Automatic Analysis of Traffic Scenario from Airborne Thermal Infrared Video

Traffic-related data analysis plays an important role in urban and spatial planning. Infrared video cameras have capabilities to operate at day and night and to acquire the scene sampled with video frame rate, but at the cost of geometric resolution. In this paper, an approach for the estimation of vehicle motion and the assessment of traffic activity from airborne IR video data is presented. This strategy is based on the separate handling of detection and tracking vehicle in the video, which differs from the common method developed to extract the object motion. The reason for it is that static vehicles are also intended to be detected. A single vehicle detector is firstly applied to find the vehicles in the image frames of video successively. Sensor movement is compensated by coregistering the image sequence under the selected geometric constraint. Afterwards, a progressive grouping concept considering temporal coherence and geometric relation is designed to recover the vehicle trajectories and classify them into static, moving and uncertain type. Image matching and the topology of trajectory are integrated into grouping process to aid the verification. Testing the algorithm on an IR video of urban area show us a promising result that 83% of moving vehicles are successfully extracted which is able to serve as basis for traffic density analysis. * Corresponding author.