Object recognition via multi-view inspection using saturation-weighted distributive hue histograms and depth information

A computer vision algorithm is presented for the detection and localization of objects within unknown environments. To search for a specific object within an image, information about the object's appearance is first extracted from database images of the object. To effectively represent the appearance of the object, the Saturation-Weighted Distributive Hue Histogram is presented which encapsulates the intrinsic color information of the object as well as the spatial arrangement of colors within the object's boundaries. These histograms are then searched for in the scene image of the environment. The use of depth information allows searching for objects at depth variant planes in order to achieve scale invariance and to obtain the 3D coordinates of match candidates. A mobile robot platform is employed to move to various locations within the environment and to inspect matching regions by capturing and processing further images.