[Pix4DMapper] [Desktop] confirming accuracy of contours to mapping standards

I am having trouble drawing the connection between the quality report and the accuracy of the contours generated from the DTM. The quality report presents a number of RMSE values and an indication of how well the processing steps went, but I feel that there is a disconnect when it comes to the DTM.

 

I understand the DTM is generated from the classified point cloud using “ground” and “road” points. An accurate DTM requires the classification to be accurate, whether done automatically or with user input. As there is no quality report for the DTM, and noise in the DTM is most evident when viewing the classified point cloud, how do I determine how accurate the contours are compared with reality without doing a ground survey and comparing the two? Mapping standards require 95% of points to be within 1/2 of the contour interval (so 6 inches for a 1ft contour map). For reference, we shot over 50 GCPs so we have quite a lot of data to use. I have read that including too many GCPs is not good, so a number of them could be used as checkpoints. Also, the area that was surveyed is approximately 2,000 ft x 1,500 ft with about 1,000 ft of relief. It is a large hillside. 

 

I tried searching the forums and google for answers but cannot find anything other than what the mapping standards require for accuracy. I haven’t found guidance on how to confirm contour accuracy within those standards, other than reference to the quality report (GSD, RMSE).

 

There are two types of accuracy in that case: the accuracy of the point classification and the accuracy of the points themselves.

  1. I would recommend to make a manual check and reclassify points if necessary with the point cloud editing tool. In this sense, the point cloud classification enables you to do the classification faster, but you still need to check the results. There is no way to tell automatically that 95% of the points were correctly classified, because you would need the ground truth of that project for that (which is the one created by manually assigning points).

2. The points that are classified as ground and road are not changed by the DTM, so if the points are accurate in the dense point cloud and DSM, they should also be accurate in the DTM. In your case, I would probably use the checkpoints to verify the accuracy of the DSM or the dense point cloud.