Robust and Accurate Objects Measurement in Real-World Based on Camera System

Abstract

Object's dimension and its proximity in real-world plays a critical role in safe navigation and collision avoidance in autonomous cars. An accurate, reliable, and cost-effective approach was developed in this paper to measure the object's dimension (distance, width, and height) in real-world solely based on camera system. Mathematical representations were derived to accurately measure object dimensions and extract extrinsic camera parameters while driving. Giving the bounding box coordinates around each object in the captured frame, the proposed approach automatically and accurately measure the object's dimension in real-world (ft.) instead of pixels. The derived models were verified and tested against the ground truth data which showed strong correlation.