【作者】 库亮；

【导师】 周治平；

【作者基本信息】 江南大学 ， 检测技术与自动化装置， 2009， 硕士

【摘要】 在雷达目标检测跟踪过程中,对于获得的远距离图像,目标成像面积小,检测到的信号较弱,特别是在复杂背景干扰下,目标被大量噪声所淹没,导致图像的信噪比很低,目标检测变得困难。因此,低信噪比条件下序列图像运动小目标的检测问题成了一个亟待解决的关键问题,探索和研究新的小目标检测理论以及如何将现有的检测理论应用于小目标仍是一项重要的课题,对现代战争以及未来战争具有深远的意义。在中高级海情的海杂波干扰背景下,脉冲多普勒机载雷达要检测海上静止或慢速运动的目标,特别是如潜艇望远镜、通气孔这样的小目标有许多难点。由于目标没有多普勒频移或者目标的多普勒频移比较小,而背景海杂波仍存在一定的多普勒频移,采用传统的PD技术对目标进行多普勒频移分辨已十分困难,需要考虑采用其他的方法来分辨小目标。由于机载雷达的特殊性,不能单纯依靠减小海杂波干扰绝对强度的办法来解决这一技术难题,须以提高信噪比与降低海杂波虚警概率为方向,解决机载雷达在中高强度海杂波背景下检测海上静止或慢速运动的小目标的技术难题。检测和跟踪弱小目标有两种方法:传统的跟踪方法和先跟踪后检测方法。传统的方法使用复杂的信号处理和跟踪方法来对接收数据取门限,然后送入单独的跟踪算法。先跟踪后检测方法结合信号处理方法和跟踪方法,目标的检测和跟踪确认是同时进行的。本文的主要工作如下:1.研究了一种有效的先跟踪后检测方法——动态规划方法应用于雷达小目标的检测和跟踪。这种方法在多帧图像数据中沿着目标轨迹积累能量,在多条可能的轨迹中,选择积累值最大的一条作为目标轨迹。通常将轨迹看成由一系列状态组成,每个状态描述了组成轨迹的每个点的相关信息。对轨迹的寻优过程就是对状态序列的优化过程。2.研究了一种跟踪杂波中弱小目标的数据关联方法——面向航迹的多假设跟踪的应用。多假设跟踪算法能够在很高虚警概率的情况下,大约是最近邻方法准确跟踪目标时虚警概率的10倍,保障准确的目标跟踪。3.结合前面两点的内容,研究了一种两阶段的跟踪方法。这个过程用动态规划方法作为第一阶段从源数据中检测到可能的航迹段,然后用多假设跟踪作为第二阶段来连接第一阶段产生的航迹段,并完成最后的航迹确认。通过这种两阶段的检测和跟踪确认过程,对于低信噪比目标的检测可以获得很好的性能。更多还原

【Abstract】 In the radar targets detection and tracking process, regarding the long-distance range image which is obtained, the image formation area of the targets is small, and the detected signal is weak, specially under the complex background disturbance, the targets are submerged by the massive noises, that causes the signal-to-noise ratio of the image to be very low, and the targets detection become difficult. Therefore, under the low signal-to-noise ratio condition, the moving dim targets detection question in the sequence images has become the key question which urgently awaits to be solved, exploring and studying the dim targets detection theory as well as how to apply the available detection theory in the dim targets will be still an important topic, which will have the profound significance to the modern warfare as well as the future war. Under the intermediate and senior sea sentiment sea noise jamming background, it is difficult that detecting marine static or the slow movement targets using the pulse Doppler airborne radar, specially such dim targets as the submarine telescope and the air vent. Because the targets do not have the Doppler shift or the targets Doppler shift is quite small, and the background sea clutter still had certain Doppler shift, it is extremely difficult to distinguish the Doppler shift of the targets using the traditional PD technology, so we need to consider other methods to distinguish the dim targets. As a result of the airborne radar approach particularity, we cannot depend upon the methods which purely reduces the sea noise jamming absolute intensity to solve this technical difficult problem, and must enhance the signal-to-noise ratio and reduce the sea clutter false alarm rate, then we can have the solution to the difficult problem that detecting marine static or the slow movement dim targets under the high strength sea clutter background using airborne radar.There are two basic approaches for tracking dim targets as conventional and track-before-detect (TBD). The conventional approach uses sophisticated signal processing and tracking methods to produce observations that, after thresholding, are sent to a separate tracking algorithm. The recently proposed TBD approach combines signal processing and tracking so that detection and track confirmation effectively occur simultaneously. The main work of this paper is as follows:1. It is studied how to apply dynamic programming algorithm, which is an effective track before detect method, in the radar dim targets detection and tracking. This method accumulates the energy in the multi-frame image data along the target trajectories, and chooses a target trajectory whose accumulation value is biggest from all possible trajectories. We can usually consider a trajectory is composed of a series of states, and each state describes the related information of each pot in the trajectory. The trajectory optimization process is also the state sequence optimized process.2. A data association method -- track-oriented algorithm which is ideal for tracking dim targets in clutter is studied. The MHT algorithm can typically extend tracking operation to a false alarm density that is at least 10 times greater than the density at which a nearest neighbor type method can operate.3. According to the content described above, a two stage tracking method is studied. This process will use dynamic Programming algorithm as the first stage to detect likely track segments in the raw data. Then MHT is used as the second stage to link the track segments produced by the first stage and confirm the final track. The ultimate performance against low SNR targets can probably be obtained using this two-stage detection and track confirmation process.更多还原