An Object Recognition Method for Indoor Robot Based on Feature-Line Flows

An object recognition method for agent, indoor mobile robot for instance, is presented based on feature-line flows. Considering the surrounding of indoor scenes, we extract the feature lines from image sequence captured through camera by using Radon Transform, and then calculate their optical flows; Secondly, process those flows through PCA, distinguish moving objects from static objects by computing residual errors; If objects can not be recognized correctly because that they are occluded by something else or similar highly to other objects, we use the secondary way-rough set processing-to do it, which is based on object features. Finally, experiments´ results are presented and they show that the method appears well on both efficiency and anti-noise.