Agisoft PhotoScan Tutorial: Aerial Data Processing

hello everyone this is a video tutorial on fall test camp program version 0.9 I'm going to show you how to process aerial images with photos camp professional edition to generate an auto photo and the digital elevation model in this tutorial we'll run through the basic steps the first thing you need to do is to adjust the program window for your project and also sure that I won't use scale bars in my project so I can close the scale bar section on the ground control pain if you want to open or close any pain you go to the View menu and click the corresponding pane for example I can open console pane if I'm going to overuse log information during the processing in tutorial we won't use it so I'll close it again one more preparation that should be done is checking of the processing settings you can use the program for the first time on the machine open tools menu and click folder scan preferences comment on general tab set the path the photos can lock file and check the depth filter and parameter set to aggressive value switch to the open CL tab and make sure that all the open cell device is available I checked folders can uses GPUs acceleration which speed ups processing significantly click ok now let's start our project the first thing you need to do is to add photos to do this switch to the workspace pane and click to on the add photos toolbar button browse to the folder containing the images you are going to process here is my dataset prepared for the tutorial select all the photos and click open on the workspace pane we see that a chunk has been created and it contains 65 cameras that is 65 photos the photos themselves can be seen on the photos pane now you need to add camera positions data sometimes we say add ground control data so you switch to the ground control pane and click import button choose specially prepared text file contingent camera positions data and click open the data should have been prepared in character separated values format my file contains X&Y coordinate and the height for each camera position you can also add columns for pitch yaw and roll values and import them into the program however in the current version the orientation data is not used during processing so it wouldn't influence the results check that the delimiter and the numbers of the columns are indicated correctly and click OK ultimately you can load the camera positions data from the exit file if available for this purpose there is import exit button on the toolbar you can see the data being loaded to the ground control pane don't forget to indicate the coordinate system click Settings button choose the corresponding coordinate system from the list and click OK the blue dots you can see in the model view represent camera positions you can check whether they correspond to the flight route now we can start the processing first you need to align photos open workflow menu and click align photos comment choose high accuracy for the best results and ground control paper selection to speed up the process of matching of the feature points detected on the photos folder scan will try to match features only on those photos which overlap according to the ground control data we have loaded before it will take certain time to do the alignment for my data set around 15 minutes if you just go in through the process and steps and don't care much about the accuracy of the results you can set the low value for the accuracy parameter this setting could be chosen for processing of a data set right in the field just to judge whether the data collected could be processed successfully finally alignment is finished camera positions were refined and there is the minute orientation can be seen in the mode of view to see the point cloud better you can switch off their blue rectangles representing cameras click the show cameras button on the toolbar photo alignment has been performed using only image data the data set I have prepared for the tutorial is quite a good one however while processing your own data you may encounter some nonlinear deformations of the model for example so-called Powell effect it will be visible on the point cloud already to remove some nonlinear deformations and improve the alignment results photoscan allows to implement ground control points with known reference coordinate if you have these data for your project you can run optimization procedure let's see how you should do it first of all you need to indicate ground control points on the photos we call this procedure marker placement to avoid time-consuming job of manual indicating of GCPs on every photo where they are visible you can use guided marker placement approach it means that photo scan will suggest where you should place a marker to be able to use programs suggestions you need to build geometry first go to the workflow menu and choose build geometry comment select point cloud geometry type and say set based count parameter to two hundred thousand polygons this is relatively low value but it is enough for the purpose it takes just several seconds and you can see the model are constructed in the purest quality to place a marker scroll to the photo where a gcp is visible double click on the photo zoom in to see the marker well then switch to the Edit Marcus mode right click on the projection of the GCP and choose create marker comment from the context menu and you market has been added onto the ground control pain selected filter folders by the marker now in the photos pane you can see only those photos with where the same GCP is visible open one of them folder scan has already played the marker for you you need to accept the suggested location or refine it if necessary drag in the marker with the mouse repeat this for every photo on the list you can see where the marker is placed on the model switch into the model view now right-click on the mark on the ground control pane and choose the rename comment from the context menu name the mark according to the disappear label now repeat the described procedure for every GCP to achieve good optimization results you should distribute around 10 G's apiece evenly around the area to be reconstructed in tutorial project we will use only 5 G's apiece now when all the markets have been placed you need to load GCPs or Markus coordinate the procedure the procedure is the same as followed in camera data click import button on the ground control plane toolbar browse to the folder where the prepare text file is selected and click open check that the limit indicated and the columns numbers and click OK you can see the loaded data on the ground control pane note that the markers will be automatically checked again don't forget to set the correct coordinate system for the GPS data in my project it differs from the coordinate system indicated for camera data so open settings dialog select the respective coordinate system from the list and click OK now you are ready to run optimization procedure first of all you need to uncheck all the cameras since the optimization will be based only on markers data if you have enough markers you can uncheck some of them to exclude from the optimization later you can use them as control points to evaluate optimization results I will use all 5 markers for optimization now open ground control settings dialog once again and set the values for the measurement accuracy parameters since your optimization is based on marker code in its data only marker occurs the parameter matters generally it is recommended to set zero value since normally disappea coordinates are measured with significantly higher compared to the GPS that is camera data click ok click optimize toolbar button and ok button again when optimization is completed you can evaluate the accuracy of gel referencing of gel referencing judging from the errors values for contour points if you have any left now you can move on to the second process and step build geometry first of all you need to adjust bounding box it indicates the region which will be reconstructed in this project in fact we don't need to adjust bounding box it was automatically set perfectly well in general the box can be resized and rotated note that red calcite indicates the plane to be treated at ground plane so it should be set under the model now you are ready to run build geometry step open workflow menu and choose build geometry comment for aerial imagery should always choose height field object type geometry type set a smooth target quality was set to medium value high values can be preferred in the projects where super detailed geometry is required yet one should be careful when choosing high geometry quality value since such operation will be very demanding in memory and processing power sources phase count value should be increased up to several millions click okay 3d model reconstruction is a computationally intensive operation it can take quite a long time so you'll have to wait for awhile to see the results finally the model has been reconstructed and you can observe it in various modes wireframe solid and shade it for more realistic and sharp visualization of the model you can generate text and apply it to the model to build texture go to the workflow menu and choose build texture comment select auto photo mapping mode and mosaic blending mode generally 4096 pixel value is enough for texture Atlas width and height but for more detail texture each value could be set to 8,192 pixels click OK once the texture is built you can see the model in text view the model itself can be exported but this tutorial our goal is to export also photo and them so how shall we do it go to the file menu select export also photo comment if you model is your references like ours in this tutorial you can export also photo in geographic projection make sure that the relevant coordinate system is selected set mosaic value for the blending mode and check the fill holes option to generate continuous auto photo leave the automatically suggested value for the pixel size split and blocks option can be helpful to export very large also photos if you want to export just a part of the reconstructed error you can indicate the boundaries of this part in the region section in our project we will export also photo of the whole scene so we click export button now we need to specify target file name and type and click Save button to generate them selected expert dem comment from the file menu again we select geographic projection and check the indicated coordinate system select crop and valid am option and leave default settings for the no data value here there are also splittin blocks and indicate the region options that can be used for the same purpose as discussed for also photo export so we click export button specify target file name and type and click Save button we can open the exported files with any viewer and expect the results here is exported also photo and here is the dam finally you can generate processing report that present summary on the project including process and results and accuracy estimate to generate the report go to the file menu select generate report comment type in the file name and click OK here is for the scan processing report on our project we can see a figure showing camera locations some summary information on input data information on gel reference and accuracy is here and finally digital elevation model that we have already seen this is the end of our tutorial illustrating the basic steps of aerial image the process and workflow in edge soft photo scan good luck in your future projects

Channel: 80lv

Views: 124,834

Rating: 4.8862557 out of 5

Keywords: PhotoScan, Tutorial (Media Genre), Software (Industry), 80.lv, Agisoft

Id: doXm3pThz3s

Length: 15min 31sec (931 seconds)

Published: Sat Aug 29 2015