Image Segmentation in Real Aperture Low-THz Radar Images

Abstract

This paper presents the proof of concept of a methodology for radar image segmentation in real aperture low-THz high resolution radar imagery, ultimately as a method to identify traversable free space for path planning for autonomous vehicles. The segmentation method, based on histogram thresholding of super-pixel statistical means is described and then applied to candidate high resolution radar images to show the potential for region finding. The subsequently segmented images are then qualitatively analysed, relevant features such as shadow and anomalous statistical regions are discussed related to identification of hazard areas for path planning.