[Documentation] [TitleIndex] [WordIndex

Only released in EOL distros:  

cob_people_perception: cob_leg_detection | cob_openni2_tracker | cob_people_detection | cob_people_tracking_filter | libnite2

Package Summary

Leg Detector using a machine learning approach to find leg-like patterns of laser scanner readings.

cob_people_perception: cob_leg_detection | cob_openni2_tracker | cob_people_detection | cob_people_tracking_filter | libnite2

Package Summary

Leg Detector using a machine learning approach to find leg-like patterns of laser scanner readings.

Introduction

The cob_leg_detection package provides tracking mechanisms for detecting people on the basis of trained data for a leg tracking process obtained from the active scanner devices. This software is based on the http://wiki.ros.org/leg_detector approach and implementation. It is furthermore adapted to the Care-O-bot platform specifications and visualization needs and is upgraded to the indigo release.

Quick Start

To use the leg tracker function to detect and track people clone the repository and required dependencies to your catkin environment and build it according to the instructions. Run the launch file:

roslaunch cob_leg_detection leg_detector.launch

Please check the settings for a laser scanner data and its topic in the cob_leg_detection/launch/leg_detector.launch file.

Detailed Description

The leg detector obtains incoming messages from the laser scanner and uses trained data to classify the groups of laser records as possible legs. To visualize the tracking data for people and their legs, please run Rviz and add a laser scanner plug-in.

Detection

According to reliability of the derived information which is set due to the parameters and configuration settings in LegDetection.cfg file (e.g. the number of points in one group) the reliability of each leg is set. If the two legs with the high reliability are found near to each other and their leg pair separation satisfies the configuration parameters (e.g. separation distance between two legs lies normally at about 10 – 50 cm), the new user will be detected and assigned a unique user id. Each new user will be saved in a special list and the user data (e.g. user position relative to a scanner frame, user velocity or user legs position) will be transferred to a tracking filter.

Tracking

The tracking filter is base on the Kalman Filter approach. It uses the incoming data from the list of the detected people to keep tracking of their estimate position state and the variance or the uncertainty of the estimate. The filter is used to reduce the inaccuracies of the user position obtained from the noisy input data of the laser scanner. If the revised data satisfy the configuration parameters of the filter, the new user data will be accordingly wrapped in a cob_perception_msgs/People message and published on a /leg_detection/people topic. Otherwise the detected user will be erased from the list and the user data will be deleted from the published message. This will lead to a new tracking process for the lost user where it gets a new user id.


2024-11-30 14:33