What are the instability factors behind machine vision inspection equipment?

When machine vision inspection equipment inspects products, it may encounter some unstable factors, which may affect the accuracy and consistency of the test results. Generally, the stability and accuracy of machine vision inspection equipment can be improved by optimizing lighting conditions, stabilizing camera installation, dealing with mechanical vibration, adjusting algorithm parameters, and so on. Resolution, accuracy, tolerance.

Resolution (Resolution)

Calculation formula: resolution = Field of View/Pixel

For example, the size of the product I want to see is 30mm*10MM, using a 2 million pixel (1600pixel*1200pixel) camera. Because the product is a long strip shape, in order to put the product into the field of view, we must consider the long side corresponding when calculating the resolution, at this time the resolution is:

Resolution = 30mm/1600Pixel = 0.019mm/Pixel

Accuracy (Accuracy)

Calculation formula: precision = resolution x effective pixel

The unit of accuracy is mm. According to the different surface and lighting conditions of the product, the number of stable pixels is identified by enlarging the image to obtain the accuracy. If the conditions do not allow the actual test observation, the general rule is that if the front lighting is used, the effective pixel is 1, and the backlight is 0.5 effective pixels.

For example, take 1 Pixel, the accuracy is 0.019mm, which is equal to 0.02mm.

How to calculate the positioning accuracy of machine vision system?

If the 300,000-pixel camera, the monitoring area is 640x480mm, its accuracy is 1mm.

In simple terms, the resolution of the 30W camera 640*480 is normally calculated as follows: use the longest side to remove the longest side of the monitoring area, so the accuracy is basically 1mm, this is the theoretical value, if you do measurement or surface scratch detection, certainly not accurate, a pixel may not be able to highlight the features.

Tolerance

In general, the corresponding relationship between accuracy and tolerance is as follows:

For a project, we first read the tolerance requirements from the drawings. Then, based on the above relationship, we can deduce how many pixels we need for the camera.

When measuring, the first few aspects to consider are: camera, lens, light source.

Such as visual project requirements: pixel accuracy 0.05mm, measurement error plus or minus 0.15mm.

Pixel accuracy: The distance a pixel represents in the real world, i.e. the shooting field of view/resolution. Such as the current use of Dahua 5 million camera, resolution 2592*2048, 100mm on the long side of the field of view, you can photograph 100mm objects, then in this direction of the pixel accuracy of 100/2592mm is about 0.0386mm.

Measurement error: The error between the distance/length measured using the algorithm and the true value.

Sub-pixel accuracy: Sub-pixel accuracy refers to the subdivision between two adjacent pixels, and the input value is usually half, third or quarter. That is, each pixel will be divided into smaller units to implement an interpolation algorithm on these smaller units.

If one quarter is selected, each pixel is counted as four pixels in both landscape and portrait. There are many more things to consider when actually measuring or detecting, such as frame rate, exposure, gain, etc.