LiDAR point cloud analysis for vehicle contour estimation using polynomial approximation and curvature breakdown

A high definition of perception sensing is significant for the advancement of driver assistance systems and autonomous driving systems. The objects detected in the field of view of perception sensors can be further analyzed to determine the shape, orientation, etc., thereby enabling possibilities of new functionalities for ADAS and integrated safety systems. This research work discusses a novel approach to precisely estimate the contour and approach angle of the oncoming vehicles, using the point cloud data from a LiDAR sensor. The actual contour is segmented into three regions and each one is represented by an arc of a circle. Hence, a combination of simple quadratic equations accomplishes the representation of a complex contour geometry. The polynomial approximation mitigates the effect of noise and curvature analysis is used to derive anchor points to split the point cloud to compute the individual radii. The test results discussed, conclude the efficiency and accuracy of this approach.