从UVI图像中提取极光椭圆边界：一种基于FLICM聚类的新方法及其评估资源-CSDN文库

106 浏览量 2021-03-01 16:47:47 上传评论收藏 772KB PDF 举报

标题《从UVI图像中提取极光椭圆边界：一种基于FLICM聚类的新方法及其评估》中涵盖的知识点主要包括： 1. 极光椭圆边界的提取：极光椭圆边界指的是极光显示在地磁极附近的环状形态，它们通常呈现为一个围绕地球磁极的环形区域，被称为极光椭圆。研究这一边界对于极光研究及空间天气预测具有重要意义。 2. UVI（Ultraviolet Imager，紫外线成像仪）数据的应用：UVI是一种常装备于卫星上的成像设备，用于捕捉地球极光的紫外图像。该数据可提供全球范围内的极光信息，与地面观测相比，它能够提供更加全面的极光观测图像。 3. 极光椭圆边界提取的新方法：该部分涉及使用基于模糊局部信息c均值（Fuzzy Local Information c-means，简称FLICM）聚类算法的迭代分割方法来提取极光椭圆的赤道侧和极侧边界。 4. 极光椭圆完整性准则：极光椭圆的迭代分割基于完整性准则，此准则旨在确保分割出来的极光椭圆形状保持完整性，不会有不必要的断层或者缺失。 5. 形状知识填补方法：在提取极光椭圆边界后，可能会出现一些间隙，利用极光椭圆的先验形状知识进行填补，保证边界的整体性和准确性。 6. 边界提取方法的评估：通过将提取出来的极光椭圆边界与DMSP（Defense Meteorological Satellite Program，国防气象卫星计划）卫星上的沉降电子边界数据进行对比，进行客观评估。DMSP卫星提供的粒子沉降数据可以用于确定极光边界，与提取方法得出的结果进行比对。 7. FLICM聚类算法：FLICM是一种聚类算法，用于图像分割和数据聚类。它是一种改进的c均值聚类算法，它考虑了数据点的局部信息，使得聚类结果更加准确。文章利用FLICM算法对极光图像进行处理，迭代地提取出椭圆边界的图像。 8. 极光观测的科学意义：极光不仅是自然现象，也是空间物理研究的重要对象。极光与太阳活动密切相关，太阳风中的高能粒子与地球磁场相互作用，进入上大气层与大气分子或原子碰撞产生光辐射。通过对极光的研究可以帮助科学家更好地理解太阳风、地球磁场、大气等自然现象的相互作用，这对于空间天气预报和地球环境变化的研究具有深远意义。 9. 卫星遥感技术：卫星遥感技术在极光观测中发挥着重要作用。通过安装在卫星上的各种传感器和成像设备，科学家可以从地球的外层空间收集到地球大气层的极光等信息，进行全球观测和研究。该研究论文通过FLICM聚类算法的应用，发展了一种从POLAR卫星上紫外线成像仪UVI获取的图像中提取极光椭圆边界的崭新方法，并通过与DMSP卫星数据的对比，评估了该方法的准确性，展示了其比传统方法更为精确的优点。同时，文章也为极光椭圆的形状特征和全球分布情况提供了科学的观测方法和分析手段。

资源推荐

资源详情

资源评论

· Article · Advances in Polar Science

doi:10.3724/SP.J.1085.2011.00184 September 2011 Vol.22 No.3 184–191

Extraction of auroral oval boundaries from UVI

images: A new FLICM clustering-based method

and its evaluation

WANG Qian

, MENG QingHu

, HU ZeJun

2∗

, XING ZanYang

2,3

, LIANG JiMin

1∗

& HU HongQiao

School of Life Sciences and Technology, Xidian University, Xi’an 710071, China;

SOA Key Laboratory for Polar Science, Polar Research Institute of China, Shanghai 200136, China;

School of Science, Xidian University, Xi’an 710071, China

Received June 28, 2011; accepted August 16, 2011

Abstract Based on the fuzzy local information c-means (FLICM) clustering algorithm, a new method is de-

veloped for extracting the equatorward and poleward boundaries of the auroral oval from images acquired by the

Ultraviolet Imager (UVI) aboard the POLAR satellite. First, the method iteratively segments the UVI image with

the FLICM clustering algorithm, according to an integrity criterion for the segmented auroral oval. Then, possible

gaps in the extracted auroral oval are ﬁlled, based on prior knowledge of its shape. To evaluate the method objec-

tively, the extracted boundaries are compared with the precipitating electron boundaries determined from DMSP

satellite precipitation particle data. The evaluation results demonstrate that the proposed method generates more

accurate auroral boundaries than traditional methods.

Keywords Auroral boundary, UVI, image segmentation, evaluation

Citation: Wang Q, Meng Q H, Hu Z J, et al. Extraction of auroral oval boundaries from UVI images: A new FLICM Cl-

ustering-based method and its evaluation. Adv Polar Sci, 2011, 22: 184–191, doi: 10.3724/SP.J.1085.2011.00184

0 Introduction

Aurorae are natural lights in the upper atmosphere that

are produced by collisions between energetic particles

precipitating from space with atoms or molecules in the

upper atmosphere. Observations from remote imager in-

struments aboard satellites usually show the lights to

have the shape of an annular ring centered on the Earth’s

magnetic poles, and they are called auroral ovals

[1−2]

Compared with local auroral observations from ground-

based cameras, auroral images acquired by the Ultravi-

olet Imager (UVI) aboard the POLAR satellite can pro-

vide global auroral information, e.g., the overall conﬁg-

uration of the oval and spatial distribution of auroral

intensity along the oval. The equatorward and poleward

boundaries of the auroral oval in UVI images, referred to

as UVI boundaries, are closely related to the energy cou-

pling between solar wind and the magnetosphere. The

boundaries expand and contract in response to geomag-

netic and solar wind activity

[3]

. For auroral study, it is

important to accurately extract the boundaries of the au-

roral oval.

To make full use of the large number of UVI im-

ages, automatic image segmentation methods for extract-

ing auroral oval boundaries are of great appeal, because

manually determining the auroral oval of so many images

is tedious and impractical. Several methods have been

∗

Corresponding authors: (Hu Zejun, email: huzejun@pric.gov.cn; Liang Jimin, email: jimleung@mail.xidian.edu.cn)

journal.polar.gov.cn

Extraction of auroral oval boundaries from UVI images 185

presented in recent decades. These include threshold-

based methods

[4−5]

, a histogram-based k-means

algorithm

[6]

, an adaptive thresholding technique

[7]

a pulse-coupled neutral network

[3]

and shape-based

methods

[8−9]

. The methods of references[4-7] do not

exploit auroral oval shape information and the spatial

relationship of pixels. Therefore, they are unable to

detect the complete auroral oval for some images, espe-

cially those with large auroral intensity variations. Cao

et al.

[8−9]

presented a shape-based method for extracting

auroral ovals. They ﬁrst segmented auroral images using

the algorithm of Li et al.

[7]

, then selected some inner and

outer boundary points from the resultant foreground,

via radial-based processing. Finally, two ellipses ﬁtted to

the points using linear least-squares randomized Hough

transform

[9]

were used as the equatorward and poleward

boundaries, respectively. Although this method pro-

duces complete boundaries of the auroral oval, it may

introduce errors if the boundaries are not strictly ellipti-

cal, or if only parts of the auroral oval are imaged.

In this paper, a new approach for extracting the

UVI boundaries is proposed, based on fuzzy local infor-

mation c-means clustering (FLICM)

[10]

. In this method,

the UVI image is iteratively segmented according to an

integrity measurement of the extracted auroral oval us-

ing FLICM, then auroral oval gaps are ﬁlled based on

prior knowledge of its shape. The method is not only

applicable to images with whole auroral ovals, but also

to images with partial ovals.

To objectively evaluate the performance of various

algorithms, we deﬁned particle precipitation boundaries

(called DMSP boundaries) derived from DMSP satel-

lite observations

[11]

as the real boundaries of the auroral

oval. The comparative experimental results demonstrate

that the method outperforms previous methods.

1 Auroral oval extraction based on fuzzy

clustering

Auroral oval boundaries are usually blurry because of

strong background noise. Fuzzy clustering techniques are

a good choice for UVI image segmentation. These tech-

niques have been studied extensively and applied suc-

cessfully in many ﬁelds

[12−15]

. In this paper, fuzzy local

information c-means clustering (FLICM)

[10]

is used to

segment the auroral oval in a UVI image. FLICM uses a

fuzzy local (both spatial and gray level) similarity mea-

sure to overcome the disadvantages of traditional fuzzy

c-means algorithms.

To extract a complete and accurate auroral oval from

a UVI image, our method contains three main steps:

image preprocessing; auroral oval segmentation using

FLICM; and gap ﬁlling, based on prior knowledge of the

auroral oval shape.

1.1 UVI image preprocessing

Eﬀective preprocessing of the original UVI images greatly

aids the accurate extraction of auroral ovals, especially

for images with strong noise. The characteristics of UVI

images include blurry and elliptical outer boundaries

[9]

blurry inner boundaries with complex shapes, and the

spatial distributions of the auroral oval centered on the

Earth’s magnetic poles, within a certain range (57.5

◦

◦ [16]

) of magnetic latitude. We considered these char-

acteristics in the UVI image preprocessing, as follows.

(1) Background removal. Pixels corresponding to

magnetic latitudes less than 50

◦

were removed.

(2) Negative pixel clearance. Pixels with negative

gray values, possibly caused by noise, were set to zero.

(3) Small bright spot smoothing. A bright spot is

deﬁned as a connected region in which the pixel gray val-

ues are greater than a given threshold. The threshold T

is deﬁned as

= µ

+ 3σ

(1)

where µ

and σ

are the mean gray value and standard

deviation of all pixels in the image, respectively. If the

bright spot area is less than a predetermined threshold

(20 pixels), the pixels inside the spot are set to the aver-

age value of pixels outside the spot and contained in the

smallest rectangle covering the bright spot. Otherwise,

the bright spot is not processed, to avoid destroying au-

roral substorm regions by mistake.

(4) Image smoothing. The pixels in each 3×3 neigh-

borhood are partitioned into two classes, according to

their gray values. The class with fewer pixels is consid-

ered the outlier, and its values are replaced by the average

intensity of the other class.

Figure 1 demonstrates an example of the preprocess-

ing steps. Figure 1(f) shows the eﬀective image region.

Only the data inside the eﬀective region of the original

image are meaningful.

剩余7页未读，继续阅读

评论收藏

内容反馈

weixin_38622475

粉丝: 0
资源: 912

从UVI图像中提取极光椭圆边界：一种基于FLICM聚类的新方法及其评估

matlab toolbox：Uvi_Wave toolbox

uvi_wave讲解

Uvi_wave工具箱

UVI:统一动词索引项目的回购

基于CEM算法的三维人脸贴图算法仿真,包含仿真操作录像,中文注释

keil uVi sion4_V9.00汉化包.part1

Show Your ToolBox——锋利的groovy

MATLAB工具箱下载地址总汇

start-fastapi：基于fastapi的轻量级Web框架

Keil+uVision4+中文教程(文档).rar_keil_keil 4_keil uvision4_keil教程_uVi

ITTVIS.IDL.v6.3.Analyst-TBE.rar_IDl

111报文的关联方式.docx

基质培模型2.0版使用说明1

Uvi_Wave300c.zip_matlab例程_matlab_

Microtones Library:适用于Sibelius，Kontakt，UVI，Max的微调音乐符号和播放-开源

基于MSP430的便携式紫外线强度检测仪设计

墨迹天气接口文档

Hard Ware Basic training

PyPI 官网下载 | adafruit-circuitpython-veml6070-3.1.7.tar.gz

toolbox:我多年来写的一堆随机脚本的工具箱

ClearCase7.2（Eclipse连接驱动包）

weather-dashboard-app

小程序报错 WAService.js:3 Uncaught Error: module ＂src/utils/utils.js＂ is not defined

rpmfusion-free-release-7.noarch.rpm

Weather-Dashboard:作业6

基于CORDIC的反正弦和反余弦计算的FPGA实现

最新资源