基于动态耦合与空间数据关联的RGB-D SLAM算法

为了解决动态环境下视觉SLAM(同步定位与地图创建)算法定位与建图精度下降的问题,提出了一种基于动态耦合与空间数据关联的RGB-D SLAM算法。首先,使用语义网络获得预处理的语义分割图像,并利用边缘检测算法和相邻语义判定获得完整的语义动态物体;其次,利用稠密直接法模块实现对相机姿态的初始估计,这里动态耦合分数值的计算在利用了传统的动态区域剔除之外,还使用了空间平面一致性判据和深度信息筛选;然后,结合空间数据关联算法和相机位姿实时更新地图点集,并利用最小化重投影误差和闭环优化线程完成对相机位姿的优化;最后,使用相机位姿和地图点集构建八叉树稠密地图,实现从平面到空间的动态区域剔除,完成静态地图在动态环境下的构建。根据高动态环境下TUM数据集测试结果,本文算法定位误差相比于ORB-SLAM算法减小了约90%,有效提高了RGB-D SLAM算法的定位精度和相机位姿估计精度。

Administrative Monopoly, Market Allocation and Manufacturing Industry Geographical Cluster in the Central Region

On the basis of the research framework of the new economic geography, this paper conducts an empirical study of the effects of administrative monopoly and market allocation on the manufacturing industry geographical cluster in the central region. The results indicate that both administrative monopoly and market allocation have significant influence on the development of the manufacturing industry cluster in the central region, but the influence of the former is bigger than that of the latter. The administrative monopoly has multiple influences on the manufacturing industry cluster in the central region, while the traditional administrative monopoly measures represented by taxation go against the promotion of industry cluster. In addition, for the changes of geographical clusters of different types of manufacturing industry,the market allocation factors playing their roles are also different. As for the geographical clusters of high level clustering industries, the effect of the money external factor among the market allocation factors is the most significant. As for the clustering industries of middle level, the effect of the scale external factor is the most significant. The effects of knowledge externality on geographical cluster of manufacturing industry in the central region are not significant.