【作者】 王雪峰；

【导师】 李琦；

【作者基本信息】 哈尔滨工业大学 ， 物理电子学， 2011， 硕士

【摘要】 激光成像雷达可以直接获得反映目标空间位置的三维几何距离像和反映目标材质属性的一维强度像,统称为四维像,具有高空间分辨率和高距离分辨率,与其它二维传感器相比,如被动红外、CCD等,获取信息丰富,适于用来进行目标检测识别。目标搜索算法是激光成像雷达检测识别的重要组成部分,由于往往采用以顺序搜索为核心的技术途径,其算法复杂度与图像分辨率成正比;随着大视场高帧频激光成像雷达的出现,传统搜索算法很难同时满足高定位精度与实时性需求。量子遗传算法是一种高速寻优算法,它是量子思想与智能算法相结合的产物。通过将量子叠加态、酉正量子门等基本量子概念引入到生物的进化过程中,使算法具有高度的量子并行性,是解决大视场激光雷达目标搜索的技术方向之一。首先,本文结合激光成像雷达四维像的特点,对量子遗传算法进行研究与改进,使它适用于激光成像雷达的图像搜索。为解决图像表述,通过坐标几率化使用波函数对图像进行了描述;为快速寻求全局最优解,分别提出了不等概率解空间初始化和自适应量子旋转门;通过使用一系列特征鲜明的测试函数,完成了算法收敛性的测试,证明了算法的可行性。其次,为验证激光雷达量子遗传算法有效性,分别对标准图像、激光雷达仿真图像和真实图像进行测试,讨论了算法的成功率、收敛性、收敛速度与稳定性,测试结果表明,量子遗传搜索算法具备实际应用的可行性。最后,本文初步描述了实现应用于激光成像雷达图像搜索的量子遗传算法的全量子方案、CPU方案与FPGA方案,并对各方案的技术要求与所能体现出的算法实时性进行了分析与讨论。更多还原

【Abstract】 Imaging laser radar can get the 3D range image which reflects the spatial location of targets and the intensity image which reflects the material of targets. These images are named 4D image of target. Compared with other 2D sensors, for example, the passive infrared sensor and CCD, imaging laser radar can get more abundant information and has high spatial resolution, so it is suitable to be used to detect and recognite targets. The target searching algorithm is an important part of detrctor and identity module of imaging ladar. Because the core approach of the algorithm always works sequencely, the complexity is ptoportional with the image resolution. Especially, with the emergence of the wide field and high frame rate laser ladar, it is difficult for the traditional algorithm to meet the demand of the high accuracy and the real-time.Quantum genetic algorithm is a kind of fast optimization algorithm, which is the combination of the intelligent algorithm and the quantum thought. Becase the basic quantum concepts, for example, the superposition states, unitary quantum gates, are introduced into the process of biological evolution, the algorithm gets the highly quantum parallelism. It is one of the key to the target searching technology of he laser radarFirstly, considering the characteristics of the 4D images of imaging laser radar, this article will research and improve the quantum genetic algorithm to make it be suitable for target searching of imaging laser radar. To state the images, the article uses the wave function to state the images by random the coordinates. To find the global optimal solution fast, the article presents the proposal of unequal-probality initialization for the solution space, and improves the quantum gate which can be adjusted adaptively with the change of the evolitional generations and the fitness respectively. In addition, a series of test functions with distinctive features are designed to test the convergence of the algorithm. The feasibility of the algorithm is proved.Secondly, in this paper, the algorithm is applied into the target searching of the standard images, the simulated images of imaging ladar and the real images of imaging ladar. The success rate, convergence and the stability of the algorithm are discussed. The results show that the quantum genetic search algorithm has the feasibility of practical application. Finally, in this paper, the preliminary proposals for realizing the established quantum genetic searching algorithm are present. They are the full quantum proposal, the FPGA proposal and the CPU proposal. Contrapose to the characteristics of these proposals, the technical requirements and the reflected real-time are analyzed and discussed.更多还原