Fast and Robust Multi-people Tracking from RGB-D Data for a Mobile Robot

Abstract

This paper proposes a fast and robust multi-people tracking algorithm for mobile platforms equipped with a RGB-D sensor. Our approach features an efficient point cloud depth-based clustering, an HOG-like classification to robustly initialize a person tracking and a person classifier with online learning to manage the person ID matching even after a full occlusion. For people detection, we make the assumption that people move on a ground plane. Tests are presented on a challenging real-world indoor environment and results have been evaluated with the CLEAR MOT metrics. Our algorithm proved to correctly track 96% of people with very limited ID switches and few false positives, with an average frame rate of 25 fps. Moreover, its applicability to robot-people following tasks have been tested and discussed

Kinect-based people detection and tracking from small-footprint ground robots

“…In these scenarios, a reasonable assumption is that at least the subject's upper body is visible: then, the detection problem is solved through simpler techniques such as: looking for obvious person markers (e.g. the head [38]); filtering point clusters matching the expected approximate dimensions of a person (most importantly, the height [9]); or using statistical classifiers trained on the whole shape of the person [32]. However, in order to ensure that the subject's upper body is visible, the viewpoint can not lie too close to the subject; moreover, in order to avoid occlusions, expecially in crowded or cluttered environments, the sensor should stay in an elevated position, at least at the level of the subjects' chest or eyes.…”

Kinect‐Based Pedestrian Detection for Crowded Scenes

“…Basso et al. () proposed a multiperson tracking algorithm for mobile platforms equipped with an RGB‐D sensor. Their approach features a point‐cloud depth‐based clustering, an HOG‐like classification to initialize a pedestrian tracker and classifier with online learning to manage the person ID matching.…”

“…Although the approach is effective in slight occlusion scene, its performance degrades increasingly in scenarios where visual and depth clutters occur simultaneously. Basso et al (2013) proposed a multiperson tracking algorithm for mobile platforms equipped with an RGB-D sensor. Their approach features a point-cloud depthbased clustering, an HOG-like classification to initialize a pedestrian tracker and classifier with online learning to manage the person ID matching.…”

People identification for domestic non-overlapping RGB-D camera networks

“…Basso et a1. [1] Similar online boosting technique, using three types of RGB-D features and the confidence maximization search in 3D space, was used to build people models for tracking by Lueber et a1. [2].…”