Processing

Please wait...

Settings

Settings

Goto Application

1. WO2020062360 - IMAGE FUSION CLASSIFICATION METHOD AND APPARATUS

Publication Number WO/2020/062360
Publication Date 02.04.2020
International Application No. PCT/CN2018/110916
International Filing Date 19.10.2018
IPC
G06K 9/00 2006.01
GPHYSICS
06COMPUTING; CALCULATING OR COUNTING
KRECOGNITION OF DATA; PRESENTATION OF DATA; RECORD CARRIERS; HANDLING RECORD CARRIERS
9Methods or arrangements for reading or recognising printed or written characters or for recognising patterns, e.g. fingerprints
CPC
G06K 2009/00644
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
KRECOGNITION OF DATA; PRESENTATION OF DATA; RECORD CARRIERS; HANDLING RECORD CARRIERS
9Methods or arrangements for reading or recognising printed or written characters or for recognising patterns, e.g. fingerprints
00624Recognising scenes, i.e. recognition of a whole field of perception; recognising scene-specific objects
0063Recognising patterns in remote scenes, e.g. aerial images, vegetation versus urban areas
00644using hyperspectral data, i.e. more or other wavelengths than RGB
G06K 9/0063
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
KRECOGNITION OF DATA; PRESENTATION OF DATA; RECORD CARRIERS; HANDLING RECORD CARRIERS
9Methods or arrangements for reading or recognising printed or written characters or for recognising patterns, e.g. fingerprints
00624Recognising scenes, i.e. recognition of a whole field of perception; recognising scene-specific objects
0063Recognising patterns in remote scenes, e.g. aerial images, vegetation versus urban areas
G06K 9/342
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
KRECOGNITION OF DATA; PRESENTATION OF DATA; RECORD CARRIERS; HANDLING RECORD CARRIERS
9Methods or arrangements for reading or recognising printed or written characters or for recognising patterns, e.g. fingerprints
20Image acquisition
34Segmentation of touching or overlapping patterns in the image field
342Cutting or merging image elements, e.g. region growing, watershed, clustering-based techniques
G06K 9/6269
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
KRECOGNITION OF DATA; PRESENTATION OF DATA; RECORD CARRIERS; HANDLING RECORD CARRIERS
9Methods or arrangements for reading or recognising printed or written characters or for recognising patterns, e.g. fingerprints
62Methods or arrangements for recognition using electronic means
6267Classification techniques
6268relating to the classification paradigm, e.g. parametric or non-parametric approaches
6269based on the distance between the decision surface and training patterns lying on the boundary of the class cluster, e.g. support vector machines
Applicants
  • 深圳大学 SHENZHEN UNIVERSITY [CN]/[CN]
Inventors
  • 贾森 JIA, Sen
  • 邓彬 DENG, Bin
  • 朱家松 ZHU, Jiasong
  • 邓琳 DENG, Lin
  • 李清泉 LI, Qingquan
Agents
  • 广州三环专利商标代理有限公司 SCIHEAD IP LAW FIRM
Priority Data
201811146961.829.09.2018CN
Publication Language Chinese (ZH)
Filing Language Chinese (ZH)
Designated States
Title
(EN) IMAGE FUSION CLASSIFICATION METHOD AND APPARATUS
(FR) PROCÉDÉ ET APPAREIL DE CLASSIFICATION PAR FUSION D'IMAGES
(ZH) 一种图像融合分类的方法及装置
Abstract
(EN)
An image fusion classification method and apparatus. Said method comprises: obtaining a three-dimensional weight matrix of a hyperspectral image by means of a support vector machine classifier (101); performing superpixel segmentation on the hyperspectral image, so as to obtain K superpixel maps, K being a positive integer (102); regularizing the three-dimensional weight matrix using the segmentation method of the superpixel maps, so as to obtain a regular matrix (103); and determining, according to the regular matrix, the category to which a sample belongs (104). Said method is able to avoid the problem in estimation of the number of superpixels, and further fuse, in a cascade manner, ground-object spatial structure information included in different superpixel maps, thereby facilitating feature-based determination.
(FR)
L'invention concerne un procédé et un appareil de classification par fusion d'images. Ledit procédé comprend : l'obtention d'une matrice de poids tridimensionnel d'une image hyperspectrale au moyen d'un classificateur de machine vectoriel de support (101); la réalisation d'une segmentation de superpixels sur l'image hyperspectrale, de façon à obtenir K cartes de superpixels, K étant un nombre entier positif (102); la régularisation de la matrice de poids tridimensionnel à l'aide du procédé de segmentation des cartes de superpixels, de façon à obtenir une matrice régulière (103); et la détermination, selon la matrice régulière, de la catégorie à laquelle appartient un échantillon (104). Ledit procédé est capable d'éviter le problème d'estimation du nombre de superpixels, et en outre de fusionner, en cascade, des informations de structure spatiale d'objet au sol comprises dans différentes cartes de superpixels, ce qui facilite la détermination basée sur des caractéristiques.
(ZH)
一种图像融合分类的方法及装置,所述方法包括:利用支持向量机分类器获得高光谱图像的三维权重矩阵(101);对所述高光谱图像进行超像素分割,以得到K个超像素图,其中,K为正整数(102);利用所述超像素图的分割法所述三维权重矩阵进行正则化以获得正则矩阵(103);根据所述正则矩阵确定样本所属的类别(104)。通过该方法,能够避免超像素个数的估计问题,进一步的,通过级联的方式融合不同超像素图包含的地物空间结构信息,显著提升了特征的判别力。
Also published as
Latest bibliographic data on file with the International Bureau