Point clouds from stereo camera not correct visualized in RVIZ after using ROS stereo calibration

edit

Hi

I have two Grasshopper3 GS3 FLIR Cameras working in Master-Slave mode as a stereo camera. The baseline between them is 50cm. I used the ROS stereo calibration tool to calibrate them. The epipolar error was 0.3pixel. This the outcome of the calibration

Left:
('D = ', [0.02737567282717913, -0.4551047182461572, -0.001137413019162705, -0.0009915982202950845, 0.0])
('K = ', [2296.961069410927, 0.0, 625.398224476376, 0.0, 2308.7971426019503, 501.86694724205375, 0.0, 0.0, 1.0])
('R = ', [0.9936699369376326, -0.009155412185781097, -0.11196532880347168, 0.007244581743852582, 0.999821294676313, -0.017461235555884403, 0.11210518481193325, 0.016539562856674725, 0.9935586899618928])
('P = ', [2553.90373019936, 0.0, 908.1497840881348, 0.0, 0.0, 2553.90373019936, 510.66265869140625, 0.0, 0.0, 0.0, 1.0, 0.0])

Right:
('D = ', [-0.12865514431050615, 0.27039839587223014, 0.00025897954088409614, 0.0027478332323184152, 0.0])
('K = ', [2323.3758802873845, 0.0, 669.4989931082697, 0.0, 2333.0361667939355, 516.0905046926046, 0.0, 0.0, 1.0])
('R = ', [0.9952103941041639, -0.008350812384399213, -0.09739884701348, 0.010010034133704436, 0.9998127795356607, 0.01655913928375224, 0.09724232969071027, -0.0174547933157972, 0.9951076622690764])
('P = ', [2553.90373019936, 0.0, 908.1497840881348, -1215.1659205593778, 0.0, 2553.90373019936, 510.66265869140625, 0.0, 0.0, 0.0, 1.0, 0.0])
('self.T ', [-0.47352832465906813, 0.0039733771083513, 0.046343077929299197])
('self.R ', [0.9998845374012215, 0.0025050295419663574, -0.014987884872093293, -0.0030114986360113027, 0.9994218681475354, -0.03386532658126264, 0.014894386254913267, 0.03390655239749609, 0.999314016194412])
None
# oST version 5.0 parameters


[image]

width
1216

height
960

[narrow_stereo/left]

camera matrix
2296.961069 0.000000 625.398224
0.000000 2308.797143 501.866947
0.000000 0.000000 1.000000

distortion
0.027376 -0.455105 -0.001137 -0.000992 0.000000

rectification
0.993670 -0.009155 -0.111965
0.007245 0.999821 -0.017461
0.112105 0.016540 0.993559

projection
2553.903730 0.000000 908.149784 0.000000
0.000000 2553.903730 510.662659 0.000000
0.000000 0.000000 1.000000 0.000000

# oST version 5.0 parameters


[image]

width
1216

height
960

[narrow_stereo/right]

camera matrix
2323.375880 0.000000 669.498993
0.000000 2333.036167 516.090505
0.000000 0.000000 1.000000

distortion
-0.128655 0.270398 0.000259 0.002748 0.000000

rectification
0.995210 -0.008351 -0.097399
0.010010 0.999813 0.016559
0.097242 -0.017455 0.995108

projection
2553.903730 0.000000 908.149784 -1215.165921
0.000000 2553.903730 510.662659 0.000000
0.000000 0.000000 1.000000 0.000000

Then I got the Q matrix with this code

#include<iostream>
#include<opencv2/calib3d.hpp>
#include<opencv2/opencv.hpp>

int main()
{   
    cv::Size imgsize;

    cv::Mat camMatrix_1 =(cv::Mat_<double>(3,3) << 2296.961069, 0, 625.398224,0, 2308.797143,501.866947, 0, 0, 1);
    cv::Mat camMatrix_2 = (cv::Mat_<double>(3,3) << 2323.375880, 0, 669.498993,0, 2333.036167, 516.090505,0, 0, 1);
    cv::Mat dist_1 = (cv::Mat_<double>(1,5) << 0.027376, -0.455105, -0.001137, -0.000992, 0);
    cv::Mat dist_2 = (cv::Mat_<double>(1,5) << -0.128655, 0.270398, 0.000259, 0.002748 , 0 );
    cv::Mat R = (cv::Mat_<double>(3 ...