fix(autoware_image_projection_based_fusion): fix bug of fov checker in pointpainting

[technolojin]

Description

When the point is projected to the image space, the point is checked if the point is in the filed-of-view of corresponding camera.
It may cut unnecessary calculation of later roi matching.

However, @a-maumau found that the current implementation of loading camera parameter is done at the initialization and do not get the camera info.

    auto fx = camera_info_map_[roi_i].k.at(0); // fx = 0
    auto x0 = camera_info_map_[roi_i].k.at(2); // x0 = 0
    tan_h_[roi_i] = x0 / fx;                   // tan_h_[0] == NaN

    if (p_z <= 0.0 || p_x > (tan_h_.at(image_id) * p_z) || p_x < (-tan_h_.at(image_id) * p_z)) {
      continue;
    }
    // p_x > (tan_h_.at(image_id) * p_z) == false
    // p_x < (-tan_h_.at(image_id) * p_z) == false

Therefore the checker of the FoV was not done correctly.

This PR will not induce any logic difference.

Special thanks to @a-maumau

Related links

Parent Issue:

• Link

How was this PR tested?

Visually confirmed that the given roi and the pointcloud class is corresponding.

Notes for reviewers

None.

Interface changes

None.

Effects on system behavior

None.

[github-actions]

Thank you for contributing to the Autoware project!

🚧 If your pull request is in progress, switch it to draft mode.

Please ensure:

• You've checked our contribution guidelines.
• Your PR follows our pull request guidelines.
• All required CI checks pass before marking the PR ready for review.

[codecov]

Codecov Report

Attention: Patch coverage is 0% with 10 lines in your changes missing coverage. Please review.

Project coverage is 29.61%. Comparing base (2127ca8) to head (355c183).
Report is 12 commits behind head on main.

Files with missing lines	Patch %	Lines
...ion_based_fusion/src/pointpainting_fusion/node.cpp	0.00%	10 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #9637      +/-   ##
==========================================
+ Coverage   29.57%   29.61%   +0.03%     
==========================================
  Files        1440     1441       +1     
  Lines      108348   108370      +22     
  Branches    41384    41389       +5     
==========================================
+ Hits        32049    32089      +40     
+ Misses      73190    73169      -21     
- Partials     3109     3112       +3     

Flag	Coverage Δ		*Carryforward flag
differential	2.62% <0.00%> (?)
total	29.61% <ø> (+0.03%)	⬆️	Carriedforward from 849e529

*This pull request uses carry forward flags. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[badai-nguyen]

@technolojin Could you make some change to check fov before calcRawImageProjectedPoint function to avoid unnecessary projection calculation for points out of FOV?

[technolojin]

@badai-nguyen
We do not know until the lens distortion is corrected.
In the case the lens warps the view inside, the more points will be inside of FoV.
In another word, if it comes before the correction, it may miss some points on the edges.

In the case the camera model is the pinhole camera and do not have lens distortion, the previous process p_x > (tan_h_.at(image_id) * p_z) will be done in the calcRawImageProjectedPoint and it means it doubles the same process anyway.

[technolojin]

well... If the lens distortion is considered when the boundary is set, we may can cut some inefficient process.. let me consider it.

[technolojin]

resolved by 355c183

to check distortion effect, additional process is needed and will be implemented by following PRs