Preferencje help
Widoczny [Schowaj] Abstrakt
Liczba wyników
Tytuł artykułu

Badanie wybranych metod matchingu do pomiaru punktów na cyfrowych obrazach bliskiego zasięgu

Wybrane pełne teksty z tego czasopisma
Warianty tytułu
Research on selected matching methods for the measurement of points on close range digital images
Języki publikacji
One of the basic aims of close-range photogrammetry is to provide an automatic mode for measurement of simple structure signalized points and discreet points with defined structure characteristics on single digital images. The automatic measurement of coordinates of signalized points centres on 2D images requires the use of matching operators. Based on matching operators analysis, four methods of measurement of points coordinates with subpixel accuracy were selected for further detailed research from the Area Based Matching and Feature Based Matching group . These methods are:. Center of gravity operator. Center weighted method . Cross correlation . Least Squares Matching A special software called .Matching. was created in Delphi 7 programming language in order to carry out the research. The above mentioned four procedures of matching were implemented in the software. The application is created for the measurement in semiautomatic mode of pixel coordinates of signalized points registered on digital images of any resolution. The matching may be carried out in greyscale [0, 255]. On colour images it is carried out in channels R, G, B and brightness I, which is calculated according to the following formula: The created application is characterized by user-friendly interface, which allows, among others, functional visualization of images and of the current work mode, defining areas of interest and naming them, determining calculation parameters, displaying measurement results in a table, recording the results on a disc in one of two formats, exporting the results to a text file etc. The application has all typical functions of Windows applications. Execution of measurements in cross-correlation and LSM methods is aided by automatic calcuation by the center weighted method or by manual determining of template coordinates. In the modified center weighted method image filtration by a gradient filter . the Sobel operator . is used. The Gauss elimination was used to solve the linearized equations system in programming of the LSM method . Research on accuracy and quality of matching methods was carried out on synthetic and real images of 2D test field with 35 targets. Synthetic colour 24 Bit digital images of 2.1K×1.5K were generated and recorded in the BMP file format. The images of the real test field with 35 targets were taken by Kodak DC 4800 (sensor CCD, 3.1 millions pixels, image resolution 2.2K×1.4K) compact type digital camera and by SLR digital camera Kodak DCS Pro 14n (sensor CMOS, 13.7 millions pixels, image resolution 4.5K×3K). I ?0.299R ??0.587G??0.114?B Research on matching methods accuracy for synthetic and real images in different orientations of the images and sizes of the targets was carried out in the brightness channel I and independently in the R, G, B channels. The results of the matching methods examined were assessed by means of mean square residual rx , ry for measured pixel coordinates. There is no significant difference between results of matching procedures for each R, G, B channel, as well as monochromatic brightness I with 8 Bit depth. In comparison with the results of the LSM method, the biggest differences, regardless of the size of sygnalized point, of the order of . +-1.3 pixels on average for synthetic images and +-1.6 pixels for real images, are achieved while using the center of gravity operator. Average diffrences in the cross correlation method amount to +-0.3 pixels for synthetic images and +--0.55 pixels for real images. There is no significant difference (š0.02 pixels) between the results obtained by center weighted method for both types of images and the pixel coordinates achieved by the LSM method. The matching of images registered with Kodak DC4800 and Kodak DCS Pro 14n cameras brings the smallest residuals in the green channel G, and the biggest in red (R) and blue (B) channels. This is the consequence of of the Bayer.s filter used and for the weights used during interpolation in defining R, G, B colors for a given pixel. The results of the research proved, that the diameter of 5÷15 pixels is the optimal size of targets. Big targets of the diameter of 25 pixels and more significantly extend the time of the matching and result in decrease of positioning accuracy. The analysis of impact of each component R, G, B on effectiveness and accuracy of matching procedures should be carried out in automatic measurements with sub pixel accuracy, as well as in elaborations combining multisensoral digital visual data.
Opis fizyczny
Bibliogr. 11 poz.
  • Katedra Fotogrametrii i Teledetekcji, Uniwersytet Warmińsko-Mazurski w Olsztynie
  • Instytutut Elektroniki, Politechnika Łódzka
  • 1. Atkinson K.B., 1996: Close Range Photogrammetry and Machine Vision. Whittles Publishing.
  • 2. Chen J., Clarke T.A., 1992: The Automatic Recognition, Location and Labelling of Targets in Digital Photogrammetric Engineering Measurement", ISPRS Washington, Com.V.
  • 3. Grün A.W., 1985: Adaptive Least Squares Correlation a Powerful Image Matching Technique. South African Journal of Photogrammetry, Remote Sensing and Cartography, 14(3), s. 175-187.
  • 4. Kowalczyk M., 2003: Badanie zakresu i stopnia możliwej automatyzacji pomiaru fotogrametrycznego bazującego na kamerze cyfrowej. Rozprawa doktorska, Politechnika Warszawska.
  • 5. Kraus, K., 1993: Photogrammetry. Dümmler Verlag, Bon.
  • 6. Luhmann T., 1996: Results of the German Comparison Test for Digital Point Operators. International Archives of Photogrammetry and Remote Sensing, Vol. XXXI Part B5/2 Com.V, s. 324-329.
  • 7. Luhmann T., 2000: Nahbereichsphotogrammetrie - Grundlagen, Methoden und Anwendungen, WichmanVerlag, Heidelberg.
  • 8. Maas H.G., 1997: Mehrbildtechniken in der digitalen Photogrammetrie. ETH Zürich, Institut für Geodäsie und Photogrammetrie, Nr 62.
  • 9. Sawicki P., 2000: Digital multisensoral video-thermal system for close range metrology applications. International Archives of Photogrammetry and Remote Sensing, Vol. XXXIII Part B5/2 Com.V, s. 691-698.
  • 10. Schenk T., 1999: Digital Photogrammetry. TerraScience.
  • 11. Trinder J.C., Jansa J., Huang Y., 1995: An Assessement of the Precision and Accuracy of Methods of Digital Target Location. ISPRS Journal of Photogrammetry and Remote Sensing, 50(2), 12-20.
Typ dokumentu
Identyfikator YADDA
JavaScript jest wyłączony w Twojej przeglądarce internetowej. Włącz go, a następnie odśwież stronę, aby móc w pełni z niej korzystać.