Target recognition of ladar range images using modified local surface patches

Abstract

In this work, a modified integrated local surface descriptor is introduced for laser radar (ladar) range image recognition. A 2D histogram binary image is added to local surface patches (LSP) as a supplementary property, and a binary dissimilarity function is proposed to further filter the matching of LSPs. Experimental results using simulated range data show that the modified descriptor, which is more stable and efficient than the original descriptor, determines the effect of projection density variation and self-occlusion on LSP. The comparison experiment using real range data demonstrates that the modified descriptor has high recognition rate in model libraries with different sizes. This work is beneficial for establishing the ladar range image recognition system with high efficiency and recognition rate.