Shape-guided superpixel grouping for trail detection and tracking

摘要: We describe a framework for detecting and tracking continuous ldquotrailsrdquo in images image sequences autonomous robot navigation. Continuous trails are extended regions along the ground such as roads, hiking paths, rivers, pipelines which can be navigationally useful ground-based or aerial robots. Our approach to single-image trail segmentation incorporates both bottom-up top-down processes. First, good grouping hypotheses efficiently generated by probabilistic clustering of superpixels based on color similarity. Second, robustly ranked with an objective function comprising shape, appearance, deformation terms. The shape term measures how well triangle, approximate template viewed under perspective, fit groupingpsilas boundary. appearance reflects visual contrast between its surroundings using between-class/within-class scatter measure. Finally, closeness fitted triangle learned distribution captures expected size, location, other degrees variation. Although detection is accurate reasonably fast variety isolated images, we introducing temporal filtering stages increases accuracy per-frame speed over sequences. Results shown varied collected from flying driving platforms, sampled Web.