NASA-AISRP

        Solar Loop Mining for the Coronal Heating Problem

 

 

Home

People

Publications

Software

Data sets

Links

   Demo

 

             DOWNLOAD IMAGE MINING TOOL!!!

HOW TO USE IMAGE MINING TOOL(documentation)

SOlar LOOp Mining System

The SOLOOM System is the software product of the research project. It is an image retrieval and mining system, which is able to sift through massive data sets downloaded from online NASA solar image databases and automatically discover the rare but interesting images containing solar loops.

The system is divided in three main subsystems:

  • Image Acquisition Subsystem
  • Multi-stage classifier training subsystem
  • Loop Mining Subsystem

The global architecture of the system is shown in the following figure:

systemStructure

 

Image Acquisition Subsystem

The objective of this subsystem is to provide tools that support the task of retrieving image files from the EIT online database, and facilitate the identification and marking of solar loops. The subsystem is composed of two main tools:

  • Automatic downloading tool: is a java program which can query and download EIT images from the NASA website.
  • Loop Interactive Marking Tool: tool for marking solar loops on Fits images using the ImageJ environment.

The architecture of the subsystem is shown in the following figure:

ImageAcquastion

Automatic Download Tool

This is a java program which can query and download EIT images from the NASA website. The user manual could be found in EITDownloadManual.

This is a snapshot of the program:

AutomaticDownloadTool

 

Loop Interactive Marking Tool

A tool for marking solar loops on Fits images using the ImageJ environment. It has been implemented as an ImageJ plugin that allows the user to:

  • Select ROIs (Regions Of Interest) containing solar loops
  • Edit the selected ROIs
  • Apply any transformation to the image using the ImageJ tools
  • Save the image with the solar loops marked rois information in Fits format
  • Load the marked ROIs of solar loops previously stored in the Fits image

 

loopMarking

 

Multi-stage classifier training subsystem

The goal of this subsystem is to produce classifier models, which are able to discriminate image block that contain loops from those that don't. We have experimented with a process consisting of two main stages depending on the features used:

  • Low level features: such as intensity, texture, etc.
  • High-level semantic structures (solar loops): such as curvature features, spatial features, Hough Transform based features

 The architecture of the subsystem is shown in the following figure:

classifierSystem

Loop Mining Subsystem

This subsystem, along with the Image Acquisition Subsystem, constitutes the production part of the overall system. It uses the classifiers trained by the Multi-stage Classifier Training Subsystem to identify and report those blocks that contain solar loops.

The architecture of the subsystem is shown in the following figure:

LoopMiningSubsystem

 

 

Testing Tool

 

This subsystem retrieves the images that have coronal loops on the limb. Users can load the images to the tool. After that, block extraction and feature extraction are applied on images. A classifier model is applied on the features of the blocks and the regions are classified into either loop region or no-loop region. For each image, we look at the predicted loop labels to decide whether the image has a loop.

 

 

Testing

 

 

Sample Output: The red rectangle shows the automatically detected loop area on the solar limb

 

Magnified Regions