Download Classification Based Multi Sensor Image Fusion and more Study Guides, Projects, Research Applications of Computer Sciences in PDF only on Docsity!
Section 1
1. Introduction
This document provides the System Requirement Specifications of the project titled as Classification based multi-sensor image fusion which will be carried out by the student of BSCIS at Pakistan Institute of Engineering and Applied Sciences as the Final Year Project. The document is divided into different sections. This section introduces the project requirements by defining the purpose, scope, goals and intended audience for the project. It also includes some definitions and abbreviations used in the document. Finally it provides an overview of the document which presents the organization of the rest of the document.
1.1 Purpose
The purpose of this document is to specify detailed requirements of the project Classification based multi-sensor image fusion. The document will serve as a monitoring tool for the supervisor throughout the project development as it contains all of the requirements of the project. The developer can use this document to track the project progress throughout t he working period. The document includes overall description of the project, its functional and non functional requirements. The document will be the first deliverable of the project. So it will serve as the first mean of communication between the developer and supervisor.
1.2 Scope of Research
Multi- sensor image fusion is the process of combining information in two or more images of a scene captured from different sensors to enhance viewing or understanding of the scene. With the availability of multi-sensor data in many fields such as remote sensing, computer vision, medical, and army applications, image fusion has emerged as a new and important research area [7]. Human eye can only sense in visible region of electromagnetic spectrum. Different sensors (e.g. thermal) are used to obtain the data (images) that human eye cannot see. The data from different sources (sensors) can be combined to produce useful information which represents features of all sources.
Image fusion is a broad field of research. Image fusion can be performed at different levels of information as mentioned below [1]. Signal level fusion Pixel level fusion Feature level fusion Symbol level fusion The developers will work on the one specific area of the field which is Classification based image fusion. However other fields will also be studied to gain comprehensive knowledge about the field. All image fusion techniques have some common objectives which are mentioned below. Extract all the useful information from the source images Do not introduce artifacts or inconsistencies which will distract human observers or the following processing. Reliable and robust to imperfections such as improper registration [4] The first step in image fusion techniques is image registration. So during the pro ject, some image registration techniques (like control point based image registration) will be studied. After image registration, an image fusion technique is applied at source images. A number of image fusion techniques have been developed during different researches. Some of them are given below. [6] Image level weighed Average (ILWA) False color technique Pyramidal techniques o Difference of low pass (DoLP) o Ratio of low pass (RoLP) o Gradient Principle component fusion Wavelet based techniques o Discrete wavel et transform method o Shift invariant DWT method
ANN Artificial Neural Networks GA Genetic Algorithms GP Gene tic Programming IR Infrared MRI Magnetic Resonance Imaging
Table1: List of Abbreviations used in the document
1.5 Definitions
Some important definitions that are relevant to the document are listed in the table below.
Term Definition
Image Fusion Imag^ e fusion is the process of combining information in two or more images of a scene captured from different sensors (or same) to enhance viewing or understanding of the scene Preprocessing Performing activities like enhancement, resizing and noise removal on an image to prepare it for further processing. Image Registration Geometric transformation process to map the images to a common coordinate system. [2] ANN An information processing system that has certain performance characteristics in common with b iological neural networks [3] SVM A tool for pattern classification. GA A search technique, learning paradigm and a machine analog of biological process.[5]
Table2: List of terms used in the document and their definitions
1.6 References
[1] : URL: http: // www.imagefusion.org [2] : Scott E Umbaugh, Computer Imaging, Digital image analysis and processing , A CRC press book. [3] : Laurene Fausett, Fundamentals of Neural Networks Architectures, Algorithms and Applications , Pearson Education. [4] : URL: http://www.ece.lehigh.edu/SPCRL/IF/image_fusion.htm
[5] : Mengj ei Zhang, Vicotria , Overview of Genetic Algorithms , COMP 422 Lecture Notes, University of Wellington , 2006. [6] : Captain Éric J.P. Lallier, Real time pixel level image fusion through adaptive weig ht averaging , Master Thesis, Royal Military College of Canada , April 1999. [7] : Yoel Lanir, Comparing Multispectral Image Fusion Methods for a Target Detection Task , Ben -Gurion University of the Negev , June 2005. [8] : Adeel Mumtaz, S.A.M Gilani, Tahir Jameel, A Novel Image Retrieval System Based On Dual Tree Complex Wavelet Transform and Support Vector Machines , CISSE 2006, Bridgeport USA, December 4, 2006. [9] : John R. Koza , Survey of Genetic Algorithms and Genetic Programming , Computer Science Dept , Stanford University
1.7 General Information about Project
1.7.1 Project team information
The project will be carried out by one student and under the supervision of one faculty member. There will be a panel of faculty members who will monitor and evaluate the work. The names of project team are given below. Student Name Aqeel Mumtaz Project Supervisor Dr. Abdul Majid Project Examiners Dr. Mutawarra Hussain Dr. Abdul Jalil Dr. Muhammad Arif Dr. Javaid Khurshid Mr. Syed Muhammad Haroon
1.7.2 Project Durat ion
The project is to be completed by July 2008.
1.7.3 Project Deliverables
The list of project deliverables is given below.
Section 2
2. Overall Description
The general factors that affect the project will be discussed in this section of the document. The section is further divided into following subsections. Research perspective System architecture User Interfaces User characteristics Product functions Constraints Assumptions and dependencies The above subsections are discussed one by one below.
2.1 Research Perspective
Multi- sensor image fusion can be performed at different levels of information as discussed in the scope section earlier. Different methods of multi-sensor image fusion will be studied and implemented, and on that basis, project is considered to be research and development project. As the title of the project suggests, the main focus will be on the classification based multi-sensor image fusion techniques. For that purpose, classification techniques like ANN, SVM, GA and GP will be studied and applied on image fusion. ( Detail of these techniques is mentioned in Appendix A) After all the research work, a final product (GUI application) will be developed which will perform image fusion of given images through all the implemented techniques. Brief description of the product is given in the product functions and user interfaces sections.
2.2 System Architecture
The system architecture is described below. Input images will be selected Images will be registered One out of different fusion techniques will be chosen
The selected f usion technique will be executed to carryout fusion process The resultant fused image will be available. Performance evaluation will be performed The pictorial representation of the system is shown below.
Figure1: Pictorial re presentation of the system
The above diagram represents the system architecture at very abstract level. As indicated, the system can accept up to N input images taken from different sensors. The input images will be of the same scene. These images will be first preprocessed. The preprocessing phase includes activities like enhancement, resizing, noise removal etc. For these purposes some sort of filters will be applied on images.
Figure 2 : Main Graphical User Interface for the project
The user can browse through his machine directory structure to select the input images. This functionality is served by the Load Image A and Load Image B push button controls. A usual file open dialog box appears which aids user to select his input image file. After the user locates the input image file, image is displayed in the corresponding axes control in the panel at the left. The following screenshot shows the input images displayed in their corresponding locations in the GUI.
Figure 3 : Input images loaded in GUI
When the user has selected all other parameters, he can perform fusion of i nput images and performance analysis. One sample of the GUI after the fusion and performance analysis process is show below.
2.4 User Characteristics
The users of the system are divided into different categories based on their occupation
2.4.1 Military Users
There are different kinds of applic ations of image fusion in military. Image fusion is used on battle fields for target detection. This task is achieved through helmet mounted display for soldiers. This display is generated through fusion of thermal and visible images. These soldiers expect fusion process to help them identifying their targets successfully. Another kind of military users of image fusion are helicopter pilots. Their requirement is to get better view of the scene in darkness or bad whether situation. Police uses image fusion for concealed weapon detection. They need to identify the suspect correctly because mostly they are dealing with more than one person at a time.
2.4.2 Medical Users
Image fusion also has its applications in medical field. Medical personnel encounters with t he situation where they need to fuse images taken from different sensors like ultra - sound and MRI images for the purpose of the disease analysis and diagnosis.
2.4.3 Robotics Users
Image fusion is also used in robotics. Robots often are designed where they need to fuse images taken from different sources.
2.4.4 Mineral Resource Detectors
They use the fusion of images to estimate the existence of the minerals inside the earth or mountains. Their requirement is to get accurate estimate as they would have to discover resources after detection, so if detection is performed incorrectly then it could result in failure and waste of time and resources.
2.5 Product Functions
The product will provide following functionalities. Input images selection Registration of t he input images Fusion method selection
Fusion method execution Displaying and saving output (fused) image Performance evaluation and comparison
2.6 Constraints
The following constraints have been identified for the project.
2.6.1 Image registration Const raint
Before any algorithm execution, input images must be properly registered.
2.6.2 Implementation Constraints
All the implementations will be carried out in Matlab. So all of the source code will be only Matlab compatible and the source files will be o f the extension .m.
2.6.3 GUI constraints
GUI can only be run through Matlab. All the associated source files must be present to get the full functionalities through the GUI.
2.7 Assumptions and Dependencies
Assumptions and dependencies of the system a re given below.
2.7.1 Hardware
Different sensors will be required to capture the multi-sensor images. The client s machine should be good enough so that Matlab can run on it.
2.7.2 Software
Matlab should be installed and functional on both the client s and developer s machine.
Drop down list C ontrols A drop down list control for each image fusion algorithm selection and image registration method selection will be provided. Parameters adjustment controls Different types of controls for parameters adjustment will be provided.
3.1.3 Image Preproc essing Functionalities
Input images can be preprocessed for noise removal, enhancement or resizing purposed. These functionalities will be provided.
3.1.4 Image Registration Functionality
There exist different techniques for image registration. System will provide different registration methods and the user can select one of them for his input images.
3.1.5 Image Fusion Algorithms
Different fusion algorithms will be available to apply on the input images. Few examples include: Image level weighed Average ( ILWA) False color technique Pyramidal techniques o Difference of low pass (DoLP) o Ratio of low pass (RoLP) o Gradient Principle component fusion Wavelet based techniques o Discrete wavelet transform method o Shift invariant DWT method
3.1.6 Development of new techn iques
New improved techniques for Classification based Multi-sensor image fusion will also be developed. These techniques will be based on the classification techniques like ANN, SVM, GA and GP. Results will be compared with existing techniques.
3.1.7 Performance Evaluation and Comparisons Functionalities
Performance of the selected fusion algorithm will be measured through some quality measurement factors. The performance comparison functionality of two algorithms applied on same input will also be avail able and the results would be shown in a graph.
3.2 Non Functional Requirements
Non functional requirements are those which do not provide any functionality in the system; however these are necessary to be met for the success of the project. Non functiona l requirements of the system are given below.
3.2.1 Reliability
The developed system should be reliable. It should handle any type of input. It should not crash. It should be able to apply all of the fusion algorithms on a given set of input images. It should perform performance evaluation and comparisons of any of the implemented methods.
3.2.2 Accuracy
The system should compute the resultant (fused) image and performance evaluation parameters correctly.
3.2.3 Security
The system should not change the input images.
3.2.4 Usability
The system should be constructed in such a manner that it should be easy to use and understand for the users. The users should not feel any difficulty in learning the system.
3.2.5 Scalability
The system should be scalable. It should be designed in such a way so that it can be easily extended in future. The extension could be the inclusion of new fusion methods or new registration techniques etc.
Section 4
4. Project Plan
The entire work of the project is semester- wised distributed. The work breakdown structure of the project is given below.
4.1 Semester 6 Activities
Literature Survey on Image Fusion o Introduction o Application o Finding Research groups o Studying Review/Survey Papers o Study & Implementing Basic/Well known techniques Literature Survey on IR Images o Introduction to IR images o Studying types of IR images o Study of IR technologies (IR cameras, hardware etc) o Iden tifying the d ifference between IR and Visible images
4.2 S ummer break Activities
Literature Survey on Multi- Sensor Image Fusion o Applications of Multi - Sensor Image Fusion o Study of well known techniques on Multi-Sensor Image Fusion and their implementation Classification Based Multi - Sensor Image fusion o Overall idea and understanding o Learning classification tools such as ANN, SVM, GA and GP o Study of Basic techniques used in this area and their implementation
4.3 S emester 7 Activities
Study of different IR and Visible Image Fusion techniques. Implementation of existing techniques. Testing results of existing techniques. Development of new algorithm. (optional)
Implementation of developed algorithm. (optional) Testing of new algorithm. (optional)
4.4 Semester 8 A ctivities
Comparison of implemented techniques. Analysis of these techniques. Research paper writing. (optional) Comparison paper. Project report writing.
