论文部分内容阅读
为了提高移动机器人同时定位与地图创建(SLAM)问题中闭环检测的准确率和实时性,提出了一种基于历史模型集的改进闭环检测算法.首先,在基于Kinect传感器的帧到模型配准模型的基础上,增加特征描述向量并使用加权方法对其进行更新,从而构建历史模型集,并利用视觉词典树(visual vocabulary tree)对历史模型集和当前帧数据进行场景描述;其次,以反比例函数代替最小值函数,使两幅图像在单个节点的相似性得分函数得以优化,从而得到改进的金字塔TF-IDF(词频-逆向文件频率)得分匹配方法.一方面,改进方法能够减少感知歧义,提高闭环检测的准确率;另一方面,利用改进方法对当前帧数据与历史模型集的从属关系进行有效判断,与传统逐帧比较方法相比,比较次数明显减少,闭环检测的实时性得到较大提高.再次,使用改进的金字塔TF-IDF得分匹配方法对当前帧数据和候选历史模型集所包含的关键帧进行相似性分析,进而提取候选闭环;最后,从时间连续性和对极几何约束两个方面剔除误正闭环.数据集和实际场景对比实验均表明,相比于IAB-MAP(incremental appearance-based mapping)、FAB-MAP(fast appearance-based mapping)和RTAB-MAP(real-time appearance-based mapping),本文的闭环检测算法具有更好的实时性,且在确保100%准确率的情况下,具有更高的召回率.
In order to improve the accuracy and real-time of closed-loop detection in simultaneous localization and mapping (SLAM) of mobile robots, an improved closed-loop detection algorithm based on historical model sets is proposed.First, based on the Kinect sensor-based frame-to-model registration model , Then adds the feature description vector and updates it by using the weighted method to construct the historical model set and uses the visual vocabulary tree to describe the historical model set and the current frame data. Secondly, with the inverse proportion function Instead of the minimum function, the similarity score function of two images at a single node is optimized, so that an improved pyramid TF-IDF score matching method can be obtained.On the one hand, the improved method can reduce the perceived ambiguity and improve On the other hand, using the improved method to judge the subordinate relationship between the current frame data and the historical model set effectively, compared with the traditional frame-by-frame comparison method, the number of comparisons is significantly reduced and the real-time performance of closed-loop detection is larger Improve.Thirdly, using the improved pyramid TF-IDF score matching method to the current frame data and candidate calendar The similarity of the key frames contained in the model set is analyzed and the candidate closed-loop is extracted.Finally, the closed-loop errors are eliminated from both the time continuity and the pole geometry constraint.The data set and the actual scene comparison experiments show that compared with the IAB -MAP (incremental appearance-based mapping), FAB-MAP (fast appearance-based mapping) and RTAB-MAP (real-time appearance-based mapping). The closed-loop detection algorithm in this paper has better real- % Accuracy rate, a higher rate of recall.