【作者】 韩洲；

【导师】 李元祥；

【作者基本信息】 上海交通大学 ， 航天工程， 2012， 硕士

【摘要】 复杂背景下的目标分割和识别一直是航空航天等相关领域研究的重要内容。本文首先对复杂背景下目标检测、跟踪和识别(DTR)系统中的空间目标分割算法进行了改进,并基于嵌入式开发平台和软件系统架构的特点对DTR系统的实际运行效率进行了优化,最后对空间目标检测、跟踪和识别硬件运行平台各模块电路进行了初步的设计。主要内容如下:1)针对由卫星仿真工具包STK产生的空间目标图像序列,提出一种基于改进先验形状CV模型的目标分割算法。该算法基于先验形状约束的变分水平集模型,在Chan模型的基础上,加入剪切及X、Y方向拉伸不变两种特性,拓展了先验形状对目标的自适应性。该算法可以较好地克服图像中出现的遮挡、阴影、噪声等干扰的影响,对于复杂背景下姿态变换较大的目标也有良好的分割效果。2)基于DaVinci系统架构的特点并利用SoC analyzer工具,对DTR系统进行优化。优化后的DTR系统运行效率有了较大的提升,基本满足系统的实时处理需求。3)针对DTR系统各算法模块的特点,基于TMS320DM6467处理器和各个片上的外设模块完成了板上系统的整体电路设计。所设计的硬件电路体现了DTR系统的特点。更多还原

【Abstract】 Detecting, tracking and recognizing (DTR) spatial objects under the cluttered background have always been an important research area of aerospace exploration. In this paper, an improved image segmentation algorithm in DTR system is firstly introduced; Based on embedded system platform and system developing architecture, DTR system’performing efficiency was then improved; Finally, a hardware platform for DTR system is redesigned. Main research results are listed as follows:1) For target image sequence generated by Satellite Tool Kits, an object segmentation algorithm based on improved prior shape and CV model is given. Given model in this paper is based on variational level set model, which introduces another two local adaptive transformation in Chan-Zhu model. New improved model in this paper overcomes not only interferences such as occlusion, shadow, noise, clutter, but also better segmentation result towards objects with complex affine transformation. In addition to that, improved model in this paper can give a better segmentation result than Chan-Zhu model. Disturbed by light and angel, locations of the building towards in Aerial remote sensing images are often uncertain. To extract interest object in such images, Chan-Zhu model will be restricted in limited self-adapting transformation towards shape prior, yet improved model in this model will have no such worry. In collusion, improved model in this paper has a better robustness.2) With the help of SEED-DVS6467 development tool chain and SoC analyzer, DTR system’performing efficiency on SEED-DVS6467 was optimized. Running speed of optimized system was improved greatly, and optimized system on platform basically satisfied our requirements.3) According to the analysis result of modules distribution in the DTR system, redesign the architecture of hardware platform, which is based on TMS320DM6467 processor and corresponding Peripherals such as video switching chips, DDR2 SDRAM, NAND flash, etc. Designed platform highlights the features of DTR system.更多还原