The technology is made of people.

Technology to find deficiencies

Three “image inspection technologies” make it possible to inspect curved surfaces, glossy surfaces, machining marks, transparent objects, and other difficult-to-inspect objects. Even minute defects are not missed.

1. Defects on curved and mirrored surfaces

Object: Curved surfaces with high gloss and strong halation, small parts with spherical surfaces, etc. (up to about 10 cm)

It is difficult to find defects in parts composed of curved mirror surfaces because light reflects off them, and even the human eye must observe them repeatedly from different angles to find them. It is also difficult to distinguish a defect from a shadow reflected on the surface. We have developed a method to detect defects by reflecting a specific light pattern onto the object to be inspected and analyzing the image for deformation. This method enables stable detection of defects as shallow as several tens of micrometers without the need for repeated observation from different angles.

2. Defects that are assimilated with processing marks

Object: Small parts, etc., precision-machined in units of several tens of microns. (10 cm or less)

For example, it is not easy to find defects of several tens of micrometers on machined parts that have machining marks on the surface, because they are assimilated with the machining marks. We are able to distinguish between the machining marks and defects by frequency analysis of the image, and can accurately detect pits as small as a few dozen micrometers. Not only pits, but also other defects can be detected.

3. Defects in transparent or translucent bodies

Object: Transparent and translucent objects

It is difficult to find defects in transparent and translucent objects because light is deflected and reflected, and conventionally these objects have been inspected by the human eye. We use multi-colored light sources and an algorithm that suppresses the effects of ambient light and individual differences in the inspected object to accurately detect the surface and internal conditions of transparent and semi-transparent objects. Any size of object can be inspected. Any material that allows light to pass through can be inspected, regardless of its hardness.

Technology to determine deficiencies Technology to achieve stable determination

Obtain stable judgments through mechanization

Visual inspections by humans are prone to human error and inconsistencies in inspectors’ judgments. In contrast, automatic visual inspection, which uses image processing of camera images to make judgments, can always make stable judgments.

Stable judgments are obtained through flexible judgment algorithms and shared judgment conditions.

Although human intuitive judgment is superior to that of machines, it is difficult to accurately convey the judgment criteria to others. We use algorithms that optimize judgment criteria such as size and length of defects as well as shape characteristics and color according to the object to enable judgment that is close to human judgment, and then convert the judgment conditions into data for shared use. This enables stable judgment even when switching inspection targets or increasing the number of inspections. Translated with www.DeepL.com/Translator (free version)

Stable judgment is obtained by setting easy judgment conditions.

Even with automation, it is sometimes difficult to set the conditions for judgment, resulting in excessive judgment, which worsens the yield rate, or conversely, NG judgment may be too lenient. We combine intermediate judgments based on machine learning with judgments based on numerical conditions to make realistic judgments that are not excessive.
Machine learning usually requires the registration of about 10,000 samples, and the registration process is a challenge.
By limiting the use of machine learning to intermediate judgments, we are able to make stable judgments with only about 100 registered good samples.

Technology to feed back inspection results to the manufacturing process

Deficiencies can be quantified and classified.

Visual inspection is an essential process, but it cannot improve quality. To improve quality, it is necessary to collect data on defects and provide feedback to the manufacturing process. By systematizing visual inspection, defects can be quantified, classified, and totaled to accumulate and utilize the data.

Quickly reflect inspection results.

Our equipment not only classifies and stores individual defect data, judgment results, classification statistics, and image data, but also statistically processes judgment results and provides quick feedback to the manufacturing process through a function that allows users to check the inspection status in real time.

Accumulated data can be easily utilized.

We provide real-time data sharing and aggregation functions using a network, and offer a centralized management system to meet your needs. This makes it easy to consolidate and utilize data even for multiple devices and a huge number of inspections.