红外图像目标识别跟踪技术及其DSP实现

【作者】 杨莉；

【导师】 李在铭；

【作者基本信息】 电子科技大学 ， 通信与信息系统， 2001， 硕士

【摘要】 双色红外玫瑰线扫描亚成像制导是弹载末制导系统的核心，也是红外热成像制导的一个主要研究方向。它吸收了红外制导与成像制导的优点，以图形识别方式代替了传统的波形相关识别方式，使系统的目标形状分辨率和探测距离都有了比较大的改进。加上它的成本低廉，便携性好，发射方式灵活，跟踪性能优，因此，具有重要的研究价值和意义。 基于对红外扫描亚成像过程和玫瑰线扫描机理的分析，本文描述了亚成像特征。对轮廓模糊、非完全采样的亚图像序列，着重分析其目标识别、跟踪特征。在此基础上，提出了红外亚图像目标识别跟踪模型。 文中简述了红外亚成像目标识别DSP系统的基本功能和工作原理，并给出在此硬件平台上编写的亚图像目标识别软件流程图。对关键模块的DSP实现进行了详细剖析。实验结果表明，系统基本能识别出不同时间、不同距离、不同气象条件下的亚图像目标。 受导引头尺寸的严格限制，硬件的高度集成化，小型化成为红外亚成像目标识别系统实用性的必要条件。为了能安放进导引头，系统还必须具有特殊的外形。据此要求、我们提出了一系列硬件改进措施，完成了“圆环缺” 目标识别系统的设计制作，并论证其实时性与可靠性。对系统“自举”的设计与实现进行了着重讨论。 针对识别处理软件的设计中存在较大冗余，本文重新设定“圆环缺”系统的功能状态，详细描述其软件优化技术。结果测定显示，“圆环缺”系统的帧处理时间由160ms减少为平均25ms，实时性得到了极大提高。 “目标的跟踪预测”是空中红外图像目标识别与跟踪系统的关键技术。本文用一套基于WIN98和VC++6.0平台的仿真系统，模拟了四种算法的跟踪预测效果。比较了它们的实时性能，以及在DSP系统上的可移植性，提出采用组合预测算法最能实时跟踪目标。更多还原

【Abstract】 The bicolor IR rosette scan guidance is the key technology of a terminal guidance system, and the important research subject of infrared imaging guidance0 It adopts the advantages of infrared and imaging guidance0 Instead of 1 pulse recognition, this technology uses image recognition to improve the resolution ratio and recognition distance 0 Additionally, being its advantages of lower cost, better portability, shooting facility, and good tracking performance, it is valuable and significant0 Based on the analysis of rosette scan principles , the sub梚maging characters are depicted in this thesis After analyzing the recognition and tracking features, we build up the model of infrared sub-maging target recognition and tracking. The functions of infrared sub-maging target recognition DSP system are introduced, and its program flow charts are given0 Also, the implementation of key modules are discussed The experimental results show that the system can work well0 The special figure and size of a seeker required that we should develop a new mini system, And so, we redesign the above system0 The mini one proves stable and efficient, which means it has the ability of real time process 0 Finally, the design and implementation of "foot Load?function in the new one is discussed0 Because the system’s software has more redundancy, we modified its functional state as well as optimized its implemental ways. The results hnw that the process time of this new system is reduced from l6Oms to 25ms0 Obviously, the system’s processing speed is largely improved0 "racking and Prediction?are key technologies in the aerial infrared target recognition and tracking system0 To compare the real-time and transplanting performance of four different algorithm, we developed a simulation software base on Windows 98 and Visual C++ 6. 0 platform0 And the conclusion is drawn that hybrid prediction method is the most suitable one to be implemented on a DSP system0更多还原