Image-Based Wheel-Base Measurement in Vehicles: A Sensitivity Analysis to Depth and Camera’s Intrinsic Parameters

Abstract

Image-based metric measurement has been widely used in industry for the past decade due to the recent advancement in processing power and also the unobtrusiveness of this method. In particular, this method is gaining attention in the realm of real-time detection, classification, and inspection of vehicles used in intelligent transportation systems for law enforcement. These systems have proven themselves as a plausible competition to under-the-pavement loop sensors. In this paper, we analyze the sensitivity in image-based metric measurement for vehicles’ wheel base estimation. Results lead to a simple guideline for calculating the optimal configuration yielding the highest resolution and accuracy. More specifically, we address the sensitivity of the metric measurements to the depth (i.e., the distance between the camera and the vehicle) and also internal calibration parameters of the visible-light imaging system (i.e., camera’s intrinsic parameters). We assumed a pinhole projection model with added barrel effect, aka, lens distortion. A 3D video simulation was developed and used as a Hardware-in-the-Loop (HIL) testbed for verification and validation purposes. Through a simulated environment, three case studies were conducted to verify and validate theoretical data from which we concluded that the error due lens distortion accounted for 0.014% of the total error whereas the uncertainty in the depth of the vehicle with respect to the location of the camera accounted for 99.8% of the total error.