This paper presents field test results for a new video vehicle detection system, Silhouette Vision. Silhouette Vision is a video vehicle detection system with side-silhouette imaging technology that utilizes two lenses and vehicle model-matching based on 3-D information. This is unlike most conventional video vehicle detection systems which only process gray-level images on a video image plane. The focus here is to evaluate the performance of Silhouette Vision under adverse conditions, such as shadows, rainy nights, twilight periods, and congestion periods where conventional video vehicle detectors do not work well. Field tests were run under these conditions and obtained 97% or more vehicle detection accuracy, confirming the system's robust and accurate vehicle detection performance in the field. (A*)
Abstract