| | SLO | ENG | Cookies and privacy

Bigger font | Smaller font

Show document

Title:VZPOSTAVITEV KORESPONDENCE PRI VELIKIH ODMIKIH DVEH KAMER ZA POLJUBNO IZBRANO TOČKO KOT PRISPEVEK K IZBOLJŠANJU RAZPOZNAVANJA SLIK
Authors:ID Podbreznik, Peter (Author)
ID Potočnik, Božidar (Mentor) More about this mentor... New window
ID Rebolj, Danijel (Co-mentor)
Files:.pdf DR_Podbreznik_Peter_2011.pdf (3,37 MB)
MD5: B35EB06B72B406B87F6637C5064C1E3A
PID: 20.500.12556/dkum/e059f040-edad-4c8a-b14a-7d68bdfe770b
 
Language:Slovenian
Work type:Dissertation (m)
Organization:FERI - Faculty of Electrical Engineering and Computer Science
Abstract:Eden izmed možnih načinov spremljanja aktivnosti na gradbišču je zajemanje slik z več kamerami, kateremu sledi razpoznavanje posameznih gradbenih elementov in ugotavljanje izgotovljenosti objekta na osnovi 4D modela. Takšen pristop zahteva vzpostavitev korespondence med slikami, zajetimi iz različnih pogledov (kamer). Osnovni problem je torej vzpostavitev korespondence za poljubno izbrano točko, ki se lahko nahaja kjerkoli na sliki (tudi na področjih slike z zelo podobno barvo ali teksturo). V doktorski disertaciji smo razvili novo robustno metodo ASIFT-SH za vzpostavitev korespondence za poljubno izbrano točko pri velikih odmikih dveh kamer, ki je mešanica metod ASIFT, segmentacije in lokalne homografije. Ta metoda sestoji iz štirih korakov: i.) določitve začetnih korespondenčnih točk na slikah obeh pogledov z metodo ASIFT, ii.) združevanja korespondenčnih točk iz prvega koraka v ustrezne podmnožice na osnovi segmentiranih regij, iii.) izračuna lokalnih homografij za posamezno podmnožico korespondenčnih točk, ki pripadajo posamezni segmentirani regiji in iv.) izračuna korespondenčne točke z uporabo ustrezne lokalne homografije. Metoda ASIFT-SH v primerjavi z metodo, ki določi korespondenčne točke na osnovi iskanja v okolici epipremice (metoda EPI), zagotavlja bistveno boljše rezultate, še posebej na področjih slike s podobno intenziteto slikovnih elementov. Povprečna napaka korespondenčnih točk, izračunanih z našo metodo, je nekaj slikovnih elementov, medtem ko je odstopanje korespondenčnih točk, vzpostavljenih z metodo EPI, nekaj sto slikovnih elementov. Kritičen korak metode ASIFT-SH predstavlja segmentacija. Metoda namreč predpostavlja, da segmentirane regije ustrezajo kriteriju koplanarnosti (tj. vse točke v segmentirani regiji ležijo v isti ravnini). V splošnem tega ni mogoče zagotoviti. V naš algoritem smo zato vpeljali še korak adaptivnega prilagajanja, s katerim lahko preoblikujemo segmentirane regije tako, da bolje izpolnjujejo zahtevani kriterij. Vpeljani korak temelji na 3D rekonstrukciji začetnih korespondenčnih točk in iskanju čim manjšega števila ravnin v prostoru, katerim te točke pripadajo. Točke, ki pripadajo posamezni ravnini, predstavljajo novo nastale podmnožice začetnih korespondenčnih točk. Rezultati meritev so pokazali, da vpeljava dodatnega koraka, ki omogoča adaptivno prilagajanje segmentiranih regij na osnovi 3D rekonstrukcije, tj. adaptivna različica metode ASIFT-SH, zagotavlja povečanje natančnosti izračunanih korespondenčnih točk. Uspešnost adaptivnega prilagajanja segmentiranih regij je neposredno odvisna od 3D rekonstrukcije, na katero pa močno vplivajo notranji in zunanji parametri kamere. Ker so kamere na gradbišču izpostavljene različnim okoljskim dejavnikom, ki vplivajo tudi na parametre kamere, smo v naši raziskavi preučili tudi vpliv temperaturnih sprememb na geometrijske lastnosti kamere. Vpliv temperature smo ugotavljali tako za eno kamero, kot tudi za sistem kalibriranih kamer, pri čemer vpliva na elektroniko kamere nismo proučevali. Najprej smo dopolnili analitični model kamere s členom, ki modelira vpliv temperaturnih sprememb na delovanje posamezne kamere. Takšen modificirani analitični model kamere omogoča kvantitativno oceno vpliva temperaturnih sprememb na delovanje kamere. Sprememba temperature neposredno vpliva tudi na natančnost sistema kalibriranih kamer. Oceno takšnega odstopanja smo določili eksperimentalno, saj analitična določitev napake ni mogoča. Na koncu smo definirali skupno odstopanje (napako) kalibriranih kamer, kjer smo upoštevali tako odstopanje zaradi numerične napake kalibracijske metode in postopka 3D rekonstrukcije, kot tudi odstopanje, ki je posledica vpliva temperaturnih sprememb. Rezultati raziskave so pokazali, da se vpliv temperature zmanjšuje z večanjem razdalje med kamero in opazovanim objektom. V predvidenih delovnih pogojih na gradbišču, se vpliv temperaturnih sprememb na sliki odraža za manj kot en slikovni element.
Keywords:spremljanje gradnje, spremljanje aktivnosti na gradbišču, ujemanje slik, korespondenčne točke, veliki odmiki kamer, poljubno izbrana točka, homografija, segmentacija, epipolarna geometrija, analitični model kamere, parametri kamere, projekcijska matrika, umerjanje kamer, 3D rekonstrukcija, temperaturne spremembe
Year of publishing:2011
Publisher:[P. Podbreznik]
Source:Maribor
PID:20.500.12556/DKUM-19329 New window
UDC:004.89
COBISS.SI-ID:256882176 New window
NUK URN:URN:SI:UM:DK:Q6KQ3BZF
Publication date in DKUM:20.07.2011
Views:2207
Downloads:203
Metadata:XML RDF-CHPDL DC-XML DC-RDF
Categories:KTFMB - FERI
:
  
Average score:(0 votes)
Your score:Voting is allowed only for logged in users.
Share:AddThis
AddThis uses cookies that require your consent. Edit consent...

Hover the mouse pointer over a document title to show the abstract or click on the title to get all document metadata.

Secondary language

Language:English
Title:Estimating any correspondence between two widely-separate views for arbitrarily-selected point, as a contribution to improving image recognition
Abstract:One possible method of monitoring activities on the building site is based on capturing images using multiple cameras, followed by the recognition of individual building components, and then identifying an object's status based on a 4D model. Such an approach requires the establishment of some correspondence between those images captured from different viewpoints (cameras). The basic problem is, therefore, the establishment of any correspondence at an arbitrarily selected point, which could be located anywhere on the image (also on image areas with uniform colour or texture). This doctoral dissertation develops a new robust ASIFT-SH method for establishing any correspondence between arbitrarily-selected points within two widely-baselined cameras, which is a combination of the ASIFT method, segmentation, and local homography. This method consists of four steps: i) determining the initial corresponding points within the images of both views by using the ASIFT method, ii.) grouping of initial corresponding points from the first step into subsets, based on segmented regions iii.) calculation of local homographies for a particular subset of corresponding points, and iv.) determining any correspondence between arbitrary points from a particular camera's viewpoint, by using a suitable local homography. The ASIFT-SH method, when compared to that of searching the area surrounding an epipolar line (EPI method), provides more accurate results, especially on surfaces with similar pixel intensities. The average error in our method comes within the order of a few pixels, whilst for the EPI method it is within the order of a few hundred pixels. The critical step of the ASIFT-SH method concerns segmentation. The segmentation presupposes that the segmented regions correspond to the coplanarity criteria (i.e., all points in the segmented region are located within the same plane). In general, this cannot be ensured. Therefore, we have introduced into our algorithm a step for adaptive adjustment, within which segmented regions are remodelled so that they better meet the required criterion. This introduced step is based on 3D reconstruction of the initial corresponding points, and a search for the minimal number of planes within the 3D space, to which these points belong. Those points that belong to a particular plane, represent a newly-created subset of the initial corresponding points. The results point out that the introduction of additional step enables an adaptive adjustment of segmented regions based on the 3D reconstruction, i.e. an adaptive version of the ASIFT-SH method, thus assuring more accurate calculation of the corresponding points. The success of adaptively adjusting segmented regions directly depends on the 3D reconstruction, which is, however, strongly affected by intrinsic and extrinsic camera parameters. Since cameras at a building site are exposed those different environmental factors which also affect the cameras' parameters, therefore, in our study we investigated any influence of temperature variations on the geometric properties of the camera. Temperature influence on the geometrical properties of both a single camera and a calibrated camera system was estimated, whereas the influence on camera electronics was ignored. Firstly, the analytical camera model was supplemented with a term, which models the influence of temperature variations over a single camera's operation. A modified analytical camera model enabled quantitative assessment regarding the influence of temperature variations on a camera's operation. Temperature variations also directly influence the accuracies of calibrated cameras. The inability to analytically-determine the calibration method error magnitude, led us to experimentally-estimate any errors regarding calibrated cameras. Finally, the total errors regarding calibrated cameras were derived at by combining the numerical errors of the calibration method and 3D reconstruction with those errors originating from temperature variations. The results show that the influence of temperature varia
Keywords:activity tracking, construction monitoring, image matching, corresponding points, large separated views, arbitrarily selected point, homography, segmentation, epipolar geometry, analytical camera model, camera parameters, projection matrix, camera calibration, 3D reconstruction, temperature variations


Comments

Leave comment

You must log in to leave a comment.

Comments (0)
0 - 0 / 0
 
There are no comments!

Back
Logos of partners University of Maribor University of Ljubljana University of Primorska University of Nova Gorica