针对移动机器人在目标识别过程中对视觉图像特征点提取慢,匹配不准确等特点,提出了一种基于SIFT算法的改进目标识别算法.通过采用组合匹配策略,将特征关键点间的距离和内积同时进行考察,根据其自身值的大小,决定对匹配相似度的贡献.组合策略的引入有效地解决了机器人在目标识别中对相同特征图像不能匹配和不同特征图像能够匹配的问题.为克服目标匹配时实性差的弱点,以关键点为根据构建K维树结构,采用最近邻点搜索,快速找出正确匹配的特征点.为实现移动机器人目标识别过程中的自主性,在特征点匹配过程中引入自适应阈值进行判断.实验表明,该方法对移动机器人目标识别准确率有较大提升,能够满足移动机器人在目标识别和跟踪过程中对视频图像处理的实时性和准确性的要求.

According to the real time and accuracy requirement to process image during target recognition for mobile robots,an improved target recognition algorithm is proposed based on SIFT algorithm.By using a combination of matching strategy,the distance and the inner product of feature vectors as considering factors are taken.According to its own values to determine their contribution to the matching similarity,the problems that the same image features does not match and the different characteristics can match are solved by the combination of strategies.In order to improve the speed of features matching,a K-dimensional tree is constructed based on a key point.By using the nearest neighbor search method,the correct matching feature points can be quickly found.To achieve the autonomy of mobile robot’s target recognition,an adaptive threshold is introduced to select during the feature point matching.It is validated that the accuracy and operability of mobile robot’s target recognition is greatly improved,and the requirements of real-time and veracity in the process of mobile robot’s target recognition and tracking is met.