Dissertation > Excellent graduate degree dissertation topics show

Research on Building Segmentation in Remote Sensing Image

Author: ZhangYan
Tutor: WangXia
School: Hebei University of Technology
Course: Communication and Information System
Keywords: segmentation of remote sensing image markov random field model waveletanalysis template matching
CLC: TP751
Type: Master's thesis
Year: 2012
Downloads: 29
Quote: 0
Read: Download Dissertation

Abstract


Building is one of typical feature of urban remote sensing images, and extraction andmeasurement of the buildings in remote sensing images are widely used in feature mapping, GISsystem updating, digitized city as well as military reconnaissance field. Segmentation of remotesensing image is the key to achieving this goal. In order to segment and extract the targetbuildings from a remote sensing image quickly and effectively, the article presents anunsupervised environment building target segmentation method from remote sensing imagebased on MRF (Markov Random Field).Firstly, the N-MRF model of image segmentation method is studied. N-MRF model regardsintensity values of each pixel of the image as random variables with certain probabilitydistribution, the mathematical model for image segmentation uses Bayesian rules, on thecondition that the observations of character field is given. Markov among pixels of the image isregarded as the prior distribution. With the equivalence of MRF and GRF, the calculation of theposterior probability is transformed into the calculation of posterior energy. The ICM algorithmis used to obtain the state of maximum probability as the optimization algorithm. Finally,parameter estimation and image segmentation go on alternately by iteration of the EM algorithm,and seek the best implementation of the implicit data.For remote sensing digital image is usually not stationary random signal, in order to expressnon-stationary characteristics of the image better, wavelet analysis and Markov random fieldmodel are combined to solve the problem. Firstly, the multi-scale sequence by waveletdecomposition is regarded as the observation of feature field on each scale, which is modeled byGaussian Mixture Model, and MRF as the prior probability distribution model of the label field.Segmentation result of coarse scale is projected directly to the most nearby fine scale as theinitial results and optimized. The final segmentation results of the original resolution areachieved by such iteration.Finally, the segmentation results are processed by mathematical morphology, and makethem match with common building shape templates by correlation, and the position of the building target is detected. Experimentations of remote sensing images demonstrate that themethod is able to complete the segmentation of the building goals in the complex backgroundand detect regular targets.

Related Dissertations

  1. Tracking Events for Food Complaint Documents Based on Ontology,TP391.1
  2. Research on Medical Image Segmentation Method Based on Markov Random Field Model,TP391.41
  3. Gesture Recognition for Traffic Control Based on Thinning Algorithm and Template Matching,TP391.41
  4. Research on Face Tracking Technologies Based on Movement Prediction,TP391.41
  5. Drivers' eyes open or closed state of the computer image recognition technology development,TP391.41
  6. Research on Method of Ship Target Detection Under Complex Harbor Background,TP391.41
  7. Research on Vision Detection Algorithm for Tracing Printing System,TP391.41
  8. Fundus Image Segmentation Based on SVM and Template Matching,TP391.41
  9. Research of the Relationship between Respiratory and Lung Tumor Displacement,TP391.41
  10. Visual Servoing Approaches Based on Parallel Mechanism,TP391.41
  11. Research on Image Matching Based Target Detection Technology and Algorithm Design,TP391.41
  12. A Dome projection system design and implementation,TP391.41
  13. Spectrum Information hyperspectral imaging target detection technique,TP391.41
  14. Infrared detection system its key technologies,TN215
  15. Mobile robot vision -based target search and tracking technology,TP391.41
  16. Technologies of Real-time Traffic Sign Segmentation and Recognition,TP391.41
  17. Development and Application of Airport Enclosure with Long-range Laser Photoelectric Alarm System,TP277
  18. Research on New Type License Plate Recognition Based on Neural Network,TP391.41
  19. On-Line Inspection System of Reversed Fitting Defect on Pinhead of One-off Injectors,TH789
  20. Study on Seal Matching and Identification,TP391.41
  21. Research and Realization of Licencse plate Recgniton Algorith,TP391.41

CLC: > Industrial Technology > Automation technology,computer technology > Remote sensing technology > Interpretation, identification and processing of remote sensing images > Image processing methods
© 2012 www.DissertationTopic.Net  Mobile