基于高斯差分的改进Harris特征点提取算法  

Feature points extraction with improved Harris algorithm based on difference of Gaussian

作　　者：王青松[1,2] 赵西安[1,2] 吕京国[1,2] 王守营 马超[1,2] 

机构地区：[1]北京建筑大学测绘与城市空间信息学院,北京100044 [2]现代城市测绘国家测绘地理信息局重点实验室,北京100044 [3]北京中铁隧建筑有限公司,北京100022

出　　处：《测绘科学》2014年第4期119-122,134,共5页Science of Surveying and Mapping

摘　　要：针对非城区影像纹理特征缺乏、特征不明显、影像特征点提取困难和图像特征附近可能出现冗余特征点问题,本文提出一种基于高斯差分的改进Harris特征点提取算法。首先用高斯函数对图像进行卷积运算生成平滑图像,将原图像与平滑图像进行差分运算生成增强纹理特征的差分图像;然后基于Harris算法进行差分图像特征点提取,进行特征点局部窗口的非极大抑制,剔除冗余特征点,提高特征点提取精度。实验证明,本文算法可以改善纹理贫乏影像区的特征提取效果,有效删除冗余特征点,并提供质量较好的特征控制点。The improved Harris feature extraction method was proposed based on difference of Gaussian in the paper. The algorithm fits for extracting features in the non-urban images, of which the texture and feature is deficient. Firstly, the difference image was built by subtracting the smoothing image filtered by Gaussian function from the original image. Then the feature points were extracted in the difference image by Harris algorithm. Finally the accurate extraction of feature points were carried out through using Non-maximum Suppression and removing redundant feature points. The results of experiment showed that the algorithm is significant for feature point extraction from the images with deficient texture and features.

关 键 词：高斯函数 差分图像 改进的Harris算子 特征提取 

分 类 号：TP75[自动化与计算机技术—检测技术与自动化装置]