Object detection via receptive field co-occurrence and spatial cloud-point data

The use of image and spatial information together in mobile robots systems it is a promising field, due to the enhanced level of discrimination and efficiency that can be gained. In this paper we employ an RGB-D camera for object detection and clustering and develop methods that combine the two strands of information: first we cluster potential objects by means of their spatial position and then link geometry and co-occurrence histograms to enable reliable object detection. Experiments and design parameters are presented for example scenarios of object detection under clutter.