基于改进BRISK算法的图像特征提取方法研究

Research on image feature extraction method based on improved BRISK algorithm

以移动机器人视觉导航为应用背景,大量研究表明,BRISK算法随着图像尺度不断增大,错误匹配急剧增加。如果使用BRISK算法作为视觉里程计的特征匹配算法,将无法为视觉SLAM后端提供准确的位置信息。针对这一问题,本文提出了改进的BRISK算法,在改进的算法中,尺度空间金字塔通过构建ui层来细化相邻图像间的尺度间隔,并使用灰度质心法为关键点分配主方向,取代了原算法利用长距离采样点对计算局部梯度的方法。实验结果表明,该算法在运行时间相差不大的情况下,尺度不变性上表现出较好的鲁棒性,特征点匹配准确率有很大提高。

With the mobile robot visual navigation as the application background,numerous studies show that,while the image scale continues to increase,the false matches of BRISK algorithm increase sharply.If the BRISK algorithm is used as the feature matching algorithm for visual odometers,it is impossible to provide accurate position information for the visual SLAM backend.In view of this problem,this paper proposes an improved BRISK algorithm.In the improved algorithm,the scale space pyramid refines the scale interval between adjacent images by constructing the uilayer,and assigns the main direction to the key points by using the gray scale centroid method.It replaces the original algorithm to calculate the local gradient using long-distance sampling points to accelerate the algorithm.The experimental results show that the proposed algorithm exhibits better robustness in scale invariance with a slight increase in time consumption,and the accuracy of feature point matching is greatly improved.