Certainly illuminators are important. But...

FlexInspector is based on the principle that instead of detecting "defects" that are difficult to define, it registers "good products" that are easy to define and detects "where they differ from good products".

What we want to consider here is the illuminator.

The "general" method of selecting illuminators has as its primary goal, for a given defective sample, to increase the contrast of the defective area. Again, this is a "defect" based approach. As a result, it is almost like "there are as many types of illuminators as there are types of defects.

IF WE BRING IN THE FI STYLE OF THINKING, THE FIRST REQUIREMENT FOR ILLUMINATORS IS THAT GOOD PRODUCTS SHOULD BE STABLY VISIBLE, AND THE SECOND REQUIREMENT IS THAT DEFECTS SHOULD BE VISIBLE ON THE IMAGE, WHICH CHANGES THE SELECTION CRITERIA FOR ILLUMINATORS SIGNIFICANTLY.

Thinking about it again, increasing the contrast of defects means that a slight variation (i.e., defect) causes a large change in the image, which is in conflict with the "stable appearance of good products. On the other hand, "fixed binarization," which is the mainstream of existing image processing algorithms, is an algorithm that expects "good products to appear stable," and there is a problem in matching "illumination" and "algorithm. An illuminator that increases the contrast of such defects should be combined with "dynamic binarization processing" or other algorithms that can follow fluctuations.

On the other hand, an illuminator that can "see good products stably" is one that can create a shadow-free state where the observation point is illuminated from all directions. An example is "coaxial dome lighting. Images taken with this illuminator are similar to those taken by normal observation with room light, and even if the direction of observation changes slightly, there is no significant change in brightness.

However, the weakness of this lighting is that it is difficult to see three-dimensional defects such as scratches. These are the type of defects that can finally be seen visually due to their position in relation to the illumination. This type of illumination uses coaxial light with strong directionality. However, this method is "unstable" and must be combined with an algorithm that can follow fluctuations.

Eventually,

1. Visual inspection by general indoor light = coaxial dome illumination + good comparison algorithm
2. Visual inspection while changing angles = coaxial illumination + dynamic binarization and other algorithms capable of following luminance variations

It seems that you can cover a great deal of ground by holding down two types of

ONE OTHER POINT I WOULD LIKE TO ADD IS THAT, AS THE TV COMMERCIALS SAID, "VISION IS ALSO REDUCED IN THE DARK". THE BEST WAY TO INCREASE THE CONTRAST OF DEFECTS IS SIMPLY TO MAKE IT BRIGHTER. A FLUORESCENT LIGHT WITH LOW DIRECTIONALITY AND HIGH LIGHT INTENSITY IS ACTUALLY A VERY GOOD CHOICE.

Certainly the illuminator is important. But it makes no sense to discuss the illuminator by itself; it is important to combine it with the appropriate algorithm for the lighting method.

SUPPLEMENTAL FI HAS AN "ALGORITHM THAT CAN FOLLOW LUMINANCE FLUCTUATIONS" IN ADDITION TO THE "GOOD COMPARISON ALGORITHM".