Support Website Contact Support Blog

Why does my 3D model look so poor?


Hoping for some guidance here. I recently did my first model (using Blenheim Palace as the subject) and have run the imagery through Pix4DMapper. I am a little disappointed with the model but I’m looking for some advice here to hopefully improve things. I used a Phantom 4 flown with Pix4D Capture app to fly the grid. In total it captured 182 images on the grid. Flight height was 220 feet, 62 deg camera angle and 80% overlap specified. I then captured around 20 images of the elevations/facades (drone at about 5m off the ground with camera facing straight on to the building façade).

I have run both sets of data (ie. with and without the elevation images) and I’m rather disappointed with both. Strangely, the model WITHOUT the elevation images actually looks BETTER. The attached screenshots from mesh view show with and without elevation images. For the version that includes the elevation images, I actually did the annotation necessary to remove the blue sky that was appearing above the roof features (you can see a sample image in the screenshot with the annotated red sky). After applying the annotation/changes the sky issue was removed but the elevation imagery is still very disappointing.

These models look even worse when uploaded to Sketchfab. I have to say that DroneDeploy and MapsMadeEasy look far better at this stage. Am I doing something wrong? Thanks in advance!


Here is a screenshot showing poor 3D model (this one uses elevation images).

Here is another screenshot, this model does not use elevation images. The model is actually better than the one shown above, but still not very good.

I think that if you use tie points, an some lines the output would be much better.

I agree - add manual tie points to tighten things up.

Jacques, the subject you are modeling is only using a very small percentage of each image. I think the camera is much too far away. All the ground and sky are ‘wasted’ pixels that could have been used for modeling. 80% overlap is a good rule of thumb for producing orthophotos and terrain models, but for architectural and engineering purposes where you want detailed modeling it should be much much higher for many projects.

Especially if you have solid objects in front of features, or your subject has great depth like the windows and the tower areas.

You could capture a very dense session using a terrestrial camera to combine with your UAV flight and I think this would be much better.