What determines the cost of an "imaging test."

In a recent comment by masa.
>We have tested each company using the same defect sample.
>I have found again that the high price is not proportional to the detection capability.
We have received the following.
The price of a system up with #FI is the same level as a mid-to-high-end image sensor.
The cost of a typical "image inspection" is
1) "Custom" or "standard"? If it is "custom", it is not surprising that it takes about 3 man-months of development. If so, it will cost at least 3 million yen. Considering that the price of "image sensors" sold as "standard products" is generally less than 2 million yen, the difference is significant.
(2) Whether or not the hardware is made in-house. Unless the company sells a large number of these products through its sales network, as it does with image sensors, the number of units shipped will not increase that much. If this is the case, the development cost of "in-house hardware" will inevitably be high because it must be paid for. Another troublesome point is that hardware advances very quickly and quickly becomes obsolete. This can be a hindrance in terms of performance.
3) Company size (number of employees). In order to support a large number of employees, you have to sell at a high price. In the case of image inspection, where "hands-off" is not an option, support personnel will be needed because "on-site" work is common....
(4) Line sensor or not. Experience shows that line sensors require three times as many man-hours. Since the transport system factors are significant, they do not work as well as area cameras.
5) Expensive camera, expensive illuminator, expensive lens. Is it really necessary? It depends on how the image processing is done.
The following are some of the factors involved.
Also, don't forget the enthusiasm of the person in charge. A "see for yourself" attitude that spares no effort goes a long way.
Imaging inspection is a substitute for work that people can do. Even if it works properly, the labor-saving effect gained is not great. Considering that automation would require handling, "less than 2 million for the imaging part" is my theory.