基于局部精确直方图匹配的无人机景象匹配导航色彩恒常算法  

Color constancy algorithm based on local exact histogram matching for scene matching navigation of UAVs

作　　者：靳珍璐 潘泉[1] 赵春晖[1] 魏妍妍[1] 马娟娟[1] 

机构地区：[1]西北工业大学信息融合技术教育部重点实验室自动化学院,西安710072

出　　处：《中国惯性技术学报》2015年第5期674-680,共7页Journal of Chinese Inertial Technology

基　　金：国家自然基金项目(61135001);国家自然基金项目(61473230;61403307);航空基金项目(2014ZC53030)

摘　　要：为了解决无人机景象匹配导航中实时图和基准图之间的色彩差异性问题,提出了一种基于局部精确直方图匹配的色彩恒常算法。将实时图分成一系列子块图像,根据景象匹配辅助导航系统预测的无人机位姿信息,获得实时图与基准图的粗略对应关系,考虑平移、尺度、旋转等补偿,选取基准图中与实时图子块对应的基准图子块,求解每对子块图像红、绿、蓝颜色通道色彩直方图的颜色映射关系,并根据建立的颜色映射关系,对实时图子块的三个颜色通道进行变换,从而降低实时图与基准图之间的色彩差异。试验结果表明,所提色彩恒常算法对光源颜色估计的欧拉角误差约降低到传统色彩恒常算法的20%,相比于不采用色彩恒常处理的景象匹配算法,将景象匹配平均误差降低了约70%,由于算法中引入了分块处理机制,能够克服多光源成像导致的色彩差异,具有并行计算的优势,便于工程应用。To address the color variability issues existing between the sensed images and the reference images in scene matching navigation of unmanned aerial vehicles（UAVs）,a color constancy algorithm is proposed based on local exact histogram matching（LEHM）. The sensed image is divided into a set of sub-images,and the erroneous correspondence between the sensed image and the reference image is obtained according to the location and attitude of UAV predicted by the scene matching aided navigation system. The shifting,scaling and rotation compensation are taken into account to select the sub-images from the reference image corresponding to the sub-images from the sensed image. Three intensity mapping relationships are established from the histograms of each pair of sub-images in red,green,and blue channels. At last,the three color channels of the sub-images from the sensed image are adjusted in accordance with the established intensity mapping relationships. This will reduce the color differences between the sensed image and the reference image. Experimental results show that the euro angle error of the estimated illuminant color using the proposed algorithm is reduced by 80% compared with those by traditional color constancy methods. Moreover,the mean scene matching error is reduced by almost 70% in comparison to the scene matching algorithm without color constancy processing. By introducing the blocking scheme,the proposed algorithm could overcome the color variability issues caused by multi-illuminant imaging and possess the potential of parallel computation,which is suitable for engineering applications.

关 键 词：色彩恒常 局部精确直方图匹配 几何校正 景象匹配导航 无人机 

分 类 号：U666.1[交通运输工程—船舶及航道工程]