A New Target-Tracking Approach in Flir Imagery Using Monocular Thermal Image Sequence

Abstract

In this paper, we present a new approach for tracking a moving target in forwardlooking infrared (FLIR) imagery taken from an anti-tank moving platform. First, the target image is captured in a frame with resolution 128×128 pixels by using IR camera, and then the image is enhanced by using a non uniformity correction algorithm to reduce the main dominant noises associated with infrared images. Second, the existing moving target is detected with time ( ) within the image by using 2-D Gabor filter kernels. Next, Gabor filter with 4 orientations applying on the window created with resolution 32×32 pixels around the predicted target center using Kalman filter estimator. Subtraction process is performed between the predicted center using Kalman filter estimator and the center of the detected target using Gabor filter, which is used to predict the next center by using Kalman filter, and to guide the detection of the target location in the tracking window of the next frame respectively. The proposed technique fasts and reduces the detection time into 1/16 of with high performance of tracking. The experiments performed on several real image data of an existing infrared imaging system. Results show the robustness of the proposed method, which combines high speed of detection and good performance of tracking even with strong ego-motion.