Object detection and localization in clutter range images using edge features

Abstract

We present an object detection technique that uses local edgels and their geometry to locate multiple objects in a range image in the presence of partial occlusion, background clutter, and depth changes. The fragmented local edgels (key-edgels) are efficiently extracted from a 3D edge map by separating them at their corner points. Each key-edgel is described using our scale invariant descriptor that encodes local geometric configuration by joining the edgel at their start and end points adjacent edgels. Using key-edgels and their descriptors, our model generates promising hypothetical locations in the image. These hypotheses are then verified using more discriminative features. The approach is evaluated on ten diverse object categories in a real-world environment. © 2009 Springer-Verlag.