Skip to content

Commit

Permalink
chore: fix spell error
Browse files Browse the repository at this point in the history
Signed-off-by: vividf <yihsiang.fang@tier4.jp>
  • Loading branch information
vividf committed Jul 16, 2024
1 parent 328d740 commit 9c77661
Showing 1 changed file with 3 additions and 3 deletions.
6 changes: 3 additions & 3 deletions calibrators/mapping_based_calibrator/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -36,7 +36,7 @@ As the tool aims to apply registration algorithms on the pointclouds from both t
To refine our calibration process, the tool first temporarily pairs `mapping lidar` keyframes with their closest `calibration lidar` frame, and then selects the final calibration pairs (`mapping lidar` keyframes and `calibration lidar` frame) based on the criteria listed below.

- The pairs have low time difference and low interpolation error (such as speed, estimated acceleration, etc.) between the keyframe and the `calibration lidar` frame.
- The pointlcoud in `calibration lidar` frame has enough features for calibration, which the tool filters out the pointcloud that has low variance in the z-axis that is mostly a plane.
- The pointcloud in `calibration lidar` frame has enough features for calibration, which the tool filters out the pointcloud that has low variance in the z-axis that is mostly a plane.

##### Data preprocessing

Expand All @@ -56,7 +56,7 @@ The first step of base-lidar calibration would be the same as [Step 1](#step-1-c

#### Step 2: Extract ground plane from the pointcloud

After constructing the map, we could get the augmented pointcloud from `mapping lidar`, which would be the same process as [Step 2](#step-2-prepare-calibration-data). Afterward, the tool utilizes PCA and a segmentation algorithm for extracting the ground plane from the augmented pointcloud.
After constructing the map, we could get the augmented pointcloud from `mapping lidar`, which would be the same process as [Step 2](#step-2-calibration-data-preparation). Afterward, the tool utilizes PCA and a segmentation algorithm for extracting the ground plane from the augmented pointcloud.

#### Step 3: Estimate transformation

Expand Down Expand Up @@ -224,7 +224,7 @@ The transformation between the lidar and the ground pose is calculated by utiliz

## Known issues/limitations

- As described in [Step 2](#step-2-prepare-calibration-data), the calibrator interpolates the pose of the `mapping lidar` at the timestamp of the `calibration lidars`. Therefore, the calibrator highly requires time synchronization between all of the lidar sensors.
- As described in [Step 2](#step-2-calibration-data-preparation), the calibrator interpolates the pose of the `mapping lidar` at the timestamp of the `calibration lidars`. Therefore, the calibrator highly requires time synchronization between all of the lidar sensors.
- It is required that the `mapping lidar` have high resolution. For the `calibration lidar`, both high and low resolutions are acceptable. Therefore, the vehicle with only low-resolution lidars could not use the calibrator.
- A good initial calibration parameters is required for the calibrator. Better initial calibration parameters could help the registration algorithm estimate a more reliable transformation.

Expand Down

0 comments on commit 9c77661

Please sign in to comment.