“In recent years, with the development of Industry 4.0 and artificial intelligence, more and more automation equipment has been widely used in the production process. Industry 4.0 is inseparable from intelligent manufacturing. In the grand plan of “Made in China 2025” proposed by my country in 2015, the first strategic countermeasure is to “promote digital networked intelligent manufacturing”. In intelligent manufacturing, the core link is Machine vision. Machine vision refers to the use of machines to simulate the function of the human eye, extracting, processing and analyzing objective information, finally realizing detection and judgment, and finally handing it over to the computer for control.China is the fastest growing country in the machine vision industry. Currently
Author: Dr. Zeng Xu
1. Background introduction
In recent years, with the development of Industry 4.0 and artificial intelligence, more and more automation equipment has been widely used in the production process. Industry 4.0 is inseparable from intelligent manufacturing. In the grand plan of “Made in China 2025” proposed by my country in 2015, the first strategic countermeasure is to “promote digital networked intelligent manufacturing”. In intelligent manufacturing, the core link is Machine vision. Machine vision refers to the use of machines to simulate the function of the human eye, extracting, processing and analyzing objective information, finally realizing detection and judgment, and finally handing it over to the computer for control. China is the country with the fastest development of the machine vision industry, and it has been widely used in many industries such as industry, aerospace, medical care, transportation, and scientific research.
Figure 1 Machine vision replaces the human eye
2. There are current problems with machine vision
Typical industrial machine vision systems include: light source, lens, camera, frame grabber, software, monitor, input/output, etc. For optical inspection, the performance of the machine vision system mainly depends on the performance of the optical-related components in the system, such as the light source, lens, and camera. In addition, the accuracy required for optical inspection is generally high, but most cameras are not specifically calibrated for optical inspection applications when they leave the factory, which often results in the accuracy of the machine vision system not meeting the requirements, and errors will occur as a result.
For example, if you take a photo of a newly manufactured industrial camera against a uniformly illuminated light-emitting surface, dark areas will often appear at the four corners of the captured image, which is mainly caused by the cosine response of the camera lens. In addition, due to the non-uniformity of the camera sensor (CCD/CMOS), when imaging a uniform light field, the image is bright and dark, and the color is uneven, as shown in the following figure. All of the above factors will cause large deviations between the detection results and the real situation in some sophisticated optical inspections (such as flat panel Display inspections).
Figure 2 Camera flat field response before calibration
In addition, the linear response of the camera to different brightness is also different. Since the signal output by the camera is a gray value, it does not have real physical meaning. Therefore, when doing optical inspection (such as brightness inspection), it is necessary to calibrate the camera’s linearity and brightness to establish the relationship curve between the camera’s gray signal and the real brightness.
Three, industrial camera calibration solution
In order to solve the above problems in the machine vision system and improve the accuracy of the machine vision system, especially the optical inspection system such as AOI, the European Machine Vision Association EMVA proposed “EMVA1288: Imaging Sensor and Camera Performance Characterization Standard”, which introduces how to A series of indicators such as spatial unevenness, sensitivity, linearity and noise of imaging sensors and cameras are characterized and calibrated. It clearly states: “The best uniform light source is an integrating sphere uniform light source”, and it is recommended that “the uniformity of the light source should be greater than 97%”.
Figure 3 The flat-field correction method of the Lanfei optical camera
When the user is in use, he only needs to point the camera at the opening of the uniform light source, take an image, and then calculate the algorithm to correct the uniformity of the camera. This process is called flat-field correction. After the uniform light source is calibrated, the uniformity of the camera can be significantly improved. The following figure shows the uniformity test results of an industrial camera before and after the uniform light source calibration of the integrating sphere. It can be clearly seen from the figure that the uniformity of the camera before correction is poor, and the response of the center field is better than that of the surroundings. The response in the camera plane is the same after correction.
Before camera calibration
After camera calibration
Figure 4 Comparison of industrial camera before and after the flat field correction of the Lanfei Optics LED uniform light source system
Fourth, the perfect integrating spherical light source
The accuracy of the industrial camera determines the detection accuracy of the machine vision system, and the uniformity of the calibrated light source determines the accuracy of the industrial camera. The more uniform the integrating sphere light source is, the higher the uniformity of the camera obtained after its calibration. According to the principle of the integrating sphere, the light incident on the integrating sphere is reflected multiple times inside the integrating sphere, and finally a surface light source with completely uniform brightness and chromaticity is obtained at the output port. The uniformity of the light exit of the integrating sphere mainly depends on the following aspects:
1. The reflection characteristics of the inner wall material of the integrating sphere. The reflection characteristics of materials can be divided into Lambertian reflection, specular reflection and mixed reflection. It can be known from the principle of the integrating sphere that the closer the reflection characteristic of the inner wall material of the integrating sphere is to the Lambertian characteristic, the higher the uniformity of the opening. In addition, when the incident light is broad-spectrum light (such as white light), the uniformity of the spectral reflectance of the material determines the uniformity of the chromaticity at the opening. The more uniform the spectral reflectance of the material, that is, the more uniform the reflectance of each wavelength. The more uniform the chroma at the opening.
2. The design of the integrating sphere. How to design the size of the integrating sphere, the position of the incident light, and the position and direction of the baffle will affect the uniformity of the opening of the integrating sphere.
Lanfei Optical Integrating Sphere Uniform Light Source LED-USS provides an ultra-uniform, high dynamic range, and brightness/color temperature can be finely adjusted surface light source. The integrating sphere light source adopts the unique high reflectivity perfect Lambertian reflective material Spectraflect of Lamfield Optics, based on Lamfield Optics’ more than 40 years of optical system development experience, and the fine integrating sphere structure design, which is the perfect solution for machine vision camera calibration. Its main features are as follows:
・ The light-emitting surface is super uniform, and the uniformity is greater than 99.5%
・ The system output is highly stable, with a stability of 0.1%
・ The brightness can be adjusted linearly, which can realize the brightness output from weak light 0.1cd/m2 to 25000cd/m2
・ The color temperature is dynamically adjustable, which can achieve output from a low color temperature of 2700K to a high color temperature of 7500K
・ Built-in brightness monitoring, real-time observation of brightness output
・The software realizes all the control of the light source and the detector, the interface is simple and easy to use, and can provide control instructions for secondary development.
・ The system can also customize various light sources with different parameters such as color temperature, brightness, monochromatic light, large field of view, etc.
Figure 5 Lamfield Optical LED Uniform Light Source System (LED-USS) and the brightness distribution of the light spot at the opening
LED-USS is currently the world’s most uniform surface light source. Its excellent performance can meet the camera uniformity, linearity, signal-to-noise ratio, dynamic range and many other parameter tests required by EMVA1288. It is an ideal calibration light source for various industrial cameras from R&D to production.