Aerial archaeology on Duvansko Polje (Bosnia and Hercegovina). CONPRA project report, 2014.
Descripción
Offficiaal Reeporrt Project: Aerial Archaeologyy in Duvvansko o Polje na) (Bossnia and Herzegovin Joze ef Chajbullin Koštial, TerraaVerita, Ltd. February - March 2014
Contents Introduction .............................................................................................................................................. 3 Imagery dataset ........................................................................................................................................ 6 Method and software solution ................................................................................................................. 7 Resection photogrammetry .................................................................................................................. 7 Matlab environment ............................................................................................................................. 7 Space resection analysis ......................................................................................................................... 10 Results ..................................................................................................................................................... 12 References .............................................................................................................................................. 16
uction Introdu Followingg document is an experttise report o of my seconddment at Lju ubljana Univversity within the frames o of CONPRA p project on FFebruary/Maarch 2014. U University’s ffield of expeertise within n this project iss mostly aerrial archaeollogy and rem mote sensin g and for th hat reason I have the un nique opportun nity to partiicipate on the academic project off aerial imaggery capturing, analysiss and interprettation in qu uite specific surroundinggs of Duvannsko Polje ((Duvno Field d) in Bosniaa and Herzegovvina (Fig. 01)).
Fig. 01. Arrea of interesst in southeasstern Bosnia a and Herzegovvina – Duvno B Basin.
The fligh ht project (practical ( te errain part) was led byy Prof. Darrja Grosman n from Dep pt. of Archaeollogy of Ljubljana Univerrsity on the 18th and 199th of March 2014. Students from both Slovenia and Croatiaa universitiess were also included (Fiig. 02, 03). A Although pro oject`s main n goal was focu used on basiic archaeolo ogical interpretation of aaerial imageery (standarrd crop, moisture and relief feature reccognition; Riiley 2009) we e have takenn a decision to extend th he possibilities of remote sensing s and d to prove the t eventuaality of terraain generatiion and mo odeling baseed on simple ae erial photos acquired du uring this cam mpaign.
Fig. 02. Ha appy Cessna fflyers before start.
Fig. 03. Prrof. Grosman during the in nstruction of ccamera handlling.
After succcessful terraain campaiggn the conse equent cabinnet analysis was proceeded. It shou uld be stated th hat cabinet w work was a kind of “blind attempt””, or “a jum mp into darkness” – we were facing the question iif there is po ossibility of generating ccomplete neew informattion from daataset that was not primariily designate ed for such u use. We hadd to choose tthe approprriate method d and e ability of im magery dataaset for satissfactory acceeptable resu ults. prove the The main goal of presented report is the extractioon of geom metric featu ures from aaerial photograaphy and mo odeling the tterrain situaation accord ing further G GIS needs. R Resulting pro oduct should be b more or less precise e Digital Elevvation Mod el1 based o on the simplle archaeolo ogical aerial photo.
1
http://en n.wikipedia.orgg/wiki/Digital_e elevation_mod del
Imagerry dataset Altogether 418 photos have be een collecte ed during thhe flight cam mpaign. Listt of all available photos (in thumbnail size) is dem monstrated in Attachme nt 01 of thiss report. e exclusivelyy due to ffact that C Cessna aero oplane used d for All phottos are of skew type photograammetry was w not equ uipped with the facillities for vvertical aerial photograaphy. Neverthe eless the only obligatoryy factors ne eeded for suuccessful pho otogrammettry processing of acquired photos werre:
Sufficient spaatial overlapping of imagges (at least 1/3 overlapping of the ttaken situatiion)
era parametter inner oriientation (id dentical cam mera lens usee and Preservation of the came sttable focal le ength preserrved)
Present orie entation maarkers on captured ssituation (ee.g. buildings with kn nown co oordinates)
Whole area of Duvansko Polje w was covered d by aerial im magery with h respective overhang w within its geographical fram me. Spatial de elimitation o of captured situation is tto be seen o on Fig. 04.
Fiig. 04. Spatiall delimitation of aerial magery extentt. im
Method and software solution Resection photogrammetry
Regarding the nature of imagery (skew images) dataset it has been obvious that the space resection methods have to be used for our purposes. Space resection in photogrammetry includes determining the spatial position and orientation of a photo based on image measurements of control points that appear in the photography (Moffitt ‐ Mikhail 1980). Whilst space resection is a nonlinear issue, developed methods involve linearisation of the collinearity condition and the use of an iterative process to determine the final solution using the least‐squares method2. The process also requires initial approximate values of the unknown parameters, some of which must be estimated by another least‐squares solution. The core word for whole resection photogrammetry is a collinearity equation (Schenk 2005). However there is a huge theoretical discussion about distinct resection methods in photometric science for over last 40 years – the Church method (most simple and straightforward; Church 1980), Hadem and Rampal method (also known as “pyramid method”; Hadem 1981) and Abdel‐ Aziz – Karala approach (cited as Direct Linear Transformation; Shih – Faig 1987) to mention those most known and used. They differ mostly by number of computed inner and outer space parameters of real and photographed objects and camera characteristics. Finally we used Direct Linear Transformation based on 11 camera parameters (6 parameters of outer and 5 parameters of inner orientation) for each image used in analysis. Matlab environment
Final functions for computational process of image alignment and model building were written in Matlab environment (Fig. 05) for the purpose of getting general algorithmic frames (E.g. Tab. 01, 02 and 03.).
2
http://en.wikipedia.org/wiki/Least_squares
Fig. 05. W Working enviro onment of Ma atlab 2009. Collinearity ge eneral functio on in Matlab.. Tab. 01. C function n [x,y]=co olinearity y_radial(c, ,xo,yo,k1,k2,Xo,Yo,Z Zo,omega,p phi,kappa,X X,Y,Z) % % %
Autho or: Archeo oConsult 2014 angle es in rad
R=R_from m_angles(o omega,phi,k kappa,0); x= xo-(c c*((R(1,1)*(X-Xo)+R( (1,2)*(Y-Y Yo)+R(1,3)* *(Z-Zo))/... (R(3,1)*( (X-Xo)+R(3, ,2)*(Y-Yo)+R(3,3)*(Z Z-Zo)))); y= yo-(c c*((R(2,1)*(X-Xo)+R( (2,2)*(Y-Y Yo)+R(2,3)* *(Z-Zo))/...
(R(3,1)*(X-Xo)+R(3,2)*(Y-Yo)+R(3,3)*(Z-Zo))));
distr2=((x-xo).^2+(y-yo).^2); distdx=(x-xo).*((k1*distr2)+(k2*(distr2).^2)); distdy=(y-yo).*((k1*distr2)+(k2*(distr2).^2)); x=x+distdx; y=y+distdy; xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
Tab. 02. Abstract function of Direct Linear Transformation written in Matlab environment.
function [P,so,VV,v,iter]=DLT(ximage,Xground,limit,maxiter) % Author: ArcheoConsult % 2014 % ximage=[x1 y1;x2 y2;...;xn yn] % Xground=[X1 Y1 Z1;...;Xn Yn Zn] for i=1:size(ximage,1) L(2*i-1,1)=ximage(i,1); L(2*i,1)=ximage(i,2); A(2*i-1,1)=Xground(i,1); A(2*i-1,2)=Xground(i,2); A(2*i-1,3)=Xground(i,3); A(2*i-1,4)=1; A(2*i-1,9)=-Xground(i,1)*ximage(i,1); A(2*i-1,10)=-Xground(i,2)*ximage(i,1); A(2*i-1,11)=-Xground(i,3)*ximage(i,1); A(2*i,5)=Xground(i,1); A(2*i,6)=Xground(i,2); A(2*i,7)=Xground(i,3); A(2*i,8)=1; A(2*i,9)=-Xground(i,1)*ximage(i,2); A(2*i,10)=-Xground(i,2)*ximage(i,2); A(2*i,11)=-Xground(i,3)*ximage(i,2); end dx=inv(A'*A)*(A'*L); xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
Tab. 03. Exporting results of resection routines. function export_resection_results(filename,im_name,control_image,exterior,err,so,v,ite r)
% Author: ArcheoConsult % October 2014
fid = fopen(filename,'wt'); fprintf(fid,'Matlab Photogrammetric Toolbox '); fprintf(fid,'\n'); fprintf(fid,'Resection Results:\n'); fprintf(fid,'\n'); fprintf(fid,'Image: %s \n',im_name); fprintf(fid,'\n'); fprintf(fid,'Exterior Orientation Parameters:\n'); fprintf(fid,'Xo = %f ± %f\n',exterior(1),err(1)); fprintf(fid,'Yo = %f ± %f \n',exterior(2),err(2)); fprintf(fid,'Zo = %f ± %f \n',exterior(3),err(3)); fprintf(fid,'omega = %f deg ± %f deg\n',exterior(4)*180/pi,err(4)*180/pi); fprintf(fid,'phi = %f deg ± %f deg\n',exterior(5)*180/pi,err(5)*180/pi); fprintf(fid,'kappa = %f deg ± %f deg\n',exterior(6)*180/pi,err(6)*180/pi); fprintf(fid,'\n'); fprintf(fid,'Iterations: %d \n',iter); fprintf(fid,'\n'); fprintf(fid,'sigma = %f \n',so); fprintf(fid,'\n'); fprintf(fid,'Image Residuals:\n'); for i=1:size(v,1)/2 fprintf(fid,'%g %f %f\n',control_image(i,1),v(2*i-1),v(2*i)); end fclose(fid); xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
Space resection analysis After this preliminary preparation and software development, further procedures were following: 1. Each couple of overlapping photos was analysed and processed via recognition routine for generating a partial geometry (Fig. 06). 2. Direct Linear Transformation scripts were applied in purpose to generate basic geometry of photographed situation. The raw result was produces in a shape of Triangular Irregular Network (TIN; Fig. 07) based on two (usually more) photos exclusively. 3. The whole process was repeatedly iterated with all photos in imagery dataset and 55 independent elevation models in TIN shape were created.
Fig. 06. Crreation of bun ndles of correelated photos in Matlab.
Fig. 07. Prrocess of Triangular Irregu ular Network ssurface generration in Mattlab.
Results Spatial and dimension independent TINs created by steps mentioned above were consequently rectified and correlated according control point targets (ground fix points) and bulked into one final geometry. This step was performed in Autodesk Map 3D environment. Resulting geometry (contoured Digital Elevation Model) is present on Fig. 08. Regarding the fact that we have basic geometry at disposal, there are no constraints for presenting it in the various visual modes. The only limitation is the possibilities of spectacular software used for that. Figures 09. – 11. show the ability of Autodesk Map to offer various kind of our single Digital Elevation Model. Benefits and disadvantages of presented model‐generated scenario could be summarized: 1. Obviously there is quite straightforward possibility for quick generation of terrain shapes from single and skills undemanding photo set. The method is software independent and process presented in this report is only one way how to seize the solution of this problem. 2. The process itself and respective results are not unique and not groundbreaking by no means. Geometry generating via the space resection photogrammetry is widely used in production environment and even popular applications. Widespread Google Earth application is one of many. 3. Although an intended goal was achieved there is no control mechanism concerning the model precision and its spatial reliability – both in absolute and relative meaning. The precision could be verified by additional geodetic surveying but such a scenario increases the time and budget demands.
Fig. 08. Diigital Elevatio on Model in th he contour sh hape. Createdd in Autodesk Map 3D softtware.
Fig. 09. Diigital Elevatio on Model like a exaggerateed triangular r surface. Isom metric view frrom SW. Crea ated in Autodesk Map 3D software.
D Elevatiion Model likke an exaggerrated trianguular surface w with draped iimagery. Isom metric Fig. 10. Digital view from m SW. Created d in Autodesk Map 3D softtware.
Fig. 10. D Digital Elevatiion Model like contour pla an of surfacee with drapedd imagery. Crreated in Auttodesk Map 3D software.
References HADEM 1981 Hadem, I.: Bundle Adjustment in Industrial Photogrammetry. Photogrammetria, 37 (1981): 45‐60.
MOFFITT ‐ MIKHAIL 1980 Moffitt, F.H. ‐ E.M. Mikhail, 1980. Photogrammetry. New York 1980. RILEY 2009 Riley, D.N.: Aerial Archaeology in Britain. London 2009. SCHENK 2005 Schenk, T.: Introduction to Photogrammetry. Columbus 2005. SHIH ‐ FAIG 1987 Shih, T.Y., W. Faig: Physical Interpretation of the Extended DLT‐model. Proceedings of ASPRS Fall Convention, Reno, Nevada 1987, 385‐394.
Attachement 01 – list of aerial imagery from Duvansko Polje
Lihat lebih banyak...
Comentarios
