Sightech Machine Vision - Self Learning Eyebot

Applications - Water Bottles - Check Presence of Lot Codes and placement of Labels

Bottling companies strive for quality when producing bottled water. It is important to check for several possible defects such as presence of a lot code and label. Using simultaneous inspection Areas, Sightech's trainable vision quickly learns how to identify the label and lot code, and signals if either are missing. Of particular interest is how well trainable vision detects the lot code when it is even difficult to see with the human eye! In this example the lot code is imprinted into the clear plastic of the bottle - there is no color used. Note how well this code is detected.

Running, the trainable vision system detects the presence of the lot code (yellow) and the missing label (pink). The decision is an OR of these two simultaneous inspections. (above)	Bottling water requires several inspections - lot codes, labels, caps and fluid level. This page demonstrates two of these inspections.

Example of a lot code - we use trainable vision to make sure they are present on every bottle. Notice how subtle the markings are. (above)	Step 1 - Setup: Two separate inspection Areas are used - the lower Area for the label area, and the upper Area for the lot code. (above)

Step 2 -Training. In Learn mode, the small orange "hits" indicate learning activity in progress. This happens very quickly - usually in a few seconds. The top Area is learning to detect the Presence of the lot code, and the lower Area is learning where a good label is placed. (above)	Step 3 - Forgetting:: For the top Area (lot code checking), we use the Forget mode when the lot code is missing. Doing this makes the system very sensitive to the Presence / Absence of the lot code. The lower Area does not require this step. (above)

Step 4 - Run: The system detects presence of the lot code and label. Since the label checking Area is in Inspection Mode, no color is a good thing. Inspection mode signals when the product differs from what was trained. This example is a PASS because the top Area shows presence of the lot code and the lower Area signals no defective label. (above)	The example shows a FAIL condition on the Lower label-checking Area. In this case, the label is completely missing. This Area also checks the label for position, up or down, errors. This example is a FAIL condition. (above)
We have shown several important things: 1) Simultaneous inspection tasks - each with its own Area 2) Easily detects presence of subtle imprinted lot codes 3) Trains in seconds 4) Combines decisions from all Areas to make a combined decision 5) NO PROGRAMMING REQUIRED!

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