A stereo range computation method using thermal infrared and visible cameras

Binocular stereo vision can provide geometric position information of the target, which enables one to track the moving object precisely. Although visible images are full of details of geometry and texture, it is difficult to detect moving objects in poor visibility. Appropriate fusion of infrared and visible images can combine the complementary information and obtain a better description of the scene, which will help in target detection and target localization. Considering the physical differences between thermal infrared cameras and visible cameras, a stereo ranging method in parallel camera configuration is proposed in this paper. The rotation center of a pan-tilt device is used as the origin of the coordinate system in the method. It could not only locate the object in poor visibility, but also determine the parameters essential in range computation in practical work, which is very difficult in traditional calibration methods. Furthermore, the effect of the relative position of a certain target and reference points is analyzed. Experiments also proved the validity of the proposed method.