To detect defects, information about the surface obtained using a laser, as well as conventional camera shots, is used.
The QuatroPBC system is installed on the production line after the press, above the conveyor. The production board passing under the device is non-contact measured by an industrial laser, at the same time, to detect defects in the tile face, 5 cameras simultaneously take a high-resolution photo of each pallet at 35 frames per second.
The resulting photos of all boards are processed using a pre-trained neural network. Every time there is a defect in the image, the network reports it to the operator interface using a color signal. Yellow frame - a tile with a defect, purple - a repeating defect.
The neural network processes photos at a speed of one board of 130 tiles in 2 seconds, which allows you to receive information online and, if necessary, stop the conveyor for further action.