摘要
针对基于图像外观的移动机器人定位中图像特征提取与匹配实时性和准确性差的问题,提出基于颜色矩的改进尺度不变特征变换分级图像匹配算法。该算法先由颜色矩来排序图像序列,再由改进尺度不变特征变换特征与排序后图像序列精确匹配实现定位。其中,改进的尺度不变特征变换算法以基于采样的迭代搜索算法检测极值点,由Sobel算子计算特征点的梯度方向和幅值,提高尺度不变特征变换算法速度及匹配精度。实验结果表明：改进的尺度不变特征变换算法降低误匹配率约9.2%,特征提取与匹配耗时减少约25.8%;分级图像匹配算法减少尺度不变特征变换特征计算代价约70%,减少总体耗时约43.3%。
The hierarchical image matching algorithm for the real-time and accuracy requirement to process image is proposed in the field of image appearance-based mobile robot localization. The improved scale invariant feature transform, based on color moment, is used for the algorithm. The algorithm is firstly performed by color moment to sort the image sequences. Following, improved scale invariant feature transform is used to match with the sorted image sequences. To improve scale invariant feature transform, the sampling-based iterative search approach is used to detect extremums, as well as the magnitude and orientation of the keypoints gradient is calculated by Sobel operator. Experimental results show that the improved scale invariant feature transform reduces the false matching rate by 9.2%, as well the time of features extraction and matching is reduced by 25.8%. The hierarchical image matching algorithm reduces the calculation cost of scale invariant feature transform by 70% and the run time by 43.3%.
出处
《计量学报》
CSCD
北大核心
2016年第2期118-122,共5页
Acta Metrologica Sinica
基金
国家自然科学基金(61201112,61172044)
河北省自然科学基金(F2013203250,F2012203169)
关键词
计量学
移动机器人
图像匹配
颜色矩
迭代尺度不变特征变换
metrology
mobile robot
image matching
color moment
iterative scale invariant feature transform
