RGB-D Human Detection and Tracking for Industrial Environments

Abstract

Reliably detecting and tracking movements of nearby workers on the factory floor are crucial to the safety of advanced manufacturing automation in which humans and robots share the same workspace. In this work, we address the problem of multiple people detection and tracking in industrial environments by proposing algorithms which exploit both color and depth data to robustly track people in real time. For people detection, a cascade organization of these algorithms is proposed, while tracking is performed based on a particle filter which can interpolate sparse detection results by exploiting color histograms of people. Tracking results of different combinations of the proposed methods are evaluated on a novel dataset collected with a consumer RGB-D sensor in an industrial-like environment. Our techniques obtain good tracking performances even in an industrial setting and reach more than 30Hz update rate. All these algorithms have been released as open source as part of the ROS-Industrial project

OpenPTrack: Open source multi-camera calibration and people tracking for RGB-D camera networks

“…Recent works ([9], [10], [1], [3], [11]) allow to avoid the sliding window approach that usually leads to analyze thousands of detection windows per image. These new methods exploit the assumption, also adopted by OpenPTrack, that people stand/walk on a ground plane and cluster algorithms on depth data to find a small number of Regions Of Interest (ROIs) which are candidates to contain people and are then classified with more robust and computational demanding algorithms.…”

“…An open source implementation of the people detection algorithms in [10] and [1] is available in the Point Cloud Library [4]. In ROS-Industrial Human Tracker repository 2 , these algorithms are combined with other people detectors in a people detection cascade which obtains even more computational efficiency and which is combined with the tracking algorithm described in [3]. Also the code for the GPU-based and the CPUbased versions of the people detection and tracking software in [11] has been recently released 3 .…”

“…• then, an Adaboost classifier based on Haar features extracted from the disparity map [3] is applied to the clusters resulting from the previous steps.…”

“…In addition to the applications listed above, OpenPTrack is also perfectly suited to be adopted in environments where a robot has to interact with a human, such as in a RoboCup@Home scenario [2] or when a human and a robot share the same workspace in a factory [3]. The constant and real time monitoring of people positions is a necessary condition to avoid any injury that an industrial robot could cause to a human.…”

Automated Tracking and Real Time Following of Moving Person for Robotics Applications

“…The motion of the human and occlusion are the common causes for false detection and lost tracking. Secondly, to share the physical environments between the human and robots, there is a need to avoid collision with robots [1]. If movement of human is detected by the robot then this information can be used by the robots to prevent the collision.…”