Skip to main content
All CollectionsManaging your dataUploading data
My dataset was not processed as I expected
My dataset was not processed as I expected

Some common challenges we see with datasets and how to avoid them with your data capture

Karen Joyce avatar
Written by Karen Joyce
Updated over a week ago

Here are some of the most common challenges we see with datasets on the platform, along with suggestions for how to improve your data capture next time. If you are seeing a different issue, or would like to know more, please get in touch.

Holes or gaps in the processed data

This can occur in areas of homogenous features (e.g. water bodies) or if there is insufficient overlap and sidelap between successive photos to allow the algorithms to detect matching features in the photos. Aim for 80% overlap and sidelap, and to reduce homogeneity, try flying at a higher altitude as this might increase the probability of detecting 'different' features in each image.

Swirly features in the processed data

Unfortunately this commonly occurs around the edges of some features (particularly trees), where the algorithm has difficulty detecting where one feature starts and another ends. Sometimes we can achieve a better result by testing flights at different altitudes, so we recommend trialing different capture parameters.

Edges of the processed data look weird

The central portion of a dataset will always return a better result than the edges. This is because there are more photos and therefore more overlap comprising the central area than at the edges. Always make sure that you plan your missions to cover an area larger than your region of interest.

Processed data covers a smaller area than that flown

This can occur in areas of homogenous features (e.g. water bodies) at the edges of your dataset or if there is insufficient overlap and sidelap between successive photos to allow the algorithms to detect matching features in the photos. Aim for 80% overlap and sidelap, and to reduce homogeneity, try flying at a higher altitude as this might increase the probability of detecting 'different' features in each image.

Alternatively we also see this if you have uploaded a dataset consisting of multiple non-adjacent flights. In this case, take care to upload separate missions as separate datasets.

Inconsistent colours across the orthomosaic

As the processing algorithms merge individual photos together to create the orthomosaic, it will also perform some colour balancing to try to maintain consistency across the scene. Unfortunately if the input images vary greatly in their colour range, the consistent effect is not always achieved. For best results, check the white balance of your camera, and ideally capture your data on either completely sunny or completely cloudy days. This will help reduce the chance of brightness variation and cloud shadow within your data capture.

Incorrect location of features within the orthomosaic

Sometimes the processing algorithms have difficulty detecting matching points to align individual images. This results in some features appearing in random places in your orthomosaic! Usually this is in areas of dense vegetation or water bodies. Aim for 80% overlap and sidelap, and to reduce homogeneity, try flying at a higher altitude as this might increase the probability of detecting 'different' features in each image.

Dataset is in the incorrect location

The algorithms to process your data rely on the positioning information in each image, derived from the global navigation satellite system (commonly called GPS) on board your drone. The accuracy of this positioning information varies between drones. The best way to obtain accurate positioning is to use a real time kinematic (RTK) enabled drone, coupled with ground control points. Talk to us if you would like to include GCPs in your processing workflow!

DTM looks the same as the DSM

In the processing workflow, the digital surface model is created first. The software then tries to detect the difference between 'ground' and 'non-ground' features before then removing the 'non-ground' to create the digital terrain model. In the even that it cannot detect this difference, the DTM will appear very similar to the DSM. This can occur in areas of dense vegetation where the ground is not detected. Unfortunately there is nothing you can do about this other than to re-capture the area using LiDAR.

The ground sample distance (GSD) is different to what my mission planning software predicted

As your flight plan is performed using the global navigation satellite system on board your drone, there will always be variations in precision and accuracy between captures. This is due to the strength or availability of satellite signal at any one time. Further, Due to the nature of the data processing workflow, small positioning differences can be introduced from dataset to dataset. The best way to obtain accurate positioning is to use a real time kinematic (RTK) enabled drone, coupled with ground control points. Talk to us if you would like to include GCPs in your processing workflow!

The ground sample distance (GSD) varies between two datasets that were captured with identical parameters

As your flight plan is performed using the global navigation satellite system on board your drone, there will always be variations in precision and accuracy between captures. This is due to the strength or availability of satellite signal at any one time. Further, Due to the nature of the data processing workflow, small positioning differences can be introduced from dataset to dataset. The best way to obtain accurate positioning is to use a real time kinematic (RTK) enabled drone, coupled with ground control points. Talk to us if you would like to include GCPs in your processing workflow!


Want to learn more? Check out our range of training options to help you take your drone mapping skills to the next level!

Did this answer your question?