A Simple and Precise Segmentation Framework for Drone Swarm via SAM Proposals and Group Similarity Analysis

Intelligent swarm drone recognition methods are becoming more and more important with the advancing of drone techniques. Most current methods treat drone segmentation independently, neglecting the usage of inner-swarm group similarity characteristics. This paper tackles the task of infrared-domain swarm drone segmentation in remote sensing images from the aspect of group similarity analysis. We use the state-of-the-art segmentation algorithm as an initial module, and incorporate the segment anything model (SAM) to generate high-quality proposals, after which a specially designed candidate selection module carries out the group similarity analysis among swarmed drones to produce a finer segmentation mask. Experimental results on the Swarm Drone Detection dataset demonstrate state-of-the-art performance for the proposed method, and validate the effectiveness of introducing group similarity into swarm drone tasks.