A fuzzy bounding box merging technique for moving object detection

In moving object detection systems, the moving objects are detected as clusters of bounding boxes based on the image differences (or motions) between frames. These differences are marked by multiple moving bounding boxes that may or may not overlapped, therefore a bounding boxing merging problem arise. The aim of this paper is to present an algorithm that derives fuzzy rules to merge the detected bounding boxes into a unique cluster bounding box that covers a unique object. In order to do this, we first define the relationships of a pair of boxes by their box geometrical affinity, by their motion cohesion, and their appearance similarity, etc. The box pair-wise relations are fuze by means of fuzzy rules and derive a fuzzy logic formulation on whether a pair of boxes can be merged or not. By considering the fuzzyness of the merging decision as a distance metric between the box pairs, the moving objects can be detected by an revised agglomerative clustering algorithm. In the experiments, we demonstrate the performance of our fuzzy moving object detection algorithm by detecting moving vehicles in aerial videos. The purpose of this note is to explore in a preliminary way the use of a fuzzy logic approach to model the uncertainty inherent in detection systems.