Multiple people detection and identification system integrated with a dynamic simultaneous localization and mapping system for an autonomous mobile robotic platform

This paper presents the integration of a multiple people detection and identification system with a dynamic simultaneous localization and mapping system for an autonomous robotic platform. This integration allows the exploration and navigation of the robot considering people identification. The robotic platform consists of a Pioneer 3DX robot equipped with an RGBD camera, a Sick Lms200 sensor laser and a computer using the robot operating system (ROS). The idea is to integrate the people detection and identification system to the simultaneous localization and mapping (SLAM) system of the robot using ROS. The people detection and identification system is performed in two steps. The first one is for detecting multiple people on scene and the other one is for an individual person identification. Both steps are implemented as ROS nodes that works integrated with the SLAM ROS node. The multiple people detection's node uses a manual feature extraction technique based on HOG (Histogram of Oriented Gradients) detectors, implemented using the PCL library (Point Cloud Library) in C ++. The person's identification node is based on a Deep Convolutional Neural Network (CNN) that are implemented using the MatLab MatConvNet library. This step receives the detected people centroid from the previous step and performs the classification of a specific person. After that, the desired person centroid is send to the SLAM node, that consider it during the mapping process. Tests were made objecting the evaluation of accurateness in the people's detection and identification process. It allowed us to evaluate the people detection system during the navigation and exploration of the robot, considering the real time interaction of people recognition in a semi-structured environment.