Interesting complaint from a mapping client...

May 6, 2014
Reaction score
If you're building the ortho from the ground surface only, then issues like this all but disappear. Pix4D isn't great at distinguishing the different surfaces (man-made, road, ground, vegetation) so it can be a crap shoot. It's one of Pix's biggest deficiencies and is why Agisoft tends to be a better option for higher level editing and terrain manipulation.
Dec 11, 2016
Reaction score
What’s up Dok! Unfortunately, I haven't had the opportunity to work with Agisoft, so I can only speak toward Pix4D. Hopefully one day I'll get a chance to try Agisoft and find that it works better.

Harley, your partner also has legitimate concerns. I have tried to capture some screen shots to illustrate my edits.

Here is a screen shot of a fence line before edits. It is a white, vinyl split-rail fence, about 4 feet tall. You can see the artifact-ing that has occurred on the fence and posts. And it is hard to determine exactly where the fence is.


By drawing a long, narrow region around the fence, I was able to find a single image that better showed the fence and posts. Depending on your image overlap, you’ll probably have a lot of images to choose from. Many do a bad job at replacement. But usually one of the images toward the top of the options does a good job. It’s a region-by-region process. 95% of the time I use the “Planar Projection” option. Long runs of fence will need to be done in shorter segments which you will have to determine, but say 100-150 regions. But your partner will be happy that it isn’t done one fencepost at a time.


Here is an example of a sign. Do you see it? Actually, I almost missed it, but caught it while reviewing the point cloud.


After editing the orthomosaic by placing a region around the sign, I was able to find an image that showed the sign face and it’s position.


Like I mentioned before, editing the ortho takes time, and I make sure to include that time for it in my proposals. But I think it improves the orthomosaic and I want it to be as useful to the end user as possible. Here is a screenshot of this project showing all the edits that were made before delivery. (Probably a two-hour process.) I clean up things like vehicles, signs, fences, columns, overhead wires, signals, and overhanging trees.


One thing that I have found is that when you draw a region, try to do it on even terrain, because Pix4D tries to warp the replaced area to match the elevation change. Say three of a region’s vertices are at ground level, but one falls on top of a four-foot-tall bush. The image replacement will warp and just never look quite right, so watch out for this. Usually I will edit the point cloud and eliminate all the above ground stuff (shrubs, trees, poles, fences, buildings) that may affect the elevations of the vertices of the regions.

Hope that helps explain my process, and maybe gives you some options to offer your client that wants to see those pesky fences.

New Threads

Forum statistics

Latest member
The scientist