Object location using edge-bounded planar surfaces from sparse range data

This paper describes a method for performing object location by applying edge detection techniques to sparse range data generated using a laser range scanner (LRS) developed by the National Research Council (NRC) of Canada [Beraldin, J-A, et al., 2000]. Range data consisted of closed-loop line scans resulting from Lissajous scanning patterns that are currently under investigation by the NRC [Blais, F, et al., 2001]. Current investigations in object detection and tracking using Lissajous scanning patterns have demonstrated the ability of the system to track objects in real-time using a simple planar representation [Blais, F, et al., 2001]. In this study we adapt traditional edge detection techniques to convert range data obtained using a Lissajous scanning pattern into sparse edge maps. We represent an object as a single planar surface by combining the original range data with object boundaries defined by the edge map. Data collected from typical LRS systems was used to develop a noise model for use with a simulated model of the LRS system [MacKinnon, DK, et al., 2003]. We then compared two edge enhancement methods to examine the effect of window size on edge sensitivity under simulated noise conditions. Edge maps were generated to approximate simple planar surfaces that were used to represent a single object in both simulated and real environments. Results show that this method is successful in locating a simple object in a static environment.