Upload
artemis-valanis
View
316
Download
0
Tags:
Embed Size (px)
Citation preview
RESTITUTION AUTOMATION FOR CLOSE-RANGE APPLICATIONS
Artemis Valanis, Andreas Georgopoulos
School of Rural and Surveying EngineeringLaboratory of Photogrammetry
National Technical University of Athens
Development of a semi- or fully- automated method for the restitution process of image products in the case of close-range applications
Applicability of the method for the case of Byzantine monuments
Trial applications on other kinds of monuments
OBJECTIVES
Construction complexity
Presence of decorative elements
Tuffstone is the basic construction material
OBJECT DESCRIPTION
Extensive research of the properties of the object of interest and trial application of several image processing techniques
Development of a method for the detection of the objects of interest
Program and interface development
Application of the proposed method for Byzantine and other kinds of monuments
Evaluation of the results
COURSE OF STUDY
Noise reduction methods (mean-, order statistics and adaptive filters)
Image enhancement in the frequency domain (FFT, Ideal filters, Butterworth filters)
Edge detection algorithms (Sobel, Prewitt, Canny, LoG, color edge detection)
Morphological processing (erosion, dilation, morphological gradient)
Segmentation and thresholding techniques
TRIAL APPLICATIONS OF VARIOUS IMAGE PROCESSING TECHNIQUES
Object complexityStrong resemblance in the appearance of the stones and jointsExistence of inclinated planes and shadowed areasErosion of the construction material Presence of moisture
PROBLEMS ENCOUNTERED IN THE DETECTION OF THE OBJECTS
Sample selectionCalculation of the mean value and standard deviation of the gray values of the pixels of the sample Region Growing
PROPOSED APPROACH
ROUTE FOLLOWED BY THE ALGORITHM
REGIONS EXAMINED BY THE ALGORITHM ACCORDING TO THE POSITION OF THE CANDIDATE PIXEL
• Connectivity criterion: At least two pixels of the examined region must belong to the object
• Homogeneity criterion: The arithmetic (mr) mean of the gray values of the candidate and the identified as object pixels of the currently examined region must belong in the confidence interval given by Equation [1]
ms - z ss mr ms + z ss [1]
CRITERIA EVALUATED BY THE ALGORITHM
DETECTION RESULTS FOR A SINGLE OBJECT
Definition of the area to be processed Sample selection for the objects that must be detected Application of an adaptive thresholding technique Improvement of the binary image which is yielded by the
thresholding process Exploitation of the improved binary image for the
automated sample selection Application of the algorithm for each one of the objects
detected
AUTOMATION OF THE PROCESS
EXAMPLE
INTERFACE OF THE PROGRAM
APPLICATIONS – BYZANTINE MONUMENTS
EXPERIMENTS FOR THE CASE OF THE DOME
EXPERIMENTS FOR THE DECORATIVE ELEMENTS
FINAL RESULTS AND COMPARISON
Tolerance: σTolerance: 2σTolerance: 3σ
EVALUATION
EVALUATION
Distance from
manually collected
boundaries(mm)
Sample 1(11559 pixels)
Sample 2(8485 pixels)
Total(20044 pixels)
Pixels percentage pixels percentage pixels percentage
d σ 8020 69% 5371 63% 13391 67%
2σ d 2σ 2251 20% 1909 23% 4160 21%
3σ d 3σ 740 6% 747 9% 1487 7%
d> 3σ 548 5% 458 5% 1006 5%
The proposed method is very flexible and fast
The program used for the application of the developed methods offers a wide range of possibilities and is user friendly
The accuracy of the restitution is objectively characterized as satisfactory for the case of Byzantine monuments
The restitution process is accelerated by a factor of at least 1.7
CONCLUSIONS
Examination of more complex properties such as texture Thorough review and further development of the fully
automated method Detailed research of the properties of other kinds of
monuments
SUGGESTIONS
Thank you for your attention!
VECTORIZATION
ARCH OF ADRIANOS
NATIONAL THEATRE
Distance from
manually collected
boundaries(mm)
Sample 1(11122 pixels)
Sample 2(9067 pixels)
Pixels percentage pixels percentage
d σ 7859 70% 5371 28%
2σ d 2σ 1934 18% 1909 22%
3σ d 3σ 382 3% 747 18%
d> 3σ 947 9% 458 32%
BYZANTINE WALL (DAPHNI MONASTERY)