Solving this problem required a more advanced setup than a traditional machine vision application - what JLI vision refers to as “hybrid vision”.
To replicate the human evaluation of the knots and resin pockets in the boards, JLI vision applied machine learning by processing 30,000 images of both approved and rejected items in order to train the system to be able to distinguish defects.
The machine learning network was incorporated into the scanner system, JLI vision built to inspect the glulam boards on both sides.
During inspection, the boards are transported on two conveyor belts that move at approximately 0.5 m/s. In the small gap between the two conveyor belts line scan cameras in combination with 3D cameras scan and map out both the front and the back of the board.
The human inspection process included touching and feeling knots to determine if they were at risk of loosening. This would obviously not be possible with machine vision, but by creating a 3D image of the surface, JLI vision was able to determine which knots would fall in that category.
If a defect is detected the board is directed to a repair line, where the defect is repaired according to the images created by the scanner system.