First: the stability of lighting
Industrial vision applications generally fall into four broad categories: Positioning, measurement, detection and recognition, among which, measurement has the highest requirement for the stability of illumination, because as long as the illumination changes by 10-20%, the measurement results may be biased by 1-2 pixels, which is not a problem of software, but the change of illumination, which leads to the change of the position of the upper edge of the image. Even the most powerful software can not solve the problem. From the point of view of system design, the interference of ambient light must be eliminated, and the luminous stability of active lighting source must be guaranteed. Of course, through the hardware camera resolution enhancement is also to improve the accuracy, a way to resist environmental interference. For example, the space size of the corresponding object of the previous camera is 10um per pixel, but after improving the resolution, it becomes 5um per pixel. The accuracy can be approximately considered to be doubled, and the interference to the environment is naturally enhanced.
Second: inconsistency of workpiece position
Generally speaking, for measurement items, whether offline detection or online detection, as long as it is fully automated detection equipment, the first step is to find the target to be measured. Every time the target to be measured appears in the shooting field of view, to be able to accurately know where the target to be measured, even if you use some mechanical fixtures, etc., can not be particularly high precision to ensure that the target to be measured every time appears in the same position, which requires the use of positioning function, if positioning is not accurate, may be the location of the measurement tool is not accurate, measurement results sometimes have a large deviation
Third: calibration
Generally, the following calibration needs to be done in high-precision measurement: optical distortion calibration (if you are not using software lens, it is generally necessary to calibrate); projection distortion calibration, that is, image distortion correction represented by the error of your installation position; and image space calibration, that is, specific calculation of the size of each pixel's counterpart space.
However, the current calibration algorithms are based on the plane calibration, if the physics to be measured is not plane, calibration will need to make some special algorithms to deal with, the usual calibration algorithm is not able to solve.
In addition, for some calibration, because the calibration plate is not used, special calibration methods must be designed, so the calibration may not be solved by all the existing calibration algorithms in the software.
Fourth: the motion speed of the object
If the object being measured is not stationary but in motion, then it is necessary to consider the effect of motion blur on image accuracy (blurred pixels = object motion speed * camera exposure time), which is also not capable of being solved by software.
Fifth: software measurement accuracy
In the measurement application, the accuracy of the software can only be considered as 1/2-1/4 pixels, preferably in accordance with 1/2, rather than 1/10-1/30 pixels as in the positioning application, because the software can extract very few feature points from the image in the measurement application.