Research on unsupervised multi-ship image segmentation algorithm of remote sensing platforms

For the complexity of the ship feature classification and sensitivity of optical sensor to weather, airspace, illumination and other marine environment, we propose a new unsupervised multi-ship image segmentation algorithm of remote sensing platforms, including separation from the target area of the adaptive multi threshold criterion, algorithm based on multiscale Ncut hierarchical, multi-objective analysis of ship motion. Finally, through verification and analysis by experiment, comparing with other relevant segmentation methods, our method has better robustness and recognition effect.