code is not successfully running for the shared data about calibanything

[vigneshgk7]

After successful build, I'm getting below error.

./bin/run_lidar2camera ./data/nuscenes/calib.json
Reading json file complete!
----------Start processing data----------
Processing data 1:
Point cloud num: 31499
Points num after downsample: 18258
Segmentation fault (core dumped)

what's wrong with my code? any idea where do I have to check. many times

[vigneshgk7]

Also, I had debugged and facing an error in calibration.cpp file line **seg.segment(indices_plane, coefficients);
Any idea why it throws segmentation fault?
Thanks in an advance.

[iceriver97]

请问你最后解决了吗?

[vigneshgk7]

Its working now, after a change made "min_plane_point_num" from 2000 to 5000 and down_sample is_valid is false in calib.json file. but not sure why I'm getting an error when down_sample flag is true.

Thank you @iceriver97

[iceriver97]

Its working now, after a change made "min_plane_point_num" from 2000 to 5000 and down_sample is_valid is false in calib.json file. but not sure why I'm getting an error when down_sample flag is true.

Thank you @iceriver97

感谢, 我发现可以将PCL库升级到1.10 并链接到 libLidarToCamera.a, 而不是直接链接到run_lidar2camera, 这样程序不再有任何问题, 我有几点疑问不知您是否能帮忙解答一下: 1. calib.json 中的相机内参为什么不是3X3 而是3X4；2. 您是否在个人数据上进行过测试, 效果怎么样? 3. 程序运行速度是不是很慢,一般需要几个小时?

[vigneshgk7]

Hello @iceriver97 ,
Yeah PCL version needs to be upgraded from 1.8 to 1.10,

1. A camera's intrinsic is typically in 3x3, capturing its internal parameters like focal length, optical centre. 3x4 matrix is usually reserved for the camera projection matrix, which combines intrinsic and extrinsic parameter for 3D-2D mapping.
2. It works for kitti and nuscenes data, but for my personal data the calibration results are not good, maybe clustering and hyperparameter tuning is required.
3. Actually for one image (4096x3000) it takes ~2minutes.

[kkk3449]

I also Segmentation fault (core dumped) error at 'seg.segment(*indices_plane, *coefficients);' and follow "min_plane_point_num" from 2000 to 5000 and down_sample is_valid is false in calib.json file solution, but is not working. same error... .

[iceriver97]

I also Segmentation fault (core dumped) error at 'seg.segment(*indices_plane, *coefficients);' and follow "min_plane_point_num" from 2000 to 5000 and down_sample is_valid is false in calib.json file solution, but is not working. same error... .

您有没有尝试将PCL库升级到1.10 并在CMakeList.txt中链接到 libLidarToCamera.a, 而不是直接链接到run_lidar2camera？

[aifpgahzw]

@iceriver97
我已经将PCL库升级到1.10 并在CMakeList.txt中链接到 libLidarToCamera.a，但还是出现那个问题，不知道您是否还做了别的改动？

[QLEOWANGQ]

@aifpgahzw 和您一样的问题，出现了段错误
Reading json file complete!
----------Start processing data----------
Processing data 1:
Point cloud num: 317
段错误 (核心已转储)

这个问题如何解决？

[QLEOWANGQ]

@vigneshgk7 排查之后同样是在这里出现了段错误，不知道您是怎样解决的？
seg.segment(indices_plane, coefficients);