Sonar for recognising the texture of pathways

Sonar is a popular sensor for robotic guidance, particularly for obstacle avoidance and for mapping. It has been used too for classifying objects and different surfaces. In this paper, we investigate further its use for surface classification. We discuss the characteristics of ultrasound reflection from surfaces and describe features which can extract from the sonar signature (a range-intensity plot) to distinguish between them. We show that a small set of features taken from the backscattering profile can be used to distinguish between several surfaces typical of those encountered by mobile robots with success rates approaching 100%. We demonstrate the methods using a data set acquired from a moving platform.