IMAGINE Discussions

Discuss and share topics of interest using ERDAS IMAGINE the world’s leading geospatial data authoring system.
Showing results for 
Search instead for 
Do you mean 
Occasional Contributor
Posts: 6
Registered: ‎10-30-2017


Hello everybody!

I extracted DSM on quarry with SGM method. WV-3 was used. Sgm settings:  min Z - (-29), max Z - 284, RPC resid threshold - 0,1, Y parallax threshold - 0,2. I think, that result is not good, because SGM created point cloud with low density on the bottom of quarry. Thera are a lot of holes and peakes, wrong elevation meanings. I attached the printscreens. Unfortunately, I closed erdas before taking session log. What can you advice to improve the result?


Best regards, Aruana

Frequent Contributor
Posts: 153
Registered: ‎10-27-2015

Re: sgm

First make sure that orientation is as good as it can be. It sounds a bit that you have now used only RPC and no ground control. Use some GCP's and improve the orientation


When orientation is as good as it can be then starting tuning up parameters for point density etc. I have no experience on those what is the most effective parameter but in principle the orientation improvement is key factor and after it is done as well as it can be you can proceed to other details.




Occasional Contributor
Posts: 6
Registered: ‎10-30-2017

Re: sgm

I used 1 GCP. After DSM creation, I did point cloud classification and filtering. The result is good on surface of the quarry. But quarry bottom has low density points, a lot of peakes and false elevation values. I think it is because of too dark histogam of stereo pair. Maybe you can give advice how to improve histogam of stereo pair and to increase the information content of images? I did rescale of each images, but when I replace the original images, connection with RPC loses.



Regular Contributor
Posts: 262
Registered: ‎11-16-2015

Re: sgm

I've had exactly the same issues that you are experiencing. I also tried enhancing and rescaling the data - you need to create a new block file with this data in order to pick up the RPCs. However, even with that I got poor matching in dark, shadowy areas. Tridicom SGM creates the epipolar pair as 8bit data; unitl as and when or if it changes to use 16 bit, I think we're stuck with noisy results with a low point density. Cheers, Johnnie.