Xinsheng IntelligentIndustry NewsIntroduction to machine vision knowledge summary

1. Machine Vision System

Industrial camera type: According to the different output signal types, it is divided into two types: analog camera and digital camera. According to different interface standards, digital cameras can be divided into 1394 cameras, USB cameras, CameraLink cameras and Gige cameras. Among them, CameraLink interface camera can solve the problem of large data transmission; Gige interface camera can solve the problem of long-distance and fast transmission; and 1394 camera and USB interface camera are easy to use and cost-effective;

Lens interface type: C interface, CS interface, U interface, etc.;

Light source type: ring light source, backlight source, coaxial light source, bar light source, point light source, spherical integral light source, etc.;

2. How to choose a camera?

1. Choose CCD or CMOS camera according to different applications

CCD industrial cameras are mainly used in image extraction of moving objects. Of course, with the development of CMOS technology, many placement machines are also choosing CMOS industrial cameras. CCD industrial cameras are generally used in visual automatic inspection schemes or industries. CMOS industrial cameras are more and more widely used due to their low cost and low power consumption.

2, the choice of resolution

Choose the resolution size according to the system requirements. First consider the accuracy of the object to be observed or measured, and select the resolution according to the accuracy. Camera pixel accuracy = size of field of view in one direction / resolution of camera in one direction. Then the resolution of the camera in one direction = the size of the field of view in one direction/theoretical accuracy. If the single field of view is 5mm long and the theoretical accuracy is 0.02mm, the resolution in one direction=5/0.02=250. However, in order to increase the stability of the system, only one pixel unit corresponds to one measurement/observation accuracy value. Generally, a multiple of 4 or higher can be selected. In this way, the camera requires a resolution of 1000 in one direction, and it is enough to choose 1.3 million pixels.

Secondly, look at the output of the industrial camera. If it is posture observation or machine software analysis and identification, high resolution is helpful; if it is VGA output or USB output, it will also depend on the resolution of the display, and the resolution of the industrial camera. No matter how high it is, the display resolution is not enough, and it is meaningless; it is also helpful to use the memory card or camera function, and the high resolution of the industrial camera is also helpful.

Application case: Assuming that the surface scratches of an object are detected, the size of the object to be photographed is 10*8mm, and the required detection accuracy is 0.01mm. First of all, assuming that the field of view we want to shoot is 12*10mm, then the minimum resolution of the camera should be selected at: (12/0.01)*(10/0.01)=1200*1000, a camera of about 1.2 million pixels, that is to say If one pixel corresponds to one detected defect, the minimum resolution must be no less than 1.2 million pixels, but 1.3 million pixel cameras are commonly used on the market, so generally, 1.3 million pixel cameras are used. But the actual problem is that if a pixel corresponds to a defect, then such a system will be extremely unstable, because any disturbing pixel may be mistaken for a defect, so in order to improve the accuracy and stability of the system, the most The area of the defect should be more than 3 to 4 pixels, so the camera we choose will be more than 1.3 million by 3, that is, the minimum should not be less than 3 million pixels, usually a 3 million pixel camera is the best (I have seen The most people hold on to sub-pixels and say that if you want to achieve a fraction of the sub-pixel, then you don't need such a high-resolution camera. For example, they say that if you achieve 0.1 pixels, it means that a defect corresponds to 0.1 pixels. The size is calculated by the number of pixels. How can the area of 0.1 pixel be expressed? These people use sub-pixels to fool people, which often shows that they have no common sense). In other words, we are only using it for measurement, so using the sub-pixel algorithm, a 1.3-megapixel camera can basically meet the needs, but sometimes due to the influence of edge clarity, when extracting the edge, just offset a pixel, Then the accuracy is greatly affected. Therefore, if we choose a 3 million camera, we can also allow the extracted edge to deviate by about 3 pixels, which ensures the accuracy of the measurement.

3. Camera frame selection

When the measured object has motion requirements, an industrial camera with a high frame rate should be selected. But generally speaking, the higher the resolution, the lower the number of frames;

4. Matching with the lens

The sensor chip size needs to be less than or equal to the lens size, and the C or CS mounts need to match.

The difference between C and CS interfaces is that the distance from the contact surface of the lens and the camera to the focal plane of the lens (the position where the CCD photoelectric sensor of the camera should be) is different:

The distance of the C-type interface is 17.5mm

The distance of CS type interface is 12.5mm

C-type lens and C-type camera, CS-type lens and CS-type camera can be used together.

A 5mm C/CS adapter ring can be added between the C-type lens and the CS-type camera to be used together.

CS-type lenses cannot be used with C-type cameras.

3. How to choose a lens?

Industrial cameras have high comparability in resolution, frame rate, light requirements, exposure methods, etc. The main component is CCD photosensitive chip.

1. Resolution: The number of pixels (Pixels) of the image captured by the camera each time. For digital industrial cameras, it generally directly corresponds to the number of pixels of the photoelectric sensor. For analog cameras, it depends on the video format. The PAL system is 768*576, NTSC system is 640*480.

2. Pixel Depth: that is, the number of bits of data per pixel. Generally, 8Bit is commonly used. For digital industrial cameras, there are generally 10Bit, 12Bit, etc.

3. Maximum frame rate (Frame Rate)/Line Rate (Line Rate): The rate at which the camera captures and transmits images, generally the number of frames per second (Frames/Sec.) for area scan cameras, and per second for line scan cameras The number of rows collected.

4. Pixel Size: The pixel size and the number of pixels (resolution) together determine the size of the camera target surface. At present, the pixel size of digital industrial cameras is generally 3 μm-10 μm. Generally, the smaller the pixel size is, the more difficult it is to manufacture, and the less easy it is to improve the image quality.

5. Exposure mode (Exposure) and shutter speed (Shutter): for line scan cameras, the exposure mode is progressive, you can choose the acquisition mode of fixed line frequency and external trigger synchronization, the exposure time can be consistent with the line period, or you can set it. Set a fixed time; area scan industrial cameras have several common methods such as frame exposure, field exposure and rolling line exposure, and digital industrial cameras generally provide the function of external trigger image acquisition. The shutter speed can generally reach 10 microseconds, and high-speed industrial cameras can also be faster.

6. Spectral Range: It refers to the sensitivity of the pixel sensor to different light waves. The general response range is 350nm-1000nm. Some cameras add a filter in front of the target to filter out infrared light. If the system needs This filter can be removed when sensitive to infrared light.

4. How to choose the light source?

The light source is an important factor affecting the input of the machine vision system, which directly affects the quality of the input data and at least 30% of the application effect. Since there is no universal machine vision lighting equipment, for each specific application instance, the corresponding lighting device should be selected to achieve the best effect. Here are some common lighting methods.

Direct lighting: The light is directed to the object to get a clear image. This type of light works well when high-contrast images of objects are required. But when it is used on bright or reflective materials, it will cause a mirror-like reflection. General lighting generally adopts ring or point lighting. Ring lights are a common form of general-purpose lighting that are easily mounted on lenses and provide adequate illumination for diffuse surfaces.

Coaxial lighting: The formation of coaxial light is the change of divergent light coming out of the vertical wall, and it hits a beam splitter that makes the light downward, and the camera looks at the object through the beam splitter from above. This type of light source is especially helpful for detecting highly reflective objects, and is also suitable for detecting objects with inconspicuous areas affected by shadows from the surrounding environment.

Darkfield lighting: Darkfield lighting provides low-angle lighting relative to the surface of an object. Use the camera to shoot the mirror so that it is in its field of view. If the light source can be seen in the field of view, it is considered to be bright-field illumination. On the contrary, if the light source cannot be seen in the field of view, it is dark-field illumination. Therefore, whether the light source is bright-field illumination or dark-field illumination is related to the position of the light source. Typically, darkfield illumination is used for illumination of raised portions of the surface or illumination of surface texture variations.

Diffuse lighting: Continuous diffuse lighting is applied to the reflective surface of objects or surfaces with complex angles. Continuous diffuse lighting applies a hemispherical uniform lighting to reduce shadows and specular reflections. This type of lighting is very useful for fully assembled circuit board lighting. This light source can achieve uniform illumination over a range of 170 solid angles.

Backlighting: Light from the back of an object with a uniform field of view. The side profile of the object surface can be seen through the camera. Backlighting is often used to measure the size and orientation of objects. Backlighting creates strong contrast. Surface features of objects may be lost when backlight technology is applied. For example, backlight technology can be used to measure the diameter of a coin, but it is not possible to determine the front and back of the coin.

In addition to the several common lighting technologies described above, there are also lighting technologies used in some special occasions, such as strip light illumination with concentrated brightness in line scan cameras; for example, parallel light illumination used in conjunction with telecentric lenses in precision dimensional measurement For example, the stroboscopic lighting technology that reduces the blur of the measured object in high-speed online measurement; another example is the structured light lighting technology that can actively measure the distance from the camera to the light source and the polarized lighting technology that reduces stray light interference.

In addition, many complex environments under test require the cooperation of two or more lighting technologies. Therefore, rich lighting technology can solve many problems of image acquisition in vision system, and the choice of light source lighting technology may be crucial to the success of a vision system.