Browse Prior Art Database

CLASSIFICATION OF OBJECTS BASED ON THE ISOMAP ALGORITHM

IP.com Disclosure Number: IPCOM000234966D
Publication Date: 2014-Feb-20
Document File: 9 page(s) / 324K

Publishing Venue

The IP.com Prior Art Database

Abstract

[0003] Embodiments of the disclosure may provide systems and methods for clustering, according to their similarity, any objects that have attributes or properties that can be measured and compared between different objects. In some examples these methods and systems may be applied to images or earth models (2D or 3D) grids; however, the systems and methods can also be extended to cluster or classify many other types of objects as medical diagnosis, investment portfolios, accident types, client, employees, software, products, etc. Further, embodiments of the disclosure may allow grouping the different attributes representing an object according to the similarity between them. In this way the analyst does not need to review all the attributes (grids) representing an object but just one representative for each cluster.

This text was extracted from a Microsoft Word document.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 23% of the total text.

Classification of Objects Based on the Isomap Algorithm

Background

[0001]          Different types of information are stored on 2D or 3D grids (satellite images, seismic attributes extractions, earth model property grids, echography or radiographic images, etc.) in different science domains.  Frequently, different properties or attributes (grids) are measured from the same object, which may enable the use of multiple attributes analysis in the classification of the objects (different seismic attributes, satellite images extracted in different bands etc.).  Each one of these properties or attributes of an object can be considered as a different dimension of the object in the analysis domain.  These different dimensions can show correlation between them and thus duplicate information. 

[0002]         The analysis of the different grids (dimensions) representing an object as to identify those objects that highlight certain features (a bright spot, a dim spot, a channel, a tumor) can take a long time because it is executed manually by an expert in the interpretation domain.

SUMMARY

[0003]         Embodiments of the disclosure may provide systems and methods for clustering, according to their similarity, any objects that have attributes or properties that can be measured and compared between different objects.  In some examples these methods and systems may be applied to images or earth models (2D or 3D) grids; however, the systems and methods can also be extended to cluster or classify many other types of objects as medical diagnosis, investment portfolios, accident types, client, employees, software, products, etc.  Further, embodiments of the disclosure may allow grouping the different attributes representing an object according to the similarity between them.  In this way the analyst does not need to review all the attributes (grids) representing an object but just one representative for each cluster.

[0004]         The present systems and methods may advance from the Principal Component Analysis and may extend Multidimensional Scaling.  In particular, the present system and methods may differentiate from these based at least partially on handling the case of nonlinear behavior of the variables.

[0005]         An example of the application of the present methods and systems is a geophysicist analyzing attributes extracted over a seismic horizon.  The geophysicist can extract 50 different attributes in a few minutes, but will need a long time to visually analyze the different images generated.  With the present method and system, the geophysicist will run a process that automatically will group the images in clusters according to the similarity between them.  The geophysicist may then analyze one image per cluster reducing the time spent on the analysis.

[0006]         Similarly, a geomodeler may create 3D earth models.  The visualization of the models is complicated and may require running a simulation on the models to determine a difference between them.  The simulation may, ...