Depth detection of trench from LIDAR data (Iphone 15)

Hi hi,

I recently received a challenge to determine the depth of a trench where an electrical cable needs to be laid down by a telecommunication company. The company wants to ascertain the depth using LIDAR data from an iPhone 15. Despite my efforts to find relevant papers on this topic, I haven’t had any luck. Could you please provide some ideas on how to make this possible?

I appreciate your help. :pray:

Hi, Shiwang.

I am not an expert on iPhone LiDAR, but I work with airborne LiDAR in my job. I hope to be helpful by providing some ideas and preprocessing the data you may need for your work.

I will proceed under the assumption that you only have LiDAR data at hand.

For preprocessing:

  • Check coordinates
  • Georeference
  • Clean noise
  • Classify ground and non-ground objects (like buildings, vegetation, etc.)

A particularly important aspect of preprocessing is determining the ground elevation.

For the main method, I suggest two idea:

Idea 1: Raster-Based Approach

  • Convert LiDAR data to DSM (Digital Surface Model) in GeoTIFF format.
  • Perform detection or segmentation for the trench.
  • Calculate the difference between ground elevation and trench elevation.

Note: DSM is a 1-band image where each pixel represents elevation. You can use any image processing technique on it.

Idea 2: Point Cloud-Based Approach

  • Segment the trench bottom surface. All supervised, unsupervised learning, and rule-based methods are worth trying.
  • Calculate the difference between ground elevation and trench elevation.

If you plan to use machine learning for point cloud data (LiDAR), please check out ‘part segmentation’ for point clouds. I believe it is similar to your task.

I apologize for my limited English skills.
Hope all of your business will be successful. :pray:

3 Likes

Hi Chiron,

Thankx a lot for sharing this valuable information. We are creating some benchmarks with your suggestions. :slight_smile:

At least we have some starting point. :wink:

Best regards,
Shiwang

1 Like