/rpg_dvs_ros

ROS packages for DVS

Primary LanguageC++GNU General Public License v3.0GPL-3.0

rpg_dvs_ros

Disclaimer and License

The RPG ROS DVS package has been tested under ROS Indigo and Ubuntu 14.04.

This is research code, expect that it changes often and any fitness for a particular purpose is disclaimed.

The source code is released under a GNU General Public License (GPL).

Package Overview

The ROS DVS package provides a C++ driver for the Dynamic Vision Sensor (DVS). It also provides a calibration tool for both intrinsic and stereo calibration. To find out more about the DVS, visit the website of the Institute of Neuroinformatics.

Authors: Elias Mueggler, Basil Huber, Luca Longinotti, Tobi Delbruck

Publications

If you use this work in an academic context, please cite the following publications:

  • E. Mueggler, B. Huber, D. Scaramuzza: Event-based, 6-DOF Pose Tracking for High-Speed Maneuvers. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Chicago, 2014. (PDF)
  • A. Censi, J. Strubel, C. Brandli, T. Delbruck, D. Scaramuzza: Low-latency localization by Active LED Markers tracking using a Dynamic Vision Sensor. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Tokyo, 2013. (PDF)
  • P. Lichtsteiner, C. Posch, T. Delbruck: A 128×128 120dB 15us Latency Asynchronous Temporal Contrast Vision Sensor. IEEE Journal of Solid State Circuits, Feb. 2008, 43(2), 566-576. (PDF)

Driver Installation

Make sure, libusb is installed on your system:

  1. $ sudo apt-get install libusb-1.0-0-dev

Only a udev rule is needed to run the DVS driver. An install script is provided in the package dvs_driver.
2. $ roscd dvs_driver
3. $ ./install.sh (needs root privileges)

You can test the installation by running a provided launch file. It starts the driver, the renderer, an image viewer, and the dynamic reconfigure GUI.
4. $ roslaunch dvs_renderer mono.launch

DVS Calibration

The calibration of a DVS is a two-stage procedure. First, the focus must be adjusted. Then, the intrinsic camera parameters are estimated.

Focus Adjustment

Adjust the focus of the DVS. One way of achieving this is using a special pattern, e.g. the Back Focus Pattern.

Instrisic Parameters

To run the intrinsic camera calibration, we use a 5x5 LED board that is blinking at 500Hz. The calibration procedure is then started using
$ roslaunch dvs_calibration intrinsic.launch
You will see an RQT interface with all necessary information. Top left is the calibration GUI, which displays the amount of detected patterns. Currently, pattern detection does not seem to work indoors. Try close to a window. Collect at least 30 samples before starting the calibration. This can take up to a few minutes and freezes your RQT GUI. Once done, the calibration parameters are shown and can be saved. The camera parameters will be stored in ~/.ros/camera_info. When you plug that DVS again, this calibration file will be loaded and published as /dvs/camera_info.

The image viewers below show the following:

  1. Accumulated DVS renderings: you should see the blinking LEDs and the gradients in the scene
  2. Detected blinking: black regions mean more detections. Once the pattern is detected, the counter in the calibration GUI should increment. The detected pattern is also visualized for a short moment.
  3. Rectified DVS rendering: once the calibration is done, you can see how well it turned out. Check if straight lines are still straight, especially in the border of the image.

Stereo DVS Calibration

Setup

Connect the two DVS from OUT (master) to IN (slave). GND must not be connected if both DVS are connected over USB to the same computer, to avoid ground loops. Time synchronization is performed automatically in the driver software. Since each DVS has a separate driver, the ROS messages might arrive at different times. Hover, the timestamps within the messages are synchronized.

Calibration

  1. Calibrate each DVS independently
  2. Use $ roslaunch dvs_calibration stereo.launch
  3. Use the same checkerboard with blinking LEDs and make sure it is visible in both cameras. Collect at least 30 samples.
  4. Start the calibration and check the reprojection error. Then save it (this will extend your intrinsic camera info files with the stereo information).

Calibration Details and Parameters

The calibration requires a board with a regular grid of blinking LEDs. In our case we have a 5x5 grid with a 0.05m distance between the LEDs. One of the rows can be turned off (to make a 5x4 grid) to avoid confusion in the stereo case. The following parameters can be tuned using ROS parameters:

  • dots_w, dots_h (default: 5) is the number of rows and columns in the grid of LEDs
  • dot_distance (default: 0.05) is the distance in meters between the LEDs

If you have your own LED board with different LEDs or blinking frequencies, you might want to tweak these parameters as well:

  • blinking_time_us (default: 1000) is the blinking time in micro-seconds
  • blinking_time_tolerance (default: 500) is the tolerance in micro-seconds to still count the transition
  • enough_transitions_threshold (default: 200) is the minimum number of transitions before searching the LEDs
  • minimum_transitions_threshold (default: 10) is the minimum number of transitions required to be considered in the LED search
  • minimum_led_mass (default: 50) is the minimum "mass" of an LED blob, i.e., the sum of transitions in this blop
  • pattern_search_timeout (default: 2.0) is the timeout in seconds when the transition map is reset (it is also reset when the LED grid was found)