机载Lidar点云与航空影像的智能三维融合技术研究

针对目前传感器种类繁多且所获取的图像各有优缺点，图像融合技术众多但缺乏普遍适用的方法，以及三维图像融合的研究刚刚起步的现状，本课题在研究Lidar点云数据的特点及预处理方法的基础上，结合像素融合法和控制点位法，研究激光雷达图像和遥感图像的三维融合方法；利用纹理贴图法，研究激光雷达图像和SAR图像的三维融合方法；从而进一步研究激光雷达图像、遥感图像与SAR图像的三维融合方法。并提出基于面向自治计算的三维图像融合方法，利用计算体的自治行为获得系统的全局最优解决方案；提出基于数据同化的三维图像融合方法，利用数据同化不断同化模型算子生成的预测图像和观测图像优势的融合图像，实现三维智能融合。融合方法效率高、鲁棒性好、自适应性强，融合后的三维图像更逼真、更接近客观事实。将三维融合和智能优化结合起来，填补了图像融合领域缺乏普遍适用的智能三维融合方法的空白，为三维立体目标的高精度检测与识别提供技术支撑。

On the basis of the characteristics of and preprocessing methods for Lidar point cloud data，this research aims to tackle current situation that there is a wide variety of sensors; the image captured has their own disadvantages and advantages; meanwhile many image fusion techniques exist but the common-used method is absent; moreover the research on 3D image fusion just starts. Our subject deals with how to fuse the Lidar image and Remote-sensing image combined with pixel method and control point method .Using the method of texture mapping, we conduct research on the image fusion of Lidar image and SAR image, thus enlarging our study on the 3D fusion of Lidar image, Remote-sensing image and SAR image.We propose the 3D fusion method based on Autonomy-Oriented Computing, utilizing the autonomy behavior to obtain the global optimal solution. And we propose the 3D fusion method based on data assimilation, which makes full use of the data assimilation to assimilate the advantageous parts of predictive image and observed image, leading to the 3D intelligent fusion. High in efficiency,good at robustness, strong in adaption, the fusion method makes the fused 3D image more reality and close to objective fact. The method combing 3D fusion and intelligent optimum fill the blank that common adaptive 3D fusion method is lacked in the image fusion field, finally providing strong technique for high-precision detecting and recognizing of 3D object.