论文部分内容阅读
针对传统图像匹配方法对光线、视角的变化敏感,约束条件单一,匹配效率低、误差率高等难题,提出了维激光扫描图像的快速实时匹配方法。首先构建激光扫描模型,搜寻尺度空间极值,提取图像特征点及其描述,并对候选特征点进行筛选;然后采用余弦相似度和距离匹配对图像进行快速实时匹配,获取图像间精确的匹配点对,利用积分图像进一步降低匹配误差,实现图像的快速实时匹配,最后实验结果表明,本文在外界噪声干扰、光照缺失等不利条件下,大幅度缩短匹配时间,有效降低了误匹配率,提高图像匹配的精度。
Aiming at the problems that traditional image matching methods are sensitive to changes of light and viewing angle, single constraints, low matching efficiency and high error rate, a fast real-time matching method of dimension laser scanning image is proposed. Firstly, the laser scanning model is constructed, the extreme value of scale space is searched, feature points and their descriptions are extracted, and candidate feature points are screened. Secondly, On the other hand, using the integral image to further reduce the matching error and achieve fast real-time image matching, the experimental results show that this paper greatly reduces the matching time under the unfavorable conditions of external noise interference, lack of illumination and so on, effectively reduces the false matching rate and improves the image Matching accuracy.