远程小目标图像特征精准匹配方法研究与仿真

Remote Small Target Image Features Accurate Matching Method Research and Simulation

研究远程图像中小目标区域的准确匹配问题。远程采集的图像中目标过小,像素与周围环境排列过于紧密,背景灰度值与目标灰度值非常相近,像素特征很难通过阀值与背景完整分割。传统的灰度图像小目标匹配的方法,当图像中的像素特征高度重合的情况下,会导致图像特征局部模糊,特征匹配的准确性不高。提出一种低频像素排序和高斯金字塔分解的弱小目标匹配方法,通过提取出图像中的像素频率信息,并对其进行排序,利用高斯金字塔分解结果将图像中的目标细化,并依此完成准确匹配。实验证明,改进的匹配方法能够准确匹配图像小目标区域,可为图像匹配优化提供参考。

Research accurate recognition of small targets in the remote images. Remote image structures are complex and small. This paper put forward a weak target matching method based on low frequency pixel sorting and Gaussian pyramid decomposition. It extracts and sorts the frequency information of the images, and uses the Gaussian pyramid decomposition, results to divide the targets and completes the exact matching. The experiment results show that the improved algorithm can extract target parameters and improve the accuracy of identification.