Artificial Intelligence Algorithms for Color and Pattern Identification


A major industrial products company sought a system for identifying more than 1,500 products that all had different colors, different patterns and textures. Their former color-only system was able to identify a good portion of the products but did not allow for a complete identification of all products, especially the ones that had very similar colors or the same color and shade but somewhat different surface finish.


MSV developed a dual camera system that identified the precise color using a VNIR hyperspectral camera and an additional low-angle line scan camera to enhance the texture differences.  The data was combined using unconventional artificial intelligence routines that were able to resolve all critical product groups. Once the algorithm was developed, the prediction of an unknown product can be done real-time at the speed of the machine producing and transporting the product. The feasibility of such a complex system shows the power of the extensive data produced by the hyperspectral cameras and the special algorithms that can produce the identification on-the-fly in industrial settings.