Retrieving images of fighter planes based on 3D model

Number of pages: 106 File Format: word File Code: 32147
Year: 2012 University Degree: Master's degree Category: Electrical Engineering
  • Part of the Content
  • Contents & Resources
  • Summary of Retrieving images of fighter planes based on 3D model

    Senior Thesis for Master Degree

    Electricity and Robotics

    Electronics Department

    Abstract

    In this thesis, the problem of retrieving images of fighter planes from a database including 10 different models Based on a query image [1], it has been investigated. Airplanes of the same model with the query image are identified in the database and presented to the user. The main challenge in retrieving images of fighter planes is the lack of the same angle of view of the camera in the images in the database and the query image, for which it is proposed to use the 3D model of the fighter planes and prepare reference images from different viewing angles by 3D geometric mapping (virtual cameras).  

    We have proposed three different methods for extracting features from images and measuring the similarity of images, two of which are time-sensitive and the other is time-resistant. In the time-sensitive methods, the first method is based on measuring the area of ??the non-overlapping area and the other is based on the gradient angle histogram. In the time-resistant method, Zernike moments are used for feature extraction. The simulation results show the superiority of the Zernike moments method in terms of recovery accuracy with an accuracy of about 80.8%, which despite using several similar classes in the database, this recovery accuracy is promising.

    Keywords: Image retrieval, 3D model, gradient angle histogram, overlapping region area, Zernike moments, virtual camera, fighter plane

    1-1 Introduction

    Content-based image retrieval systems (CBIR) [1], include a set of methods that perform the retrieval process to determine the user's desired images based on low-level visual characteristics such as Color, texture, shape, or high-level semantic features from the image database. In these systems, recovery is highly correlated with low-level visual features. Due to the fact that the use of these systems is increasing today, so the need for techniques that can perform the recovery process as accurately as possible seems essential.

    Since the two main steps in all CBIR systems are visual feature extraction and similarity measurement, researchers in order to increase the accuracy of the recovery process, today, various methods in the field of CBIR, in order to determine They presented efficient features to display the content of images and adaptive techniques to determine the degree of similarity between images as efficiently as possible.

    Human understands the real world semantically, but CBIR systems understand images based on low-level features such as color, texture, and shape. Therefore, there is a distance between system understanding and human understanding, which is called "semantic distance".

    Therefore, one of the basic problems in content-based image retrieval systems is the semantic distance between system understanding and human understanding, so that the more the methods provided in retrieval systems can reduce this parameter in determining the characteristic vector and similarity between images, the more suitable methods

    Although researchers today have presented various techniques based on low-level features to determine the semantic feature vector that is very close to human perception, but still, in many cases, the resulting vector has a great semantic distance from human perception [1].

    1-2 Some applications of CBIR systems

    Applications Image recovery systems are increasing day by day and there are many applications in this field. In this section, some of its important applications are mentioned.

    1-2-1 Searching web pages

    Most of the applications of CBIR are for searching web pages.. There are a number of search engines such as: Yahoo Simplicity, Netra, Qbic and Google image search that have simplified the search for images from web pages.

    1-2-2 Law enforcement

    CBIR has various applications in law enforcement and crime prevention, such as fingerprint recognition, facial recognition, footprint recognition, and surveillance systems. Many people use the Internet to sell and display their illegal goods, such as drugs, weapons, etc. they use The use of CBIR can help to identify them.

    1-2-3 Medical profession

    In the medical profession, X-ray image database and scanned images are used for diagnosis and monitoring and research purposes.

    1-2-4 Architecture and engineering design

    In architectural and engineering designs, there are image databases for design projects, finished projects, and parts and machines.

    1-2-5 Fashion and Publishing

    In publishing and advertising, journalists use image databases for various events and activities such as sports, buildings, personalities, national and international events, and Advertisements have products.

    1-2-6 Historical research

    In historical research, there are image databases in fields such as art, sociology, and medicine.

    1-2-7 Remote sensing

    Applications of remote sensing include the analysis of aerial and satellite images. (taken from geographical areas) that are useful in mapping, agriculture and meteorology can be mentioned.

    1-2-8 Some other applications

    Other important applications of image recovery can be applications such as digital libraries, or military affairs such as the detection of attacking aircraft, museum archives, natural resource management, weather forecasting and so on.

    1-3 Research Objectives

    Retrieving images of fighter planes is one of the important branches of CBIR, which can be used in military applications such as detecting attacking aircraft, obtaining images of various models of fighter planes at different angles and so on. be useful As far as we know, no research has been done in this field (if any research has been done, it has not been published due to confidentiality). Therefore, considering the applicability of this field of image recovery in defense systems, it seems necessary to conduct this research. The proposed recovery system, using the 3D model of fighter planes and space maps, has provided the ability to recognize images of similar planes at different angles.

    Of course, the methods used in this research, in addition to being used in military applications, can also be used for other similar applications such as recognizing the model and type of cars, tanks, and armor.

     

    Abstract

    Today, In this thesis, the problem of image retrieval for fighter aircrafts from a dataset consists of 10 different fighter classes is studied. In this project, a CBIR system is developed which found similar aircrafts to a query aircraft picture from database.  The difference of viewpoint in query picture and search domain pictures is the main challenging problem in image retrieval systems. We tackle this problem using 3D models of different fighter aircraft classes. We use the concept of virtual cameras around a surrounding spherical surface, in order to prepare a lot of reference pictures of the aircrafts from different viewpoints.

    We propose three different methods for feature extraction and similarity evaluation: 1) cross correlation of aircraft silhouette, 2) histogram of oriented gradients, and 3) Zernike moments, which the latter is robust to rotation.

    Simulation results show that the best retrieval accuracy (about 80.8%) is achieved by using Zernike moments method.

  • Contents & References of Retrieving images of fighter planes based on 3D model

    List:

    Chapter 1: Introduction. 1

    1-1 Introduction. 3

    1-2 Some applications of CBIR systems. 4

    1-2-1 Search web pages. 4

    1-2-2 Law enforcement. 4

    1-2-3 medical profession. 5

    1-2-4 architecture and engineering design. 5

    1-2-5 fashion and publication. 5

    1-2-6 historical research. 5

    1-2-7 remote sensing. 5

    1-2-8 Some other applications. 6

    1-3 research objectives. 6

    1-4 research questions. 6

    1-5 research results. 7

    1-6 thesis structure. 8

    Chapter 2: Basic concepts. 9

    2-1 Introduction. 11

    2-2 Objectives of CBIR systems. 12

    2-3 different techniques in target search method. 13

    2-3-1 Target search by visible sample. 13

    2-3-2 Target search based on drawing. 13

    2-3-3 target search based on the draft. 14

    2-4 structure. 14

    Chapter 3: Past works. 17

    3-1 Recovery of 3D models. 19

    3-2 Retrieving images based on content 20

    3-2-1 Color. 20

    3-2-1-1 color histogram. 21

    3-2-1-2 moment of color. 21

    3-2-1-3 circular color histogram. 22

    3-2-2 texture. 23

    3-2-2-1 Different statistical methods to determine texture characteristics. 24

    3-2-2-2 structural methods to determine texture characteristics. 24

    Figure 3-2-3. 25

    3-2-3-1 Determination of shape characteristics using edge-based methods. 25

    3-2-3-2 Determining the characteristics of the shape using methods based on the edge of the contour or border. 26

    3-3 criteria of similarity. 26

    3-3-1 possible similarity criteria. 27

    3-3-1-1 Multivariate Gaussians (MVG) 27

    3-3-1-2 Independent Adaptive Distributions (FIT) 28

    3-3-1-3 Combination of Gaussians (GMIX) 28

    3-3-1-4 Use of logical regression. 29

    3-3-2 criteria of geometric similarity. 29

    3-3-3 histogram similarity criteria. 29

    3-3-3-1 exploratory histogram interval. 29

    3-3-3-2 Non-parametric statistical tests. 30

    3-3-3-3 divergence of scientific information. 30

    3-3-4 Creating a new similarity criterion based on the combination of several criteria. 31

    Chapter 4: Suggested methods. 33

    4-1 Steps of the proposed image recovery system. 35

    4-2 Preprocessing. 36

    4-3 Alignment of 3D models. 37

    4-4 2D maps of 3D model. 38

    4-5 Reducing the number of views 40

    4-5-1 Reducing views in time-sensitive methods. 41

    4-5-2 Reducing views in time-resistant methods. 42

    4-5-3 The final number of reduced views and their comparison 43

    4-6 Shadow view extraction from database images and query image. 43

    4-7 Alignment of shadow images..44

    4-8 feature extraction 45

    4-8-1 feature extraction with the method of non-overlapping region area. 45

    4-8-2 feature extraction with gradient angle histogram method. 48

    4-8-3 feature extraction with Zernike moments method. 50

    4-9 Measuring similarity based on Euclidean distance. 54

    4-10 Retrieving images based on similarity. 54

    Chapter 5: Simulation results. 57

    5-1 Database. 59

    5-2 Simulation results. 62

    5-3 test results. 63

    4-5 An example of the results. 84

    5-5 Conclusions and future work. 88

    References. 90

    Source:

    Pourjunidi M, (2008), senior thesis, "Presentation of a new method based on color and fuzzy edge characteristics in image recovery", Faculty of Electrical, Computer and Information Technology, Qazvin Islamic Azad University.

    [1]

    [2]

    M. Mehrdad, and H. Ebrahimnezhad, "3D model retrieval using linear prediction coding descriptor", Iranian Conference on Electrical Engineering, (ICEE), pp. 721-725, 2012.

    [3]

    D. Frejlichowski, "A three-dimensional shape description algorithm based on polar-fourier transform for 3D model retrieval", Image Analysis, Springer, pp. 457-466, 2011.

    [4]

    A. Khatun, W. Chai, and M. Islam, "An Ellipsoidal 3D Shape Representation and Wavelet Transform Feature Descriptor For 3D Shape Retrieval", Asian Jornal of Information Technology, Vol. 9, pp. 101-106, 2010. [5] W. Mohamed, and A. Hamza, "Reeb graph path dissimilarity for 3DHamza, "Reeb graph path dissimilarity for 3D object matching and retrieval", The Visual Computer, Vol. 28, pp. 305-318, 2012. [6] K. Zou, W. Ip, C. Wu, Z. Chen, K. Yung, and C. Chan, "A novel 3D model retrieval approach using combined shape distribution", Multimedia Tools and Applications, pp. 1-20, 2012.

    [7]

    L. Li, H. Wang, T. Chin, D. Suter, and S. Zhang, "Retrieving 3D CAD models using 2D images with optimized weights", In Image and Signal Processing(CISP), 3rd International Congress on IEEE. Vol. 4, pp.1586-1589, 2010.

    [8]

    C. Tangelder, and R. Veltkamp, ??"A survey of content based 3D shape retrieval methods", Multimedia Tools and Application, Vol.39, pp. 441-471, 2008.

    [9]

    A. Masaki, and H. Iwabuchi. "3D Shape Retrieval from a 2D Image as Query", Signal & Information Processing Association Annual Summit and Conference (APSIPA ASC), pp. 1-10, 2012.

    [10]

    T. Napoleon, and I. Sahbi, "From 2D Silhouettes to 3D Object Retrieval: Contributions and Benchmarking", EURASIP Journal of Image and Video Processing, Vol. p. 17, 2010.

    [11]

    T.F. Ansary, J. Vanderborre, and M. Daoudi, "3D-Model Search Engine from Photos", ACM International Conference on Image and Video Retrieval(CIVR), pp. 89–92, 2007.

    [12]

    B. Bustos, D. Keim, D. Saupe, and T. Schreck, "Content-Based 3D Object Retrieval", IEEE Computer Graphics and Applications, Vol. 27, pp. 22–27, 2007. T. F. Ansary, M. Daoudi, and A. Vandeborre, “A Bayesian 3D Search Engine using Adaptive Views Clustering”, IEEE Transactions on Multimedia, Vol. 9,

    No. 1, pp. 78–88, 2007.

    [14]

    J. Muwei, D. Junyu, and T. Ruichun, "Image Combining Color, Texture and Region Whit Objects of Users Interest for Content-Based Image Retrieval", Eighth ACIS International Conference on Software Engineering, Artificial Intelligence, Networking and Parallel/Distributed Computing, IEEE Computer Society, Vol. 1, pp. 764-769, 2007. [15] Z. Huang, P.K. Chen, W.Y. NG, and S. Yeung, "Content-Based Image Retrieval Using Color Moment and Gobor Texture Feature", Proceeding of the Ninth International Conference on Machine Learning and Cybernetics, Qingdao, Vol. 2, pp. 719-724, 2010.

    [16]

    P.K. Manesh Kokare, and B.N. Biswas, "Texture - Image Retrieval Using New Rotated Complex Wavele - Filters", IEEE Transactions on Systems, Vol. 35, No. 6, pp. 1168-1178, 2005.

    [17]

    L. Chen, G. lu, and D. Zhang, "Effects of Different Gobor Filters Parameters on Image Retreval by Texture", In Multimedia Modeling Conference, pp. 273_278, 2004.

    [18]

    F. Long, H. zhang, and D. Feng, "Fundamentals of Content-Based Image Retrieval", in Multimedia Information Retrieval and Management - Technological - Fundamentals and Applications," Springer - Verlag, pp. 1-26, 2003.

    [19]

    R. Alaoui, S.O.E. Alaoui, and M. Meknassi, "An Efficient Similarity Measure(PMM) for Color-based Image Retrieval", IJCSNS International Journal of Computer Science and Network Security, VOL. 8, No. 7, 2008. D. Zhang, and G. "Study and Evaluation of Different Fourier Methods for Image Retrieval, pp. 33-49, 2005. R.C. Veltkamp, ??and M. Hagedoorn, In Principles of Visual Information Retrieval, M. Lew, pp. 87-119. 2001. [23] D. Zhang, and G. Lu, “Review of Shape Representation and Description Techniques”, Pattern Recognition, pp. 1-19, 2004.

    [24]

    S. Aksoy, and R.m.

Retrieving images of fighter planes based on 3D model