Dissertation > Excellent graduate degree dissertation topics show

The Study of Wavelet Analysis Based Target Recognition Methods of High Resolution Radar

Author: ZhaoBingAi
Tutor: FanXiaoHong;SuHui;QiuZhiMing;HuXinSheng
School: Harbin Engineering University
Course: Management Science and Engineering
Keywords: Synthetic Aperture Radar (SAR) Target recognition Wavelet Analysis Radial basis function (RBF) neural network Mathematical Morphology
CLC: TN957.52
Type: PhD thesis
Year: 2003
Downloads: 1160
Quote: 6
Read: Download Dissertation

Abstract


Radar target recognition system applied on the battlefield for the development of modern warfare has great significance to enhance target identification system classification algorithm for dynamic and adaptive environment, improve the intelligent recognition system is a whole field of radar automatic target recognition Target Recognition (ATR) is a key problem areas. Synthetic aperture radar (SAR) technology continues to evolve, spaceborne imaging with a great number of airborne SAR, SAR image analysis applications in the ATR technique to strengthen the high-resolution radar image target recognition classification of great value. Wavelet decomposition of the signal analysis theory has good localization properties, the use of wavelet analysis theory in multi-scale analysis of target features, in line with the human brain information processing of the basic characteristics of a radar target signal processing powerful mathematical tool. SAR image speckle noise on the identification of a greater impact in target recognition required before the spots removed from the image. The traditional filtering techniques under certain conditions, it is difficult to meet the high performance requirements of filtering, and the use of filtering method based on wavelet analysis edge features while maintaining target has a great advantage. Face a lot of filtering algorithm, depending on the image type and task requirements, use of different filtering method urgent in-depth studies. In this paper, using the spatial domain filtering algorithm used, the combination of time and frequency domain filtering algorithms filter out noise in SAR images results were compared. Were selected to enhance the Lee filter algorithm unbiased GMAP algorithm, wavelet coefficients compression filtering algorithm, wavelet transform combined with Wiener filtering, wavelet transform and spatial filtering methods combining four different SAR image filtering, indicate whether the filtering algorithm The advantages and disadvantages of different image processing are given in different application environments, choose to use a different filter opinions. Based on the analysis of SAR image wavelet variance at all levels of detail subgraph approximately linear relationship exists between the threshold and experience to improve the common scale factor wavelet filtering method, on the basis of ensuring the accuracy improves the processing speed. Dimensional wavelet transform of the image target characteristics revealed at different resolutions, the resulting image detail images and approximate reflected in the distribution of targets under multi-resolution texture and other characteristics, with more class separability of SAR image analysis is very sense. Traditional dyadic wavelet transform is missing the required shift invariant pattern recognition feature, you need to determine the appropriate shift invariant pattern recognition feature with two-dimensional wavelet transform. In this paper, the traditional two-dimensional wavelet transform and wavelet transform completely over the differences and compare; and conventional 2D wavelet transform without downsampling algorithm for wavelet transform and wavelet transform to extract the porous texture energy characteristics were studied compared with the control. This is the first use of the image gray value with no down-sampling wavelet detail image texture feature vector composed of energy, so that the differences between the objectives texture than other methods Seisuke Fukuda proceeds more obvious. In comparison with traditional wavelet transform and wavelet algorithm over completely SAR image texture energy feature extraction The main difference between the process, found this paper had completely wavelet transform to extract texture features easy it is appropriate, SAR image can be used as the statistical characteristics of the target surface. Artificial neural network is an important tool for pattern recognition, this paper were using BP neural network, radial basis function (RBF) neural network, self-organizing feature map neural network for surface targets SAR images were analyzed, the choice of gray value, mean, wavelet Texture characteristics of the different feature as the input vector, to obtain a high classification accuracy. This is the first wavelet decomposition will be too entirely new texture feature extraction (O recognition feature to find rF) vector and BP, RBF neural network to classify the combination of target identification, classification results show that, OWAI next feature vector combined with the RBF neural network processing better results. Traditional K-means algorithm, fuzzy C-means clustering algorithm and self-organizing feature map neural network image classification method in the automatic classification process to pixel gray for the analysis object, which is the use of these methods classify the root cause of the slow speed of operation. This is an important factor as the histogram into count Ting, in the iterator object from the gray value of each pixel into a gray level of the image exists, the arithmetic magic fell to the original algorithm (M only N), due to accounting for the amount of memory corresponding smaller, computing speed further increased in a very short}} completed within inches of image segmentation. In this paper, three color SAR images or film scanning SAR images, using a saturation-based experience from the K-means classification algorithm for analysis, and with the gray value for the image analysis results can be compared. By comparison, based on experience from the saturation of the K-means algorithm classification result reflects the various regions of similar consistency between the objectives, the visual effect is better. In the image target recognition, based on statistical pattern recognition and artificial intelligence-based methods are not applicable] fl \For strong noise background airports SAR image target recognition problem, use two wooden paper were morphological gradient values ??and the gray-value method of analysis, the method to maintain the target shape and highlight the important goal of strong noise background information letter J, is superior to conventional algorithms. This article will discuss the morphological system - with the direction of the wavelet filter a combined extracts SA weak ruler image edges in the image nicely segmented \and K-S distance sea SAR image target detection, less ram made a comprehensive analysis of the whole image and determine the closing value of the target area image method method has been applied for image binarization is better than the conventional method, the target for the waters The classification ready.

Related Dissertations

  1. Research on Testing and Analyzing Technology for Time Parameter of Aerospace Relay,TM58
  2. ISAR Imaging Simulation of Space Targets and Target Recognition Based on ISAR Images,TN957.52
  3. Tongue Feature Extraction and Research of Fusion Classification,TP391.41
  4. Study on Virtual Detector of Infrared Hyper-Spectral Image,TP391.41
  5. Research on the Image Real-Time Acquisition, Storage and Image Processing System,TP391.41
  6. The Properties of Laser Speckle Based on the Mathematical Morphology,O29
  7. Study of Event Related Potentials Based on Chinese Auditory Cognition,R318.0
  8. Road extraction algorithm based on region segmentation of remote sensing image,TP751
  9. Technology of Blood Vessel Diameter Measurement Automatic Based on Digital Image Processing,R310
  10. Research for Infrared Image Target Identification and Tracking Technology,TP391.41
  11. The Compression and Fusion Technique Research of Underwater Target Feature,TN911.7
  12. The Application of Ant Colony Algorithm in Meteorological Satellite Cloud Pictures Segmentation,TP391.41
  13. Based on high-resolution remote sensing data mining houses information extraction,TP751
  14. Research Implementation of Crack Detection Method Based on Its Acoustic Property,TP274
  15. The Study of Palmprint Image Identification Based on Wavelet Analysis,TP391.41
  16. Study of GPS Data De-noising Method Based on Wavelet and Kalman Filtering,P228.4
  17. Research on Subimage Selection and Mathching Method for Synthetic Aperture Radar(SAR) Target Recognition,TN957.52
  18. Design and Implementation of Intelligent Transportation Video Monitoring System Based on 3G Network,TP391.41
  19. Research on Specific Target Recognition Algorithms in Video Surveillance,TP391.41
  20. Development and Application Research on a Video-based Motion Tracking and Analysis System for Animals,TP391.41
  21. Integration of Spatial Information Bag of Feature in Image Annotation,TP391.41

CLC: > Industrial Technology > Radio electronics, telecommunications technology > Radar > Radar equipment,radar > Radar receiving equipment > Data,image processing and admission
© 2012 www.DissertationTopic.Net  Mobile