Support Website Contact Support Blog

Getting from DSM to DTM

Novice user would appreciate perspective. It looks like Pix4d released an algorithm-generated dig terrain model about 3 years ago, and prior to that the user edited the point cloud to remove man-made artifacts like equipment and buildings. Subsequently the point cloud editing capability was retained.

Are there typical conditions where the system-generated DTM will likely do a great job of eliminating artifacts, and conditions that would likely need point cloud editing?

Just trying to understand. Thanks for any input… R

Hi @rrussell,

It depends on the dataset, on how accurately the algorithm can classify the point cloud. Once it’s correctly done, the software won’t have much problem generating a good quality DTM.

It’s hard to speak about the conditions. The algorithm differentiates four classes: terrain (ground), vegetation, building and unknown focusing on the geometry and colour of the objects. Based on the complexity of the reconstructing area, the accuracy of the algorithm varies. In other words, if in the dataset there are a lot of overlapping environmental objects (e.g. there is a dirt road partially covered by grass) the software will have a problem assigning it to the proper category. Then, some artefacts may occur. In this example, the manual edition would be necessary.

I hope it helps you understand the process.


That does help. Thanks.

A refinement of the question – When the algorithm removes objects from the DSM to make the DTM, does is always fill in terrain where the object was, or will it leave a hole sometimes ? An example: A large pile has trees growing out of the flat top and I want to know the volume of the pile. If the algorithm can identify the trees, will the volume of the pile reflect holes where the trees were removed, or will it assume terrain where the trees were removed? A similar example is a construction site with a lot of cut and fill being done. If the algorithm can successfully identify the earth moving equipment, will it put terrain beneath the removed equipment, or will it leave a hole where the equipment was removed?

I’m probably asking for absolute answers to questions that vary on a case-by-case basis, but that’s why I’m trying to understand what the algorithm will attempt to do and what to expect for its limitations so I’ll know how to answer a client who asks if I can do a job with drone mapping. I’d rather decline a job than accept one I can’t deliver good results.

You’ve already helped, so thanks for your patience. Regards… Bob R.


When the algorithm removes objects from the DSM to make the DTM, does is always fill in terrain where the object was, or will it leave a hole sometimes?

That isn’t the way the algorithms works. It doesn’t remove objects but smooths the entire class based on the strength applied to a specific category.

The examples you described aren’t related to the DTM but to how to measure volumes where other objects occur on the stockpile, e.g. trees, excavator.

The volume is computed based on the DSM. Therefore the DSM should be generated without considering those objects. Have a look at our instruction on How to Remove Objects from a Stockpile for the Volume Computation. Ina in this video showed exactly how you can remove unwanted items from the point cloud to prepare the data for the volume computation. For your comfort, I adjusted the video at the right time.

Let me know whether you have other questions. It’s essential to be confident you delivered the best results to your clients.

Cheers :slight_smile: