Problem with pix4D mapper processing camera not aligned

Regards, I tried to process in PIX4D Mapper a larger volume of pictures (~900 )
taken with a DJI Phantom 4 RTK, I managed to complete the processing but the results are not according to the flight, the flight was made with the camera set at 90 degrees to ground throughout the flight. I tried both Standard 3D Maps and Rapid 3D Maps Rapid / Low Res processing and the results are poor. Initially I thought it was a PC problem but I tried it on other PCs and the problem persists as well. I tried to use other processing programs (pix4d matic, dji terra) but there the processing is done without problems and the results are as expected. I have attached pictures and the file with the report provided by Pix4D Mapper.

Hi razvan,
Can you share your quality report so someone can assess the issue?


Thank you for sharing the quality report.

Given this is an RTK equipt drone I would like to assume that the image geolocation is accurate. According to the overlap map and ATP per pixel average diagram, there appears to be gaps in the image content.

It would be better to see solid green throughout the subject area. These gaps seem to indicate areas with low texture or a general lack of overlap.

It would be better to see solid white here. Black specs indicate the software is unable to find ATPs in those areas.

I see that you have used Rapid, Image Scale: 0.25, which might not be helping us here. In this case, I think it might be better to give the software more image content to work with. I would suggest an approach that gives PIX4Dmapper more image content to work with while also suggesting that it should put more confidence into the initial camera positions.

This would start with setting the Keypoints Image Scale to Full.

Then in the Calibration tab: set Internal Parameters OptimizationAll Prior, and RematchCustom, Rematch.

You could also consider the Alternative Calibration Method, however, I do not have enough familiarity with the dataset to make this determination. There are certain requirements the imagery needs to follow(the majority of images need to be taken from nadir, among others). The following documentation explains the option if you care to explore it. Menu Process > Processing Options… > 1. Initial Processing > Calibration From what I can see in the orthomosaic this dataset appears to be toeing the line between flat terrain and low texture.

I would be curious to see the quality report after running with these processing options, I would hope this will give you a better reconstruction, otherwise should point us in a better direction. I look forward to hearing the results.