Sometimes projects that you process with ArcGIS Drone2Map will not have perfect results. Many factors can cause poor-quality reconstruction. In the past, the solution to many of those issues was to fly the same project area again. With improvements to how products are generated in Drone2Map, there are now multiple methods to improve the quality of your project outputs. These options apply to both 2D and 3D output products and correct areas of poor reconstruction. Additional options have been added that provide more control over how mosaic datasets are displayed in the program. The combination of these methods and options makes it more likely for you to achieve accurate results without the need for postprocessing software.
The Adjustment step in Drone2Map is when the software attempts to determine the correction position and orientation in 3D for each image in a series of aerial images so they can be compiled and mapped to the surface of the earth. Without a well-matched adjustment, your products will not be as accurately mapped to the earth's surface or mosaicked into clean products. Drone2Map allows you to control the input settings used for this adjustment to mitigate common issues.
Tie point options have been updated to increase control over the matching neighborhood used to generate tie points between images. This can be useful when you know the flight may not have had consistent overlap or for free flight projects. A refine adjustment option has also been added that can be turned on to run an extra pass on tie point generation at an increased image size. This typically results in many more tie points being generated even on lower image scale templates such as the rapid template. After an initial adjustment is run on the project, images are classified as either enabled or uncalibrated. If you receive a high number of uncalibrated images, you can adjust the tie point settings to larger neighborhoods and image scale to potentially obtain enough matches for the uncalibrated images.
Control over camera calibration options has also been expanded in ArcGIS Drone2Map. All aspects of the camera parameters are now visible and can be edited manually for custom camera settings. You can also make adjustments to the automatically calculated values provided by Drone2Map when you run initial processing. The changes made to the edit camera pane have also simplified access to principal point and distortion coefficient settings. If you receive poor-quality output and have a flight that was collected using general best practices for the subject matter, the first troubleshooting step is to ensure that the camera model information provided to Drone2Map is correct.
2D product reprocessing
The options for refining 2D products depend on your license level.
New settings have been added to the orthomosaic processing options that allow you to define the elevation source that is used for orthorectifying the imagery. The Orthorectification Method drop-down menu allows you to choose whether to rectify an orthomosaic using solution points or nothing. Choosing None for the orthorectification method can be beneficial when you are working with flights that do not require high accuracy in an elevation source. If you are processing thermal imagery or an agricultural area that does not contain much elevation variance, using the None option is faster than using the sparse point cloud option. When the sparse point cloud option is set as the orthorectification method, additional steps are required, a point cloud is generated, and an elevation surface is interpolated from it. The sparse point cloud is the best option for a project area that has large structures or significant changes in elevation.
In ArcGIS Drone2Map, orthomosaics are generated as tiles and then mosaicked together. The existence of these tiles allows you to reprocess portions of the imagery rather than the entire product. This saves time, and you can iterate corrections on the output orthomosaic to achieve the best results. When you draw a correction feature or waterbody mask over an existing orthomosaic, only the tiles that contain those features are reprocessed. This significantly reduces the time it takes to generate a new orthomosaic and refine features in the project area.
There are two methods of applying the tile-based processing workflow. The first method allows you to reprocess tiles but maintains the original tile in a mosaic dataset. This means you can select which tile you want to keep on a tile-by-tile basis. The second method automatically overwrites the tiles that have correction features or waterbody mask features and writes a new orthomosaic. The original orthomosaic is maintained in the project folder as a backup. To learn more about tile-based processing, see Correction features.
3D product reprocessing
If you are using ArcGIS Drone2Map with an advanced license, you can also reprocess 3D products. Correction features can be drawn over the 3D products to modify the definition of features. Similar to using correction features on 2D products, each vertex receives its z-value from the underlying elevation surface. The z-value can be modified when drawing the features in 3D. For more information, see Specify an elevation for 3D features.
Any processing option changes to the point cloud density require the dense processing step to be rerun. If you are creating a 3D mesh after you have generated a 3D point cloud, the dense processing step will be skipped, and only the 3D steps will run. If you want to save time and will be continually adjusting 3D mesh features with correction features, it is recommended that you only process the 3D point cloud when necessary. Alternatively, processing the point cloud on Ultra quality means that subsequent processing runs do not need to redo the dense processing step and can save time beyond the initial processing run. Low and medium point clouds are also typically sufficient for modeling areas that do not contain a wide variety in elevation or features, such as a field of crops. Consider the subject matter of the project when determining the level of point cloud density, since it is a resource- and time-intensive product to generate.