Machine vision combined with neural networks and artificial intelligence can solve previously unsolvable problems. Our neural network-enhanced vision systems for the most demanding applications think like a human in principle, but evaluate product quality with the extreme speed, accuracy and relentlessness of a machine.
The control system for evaluating the quality of a given product is trained in a similar way to training a new employee. The machine vision system learns what is considered a defect or defect in the product, what colors or shapes this defect may have based on the product samples that are presented, which are both high and low quality, and then can reliably detect these defects in various places in the product. For each product, the system evaluates the size, location and type of defect. Then, based on the tolerances you specify, a decision will be made as to whether the product is of good or poor quality.
Neural networks are most commonly used for:
Neural networks and artificial intelligence are absolutely perfect tools for checking the quality of the wood surface, where you can check, for example, the presence of knots, staining, chipping or any other defects.
In the picture below you can see an example of locating knots and other dark spots on wooden planks. In the following quality control steps, the location and size of the defects would be evaluated and based on these parameters, the quality classes would be sorted according to the tolerances you set.
Another example of quality control of unpredictable surface defects is the detection of scratches on the surface of a steel plate, which can be seen in the figure below. After complete localization, the location and size range of the defects are again evaluated and sorted into quality classes according to the tolerances you set.