Dissertation > Excellent graduate degree dissertation topics show

BP neural network based high-resolution remote sensing image classification

Author: JiangJieFeng
Tutor: ZhangJing
School: Capital Normal University
Course: Cartography and Geographic Information Systems
Keywords: Quickbird high resolution remote sensing image classification Improved BP neural network algorithm Maximum Likelihood Minimum distance method ISODATA method Classification accuracy assessment method
CLC: TP751
Type: Master's thesis
Year: 2011
Downloads: 215
Quote: 3
Read: Download Dissertation

Abstract


With the development of science and technology , continuous improvement satellite sensors , remote sensing satellites have been able to capture this stage Yami -level high-resolution remote sensing image data . How to extract the rich spectrum of high-resolution images with texture feature information , into a scientific research and national defense , economic development services , remote sensing image classification has become a key area of ​​research topics . This paper discusses the multi-layer back propagation (BP) neural network classification algorithm is applied to high resolution remote sensing image classification . BP neural network classification algorithm compared to other statistical classification algorithm has strong ability to learn and easier to integrate remote sensing image texture , spectral , slope, aspect and other information on the feature class information extraction, in the field of pattern recognition has been widely applied . But BP neural network has a network structure is not easy to determine their own learning speed is slow , the training plane easy to fall into local minimum error shortcomings , this paper proposed to use for these shortcomings genetic algorithm to establish the network structure, the use changing the learning rate and momentum factor method to accelerate training learning speed , the BP neural network algorithm to be improved in order to enhance the high-resolution remote sensing image classification accuracy. In this paper, Beijing Fourth Ring Quickbird high resolution remote sensing image as the basis of experimental research data, using MATLAB software improved BP neural network program development, research area for image classification of surface features , and uses the results of the classification error matrix accuracy assessment , then the accuracy of the results and the accuracy of other classification methods were compared , the result is improved BP neural network classification algorithm 's overall classification accuracy was 93.4 %, compared with the traditional BP neural network increased by 2.4 %, compared with the maximum likelihood method increased by 5.5% than the minimum distance method increased by 10.2 %, compared with 20.2 per cent increase ISODATA method . In summary, based on improved BP neural network high-resolution remote sensing image classification method to quickly classify images and get a higher classification accuracy. The method was applied to classify urban high resolution image to obtain a similar surface features suitable for urban areas, the rapid classification methods to aid urban planning departments easily and quickly understand the development trend of urban social and economic activities of urban land resources planning and rational use of guidance.

Related Dissertations

  1. Based on statistics of the lognormal distribution heteroscedasticity model inferred,O212.1
  2. Mixed Exponential Distribution under Censored Accelerated test of quadratic estimates,O211.3
  3. Based on RFID Prison Intelligent Management System Research and Implementation,TP315
  4. Research and Design of Electronic Equalization Based on Most Likelihood Sequence Estimation,TN911.5
  5. The Non-Linear Mixed Effect Models in Population Pharmacokinetics and Its Discussion Based on Stochastic Differential Equations,R911
  6. NVD CD PRML read channel design and simulation,TP333.4
  7. Improved algorithms of the network topology tomography,TN915.02
  8. A Random Weighted Linear Estimator of the ARCH Parameters,F830
  9. Research on Signal Detection Techniques for MIMO Communication Systems,TN92
  10. On the Robust AMT Impedance Estimation of Time-series Data from STRATAGEM,P631.325
  11. Research on Algorithms for Near-Optimal Detection for V-BLAST Systems,TN929.5
  12. Research on Land Cover Classification in Mongolian Plateau Based on MODIS Data,P237
  13. MIMO System Equalization Technology Based on Sphere Decoding Algorithm,TN919.3
  14. Research of the Constant Envelope Signals Detection Algorithm of the Long Distance Communication,TN911.23
  15. Study of Quaternion Space Time Block Codes and Their Performance,TN911.2
  16. Research on Indoor Wireless Location,TN929.5
  17. DFT-S-OFDM detection technology,TN919.3
  18. Research of New Approach of Predicting Financial Distress Using Weighted Maximum Likelihood to Estimate Logit Model,F275
  19. Step drop and sequencer lowering stress accelerated life testing,O213.2
  20. Parameter Maximum Likelihood Estimations from Incomplete Data in Generalized Linear Models,O212.1
  21. Study on Synchronization Algorithm of OFDM System,TN919.3

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