中文摘要：

针对移动机器人导航过程中基于尺度不变特征变换（SIFT）算法图像匹配速度较慢，提出了基于减法聚类和特征描述符二值化的改进SIFT算法。通过减法聚类消除大量特征点中的冗余特征点，在不影响原SIFT算法稳定性的前提下有效降低了特征点数量，然后将生成的特征描述符进行二值化，依据Hash函数生成索引，以汉明距离作为度量准则。实验结果表明：与原SIFT算法相比，改进的SIFT算法中特征点数量下降30％～40％；匹配对数基本维持不变；匹配率上升6％～12％；匹配时间下降60％～70％。与基于颜色矩的改进SIFT分级图像匹配算法相比，改进的SIFT算法中特征点数量下降15％～25％；匹配对数基本维持不变；匹配率上升5％～10％；匹配时间下降45％-55％。

英文摘要：

In the process of mobile robot navigation based on SIFT algorithm, the speed of image matching is slow, the improved SIFT algorithm based on subtractive clustering and the binarized feature descriptor are proposed. Firstly, subtractive clustering is used to reduce the redundant points of feature points, which effectively reduces the feature number without affecting the stability of the original SIFT algorithm. The generated feature descriptors are binarized, indexes are produced by the Hash function, and Hamming distance as the metric. Experimental results show that compared with the original SIFT algorithm, in the improved SIFT algorithm , the number of feature points is decreased by 30% -40% , the logarithm of the matched is basically unchanged, the matching rate increased by 6%- 12% and the matching time decreased 60% -70%. Compared with the improved SIFT algorithm which based on color moment and hierarchical image matching, in improved SIFT algorithm the number of feature points is decreased by 15% -25% , the logarithm of the matched is basically unchanged, the matching rate increased by 5% - 10% and the matching time decreased 45% -55%.