Detecting Cable Channel Edges for Autonomous Robot Control

Hardware
1

Hi everyone,

For my proof-of-concept project, I am developing a mobile robot that can autonomously drive along cable channels. A key part of the project is detecting the edges of these channels to feed positional data into the robot’s control system. The goal is to maintain a precise lateral distance to the channel edges while driving.

I have a few technical questions regarding the use of ZED cameras for this task:

  1. Are ZED cameras suitable for reliably detecting the lateral distance to channel edges in outdoor environments (concrete channels with surrounding vegetation)?

  2. Which ZED camera models would you recommend for this type of edge detection and real-time position measurement?

  3. Are there specific tools, SDKs, or pretrained AI models provided that could help with edge detection and distance estimation?

I have attached two photos to illustrate the situation and the kind of environment where the robot will operate. Any advice on camera setup, sensor placement, or software workflow would be greatly appreciated.

Thank you very much for your help!

Screenshot 2025-09-23 104401
Screenshot 2025-09-23 104401838×597 46.5 KB

Screenshot 2025-09-23 112906
Screenshot 2025-09-23 1129061920×1398 133 KB

2

Hi @Semi
Thank you for describing your interesting project.

The ZED X cameras will suit your design. I recommend considering the ZED X Mini with a 4 mm focal length. The narrow field of view and the short baseline allow you to focus your analysis on the front of the robot with good accuracy.
Furthermore, the global shutter sensor of the ZED X allows you to obtain stable images even if the robot is vibrating on the ground.
Finally, I would add the polarizer filter to improve the quality of the images under the sunlight.

The ZED SDK does not provide ready AI models to detect the grass and segment the areas, but you can easily use your custom AI models:

3

Hi @Myzhar

Thank you very much for your recommendations. I will see how I implement my project and whether I will use a Stereolab camera.