Depth Map Overlay

Objective

Overlay a graphical representation of distance readings from a 2D LiDAR scanner onto a POV image taken of a mobile robot's FOV.

Install

pipenv install --dev

Usage

Inputs (input/)

• An image of the required scene in your preferred format matching the configured resolution below.
• A Python dictionary of raw LiDAR distance measurements.

Outputs (output/)

• combined.jpg - Contains a grayscale version of the scene image with the relative distance measurements overlayed as red dots. The more solid a red dot is the closer that point is to the camera.
• points_only.jpg - Contains only the red dot points used in the combined image.

Capture image from webcam (optional)

pipenv run python tools/capture_scene_image.py --camera_source 0 --filename input/scene.jpg

Generate composite image

pipenv run python generate.py --image input/scene.jpg --data input/data.py

Components

• 2D RPLiDAR scanner (RPLiDAR C1)
• Generic USB webcam (Hikvision DS-U02)

Configuration Settings

Adjust in generate.py.

• CAMERA_HORIZONTAL_FOV_DEGREES = 86
• CAMERA_VERTICAL_FOV_DEGREES = 55
• CAMERA_HORIZONTAL_RESOLUTION_PX = 640
• CAMERA_VERTICAL_RESOLUTION_PX = 480
• CAMERA_VERTICAL_CENTREPOINT_IN_FRAME_MM = 130
• LIDAR_VERTICAL_OFFSET_MM = 45
• LIDAR_DISTANCE_OFFSET_MM = -120
• CAMERA_FOCAL_LENGTH_MM = 3.6

Limitations

• The format and parsing of the LiDAR data file is not ideal (working with data generated for another project).

Changelog

22 October 2024

• Correct interpolation to the vanishing point.
• Change dot colour to red.

21 October 2024

• CAMERA_FOCAL_LENGTH_MM setting is now required (changed from reserved).
• Depth map overlay is more tightly fitting the scene. No longer just a straight line across the scene (using the camera's focal length to approximate the vanishing point).