This post has NOT been accepted by the mailing list yet.
Hi, I am Martina Porfiri, an environmental engineering, and at the moment I am PhD student at Geodesy and Geomatics Division, University of Rome “Sapienza”.
I am thinking about an idea to take part in the incoming GSoC 2013. Looking at the proposed Opticks thematic I am oriented to the project “accuracy assessment tools”. I will share with you the following idea concept. I would like to create an Opticks plugin for the “Accuracy Assessment” able to characterize not only the classified images but also the Digital Surface Models. In this way it is possible to compute crossed statistics taking into account both land cover and terrain morphology.
The main plugin features could be summarized as follows:
1. CLASSIFICATION ACCURACY ASSESSMENT: a tool that allows to compute the Error Matrix and its statistical parameters (as Overall accuracy, Variance, KHAT, etc). The main steps could be:
• input data: known reference data (ground truth) of the AOI/ROI and corresponding results of an automated classification previously produced (for example by the already developed classification algorithms implemented during the previous GSoC into Opticks plugin)
• main functions: comparison, on a category-by-category basis, of the relationship between the input data and computation of the parameters of interest, using the expressions known from literature
• output data: visualize the resulting validation and print an ASCII file that contains all processing information and results
2. DSM CLASSIFICATION ASSESSMENT: a tool that allows to compute the height discrepancies (Δz) between two Digital Surface Models (DSMs) and to perform the subsequent statistical analysis according to the classification. The main steps could be:
• input data: known reference DSM, corresponding analysed DSM (or also point cloud) and classified data (vector or raster)
• main functions: computation of the Δz between two DSMs, determining the reference points corresponding to the file for analysis by several interpolation methods (e.g. nearest neighbour, linear, bilinear) and results analysis considering the classification data
• output data: a file ASCII that contains the height discrepancies value and several plots that shows the Δz trends with respect to the land cover
I am not proficient in Opticks developing, but looking at the Opticks core and its C++ APIs I think that this plugin could be conveniently implemented.
I apologise for the delay in my proposal and I look forward to have all your comments and suggestions.
Thank you in advance for your availability.
o Lillesand T.M., Kiefer R.W., Chipman J.W, Remote sensing and image interpretation
o Foody G.M, Status of land cover classification accuracy assessment
o Passini R., Jacobsen K., Accuracy analysis of SRTM height models