【作者】 张秀丽；

【导师】 李琦；

【作者基本信息】 哈尔滨工业大学 ， 物理电子学， 2009， 硕士

【摘要】 CO₂相干激光雷达有较高的分辨率,能成距离像和强度像,这两种图像都含有目标大量的本征信息,易于目标识别,由于其采用的是主动成像模式,因此扫描视场较小、作用距离近、容易暴露;而被动红外图像由于采用的是被动成像模式,作用距离较远、不易暴露自身位置,但与此对应,其抗干扰能力差、图像不稳定且探测虚警率高。CO₂相干激光雷达和长波红外被动成像复合输出信息具有互补性和兼容性,可以扩展系统探测目标的空间和时间覆盖范围,提高系统空间分辨率、全天候工作能力及目标识别和抗干扰能力。本文主要研究数据融合在红外主被动复合成像分割中的应用,分析了主被动成像的特点,通过被动红外系统对目标进行粗略定位,用激光雷达距离强度像进行图像分割,用分割后的图像提取相应的红外图像的融合方法。重点对激光雷达距离像的分割进行研究。本文主要分为三个部分。首先介绍了被动红外图像和激光雷达图像的仿真原理,包括激光雷达的系统组成和基本原理,激光雷达距离像噪声产生机理和统计模型。进而介绍了用于主被动图像分割的三种方法,即卡尔曼滤波阈值分割方法、背景抑制分割方法和脉冲神经网络分割方法。随后进行了激光雷达距离像与被动红外图像融合的分割实验,对比分析三种方法的分割结果,得出了卡尔曼滤波阈值分割方法能够得到较好的分割效果。将分割后的主被动图像进行三维视觉处理,通过激光雷达距离像与强度像融合与其对比,得出主被动复合的优势。最后研究了卡尔曼滤波阈值分割方法的CNR适用范围,并给出了实验结果。更多还原

【Abstract】 CO₂ coherent laser radar provides range image and intensity image with high resolution, both of which have a plenty of latent information of the targets, and are suitable for target recognition. As an active imaging sensor, laser radar has the disadvantages of small field of scan view, near interactive distance and large exposure chance. Passive IR image senor has long interactive distance and could not be easily found. However, it has poor anti-interference ability, instable image and high rate of false detection alarm. The output combined information of two sensors could be complement and compatible. The data fusion of CO₂ coherent laser radar and passive IR image could extend the space and time coverage of detecting target, improve the spatial resolution, target identification and anti-interference ability.Data fusion application in infrared active-passive image compound segmentation is investigated in this paper. Analyzing the characteristic of active passive image, use passive IR image to search targets and laser radar range image to segment the target and background. And then the target in the IR image is extracted by the worked range image. Range image segmentation is the emphasis in this paper.There are three parts in this paper. Firstly, the simulation principles of passive IR image and laser radar image are introduced, including laser radar system, basic principle and noise statistic model. Secondly, kalman filter threshold segmentation, background suppression and PCNN, which are used to laser radar range image segmentation, are introduced. After that the experiments on active-passive compound image segmentation are simulated. Comparing the results of three methods above, kalman filter threshold segmentation is proper for range image segmentation. The fusion results of active-passive image are better than the fusion of intensity image and range image. At last, the CNR range for kalman filter threshold segmentation is studied and the result are obtained.更多还原