Thank you very much for your hard work, and it is very meaningful in sensor calibration! But when I tried this project, I encountered some problems and needed your help!
Due to the lack of actual LiDAR and calibration office, I attempted to calibrate it in Gazebo simulation. The LiDAR used is Velodyne HDL-32E. The camera parameters used are as follows:
Height: 512
Width: 640
Disruption_model: "plumb'bob"
D: [0.0, 0.0, 0.0, 0.0, 0.0, 0.0]
K: [554.254691191187, 0.0, 320.5, 0.0, 554.254691191187, 256.5, 0.0, 0.0, 1.0]
R: [1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 1.0]
P: [554.254691191187, 0.0, 320.5, -38.79782838331, 0.0, 554.254691191187, 256.5, 0.0, 0.0, 0.0, 1.0, 0.0]
Binning_x: 0
Binning_y: 0
ROI:
X_offset: 0
Yyoffset: 0
Height: 0
Width: 0
Do-rectify: False
When I follow the readme tutorial for calibration, I can correctly capture the calibration board in the camera and the calibration board in the LiDAR. As shown in the following figure:
I collected a total of eight samples and the final results are as follows:
---- Calculating average reprojection error on 8 samples ----
1/ 8 | dist= 0.000m, dimerr= 70.300mm | error: 0.258pix --> 3.262mm
2/ 8 | dist= 0.000m, dimerr= 61.920mm | error: 0.040pix --> 0.499mm
3/ 8 | dist= 0.000m, dimerr= 58.280mm | error: 0.208pix --> 2.156mm
4/ 8 | dist= 0.000m, dimerr= 88.620mm | error: 1.001pix --> 14.972mm
5/ 8 | dist= 0.000m, dimerr= 69.050mm | error: 0.221pix --> 2.957mm
6/ 8 | dist= 0.000m, dimerr= 77.000mm | error: 0.192pix --> 2.141mm
7/ 8 | dist= 0.000m, dimerr= 77.650mm | error: 0.205pix --> 2.371mm
8/ 8 | dist= 0.000m, dimerr= 74.440mm | error: 0.291pix --> 3.116mm
Although the calibration result does not seem to match the actual pose relationship of my sensor, I still tried the final reprojection and the final result is as follows:
---- Calculating average reprojection error on 8 samples ----
1/ 8 | dist= 0.000m, dimerr= 70.300mm | error: 0.258pix --> 3.262mm
2/ 8 | dist= 0.000m, dimerr= 61.920mm | error: 0.040pix --> 0.499mm
3/ 8 | dist= 0.000m, dimerr= 58.280mm | error: 0.208pix --> 2.156mm
4/ 8 | dist= 0.000m, dimerr= 88.620mm | error: 1.001pix --> 14.972mm
5/ 8 | dist= 0.000m, dimerr= 69.050mm | error: 0.221pix --> 2.957mm
6/ 8 | dist= 0.000m, dimerr= 77.000mm | error: 0.192pix --> 2.141mm
7/ 8 | dist= 0.000m, dimerr= 77.650mm | error: 0.205pix --> 2.371mm
8/ 8 | dist= 0.000m, dimerr= 74.440mm | error: 0.291pix --> 3.116mm
Calibration params (roll,pitch,yaw,x,y,z): -1.5710,0.0008,-1.5709,0.2440,-0.0096,-0.1140
Mean reprojection error across 8 samples
- Error (pix) = 0.302 pix, stdev = 0.292
- Error (mm) = 3.934 mm , stdev = 4.544
[ INFO] [1714097377.784034410]: Projecting points onto image for pose #3
This looks very strange. I have tried many times, and each time I get the same extrinsic parameters, and the reprojection is also similar. This result is very strange and bothers me a lot. I couldn't find what problem led to this result. I wonder if you have encountered a similar situation? Can you provide me with some help?
Thank you very much for your hard work, and it is very meaningful in sensor calibration! But when I tried this project, I encountered some problems and needed your help!
Due to the lack of actual LiDAR and calibration office, I attempted to calibrate it in Gazebo simulation. The LiDAR used is Velodyne HDL-32E. The camera parameters used are as follows:
Height: 512 Width: 640 Disruption_model: "plumb'bob" D: [0.0, 0.0, 0.0, 0.0, 0.0, 0.0] K: [554.254691191187, 0.0, 320.5, 0.0, 554.254691191187, 256.5, 0.0, 0.0, 1.0] R: [1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 1.0] P: [554.254691191187, 0.0, 320.5, -38.79782838331, 0.0, 554.254691191187, 256.5, 0.0, 0.0, 0.0, 1.0, 0.0] Binning_x: 0 Binning_y: 0 ROI: X_offset: 0 Yyoffset: 0 Height: 0 Width: 0 Do-rectify: FalseWhen I follow the readme tutorial for calibration, I can correctly capture the calibration board in the camera and the calibration board in the LiDAR. As shown in the following figure:
I collected a total of eight samples and the final results are as follows:
Although the calibration result does not seem to match the actual pose relationship of my sensor, I still tried the final reprojection and the final result is as follows:
This looks very strange. I have tried many times, and each time I get the same extrinsic parameters, and the reprojection is also similar. This result is very strange and bothers me a lot. I couldn't find what problem led to this result. I wonder if you have encountered a similar situation? Can you provide me with some help?