Thanks for sharing Christoph, It’s an interesting paper and I will read it more carefully. Another approach could be to make a point cloud with the drone images and to work with a high resolution DEMs. Depending in the drone camera and flight altitude, you could have a really good cm level DEM. That could help with both semi automatic classification methods, and deep learning. It could make it much easier also to identify cracks visually by building, for example, a shade model.
The time spent doing this automatically is likely going to be bigger than the time spent doing it manually. It depends on the scale and goal of the project of course. You could always use the KISS method (Keep it simple stupid!) and just spray paint the cracks before flighting the drone. Depends on what you want to achieve. Cheers! Nicolas > Le 14 déc. 2019 à 13:57, Christoph Jung <[email protected]> a écrit : > > > Hello Doug, > > Automatic crack detection is not an easy problem. A semi automatic > classification will not find all cracks, because cracks can have very > different appearances. A state of the art approach would be deep learning. At > the German university TU Ilmenau they developed a robust neuronal network for > the detection and classification of distresses in road surfaces. Here is a > Link to one of the papers of this project: > > https://www.tu-ilmenau.de/fileadmin/media/neurob/publications/conferences_nat/2017/Eisenbach-ATIM-2017.pdf > > Sincerely, > Christoph > >>> Am 14.12.2019 um 07:05 schrieb Doug <[email protected]>: >>> >> >> I have imported a set of drone pictures to QGIS. They show the streets in my >> subdivision. I want to “record” the cracks in the pavement to be able to >> compare them over time. I want to be able to segment the streets in >> arbitrary lengths. For each segment, I want to “draw” the cracks (in layers) >> to record where they are. Every year or two I will repeat the process. The >> purpose is to be able to compare the set of cracks in successive snapshots >> to see how the asphalt is holding up and identify underlying problems. I >> would love to have some software to find the cracks for me and generate the >> crack layer, if anyone knows of such software. >> >> First problem: I can trace the outline of the streets to generate a layer >> which represents to road surface. Is there a way to slice that layer by >> picking two points to “slice” the layer and then have two layers? And then >> repeat the process so that I get the segments that I want? Alternatively, is >> there a way to create a segment and “snap” to next segment so there is not >> overlap or missing space? >> >> Second problem. Given that I have a layer representing a segment of the road >> surface, can I “draw” the cracks for that segment such that every line in >> the cracks layer is related (as in SQL) to the segment it reside in? >> >> Once I have the road surface segments, and the associated cracks, I can then >> use SQL to generate all sorts of reports on the cracks: average and median >> length, number of horizontal and vertical cracks, total crack length per >> road surface segment (or square feet), etc. >> >> Thanks. >> Doug >> >> >> _______________________________________________ >> Qgis-user mailing list >> [email protected] >> List info: https://lists.osgeo.org/mailman/listinfo/qgis-user >> Unsubscribe: https://lists.osgeo.org/mailman/listinfo/qgis-user > _______________________________________________ > Qgis-user mailing list > [email protected] > List info: https://lists.osgeo.org/mailman/listinfo/qgis-user > Unsubscribe: https://lists.osgeo.org/mailman/listinfo/qgis-user
_______________________________________________ Qgis-user mailing list [email protected] List info: https://lists.osgeo.org/mailman/listinfo/qgis-user Unsubscribe: https://lists.osgeo.org/mailman/listinfo/qgis-user
