Collision and Reaction Model of Feature Fusion: Mechanism and Realization

Duplicate data are commonly seen and need to be properly handled in the process of an intelligent search/surveillance/tracking etc. For better detecting and fusing them, we take a brand new perspective on feature fusion based on the “collision and reaction” mechanism of data objects. This mechanism can greatly facilitate differentiating duplicate data and non-duplicate data in fusion (reaction) probability. With this perspective, the fusion (reaction) probability of a feature sample in a dataset is the cumulative effect of the microscopic collision (interaction) results between this sample and other samples. The operations on duplicate detection and feature fusion depend on the fusion probability and are finally realized when the collision process is over. This new quantum inspired feature fusion model gives interesting experimental results and could be of great implications in some intelligent systems.