复杂杂波背景下的多策略目标检测方法研究

发布时间：2018-05-07 16:00

本文选题：高频地波雷达 + 目标检测　；参考：《哈尔滨工业大学》2014年硕士论文

【摘要】：高频地波超视距雷达（High Frequency Surface Wave Radar, HFSWR）利用高频电磁波在海面绕射原理，有效的解决了传统微波雷达的隐身技术所带来的检测性能损失，实现了大范围、全天候远程海态监测和关注于舰船目标检测的海事监控。由于大量的外部噪声和多种类型的杂波占据了部分HFSWR检测背景，使其表现出时变、多源、起伏的复杂特性，仅应用经典的恒定虚警概率（Constant False Alarm Rate，CFAR）算法易于导致较高的虚警和较差的检测性能，使得HFSWR舰船目标检测成为一个难点问题。为有效地改善多目标共存环境下的舰船目标的检测性能，本文研究了一种基于背景识别分类的自适应多策略目标检测方法。具体包括检测背景分析、背景分割处理、背景数据指数归一化等部分，主要研究内容如下： 1，深入研究HFSWR检测环境，介绍其不均匀，多杂波的复杂特性，并且着重分析了几种主要杂波类型的产生原理和物理特征。为后文基于杂波类型的背景分割做理论铺垫，得出了进行背景感知和信息提取的必要性结论。 2，针对检测背景的多源、不均匀的复杂特征，对检测背景做分割处理：利用K-L散度原理，将检测背景分割为均匀部分和分均匀部分；按照杂波类型分割雷达检测背景为四个杂波区域，基于统计理论分析每一个分割区域，得出各分割区数据服从不同参数的威布尔分布的结论，通过上述研究，实现对检测背景有用信息和知识的提取，用于目标检测策略的设计和参数选择，以完成利用检测环境感知技术的多策略检测系统。 3，为进一步提高检测策略的性能，研究HFSWR检测背景的预处理技术，利用曲线回归理论去除背景中影响数据统计和背景区域分割的大功率点。介绍了背景数据的指数归一化技术，把服从威布尔分布的背景数据归一化为服从同一参数的指数分布数据，实现检测背景的均匀化处理，有效解决检测背景数据统计差异所带来的检测性能损失，增强恒虚警检测的性能。在以上研究基础上，提出一种基于预分割的双参数CFAR检测方法，按照分割结果分别对各分割区域进行指数归一化，并分别进行经典的CFAR检测处理，有效的改善了检测性能。 4，综合以上研究成果，提出了一种针对于HFSWR复杂检测背景的多策略目标检测策略，，介绍该策略的详细算法流程。将背景分割技术、大功率目标去除技术、背景数据指数归一化技术综合起来，运用到检测策略中。并使用半仿真半实测数据进行检测性能分析，实验表明，与其他几种广泛应用的检测算法相比，该多策略目标检测方法的弱目标发现能力更强。本文提出的复杂杂波背景下的多策略目标检测方法，有效地利用背景在线感知信息，在多源杂波背景下对弱目标检测具有很好的性能，为未来雷达智能处理方面的研究提供了参考。
[Abstract]:The high Frequency Surface Wave Radar, HFSWR) of high frequency ground wave over-the-horizon radar utilizes the high frequency electromagnetic wave diffraction principle in the sea surface, effectively solves the detection performance loss caused by the traditional microwave radar stealth technology, and realizes a wide range. All-weather remote sea monitoring and marine monitoring focused on ship target detection. Because a large number of external noise and various types of clutter occupy part of the background of HFSWR detection, it shows the complex characteristics of time-varying, multi-source and undulation. Only using the classical constant false alarm probability constant False Alarm algorithm can easily lead to high false alarm and poor detection performance, which makes HFSWR ship target detection become a difficult problem. In order to effectively improve the performance of ship target detection in multi-target coexisting environment, an adaptive multi-strategy target detection method based on background recognition and classification is studied in this paper. It includes detection background analysis, background segmentation processing, background data index normalization and so on. The main research contents are as follows: The main contents are as follows: 1. The HFSWR detection environment is deeply studied, and the complex characteristics of its heterogeneous and multi-clutter are introduced, and the generation principle and physical characteristics of several main clutter types are emphatically analyzed. The background segmentation based on clutter type is the theoretical foundation, and the necessity of background perception and information extraction is obtained. 2. Aiming at the complex features of multi-source and uneven detection background, the detection background is divided into uniform part and homogeneous part by K-L divergence principle. The radar detection background is divided into four clutter regions according to the clutter type. Each segmentation region is analyzed based on the statistical theory, and the conclusion that the data of each segmentation region is distributed according to different parameters is obtained. It can extract useful information and knowledge from detection background and be used in the design of target detection strategy and parameter selection in order to complete a multi-strategy detection system using detection environment awareness technology. 3. In order to further improve the performance of the detection strategy, the pre-processing technology of HFSWR detection background is studied, and the high-power points which affect the data statistics and background region segmentation in the background are removed by the curve regression theory. This paper introduces the exponential normalization technology of background data, normalizes the background data from Weibull distribution to exponential distribution data from the same parameter, and realizes the uniform processing of the detection background. The loss of detection performance caused by statistical difference of background data is effectively solved, and the performance of CFAR detection is enhanced. On the basis of the above research, a two-parameter CFAR detection method based on pre-segmentation is proposed. According to the segmentation results, the segmentation regions are exponentially normalized, and the classical CFAR detection processing is carried out respectively, which improves the detection performance effectively. 4. Based on the above research results, a multi-strategy target detection strategy for HFSWR complex detection background is proposed, and the detailed algorithm flow of the strategy is introduced. Background segmentation technology, high power target removal technology and background data index normalization technology are integrated and applied to detection strategy. The detection performance is analyzed with semi-simulated semi-measured data. The experimental results show that the multi-strategy target detection method has stronger weak target detection ability than other widely used detection algorithms. The multi-strategy target detection method proposed in this paper can effectively utilize the background on-line sensing information and has good performance for weak target detection in multi-source clutter background. It provides a reference for the research of radar intelligent processing in the future.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN958.93

【参考文献】