Cooperative object recognition method of multi-UAVs based on decision fusion

In this paper, a cooperative object recognition algorithm of multi-UAVs based on decision fusion is proposed. Firstly, the SURF operator is used to get the coordinate transformation matrix and stitch the images. Then, the detection results of each UAV are projected to the coordinate system of the stitched images, and the spatial coordinates of the detection information are unified. Using Jaccard overlap and Hellinger distance to synthesize position and attribute information. Then constructing the probability matrix of the detection box association. The nearest neighbor association rule based on the global optimization is used to operate multi-objective association. Finally, a dynamic switching strategy based on Jousselme distance is adopted, which can adaptively select the DST or DSmT, to fuse the information of association detection bounding boxes according to the degree of conflict. Experiments are performed on the constructed multi-UAVs cooperative object detection datasets, and the performance of several algorithms is compared. The results show that the proposed algorithm can not only increase the patrol range, but also improve the recognition accuracy of the UAV patrol system.