Sightech Machine Vision - Self Learning Eyebot

Applications - Detect Exposed Fruit, Damaged Packaging, Defective Labeling in Raison Boxes

Packaging inspection can be challenging for ordinary machine vision because of the great variety of defects that can occur and the rather uncontrolled presentation. Self-learning vision overcomes many of these obstacles and can provide an effective solution. This raison box example shows how easily a trainable vision system can be setup and deployed. It also shows how well it can detect normal defects encountered in an everyday packaging application.

The trained vision system detects protruding fruit - the green "Hits" show the location of the defect. (above)	Protruding fruit, label defects, and damaged packaging can all be easily detected. (above)

Step 1: Setup the camera, learning mode, and video filtering. (above)	Step 2 Learn:: Train for about 1 minute on production packages. The orange "Hits" on the image show the learning activity. (above) As learning progresses, the quantity of "Hits" diminish.




Step 4 Recognize: Operate - the green "Hits" show where the defect is detected. User settable decision thresholds, "Sensitivity" and "Region", allow tailoring decision sensitivity to the needs of an application. (above)
In summary, we have shown that the PC-Eyebot can easily be trained to inspect packages for protruding materials, physical damage, flaps not tucked in properly, and labeling defects.What would normally be a difficult and complicated programming process is replaced, using self-learning vision, with a simple, painless, and brief training process. This demonstrates the flexibility and power of trainable vision.

Home | About Sightech | Contact | Products | Applications | Support | Technology | Careers | Site Map | Resources