PhotoScan_Analysis

With PhotoScan_Analysis you can inspect the interior accuracy of a Agisoft PhotoScan project.
It can do three things to help you to improve your results with PhotoScan

• Calculate the standard deviation the reprojection error for each image and the Root Mean Square (RMS) over all images.
• It visualize distribution of feature points over all image
• calculate and export error ellipsoid in a 3D plot to evaluate the camera configuration

Report

 Camera Name   Projections  SIGMA x  SIGMA y  SIGMA P    MAX x    MAX y
IMG_4287.JPG           297  0.32126  0.35394  0.47799  1.26451  1.10858
IMG_4288.JPG           231  0.36953  0.39133  0.53823  1.23350  1.23678
.......
IMG_4335.JPG           612  0.19908  0.28742  0.34963  1.08133  1.27099
IMG_4336.JPG           367  0.27735  0.35559  0.45096  1.09968  1.40987

                      RMS:  0.39662  0.52912

• Projections gives you the count of 3D feature points for the image
• SIGMA x,y (standard deviation of the feature points)
• SIGMA p ( sqrt(SIGMA x^2 + SIGMA y^2) )
• RMS (Root Mean Square of the standard deviation over all images)

Visualization of the image measurements [top left] image measurements (feature point) with error vector over all images
[bottom left] image measurements (feature point) with error vector for one image
[top right] overview of image measurements in a grid. The center point of a cell shows the error vector of all measurements per cell. the color scale indicates the number of measurements.
[bottom right] same as top right but the color scale indicates the average number of ray intersection per cell.

Error ellipsoids of the object points Image of error ellipsoids which can be useful for evaluate the camera configuration

How To

To use the analysis.py script with PhotoScan make sure the the following files are located in the same folder:

• sp_exp_for_test.stl
• sphere_aus_mashlab.stl
• STL_Writer.py
• svd.py

To run the analysis.py file in PhotoScan via Tools->Run Script.

You can use the following command line arguments:

PhotoScan Analysis v0.4.1

HowTo:
Command Line Arguments:
-report_out [filename]    Creates a report file. Options: filename (default: report)
-svgout [filename]    Creates a SVG-Image with image-measurements Option: filename (default: image_measurements
-svgfactor [factor]       Magnification factor of the error-vector for the SVG-File (default: 40)
-svgcols [columns]    The number of columns used to generate the overview image (default: 20)
-stlout [filename]    Create a STL-Mesh with Point-Error-Ellipsoids. Option: filename (default: stl_export)
-stlfactor [factor]       Magnification factor of the ellipsoid-axis (default: 100)
-export_ellipsoids [filename]      Export a ellipsoid file


Sample:
-rout reportname -svgout svgname -svgfactor 12 -svgcols 10 -stlout stlname -stlfactor 12

GUI
You can also use the GUI by choosing the argument '-useGUI'

Result

Depending on the your command line options the results are several files:

• .txt Report
• .svg Visualization of the image measurements (open with a web browser)
• .stl error ellipsoids (open with mashlab)
• .ell error ellipsoids parameter

Changelog

v0.2 [added] export of the ellipsoid parameter (.ell)
v0.3 [added] overview with average number of ray intersection per cell
v0.4 [added] transformation from chunk to world coordinates
v0.4.1 [update] photoscan 1.2 only support multiplication with a vector

Used Source Code

• svd.py GNU-License (http://stitchpanorama.sourceforge.net/Python/svd.py)
• STL_Writer.py MIT-License (https://github.com/jv4779/spin_mirror/blob/master/STL_Writer.py)
• pysvg BSD-License (http://codeboje.de/pysvg/)