Classification of Ground points

[Herman]

Hi

Have experienced problems with the classification in the sense that when building the mesh and selecting only the ground points to be used , the resulting DEM would still include buildings & trees.

The automatic  classification generates "Unclassified"  , "Ground" & "Low Point (noise)" with the ground showing brown and all other (buildings and trees)  showing in white. Using this classification and building the mesh it seems the buildings and trees are not removed and still included in the final DEM
When manual classification is done and building & trees are defined as such they will change from being shown in white to the resulting Red & green and the resulting DEM from the mesh would would have these building and trees excluded.

Also when writing a LAS file and selecting only the ground points to be exported the resulting DEM generated from the LAS still include buildings & trees

Regards
Herman

We are using ver 1.1.6 64 bit , previously the automatic classification successfully managed to remove all buildings & trees.

[Diego]

People still do not understand that this is photogrammetry.Classification aid will never be 100% clean, it depends on the type of terrain, in some cases work well and others do not believe the problem is that this makes magic. For precision work where you need the bare ground. You need to use stereoscopic observation photogrammetric stations and specialized personnel, ie photogrammetrists, to complement the initial classification of software, even with lidar can expect a 100% successful classification. I think many people who are producing maps with photographs they forget that this is still photogrammetry, and I recommend some training to get the best results, not only depends on a program, get trained, and often manual work intense to achieve quality results.

[Herman]

Hello Diego

Thank you for the reply , maybe I did not explain too well.

 I appreciated that one cannot expect a 100% clean result, but what we find is that it seems that the classification  seems to be 100% ignored when the DEM is created form a surface where only the ground classification is selected.

The level too which Agisoft manages to classify is amazing but you will allays need some human intervention , the only problem is that its seems that Agisoft was completely ignoring the classification.

However the problem was found , with some good advice from Dmitry , the classification parameters was just increased.
The initial problem was that the classification seems to work perfectly when visibly observed , once the DEM is tested their is still 2m height differences  present  which  seems to be removed when the visible classification is inspected.

The increased parameter settings removes some other detail , this will be manually re-classified - basically just underwriting the comments form Diego.

Diego thanks for your comments and once again thanks a lot to the Agisoft team for a great product and amazing support

Regards
Herman