《基于TDSDM642空中目标识别与跟踪技术研究》由会员分享，可在线阅读，更多相关《基于TDSDM642空中目标识别与跟踪技术研究（80页珍藏版）》请在金锄头文库上搜索。
1、南京航空航天大学硕士学位论文 I 摘 要 基于图像的目标识别与跟踪在国民经济和军事领域的许多方面都有着广泛的应用，对它的研究受到日益广泛的关注。本文讨论了基于 DSP 的空中目标识别与跟踪方法，主要做了以下几个方面的研究工作： 首先介绍了 DSP 技术的发展，着重探讨了 DSP 在图像处理应用过程中程序结构的设计方法及大量数据传输的瓶颈问题。 其次，重点研究了空中目标的分割技术。对于灰度分割，研究了基于二维二进小波变换的多尺度边缘检测方法及其改进算法；对于彩色的分割，探索了基于四元数描述的彩色边缘检测方法和基于色彩相似性的彩色聚类方法。通过仿真实验表明，这些分割方法均取得了不错的效果，为目标识
2、别作好了准备。 接着研究了对分割目标的特征提取，主要进行了边界不变矩特征的提取，同时为了后续的目标跟踪，也提取了图像的角点特征。 然后是对目标进行识别，将目标特征样本输入到分类器中进行训练，再运用测试样本进行测试，得到了良好的识别结果。识别方法主要是运用了小波神经网络和模糊神经网络。小波神经网络主要研究了网络参数的初始化问题，模糊神经网络主要是对模型结构进行改进。 在运动目标检测与跟踪阶段，对于静态背景和动态背景，分别研究了基于混合高斯模型背景建模和双帧差法的目标检测方法；然后确定出运动目标最小外接矩形，提取区域内角点特征，在相邻帧间进行角点匹配，完成目标跟踪。 最后简单地总结了本课题的内容，
3、 并从硬件及软件方面对本课题的研究方向进行了展望。 关键词： 关键词： DSP，图像分割，目标检测，目标识别与跟踪 基于 TDSDM642 空中目标识别与跟踪技术研究 II ABSTRACT Target recognition and tracking based on image has been increasingly drawing more attention in the past decades in the areas of national economy and military. This paper mostly has discussed target recogn
4、ition and tracking technique of aero targets based on the DSP, it mainly includes the following several aspects: Firstly, Technical development about DSP has been introduced; mainly the design means and data-transmission of massive. Secondly, segmentation techniques of aero target have been studied
5、emphatically. For gray image, multi-resolution edge detection algorithm and its corresponding improved algorithm based on wavelet transform have been introduced. For color image, color edge detection algorithm based on quaternion representation and color clustering algorithms based on color similari
6、ty coefficient have been introduced. Computer simulation shows that good effect has been achieved and made preparation for the following target recognition. Thirdly, feature extraction means have been studied, mainly the edge invariant moments algorithm has been extracted. And the corner features of
7、 images have also been extracted for tracking. Fourthly, target recognition has been studied in which target samples are first input to interpolator to be trained, then good recognition results have been gotten by using testing samples. Wavelet neural network and fuzzy neural network recognition alg
8、orithms have been provided. In the wavelet neural network, the initial algorithms of network parameters are studied. In the fuzzy neural network, mainly the improving of model structures has been introduced. In the final phase of moving target detection and tracking, the algorithm based on mixture g
9、auss model background modeling for static background and the algorithm based on twice-frame-difference for dynamic background have been studied. Then minimal outer rectangle of moving target has been fixed in which corner features are extracted and matched in the neighbored frame to accomplish the t
10、arget tracking procedure. Finally, primary content of this task has been simply introduced and research aspect has been expounded. Key Words: DSP, image segmentation, target recognition and tracking 基于 TDSDM642 空中目标识别与跟踪技术研究 VI 图图 清清 单单 图 1.1 硬件发展示意图. 5 图 2.1 系统硬件功能框图. 7 图 2.2 图像处理平台内部硬件结构图. 8 图 2.3
11、 TMSDM642 内部主要结构图. 9 图 2.4 视频模块框图. 10 图 2.5 视频设备驱动程序初始化 3 个视频帧缓冲区. 13 图 2.6 应用程序从驱动程序获得视频帧 A. 13 图 2.7 视频帧 A 与视频帧 B 交换 . 13 图 2.8 视频帧 B 与视频帧 C 交换. 13 图 3.1 四连通成分示意图. 17 图 3.2 多尺度边缘提取实验结果对比. 20 图 3.3 单纯背景下的多尺度边缘提取效果. 21 图 3.4 复杂背景下的多尺度边缘提取效果. 21 图 3.5 彩色边缘提取效果 1. 25 图 3.6 彩色边缘提取效果 2. 26 图 3.7 彩色边缘提取效果 3. 26 图 3.8 彩色相似性检测效果 1. 30 图 3.9 彩色相似性检测效果 2. 30 图 4.1 不变矩对比曲线图. 35 图 4.2 角点提取效果图.
