Interval Type-2 Fuzzy Control System and Method Applying Active Contour Models

Aurora Oval Image Segmentation System Combining Type-2 Fuzzy and Active Contour Models

This technology is an image segmentation system that distinguishes between aurora regions and background areas in oval-shaped aurora images captured by imaging equipment. It calculates the upper and lower membership values for each pixel using interval type-2 fuzzy logic, performs type reduction, and then uses an active contour model to perform hard partitioning of the pixels.

Aurora images are vital data for ionospheric and magnetospheric research, as well as for analyzing impacts on power grids and communications. However, they have high uncertainty due to blurred boundaries and significant brightness variations. Conventional pixel intensity-based detection methods have struggled to effectively distinguish between the background and the aurora oval.

This technology represents boundary ambiguity using the uncertainty membership values of type-2 fuzzy sets and incorporates defuzzified prototypes into a fuzzy energy function, resulting in segmentation that is less sensitive to initial contour settings. It can be applied to space weather analysis based on satellite and ground observations, as well as image analysis software for polar atmospheric research, enabling consistent extraction of aurora regions with blurred boundaries.

‍

‍

Key Features:

• An image acquisition unit that captures oval-shaped aurora images using imaging equipment and determines the region membership of each pixel
• A fuzzy logic unit that calculates the degree to which each pixel belongs to the aurora oval region and the background region as first and second membership values using membership functions
• A type reduction processing unit that performs type reduction on the first and second membership values and defuzzifies them by estimating the right and left values of the prototype
• An active contour model unit that performs hard partitioning of pixels into aurora oval and background regions based on membership values and distinguishes the image using a fuzzy energy function

‍