Combining advanced image processing and automation technologies, DECSYS visual inspection equipment

► Semi-automatic inspection system SA-C series

Learning and optimization of inspection criteria thresholds Easily improves inspection quality by comparing product images with good models. Operators only need to manually insert workpieces. Only the inspection is automated, which reduces costs while eliminating human defect outflow and enabling quantitative inspection. Quantitative inspection is now possible.

Just by putting the workpiece into the inspection unit, Anyone can perform quantitative inspection

No operator visual inspection training required. Camera inspection ensures consistent quality.

Stable detection of "flaws" that are difficult to detect even by visual inspection

Highly accurate inspections can be performed by the multiple imager. (Patented technology)

On a table-top machine that does not require a PLC, Low cost due to semi-automatic operation

No technical knowledge is required to use this system. Motor and lighting control with dedicated software

DECSYS' Only One Technology! Multiple Imager

Outline of equipment

W600 x D400 x H800

target work

・Shaft length: up to 150 mm, diameter: 3 to 10 mm
・Cylindrical parts Height: ~50mm Diameter: 10 to 30mm

inspection capability

1sec to 2sec per piece (including rotation and processing time)

detection capability

Scratches, dents, foreign objects 0.15mm～

► Automatic sorting machine for appearance of all shaft circumference and both end faces

DECSYS, the only company that combines image processing and automation technologies, DECSYS is a one-of-a-kind company that combines image processing and automation technologies.

Detects defects such as foreign matter, dirt, scratches, and dents at the level of 0.2 mm square minimum.

► Linear visual inspection system

High-speed inspection of linearly conveyed products with up to 5 cameras (1 top camera and 4 side cameras)! Dirt, scratches, and dents are detected without omission.

Applicable to various products such as metal parts, resin molded products, and rubber products.

Ideal for top or side full-surface inspection (Number of cameras: up to 5 cameras)

Because a fixed quantity can be inspected automatically, One person can perform multiple operations with other equipment. Multiple operations are also possible.

Applicable to various products such as metal, resin molding, rubber products, etc. Applicable to various products

target work

Scratches, dents, short molds, foreign matter 0.2sq.～

throughput

40-60 pieces/minute (varies depending on workpiece shape)

Target size

～50mm dia. x 50mm H (Other sizes are available upon request.)

With AI image inspection using Toshiba's unique good quality learning method, Automation of inspection and improvement of inspection accuracy in a short period of time

Challenges in AI learning methods

Issues in Inspection Equipment

​

​

​

Image inspection for on-site use based on Toshiba's technology and know-how

► Functions provided / Form of provision

● Learning Tools

Good model is created by inputting training images and setting parameters on the screen.

● Good/failure judgment engine

Compares the good model with the image to be inspected to determine whether the product is good or bad.

● Image preprocessing

Performs pre-processing such as position correction of the image to be inspected

● Camera I/F

Receives images to be inspected taken from the camera and passes them to image preprocessing.

● Judgment result display screen

This screen displays real-time and historical results of the good/fail judgment.

The software package for the above functions is available as a subscription-type annual license. Hardware other than software, such as cameras and lighting, is not included in the package. (Individual quotation)

► Good learning method / Application examples

◆ Compare product images with a model of good products learned and optimized for inspection criteria thresholds.

Statistical learning of acceptable threshold values for good products based on variations in images of good products, and judging products that deviate from the threshold as defective.

◆ Threshold optimization method to reduce false positives (*Patent pending)

Detects areas prone to false positives and iteratively trains to optimize the threshold of the good model to prevent false positives from occurring in those areas

◆ Welding inspection at an automotive parts manufacturer

Applied to the inspection of welding, which is considered difficult to inspect without skilled workers due to the variability in workmanship, even for good products.

◆ Automation of visual inspection by inspectors

Automates visual inspections in the inspection process, where inspectors used to perform visual inspections and defect corrections, reducing inspection workload while maintaining inspection quality.