Image Segmentation of Cross-Country Scenes Captured in IR Spectrum

Abstract

Computer vision becomes a major source of information for autonomous navigation for different types of vehicles including military robots and mars/lunar robot rovers. Nevertheless, the attention is mostly focused on texture descriptors with fixed scales. Besides, the majority of work is done in analyzing images captured in visual spectrum. In this chapter we elaborate on the problem of segmenting cross-country scene images in IR spectrum using salient features. Salient features are robust to variations in scale, brightness and angle of view. As for salient features we choose Speeded-Up Robust Features (SURF). We provide a comparison of two SURF implementations. SURF features are extracted from input imaged and for each of features terrain class membership values are calculated. The values are obtained by means of multi-layer perceptron. The features class membership values and their spatial positions are then applied to estimate class membership values for all pixels in the image. The values are used to assign a terrain class to each pixel in the image. To decrease the effect of segmentation blinking and speed up segmentation, we are tracking camera position and predict positions of features.