Richard House Posted January 12, 2016 Report Share Posted January 12, 2016 I've been following CHI's Diego Rivera Mural project and found myself in a situation trying to photograph a wall drawing 37'+ feet long. I captured images for Photogrammetry while I was at it and am wondering if anyone has experience with the upper end of Orthophoto exporting at a much higher image resolution. Do you just enter your desired pixel width when creating the Texture? That dialog box seems to match the output file I am getting and I'm going to try to push that up to a much higher image resolution if I can. Thanks for any tips. Rich House Senior Photographer Yale University Art Gallery richard.house@yale.edu artgallery.yale.edu Link to comment Share on other sites More sharing options...
Taylor Posted January 13, 2016 Report Share Posted January 13, 2016 Hi Richard, It sounds like a really interesting project. I have a mural project on-going as well and hope we can exchange information. If you haven't already, I'd recommend you update to the latest version of Photoscan (currently v. 1.2.1). It seems to produce orthomosaics with fewer artifacts and better resolution, based on the results I've seen so far. It's not necessary to generate texture for the mesh in order to generate an orthophoto. However, the resolution of the orthomosaic is partly dependent on the quality of the mesh. The size of the pixels in the "Export Orthomosaic" dialogue box is set by default to the highest possible resolution, so I wouldn't change these values. If the image is too large, it might help to break the orthomosaic into blocks, which you can choose in the Export Orthophoto dialogue box. First, I'd generate the highest quality point cloud that your computer's memory is capable of using to generate a mesh (it's the step most demanding of RAM). It's better to start with the most points possible in the dense point cloud and then decimate the mesh if necessary, rather than to start with a less dense point cloud and generate the mesh at the highest possible settings. After you've generated the mesh using the highest quality settings your computer's memory will allow, I usually use Tools->Mesh->View Mesh Statistics->Fix Topology to repair defects in the mesh. If there are holes in the mesh or occlusions, you can also use Tools->Mesh->Close Holes. Then you can export the orthomosaic using the best quality settings. I've found the tutorial on generating orthophotos is a useful reference, also. Best of luck, Taylor Link to comment Share on other sites More sharing options...
Richard House Posted January 13, 2016 Author Report Share Posted January 13, 2016 "After you've generated the mesh using the highest quality settings your computer's memory will allow" Hi Taylor, Thanks for the link and your thoughts about this. I am still struggling with building meshes in that I often get the message that my computer has run out of system memory and Photoscan is paused. I don't have other programs running and it's a beefy mac pro tower I'm working with. SO frustrating to try to figure out just how high I can go before the system fails. I'll leave it for an hour with everything looking like it's going well only to return to find the system stalled out. Photoscan won't ever continue once it's at that point and I need to force quit photoscan in order to start again with lower settings for the mesh. I'll let you know how this works... I'm currently running it again now. Rich Link to comment Share on other sites More sharing options...
Taylor Posted January 13, 2016 Report Share Posted January 13, 2016 Hi Rich, It's a very common problem running out of memory while processing meshes in Photoscan. Once the program freezes due to lack of memory, you really can't proceed with it. FWIW, I usually monitor the memory using the Activity Monitor provided as a utility in OS X. If the memory demand moves into the yellow and red zones, I generally quit and try generating the dense point cloud at the next-lower quality setting. Occasionally, I can squeak by with a few excursions into the red, but usually this means the dense point cloud is just too big for the available memory and I go back and generate a less-dense point cloud. For a given size of point cloud, the mesh generation step takes as much memory no matter what settings you use in the Generate Mesh step. For example, if you select medium quality, Photoscan generates the mesh at the highest possible number of polygons, then decimates the mesh to the size you select, so it's just as memory-intensive. Another option you could consider is to make copies of the point cloud as new chunks and reduce the size of the bounding box to overlapping areas of the point cloud. You can then process the mesh for each chunk at higher settings, produce orthophotos, and then stitch the orthophotos together in Photoshop or other software. This alternative is now easier with the new file format (PSX) in v. 1.2.1, since it doesn't dramatically increase the size of the file like the older (PSZ) format. My 2012 Mac 12-core currently has 32Gb of RAM and I can sometimes generate a mesh at the highest quality settings (ultra-high point cloud and high-polygon mesh) for about 200 images that are ~18 Mb TIFFs. I'm upgrading it now to 128 Gb of RAM and an Nvidia GTX 980Ti GPU modified for Mac. It will be interesting to see how much this helps with the mesh generation and I'll let you know when it's running. I could try processing a mesh for you if it would help. Best, Taylor Link to comment Share on other sites More sharing options...
ozbigben Posted January 15, 2016 Report Share Posted January 15, 2016 When I butt up against the RAM limit for generating a mesh in Photoscan I export the dense point cloud and do the surface reconstruction (Poisson) in CloudCompare. You can do it in Meshlab as well but I find CloudCompare is more robust when handling large point clouds. It takes a bit of juggling of settings to get the best mesh for the available RAM but it saves generating a smaller point cloud. CloudCompare also has a nice feature where you can set the vertex density to a scalar field and then use that to filter out low quality polygons. Then import it the model back into Photoscan to generate the texture. For really large textures I export tiled image sets and recombine them later in GlobalMapper (but then I have GM because I'm a map nerd). GIS applications are good at handling huge amounts of data with relatively low memory requirements and are quite useful for converting to other formats. Link to comment Share on other sites More sharing options...
Recommended Posts
Archived
This topic is now archived and is closed to further replies.