Using general master equation for feature fusion

Abstract The rational division of subsets is a key issue for feature fusion, which often requires that the feature data units in different subsets can be differentiated easily. Regarding this, this paper uses the transformation effect between microscopic and macroscopic of general master equation to widen the differences of fusion probability between the feature data units in different subsets. Then, based on the more differentiable feature data units with widened fusion probabilities, this paper proposes a new dynamic quantum inspired feature fusion method, which uses the Wootters statistical distance in probability space to detect the duplicate feature data and uses the weighted median operation to fuse the detected duplicate feature data. The experimental results show that the fusion performances on fusion rate, relative completeness, and conciseness of the proposed feature fusion method are encouraging.