appropriate result

We have had users from a variety of industries evaluate FlexInspector. Many of them said, "It was easy to set up compared to other companies' products, so I was able to evaluate it thoroughly. The detection capability is the best we know of. This is the best detection capability that we know of.

This is a natural result.

This is because FlexInspector performs statistical processing by sampling actual good images and automatically determines the "good range for each pixel, the smallest unit of the image," at the "very edge of the line where yield can be secured. In other words, as long as the imaging conditions are the same, this method cannot further improve the detection capability.

Furthermore, it is difficult to "go further" because "all" of the image information is used as a method to select defect "candidates" to be extracted in the above.

THE ONLY METHODS LEFT ARE TRICKY METHODS SUCH AS AI METHODS, AND IT IS DOUBTFUL THAT THEY CAN BE APPLIED IN VISUAL INSPECTIONS, WHERE "NG MUST BE 100% DETECTABLE.

As for us, we are stuck with the algorithm for (high-precision) inspection. It is impossible to go any further with software alone. Therefore, we are now in the stage of considering ways to improve the image itself, including lighting, cameras, lenses, etc.