Feature-based Classification for Image Segmentation in Automotive Radar Based on Statistical Distribution Analysis

Segmentation and potential classification of surface and obstacle regions in automotive radar imagery is the key enabler of effective path planning in autonomous driving. As opposed to traditional radar processing where clutter is considered as an unwanted return and should be effectively removed, autonomous driving requires full scene assessment, where clutter carries necessary information for situational awareness of the autonomous platform and needs to be fully assessed to find the passable areas. In this paper, the statistical distribution features of the radar intensity data of several road-related scenes including asphalt, grass, shadow and target object areas are investigated. The algorithm of classification is developed based on distribution feature extraction and a multivariate Gaussian distribution (MGD) model. Under test dataset recorded by multi-sensor suit was used to evaluate the confusion matrix and F1 score of this classification algorithm.