复杂背景图像的多尺度Harris特征点检测算法

Multi-scale Harris Feature Point Detection Algorithm for Complex Background Images

为了复杂背景光照映射图像的准确检测能力,需要进行多尺度Harris特征点检测,提出一种基于特征点模糊边缘滤波的复杂背景光照映射图像多尺度Harris特征点检测算法。采用多尺度Retinex算法构建复杂背景光照映射图像的边缘轮廓特征检测模型,结合对复杂背景光照映射图像多尺度Harris角点分布区域进行特征重组,根据对图像的边缘信息特征分解结果,采用边界化的特征分布式融合增强和融合算法,提取复杂背景光照映射图像多尺度Harris角点分布区域的高斯特征分块组合信息,采用模糊度统计参数跟踪识别方法,建立复杂背景光照映射图像多尺度Harris角点分布区域的匹配滤波检测模型,采用分组特征映射和模板匹配技术,实现对复杂背景光照映射图像多尺度Harris角点分布区域增强和多尺度Harris特征点检测。仿真结果表明,采用该方法进行复杂背景图像的多尺度Harris特征点检测精度较高,自适应性较好,提高了图像特征点的定位识别能力。

For the accurate detection ability of complex background illumination map image,multi-scale Harris feature point detection is needed. This paper proposes a multi-scale Harris feature point detection algorithm for complex background illumination map image based on feature point fuzzy edge filtering. The multi-scale Retinex algorithm is adopted to construct the edge contour feature detection model of the complex background illumination map image,and combined with the multi-scale Harris corner distribution area of complex background illumination map image,the feature reconstruction is carried out. According to the edge information feature decomposition result of the image,the bordered feature distributed fusion enhancement and fusion algorithm is adopted to extract the Gaussian feature block combination information of the multi-scale Harris corner distribution area of the complex background illumination map image. The matching filter detection model of multi-scale Harris corner distribution area in complex background illumination map images is established by the fuzzy statistical parameter and recognition method,and the multi-scale Harris corner distribution area enhancement and multi-scale Harris feature point detection of complex background illumination map images are realized by using grouping feature mapping and template matching technology. The simulation results show that this method has high accuracy and better adaptability in multiscale Harris feature point detection of complex background images,and improves the ability of image feature point location and recognition.