Speed detection of moving vehicles from one scene of QuickBird images

Abstract

A new method is developed to extract moving vehicles and subsequently detect their speeds from a pair of QuickBird (QB) panchromatic (PAN) and multi-spectral (MS) images automatically. Since PAN and MS sensors of QB have a slight time lag, about 0.2 seconds, the speed of moving vehicles can be detected by the movement between PAN and MS images in the time lag. From a PAN image with 0.6m resolution, vehicles can be extracted by an object-based approach. But it is difficult to extract the accurate position of vehicles from a MS image with 2.4m resolution. Thus an area correlation method is proposed to estimate the location of vehicles from MS images in a sub-pixel level. Using the results of the vehicle extraction, the speed of moving vehicles can be detected. The approach is tested on several parts of the QB image covering the central Tokyo, Japan, and the accuracy of the result is demonstrated.