基于相关面后处理的可能误匹配点剔除方法

Method for discarding the possible mismatched points based on post-processing of correlation plane

在景象匹配过程中,各种成像畸变使真实匹配位置往往偏离相关面的最大峰值处,导致传统的"最大峰值决策"匹配算法产生误匹配。本文通过分析成像畸变对相关面的影响及真实匹配位置处相关峰所具有的特征,提出了一种基于相关面后处理的可靠景象匹配方法。该方法以相关峰的融合特征作为匹配依据,通过计算相关面上多个局部极大峰融合特征值的差异程度来判断实测图的可匹配性,放弃易被误匹配的实测图,保留可信度高的实测图的匹配结果。在红外、可见光、SAR图像上的仿真结果表明,所提方法在丢弃少量帧的情况下,能够剔除传统"最大峰值决策法"的多数误匹配点,正确匹配概率有了明显的提高。

During the scene matching process,imaging distortion often makes the actual match point deviate from the maximum peak of correlation plane,inducing a mismatch of the traditional decision making method based on maximum peak（＂MPDM＂for short）.By analyzing the influence of image distortion on correlation plane and the peak properties corresponding to the actual match point,we develop an approach to eliminating possible mismatched points based on postprocessing of correlation plane.The approach takes the fusion feature of correlation peak as the matching criteria,and evaluates the real-time image ability to obtain correct matching by calculating the fusion feature difference among several local maximum peaks.So we can abandon those liable mismatched real-time images,and preserve the matching results with high reliability.Simulation results which are performed with visible image,infrared image and SAR image show that the approach proposed can eliminate most of the mismatched points produced by the MPDM method with a few frames abandoned.Thus acquisition probability is improved.