L-CAS RGB-D-T Re-identification Dataset


 


L-CAS RGB-D-T Re-Identification Dataset

Collected by the Kinect and the Optris PI-450 thermal imager

Description

This dataset contains RGB-D(epth)-T(hermal) images of different people acquired in the Lincoln Centre for Autonomous Systems (L-CAS) at the University of Lincoln, UK. Data were recorded into different rosbag files, each containing RGB, Depth, and Thermal images of a person, together with 2D laser data. The dataset consist of four parts: i) a person standing still and turning around at 2m, 3.5m, 5m., ii) a person walking freely, iii) a person sitting on a sofa while the robot observes the person at front, left side and right side, and iv) a person turning around while wearing hat, glasses, and a scarf occluding different parts of his/her face. 

Contributions

This dataset provides:

  1. Robot Operating System (ROS) rosbags. Each rosbag contains about 1,000 continuous thermal, color, and depth images.
  2. Ground truth
  3. Calibration of RGB-D and thermal cameras

Note: the RGB-D images and calibration files can be provided only for research purposes according to Privacy Rule, please send us your request by email.

Citation

If you are considering using this data, please reference the following:

S. Cosar, N. Bellotto, Human Re-Identification with a Robot Thermal Camera using Entropy-based Sampling, Journal of Intelligent & Robotic Systems, 2019.

@Article{Cosar2019,
  Title                    = {Human Re-Identification with a Robot Thermal Camera using Entropy-based Sampling},
  Author                   = {S. Cosar and N. Bellotto},
  Journal                  = {Journal of Intelligent \& Robotic Systems},
  Year                     = {2019},
  Doi                      = {https://doi.org/10.1007/s10846-019-01026-w}
}

Recording platform

  The Optris PI-450 thermal imager was mounted at 1.3m from the floor, on the top of a Kompaï robot. The Kinect was mounted at 1.1m. The distance between the robot and the face was about 1.3m. Thermal images were recorded using the Optris official driver. Temperature data is encoded as: float t = (float)(data - 1000) / 10.f.

Optris PI-450 thermal imager parameters
Optical resolution: 382 x 288 pixels
Frame rate: 10 Hz
Measurement range: 20 – 100 °C
Optics field of view: 38° x 29° / f = 15 mm oder
Kinect parameters
RGB resolution: 640 x 480 pixels
RGB Frame rate: 30 Hz
Depth resolution: 640 x 480 pixels
Depth Frame rate: 30 Hz

Download

The full dataset is available here. Please send an email to scosar@lincoln.ac.uk or nbellotto@lincoln.ac.uk if you have any problem downloading it.

 

Experiments on a mobile robot

Below video shows the test of our thermal re-identification method running on a TIAGo mobile robot

License

Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 3.0 Unported License.
Copyright (c) 2016 Serhan Cosar, and Nicola Bellotto.

Funding

This work was funded in part by the EU Horizon 2020 project ENRICHME, H2020-ICT-2014-1, Grant agreement no.: 643691.