Does changing the keypoint image scale from 1 to 1/2 effect the final orthomosiac resolution?
Yes the expectation would be if you decrease the image scale the final mosaic resolution would be lower. You should not get the same resolution from the image scale 2 and image scale 1/8.
When changing the keypoints images scale from 1/2 to 1, more keypoints will be extracted as the software will have a larger search area. This will have a positive impact on the accuracy of the point cloud and consequently on the accuracy of all the results.
Re.reply from Tech Support Engineer: I understand that 1/2 to 1 will have a positive impact on the accuracy of the point cloud and all results. But from your answer it is unclear whether going from 1/2 to 1 will also affect the final orthomosaic resolution? If so, how will the resolution change?
(Put another way, I’m able to create Ortho in step 3. without creating a point cloud in step 2. , so I didn’t know the two were linked.)
Changing the image scale from 1/2 to 1 will also impact the level of detail, number of keypoints and the resolution in your othomosaic. You would have more detail with image scale 1 in your orthomosaic compare to image scale. 0.5. You are able generate an orthomosaic even if you don’t generate dense cloud. However the level of detail for your othomosaic for not including the dense cloud wouldn’t be the same if you were to do image scale 2, and the highest dense cloud settings
Hi Nicholas and Selim,
Having less keypoints generated will probably affect the calibration result (depending always on the quality of the images). This might affect the average GSD that is calculated during step 1, but the change is not expected to be significant.
If the resolution of the orthomosaic is 1xGSD, the pixel size will be the same but the level of details will be changing.
Indeed, as Selim mentioned it is possible to generate the orthomosaic without generating the point cloud but then the level of details will be lower and the orthomosaic will be less accurate as the DSM will be more interpolated.