三维激光影视图像归一化互信息特征配准技术

Normalized mutual information feature registration of 3D laser film and television images

以提升三维激光影视图像配准过程中特征点配准精度为目的,研究三维激光影视图像归一化互信息特征配准技术。采用Harris算法提取三维激光影视图像特征点,通过非最大约束删除冗余特征点;在确保互信息波动曲线不变的条件下,通过忽略归一化互信息测度过程中的冗余项优化归一化互信息,将优化后的归一化互信息作为配准测度标准构建初始匹配特征点集;采用松弛算法在初始匹配特征点集内确定最优的一一对应的特征点,精确匹配特征点并完成均匀化处理;针对均匀化处理后的匹配点对集合,利用最小二乘法拟合仿射变换参数处理待配准图像内的特征点,通过双线性内插法重采样特征点,最终实现三维激光影视图像配准。应用结果显示该技术特征点提取结果的对应率达到98%以上,特征点一次性正确配准率达到97.8%。

In order to improve the accuracy of feature point of 3 D laser films and television images, the normalized mutual information features are studied. The Harris algorithm is used to extract the feature points of the 3 D laser film as well as the television image, and the redundant feature points are deleted by non maximum constraint. Under the condition that the fluctuation curve remains unchanged, the normalized mutual information is optimized by ignoring the redundancy in the process of normalized mutual information measurement, and the optimized normalized mutual information is used as the standard to construct the initial matching feature point set. The relaxation algorithm is used to determine the optimal one-to-one corresponding feature points in the initial matching feature point set, and to accurately match the feature points and complete the homogenization processing. For the matching point pair set after homogenization, the least square method is used. This fits the affine transformation parameters to process the feature points in the image, and the feature points are resampled by bilinear interpolation, so as to realize the 3 D laser film and television image registration. The application results show that the correspondence rate is more than 98%, and the one-time correct registration rate is 97.8%.