基于MSA直方图的景象匹配算法研究

提出了一种基于MSA直方图的景象匹配算法，用来解决景象匹配辅助导航系统中，由于拍摄的实时图背景复杂、存在畸变、易受遮挡等因素所带来的匹配困难。该算法利用灰度信息及其空间分布信息，分别计算目标区域及模板的MSA直方图，并计算匹配度，采用金字塔算法及中途停止策略进行搜索，对目标进行识别和定位。用cameraman图像、人脸图像和实际卫星拍摄图像进行了实验，实验结果表明：该算法对仿射变换、轻微透视投影变换、噪声、遮挡及视角变化具有鲁棒性，且比灰度直方图具有更好的性能，在实际拍摄的图像中能对目标进行准确的识别和定位。

Study on scene matching algorithm based on MSA histogram

A scene matching algorithm based on MSA histogram is presented to solve the matching difficulties induced by complex background,distortion and occlusion of real-time images.Combining gray and its distribution,MSA histograms of object region and template and the matching degree between them are computed.Then pyramid searching method and stop-in-middle strategy are adopted to recognize and position the object.Experiments are done with cameramen image,face images and real scenes,the results show that our method is robust to affine transform,light projective transform,noise,occlusion and change of angle of view,and can accurately recognize and position object in real scenes.