Real-time people detection and mapping system for a mobile robot using a RGB-D sensor

In this paper, we present a robotic system capable of mapping indoor, cluttered environments and, simultaneously, detecting people and localizing them with respect to the map, in real-time, using solely a Red-Green-Blue and Depth (RGB-D) sensor, the Microsoft Kinect, mounted on top of a mobile robotic platform running Robot Operating System (ROS). The system projects depth measures in a plane for mapping purposes, using a grid-based Simultaneous Localization and Mapping (SLAM) approach, and pre-processes the sensor´s point cloud to lower the computational load of people detection, which is performed using a classical technique based on Histogram of Oriented Gradients (HOG) features, and a linear Support Vector Machine (SVM) classifier. Results show the effectiveness of the approach and the potential to use the Kinect in real world scenarios.