基于似然域背景差分的行人检测和跟踪方法

基于激光雷达的室内机器人行人检测、跟踪容易受到复杂背景的影响。针对这种情况,提出一种基于似然域背景差分的行人检测、跟踪和跟随系统。利用即时定位与地图构建算法获得陌生环境的二维栅格地图,通过蒙特卡洛定位获得机器人在地图中的后验位姿,利用似然域模型分割出前景对应的激光雷达数据后,进行行人的检测、跟踪以及跟随。实验结果表明,该系统使行人检测准确率提升3.49%,平均检测时间缩短近32%,有效降低复杂背景对多行人检测与跟踪的影响,实现机器人对目标行人的实时跟随。

Background subtraction method based on likelihood field model for people detection and tracking

Laser radar based indoor robots for pedestrian detection and tracking are easily affected by complex backgrounds.In view of this,a pedestrian detection,tracking and following system with background difference method based on likelihood field model was proposed.Simultaneous localization and mapping were used to get 2D grid maps of strange environments,and the posterior pose of the robot in the map was obtained through Monte Carlo localization.After extracting the laser radar’s data corresponding to interesting foreground using likelihood field model,pedestrian detection,tracking and following were performed.Experimental results show that,the system improves the pedestrian detection precision by 3.49%and reduces the average detection time by nearly 32%,effectively reducing the impact of complex backgrounds on multi-pedestrian detection and tracking,enabling robots to follow the target pedestrians in real time.