Detecting regions of change while reducing unimportant changes in multiple outdoor images of the same scene containing fence wires (i.e., a chain-link mesh fence) acquired by a mobile camera from slightly different viewing positions, angles and at different times is a very difficult problem. Regions of change include appearing of new objects and/or disappearing of old objects behind fence wires, breaches in the integrity of fence wires and attached objects in front of fence wires. Unimportant changes are mainly caused by camera movement, considerable background clutter, illumination variation, tiny sizes of fence wires and non-uniform illumination that occurs across fence wires. There are several issues that arise from these kinds of multiple outdoor images. The issues are: (1) parallax (the apparent displacement of an object as seen from two different positions that are not on a line with the object) among objects in the scene, (2) changing in size of same objects as a result of camera movement in forward or backward direction, (3) background clutter of outdoor scenes, (4) thinness of fence wires and (5) significant illumination variation that occurs in outdoor scenes and across fence wires. In this dissertation, an automated change detection method is proposed for these kinds of multiple outdoor images. The change detection method is composed of two distinct modules, which are a module for detecting object presence and/or absence behind fence wires and another module for detecting breaches in the integrity of fence wires and/or attached objects in front of fence wires. The first module consist of five main steps: (1) automated image registration, (2) confidence map image production by the Zitnick and Kanade algorithm, (3) occlusion map image generation, (4) significant or unimportant changes decision by the first hybrid decision-making system and (5) false positives reduction by the template subtraction approach. The second module integrates: (1) the Sobel edge detector combined with an adaptive thresholding technique in extracting edges of fence wires, (2) an area-based measuring in separating small and big objects based on their average areas determined once in the calibration process and (3) the second hybrid decision-making system in classifying objects as significant or unimportant changes. Experimental results demonstrate that the change detection method can identify and indicate approximate locations and possible percentages of significant changes whilst reducing unimportant changes in these kinds of multiple outdoor images. The study has utilized occluded regions in a confidence map image produced by the Zitnick and Kanade algorithm as potential significant changes in the image change detection research. Moreover, the study proves that the use of the Sobel edge detector combined with an adaptive thresholding technique is applicable in extracting edges of outdoor fence wires. In the future, the method could be integrated into patrol robots in order to provide assistance to human guards in protecting outdoor perimeter security. / http://proxy.library.adelaide.edu.au/login?url= http://library.adelaide.edu.au/cgi-bin/Pwebrecon.cgi?BBID=1522689 / Thesis (Ph.D.) -- University of Adelaide, School of Mechanical Engineering, 2010

Context: We identify a material or thing that can be seen and touched in the world as having structures at both coarser and finer levels of scale. Scaling problem presents in a branch of science concerned with the description, prediction understanding of natural phenomena and visual arts. A moon, for instance, may appear as having a roughly round shape is much larger than stars when seen from the earth. In the closer look, the moon is much smaller than the stars. The fact that objects in the world appear in different ways depending upon the scale of observation has important implications when analyzing measured data, such as images, with automatic methods [1]. The type of information we are seeking from a one-dimensional signal or two-dimensional image is only possible when we have the right amount of scale for the structure of an image or signal data. In many modern applications, the right scale need not be obvious at all, and we all need a complete mathematical analysis on this scaling problem. This thesis is shown how a mathematical theory is formulated when data or signal is describing at different scales. Objectives: The subtle patterns deforming in data that can foretell of a scaling problem? The main objectives of this thesis are to address the dynamic scaling pattern problem in computers and study the different methods, described in the latest issue of Science, are designed to identify the patterns in data. Method: The research methodology used in this thesis is the Fractional Fourier Transform. To recognize the pattern for a different level of scale to one or many components, we take the position and size of the object and perform the transform operation in any transform angle and deform the component by changing to another angle which influences the frequency, phase, and magnitude.  Results: We show that manipulation of Fractional Fourier transform can be used as a pattern recognition system. The introduced model has the flexibility to encode patterns to both time and frequency domain. We present a detailed structure of a dynamic pattern scaling problem. Furthermore, we show successful recognition results even though one or many components deformed to different levels using one-dimensional and two-dimensional patterns. Conclusions: The proposed algorithm FrFT has shown some advantages over traditional FFT due to its competitive performance in studying the pattern changes. This research work investigated that simulating the dynamic pattern scaling problem using FrFT. The Fractional Fourier transform does not do the scaling. Manipulating the Fractional Fourier transform can be helpful in perceiving the pattern changes. We cannot control the deformation but changing the parameters allow us to see what is happening in time and frequency domain.

This thesis proposes a Visual Scene Change Detector(VSCD) system which is a system which involves four parts, image retrieval, image registration, image change detection and panorama creation. Two prestudies are conducted in order to find a proposed image registration method and a image retrieval method. The two found methods are then combined with a proposed image registration method and a proposed panorama creation method to form the proposed VSCD. The image retrieval prestudy evaluates a SIFT related method with a bag of words related method and finds the SIFT related method to be the superior method. The image change detection prestudy evaluates 8 different image change detection methods. Result from the image change detection prestudy shows that the methods performance is dependent on the image category and an ensemble method is the least dependent on the category of images. An ensemble method is found to be the best performing method followed by a range filter method and then a Convolutional Neural Network (CNN) method. Using a combination of the 2 image retrieval methods and the 8 image change detection method 16 different VSCD are formed and tested. The final result show that the VSCD comprised of the best methods from the prestudies is the best performing method. / Detta exjobb föreslår ett Visual Scene Change Detector(VSCD) system vilket är ett system som har 4 delar, image retrieval, image registration, image change detection och panorama creation. Två förstudier görs för att hitta en föreslagen image registration metod och en föreslagen panorama creation metod. De två föreslagna delarna kombineras med en föreslagen image registration och en föreslagen panorama creation metod för att utgöra det föreslagna VSCD systemet. Image retrieval förstudien evaluerar en ScaleInvariant Feature Transform (SIFT) relaterad method med en Bag of Words (BoW) relaterad metod och hittar att den SIFT relaterade methoden är bäst. Image change detection förstudie visar att metodernas prestanda är beroende av catagorin av bilder och att en enemble metod är minst beroende av categorin av bilder. Enemble metoden är hittad att vara den bästa presterande metoden följt av en range filter metod och sedan av en CNN metod. Genom att använda de 2 image retrieval metoder kombinerat med de 8 image change detection metoder är 16 st VSCD system skapade och testade. Sista resultatet visar att den VSCD som använder de bästa metoderna från förstudien är den bäst presterande VSCD.