Low-Complexity and Low-Delay Structure from Motion Approach for Advanced Driver Assist Systems

摘要

Structure from Motion (SfM) is a fundamental capability in Advanced Driver Assist Systems (ADAS) which require accuracy and performance. This paper presents an improved solution to depth estimation based on a low-complexity and low-delay (LOCAD) 3D SfM approach that requires as few as two images from a calibrated camera and that makes use of a new Multi-scale Fast Feature Point Detector (MFFPD). Results are presented to illustrate the improved performance of the proposed method as compared to existing depth estimation methods.