Coronal loop detection from solar images and extraction of salient contour groups from cluttered images.

Curve tracing;Feature extraction;Contour grouping;Pattern recognition;Coronal loops;Image retrieval

Nurcan Durak

University:University of Louisville

Department:Computer Engineering and Computer Science

关键词: Curve tracing;    Feature extraction;    Contour grouping;    Pattern recognition;    Coronal loops;    Image retrieval;   

Others  :  https://ir.library.louisville.edu/cgi/viewcontent.cgi?article=1385&context=etd

美国|英语

来源: The Universite of Louisville's Institutional Repository

【 摘 要 】

This dissertation addresses two different problems: 1) coronal loop detection from solar images: and 2) salient contour group extraction from cluttered images. In the first part, we propose two different solutions to the coronal loop detection problem. The first solution is a block-based coronal loop mining method that detects coronal loops from solar images by dividing the solar image into fixed sized blocks, labeling the blocks as "Loop" or "Non-Loop", extracting features from the labeled blocks, and finally training classifiers to generate learning models that can classify new image blocks. The block-based approach achieves 64% accuracy in IO-fold cross validation experiments. To improve the accuracy and scalability, we propose a contour-based coronal loop detection method that extracts contours from cluttered regions, then labels the contours as "Loop" and "Non-Loop", and extracts geometric features from the labeled contours. The contour-based approach achieves 85% accuracy in IO-fold cross validation experiments, which is a 20% increase compared to the block-based approach. In the second part, we propose a method to extract semi-elliptical open curves from cluttered regions. Our method consists of the following steps: obtaining individual smooth contours along with their saliency measures; then starting from the most salient contour, searching for possible grouping options for each contour; and continuing the grouping until an optimum solution is reached. Our work involved the design and development of a complete system for coronal loop mining in solar images, which required the formulation of new Gestalt perceptual rules and a systematic methodology to select and combine them in a fully automated judicious manner using machine learning techniques that eliminate the need to manually set various weight and threshold values to define an effective cost function. After finding salient contour groups, we close the gaps within the contours in each group and perform B-spline fitting to obtain smooth curves. Our methods were successfully applied on cluttered solar images from TRACE and STEREO/SECCHI to discern coronal loops. Aerial road images were also used to demonstrate the applicability of our grouping techniques to other contour-types in other real applications.