الفهرس 
Problem StatementOnly 14 pages are availabe for public view
 |  | from | 97 |  |  |
| | | from | 97 | | |
Abstract
The phrase ”anomaly detection techniques” is often used to describe any technique that looks for samples that differ from expected patterns. Depending on availability of data labels, types of abnormalities and applications, many anomaly detection techniques have been developed. This thesis presents different approaches for anomaly detection from images as well as other patterns. It also describes the fundamental anomaly detection techniques, as well as their modifications and importance. Landmine detection and brain tumor detection are two different applications that are discussed for anomaly detection techniques. The first approach in this thesis is concerned with two algorithms for landmine detection from Ground Penetrating Radar (GPR) images. The first algorithm depends on a multi-scale technique. A Gaussian kernel with a particular scale is convolved with the image, and after that, two gradients are estimated; horizontal and vertical gradients. Then, histogram and cumulative histogram are estimated for the overall gradient image. The bin values on the cumulative histogram are used for discrimination between images with and without landmines. This algorithm shows a promising landmine detection performance with a 92% success rate. The results reflect the possibility of detecting landmines with histogram bins. Some missing landmines are attributed to the close values of bins at the start and end of cumulative histograms of images with and without landmines. The second algorithm is based on scale-space analysis with the number of Speeded-Up Robust Feature (SURF) points as the key parameter for classification. selecting an appropriate threshold for the number of SURF points can make the detection process easy with a success rate of 100%. No false alarms are recorded with this algorithm. In addition, this thesis presents a framework for size reduction of GPR images based on decimation for efficient storage. An interpolation scheme can be used to reconstruct the landmine images with their original sizes prior to any landmine detection process. Simulation results reveal that landmine detection is not affected by the decimation and interpolation processes.