Parallax Error in Video-Image Systems

Abstract

Parallax error is a distance-related measurement error when the measurement has to rely on field of view, such as video-image systems. This paper provides a detailed investigation of the geometric elements associated with parallax error in video-image systems. A field-of-view calibration model is presented based on a study of various geometric elements, including camera height, camera location, view of image, and vehicle dimension. The study found that parallax error ranges between 10 and 50% based on normal camera setup. The model presented provides a better understanding of the parallax error; thus the accuracy of distance-related measurements could be improved when video-image systems are deployed. The proposed model has special application values at signalized intersections where video-image systems are deployed. One of the applications is to identify the range of possible blockage by a leading vehicle, which is one of the major causes of missed detections. Another application is to identify the conditions that result in vehicle occlusion, which is a source of false detections.