CAJ | 학술논문

针对SIFT算法在图像匹配时会产生很多误匹配的问题，本文利用霍夫变换有效的删除了SIFT所产生的误匹配。首先，由SIFT算法粗略的完成图像匹配；然后利用霍夫变换形成等分的霍夫单元，将所有匹配按照匹配参数分配到这些霍夫单元中，把含有较少匹配对的单元格中的匹配删除。实验表明，该方法能有效的提高特征点匹配的正确率，为后续的机器人视觉导航奠定了基础。
침대SIFT산법재도상필배시회산생흔다오필배적문제，본문이용곽부변환유효적산제료SIFT소산생적오필배。수선，유SIFT산법조략적완성도상필배；연후이용곽부변환형성등분적곽부단원，장소유필배안조필배삼수분배도저사곽부단원중，파함유교소필배대적단원격중적필배산제。실험표명，해방법능유효적제고특정점필배적정학솔，위후속적궤기인시각도항전정료기출。
In view of the SIFT algorithm in image matching will produce a lot of mismatches, a method based on hough transform can effectively remove the SIFT matching error . Firstly, by the SIFT algorithm roughly finish image matching. Then, uniform hoff units, to which all match according to matching parameters are assigned, are galned through hough transform, the matching units contalning the least match are deleted. Experimental results show that the method can effectively improve the matching accuracy of feature points for visual navigation of a mobile robot.