Show EOL distros: 

people: face_detector | leg_detector | people_msgs | people_tracking_filter | people_velocity_tracker | social_navigation_layers

Package Summary

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

people: face_detector | leg_detector | people_msgs | people_tracking_filter | people_velocity_tracker

Package Summary

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

people: face_detector | leg_detector | people_msgs | people_tracking_filter | people_velocity_tracker

Package Summary

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

people: face_detector | leg_detector | people_msgs | people_tracking_filter | people_velocity_tracker

Package Summary

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

people: face_detector | leg_detector | people_msgs | people_tracking_filter | people_velocity_tracker

Package Summary

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

people: face_detector | leg_detector | people_msgs | people_tracking_filter | people_velocity_tracker

Package Summary

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

The Leg Detector

This leg detector package takes sensor_msgs/LaserScans as input and uses a machine-learning-trained classifier to detect groups of laser readings as possible legs. Sadly, the training dataset has been lost to Willow Garage history (it wasn't even available before they closed). The code is in the repository for retraining, but is unsupported at this time.

This node will publish people_msgs/PositionMeasurementArrays for the individual legs, and it can also attempt to pair the legs together and publish their average as an estimate of where the center of one person is as a people_msgs/PositionMeasurement. The node will also optionally publish visualization Marker messages to indicate where detections happened.

Possible Usages

There are two ways to use the leg_detector: seeded and unseeded. In the unseeded mode, the only input is the received sensor data. This can provide many false positives, but works in many scenarios.

In the seeded mode, the algorithm will use another source of PositionMeasurement messages to guide the algorithm to possible locations for people. Historically, this has been used with a face detection algorithm that alerts the leg_detector that there is probably a pair of legs underneath. This mode is enabled using the use_seeds parameter.

leg_detector

Subscribed Topics

scan (sensor_msgs/LaserScan)
  • The laser scan to detect from.
people_tracker_filter (people_msgs/PositionMeasurement)
  • PositionMeasurements to use as the seeds (see above)

Published Topics

leg_tracker_measurements (people_msgs/PositionMeasurementArray)
  • Estimates of where legs are
people_tracker_measurements (people_msgs/PositionMeasurementArray)
  • Estimates of where people are
visualization_marker (visualization_msgs/Marker)
  • Markers where legs and people are

Parameters

use_seeds (boolean, default: false)
  • Whether or not to require additional seeds
connection_threshold (double, default: 0.06)
  • Groups of connected laser reads must be within this many meters of each other to be considered a group
min_points_per_group (int, default: 5)
  • Number of points required for a group of laser reads to be considered a group
leg_reliability_limit (double, default: 0.7)
  • Only publishes legs with reliability higher than this limit
publish_legs (boolean, default: True)
  • Whether to publish leg_tracker_measurements
publish_people (boolean, default: True)
  • Whether to publish people_tracker_measurements
publish_leg_markers (boolean, default: True)
  • Whether to publish legs as visualization markers
publish_people_markers (boolean, default: True)
  • Whether to publish people as visualization markers
no_observation_timeout (double, default: 0.5)
  • Forget a cluster after this many seconds
max_second_leg_age (double, default: 2.0)
  • seconds
max_track_jump (double, default: 1.0)
  • x
max_meas_jump (double, default: 0.75)
  • x
leg_pair_separation (double, default: 1.0)
  • Pairs of legs can be at most this far from each other
fixed_frame (string, default: odom_combined)
  • Topics are published relative to this frame
kalman_p (double, default: 4)
  • Used to smooth measurements
kalman_q (double, default: .002)
  • Used to smooth measurements
kalman_r (double, default: 10)
  • Used to smooth measurements

Wiki: leg_detector (last edited 2014-10-24 21:43:35 by DavidLu)