Human Target Detection And Tracking With Ultra-wideband Radar

Short-range human tracking is very important for safety applications, such as airport security, through-wall sensing of terrorists and disaster rescue. Ultra-wideband(UWB)radar has good range resolution and penetrating ability, which make it suitable for short-range human tracking. However, because of the peculiar working environment, target detection, identification and tracking with UWB radar need higher performance. Thus, this paper takes human target detection and tracking as the focus of the study.First of all, this paper introduces an ultra-wideband radar system, signal waveform and signal processing technique, including four kinds of usual UWB radar signals, coupling alignment, background cancellation, and multipath suppression. The coupling alignment can improve the performance of suppressing strong coupling clutter of background cancellation by compensating system waveform jitter. Index weighting method has stable performance for different motion states. Finally, time gating is used to separate the unwanted multipath and clutter from the target reflections.After completing moving target indicator, target detection’s main purpose is to make a decision that whether there is target or not. When it comes to the single human target detection, in order to reserve the target information better, the traditional CFAR methods often have high false alarm probability. At the same time, there is more background clutter. Otherwise, when it comes to multiple human targets which are very close to each other, even cross over or overlap, there will be severity target shadowing phenomenon. Therefore, this paper uses a CLEAN method which is usually used in image processing. Compared to CFAR method, it has a more stable performance on the detection of the target in different motion states and multiple targets situation, which can not only suppress clutter, multipath, and target shadowing phenomenon effectively, but also exactly retain target information, extract human multi-scatters and record the arrival delay of every scatter.Based on the returned UWB scattered waveform model from human to extract human multi-scatters through the CLEAN algorithm, which combines with the Nearest Neighbor Data Association(NNDA) algorithm and Joint Probabilistic Data Association(JPDA) algorithm, the paper can effectively track the range of moving human targets. With the increase of measurements, the load of Joint Probabilistic Data Association algorithm computation grows rapidly. Aiming at this defect, the paper proposes a improved Joint Probabilistic Data Association algorithm, which can effectively track human targets, even though the targets overlap in range. At the same time, the performance of algorithm is almost the same with Joint Probabilistic Data Association algorithm, but the computation is highly decreased.