XB3

xb3: xb3 driver

This package is a Ros based driver for Bumblebee xb3 stereo cameras.

Homepage: http://ros.org/wiki/xb3

This is a ros driver package for Bumblebee xb3 stereocameras. This node receive images from camera and publish image topics and camera service info.

Author:

Tiago Talhada

Date:

May 2012

Table of contents

1. Published topics

2. How to use the package?

3. How to process stereo with stereo_image_proc?

4. Stereo image processing parameters

   1. Pre - Filtering

   2. Post - Filtering

   Published Topics

   This package publishes the following topics
   

   /snr/scam/short/left/image_raw
   /snr/scam/short/left/camera_info
   /snr/scam/short/right/image_raw
   /snr/scam/short/right/camera_info
   

   /snr/scam/wide/left/image_raw
   /snr/scam/wide/left/camera_info
   /snr/scam/wide/right/image_raw
   /snr/scam/wide/right/camera_info
   

   How to use the package?

   Here is an example on how to use xb3 package.

   <launch>
     <!-- This is an arpua lauch file for starting up xb3. -->
     <group ns="/snr/scam">
       <node name="xb3" pkg="xb3" type="xb3">
           <remap from="/xb3_optical_frame" to="/atc/camera/xb3/left"/>  
     </node>
           <param name="width" type="int" value="640"  />
           <param name="height" type="int" value="480"  />
   </group>
   </launch>
   

   You can use the full camera resolution 1280x960 or 640x480.
   If you change your resolution, remember to change also your calibration files.

   How to process stereo with stereo_image_proc?

   You can use the following as a launch file to run stereo_image_proc for both baselines.

   <group ns="/snr/scam/short">
         <node name="stereoproc" pkg="stereo_image_proc" type="stereo_image_proc" 
               args="_prefilter_size:=255 _prefilter_cap:=63 _correlation_window_size:=31 _min_disparity:=0 _disparity_range:=32 
               _uniqueness_ratio:=10 _texture_threshold:=100 _speckle_size:=400 _speckle_range:=15">
         </node>
   </group>
   
   <group ns="/snr/scam/wide">
         <node name="stereo_image_proc" pkg="stereo_image_proc" type="stereo_image_proc" 
               args="_prefilter_size:=255 _prefilter_cap:=63 _correlation_window_size:=31 _min_disparity:=0 _disparity_range:=32 
               _uniqueness_ratio:=10 _texture_threshold:=100a _speckle_size:=400 _speckle_range:=15 __name:=xb3_wide_stereo_proc" />
   </group>
   

   Stereo image processing parameters

   Here is a brief description on what does each parameter from opencv block match stereo processing.
   I take this description from the book: Learning OpenCV (page: 438:444).
   There are three stages to the block-matching stereo correspondence algorithm.

   1. Pre-filtering to normalize image brightness and enhance texture.

   2. Correspondence search along horizontal epipolar lines using an SAD ("sum of absolute difference") window.

   3. Post-filtering to eliminate bad correspondence matches.

   Pre-filter parameters

   1. prefilter_size (5, 255) (5x5, 21x21)
      -Reduce lighting differences and to enhance image texture.
      -The center pixel Ic under the window is replaced by min[max(Ic - I , Icap), Icap],
      where I is the average value in the window and Icap is
      a positive numeric limit whose default value is 30 (prefilter cap).
      

      • Default 9x9.
        

   2. prefilter_cap (1, 63) (about 1/4 of image range (0-255))
      -definition above (prefilter_size)
      

      • default 30.
        

   3. correlation_window_size (5, 255)
      -With smaller windows the processing speed is much higher
      -SAD matching window size. (5x5,...,21x21).
      

      • default 15.
        

   4. min_disparity (-128, 128)
      -Minimum disparity (=0)
      -Minimum value where the matching search will start
      

      • default 0.
        

   5. disparity_range (32, 128)
      -Reducing the disparity range to be searched can help cut down computation
      time by limiting the length of a search for a matching point along an epipolar line.
      
      • default 64
        

   Post filtering parameters

   1. uniqueness_ratio (0, 100)
      -The minimum margin in percents between the best (minimum) cost function value and
      the second best value to accept the computed disparity.
      

      • UniquenessRatio > (match_val-min_match)/min_match.
        

   2. texture_threshold (0, 100)
      -Minimum value for texture to compute disparity.
      -That is, if the sum of absolute values of x-derivatives
      computed over SADWindowSize by SADWindowSize pixel neighborhood
      is smaller than the parameter, no disparity is
      computed at the pixel.
      

      • default 10.
        

   3. speckle_size (0,1000)
      -The maximum area of speckles to remove (set to 0 to disable speckle filtering)
      -Set a speckle detector over a speckle window to prevent speckles.
      

      • default 100
        

   4. speckle_range (0,31)
      -Acceptable range of disparity variation in each connected component.
      -As long as the minimum and maximum detected disparities are within speckleRange,
      the match is allowed.
      
      • default 4.