Close range inspection using novelty detection results

Abstract

This paper presents a close range inspection strategy which requires minimum information from mobile novelty detection results. We propose the use of the estimated position of the detected anomaly to assist path planning for close range inspection. This strategy allows the robot to have more inspection coverage of the surrounding perimeter of the object which in turn will increase the performance of detection using sensors with limited work range. A down sampled laser range finder measurements are used to show that the strategy is robust and work with any range sensors including one that is noisy, inaccurate and has low angular resolution. Experimental results show that the close range inspection brings the robot near the perimeter of the anomalous object and also achieved total close range inspection coverage. © 2009 Springer-Verlag Berlin Heidelberg.