lidar_colorization and mast3r reconstruction codes added

GEMstack/offboard/lidar_colorization/README.MD

@@ -0,0 +1,87 @@
+# Point Cloud Colorizer from Multi-View Images
+
+This script projects LiDAR point clouds into multi-view camera images and colorizes the 3D points using RGB values interpolated from the corresponding camera views. It supports multiple camera types and exports both raw and colorized point clouds in PLY format.
+
+## Features
+
+- Projects LiDAR point clouds into camera frames using calibrated extrinsic and intrinsic parameters
+- Handles undistortion of input images using OpenCV
+- Samples RGB values at subpixel precision from multi-view images
+- Merges color information into a unified point cloud and saves as `.ply`
+- Supports multiple camera types: front-right (`fr`), rear-right (`rr`), front-left (`fl`), rear-left (`rl`)
+
+## Installation
+
+Make sure you have the following dependencies installed:
+
+```bash
+pip install numpy opencv-python open3d
+```
+### Usage: 
+
+```bash
+python path/to/colorize_pointcloud.py \
+    --folder_path /path/to/your/pointcloud_and_image_data \
+    --output_path /path/to/save/output \
+    --camera_types fr,rr
+```
+##### 1. Arguments
+```bash
+    --folder_path: Directory containing .b files (NumPy arrays) with point cloud data and corresponding images ({camera_type}_{index}.png)
+
+    --output_path: Directory where the resulting .ply point clouds will be saved
+
+    --camera_types: Comma-separated list of camera types to project and colorize from. Supported values: fr, rr, fl, rl
+```
+
+###### 2. Data Format Requirements
+- Point cloud files: .b files that store N x 3 or N x 6 NumPy arrays of 3D points (with optional RGB). You can rename your saved pointcloud in NumPy arrays format to .b and then use this script.  
+<br>
+- Images: RGB .png images named like fr_001.png, rr_001.png, etc., located in the same directory as the .b files
+<br>
+
+###### 3. Calibration: 
+Calibration parameters must be accessible via the settings module:
+
+- Extrinsic:
+
+    ```python
+    settings.get("calibration.<camera_name>.extrinsics.rotation")
+    settings.get("calibration.<camera_name>.extrinsics.position")
+    ```
+    ```python
+    settings.get("calibration.top_lidar.rotation")
+    settings.get("calibration.top_lidar.position")
+    ```
+
+- Intrinsic:
+    ```python
+    settings.get("calibration.<camera_name>.intrinsics.focal")
+    settings.get("calibration.<camera_name>.intrinsics.center")
+    settings.get("calibration.<camera_name>.intrinsics.distort")
+    settings.get("calibration.<camera_name>.intrinsics.skew")
+    ```
+###### 4. Outputs
+For each .b file:
+
+    pure_pointcloud_<filename>.ply: raw point cloud without color
+
+    colored_pointcloud_<filename>.ply: point cloud with interpolated RGB from visible camera views
+
+##### Example
+```python
+python colorize_pointcloud.py \
+    --folder_path data/session1 \
+    --output_path data/colored_output \
+    --camera_types fr,rr
+```
+
+## Notes
+    Only the 3D points that are visible in at least one selected camera are assigned color.
+
+    If no image exists for a particular view or projection fails, that camera is skipped.
+
+    Requires accurate and synchronized calibration between LiDAR and camera systems.
+
+    Author: Henry Yi
+    License: MIT