wheel alignment, parking lot, electronic police, etc., use cameras, probes, industrial cameras, all belong to the category of machine vision.
