Modeling of visual flotation froth data

Abstract In this paper, the principles of sensor fusion are presented. A new sparse coding method based on a generalization of the generalized Hebbian algorithm (GGHA) is presented. The algorithm is realized using a modification of the Kohonen network. The method is tested on an image analysis of flotation froth, in order to find features corresponding to the poisoning phenomenon in a flotation cell. The features found are capable of predicting the poisoning earlier than the ordinary process instrumentation.