Optical sensor-based object detection for autonomous robots

In this paper, we introduce objects detection optical sensors (CCD sensor: Charge Coupled Device). For autonomous robot, the location of around-object is very important because robot should avoid it for driving. In the field of computer vision research, the various object detection methods have been used by research engineers. In particular, the combination of Haar-like feature and AdaBoost classifier is a popular method for object detection. It has been used for face detection, but performs well for other object detection too. So it has become the choice of many researchers in the intelligent autonomous robot field. It is prone, however, to yield many false-positive results and use excessive processing time. We propose a solution for overcoming this limitation. We begin by normalizing the image database to improve the accuracy of classification. And optimizing AdaBoost training allows us to get the short computing time and accurate detection. Our experiments prove the superiority of the proposed method.