WO2011142651A1 - Système et procédé de classification d'images - Google Patents

Système et procédé de classification d'images Download PDF

Info

Publication number
WO2011142651A1
WO2011142651A1 PCT/MY2010/000265 MY2010000265W WO2011142651A1 WO 2011142651 A1 WO2011142651 A1 WO 2011142651A1 MY 2010000265 W MY2010000265 W MY 2010000265W WO 2011142651 A1 WO2011142651 A1 WO 2011142651A1
Authority
WO
WIPO (PCT)
Prior art keywords
classifier
classification
engine
image input
result
Prior art date
Application number
PCT/MY2010/000265
Other languages
English (en)
Inventor
Ern Syn Ang
Yew Seng Tan
Dickson Lukose
Original Assignee
Mimos Berhad
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Mimos Berhad filed Critical Mimos Berhad
Publication of WO2011142651A1 publication Critical patent/WO2011142651A1/fr

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
    • G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
    • G06F16/583—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/254—Fusion techniques of classification results, e.g. of results related to same input data
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/94—Hardware or software architectures specially adapted for image or video understanding
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/96—Management of image or video recognition tasks

Definitions

  • This invention relates generally to a method and system for use in image classification, and more particularly to a method and system for use in image classification which produces a merged classification output or result based on an input image .
  • Image classification is generally a method used in classifying an image into its most suitable class which is typically subjected to a provided training set.
  • the current developments in image classification are focused into two primary categories; these are supervised classification and unsupervised classification.
  • Classifying images can be considered as a significant element in image related advancements and it typically involves various fundamental factors to be accordingly evaluated by the user for each project. For instance, classification of captured images permits users to obtain or produce results which can be used for various applications, particularly, object recognitions, landscaping for scientific researches and assessment purposes. It is therefore essential to develop and thereby incessantly improve the available systems to as to extract accurate results.
  • a method for use in classification of subjects into a plurality of classes enables a form of output which includes an associated output class, output score and ranking.
  • classifier outputs are combined and the system relies on a voting algorithm for determination of output class.
  • the system as disclosed allows combined output, the system does not provide the ability to add or remove classifiers dynamically during run-time which is a disadvantage, particularly in the event that the user is required to exclude classifiers which may hot be necessary for certain type of images. Further, this disclosure does not teach the integration of different classifiers to support any form of multi-level classification
  • a method for use in multi-level image- classification system comprising the steps of: starting up the classification system; receiving image input; delegating the classification task based on said receive image input; determining at least one suitable classifier for classification of said image input; invoking the selected classifier for said image input; classifying the image input; analyzing the results obtained based on the classification; merging the analyzed classification results.
  • a system for use in multi-level image classification incorporating at least one classifier, said system comprising: at least one means (100) configured for accepting image input, handling of said image input and determining the suitable classifier to be invoked based on said image input; at least one means (200) configured for analyzing results obtained from at least one selected classifier; normalizing the result if required and merging of results of classification so as to provide a uniformed output; at least one means (300) configured for determining the availability of engine plugin(s) for each selected classifier and thus loading engine (s) based on the selected classifier ( s ) ; at least one means (30) configured for accepting an image as input and generating class scores based on the association of the received image input; at least one means for analyzing the results (600) at least one means (700); configured for merging and thus providing a merged classification result (s) based on the classification results obtained.
  • FIG 1 illustrates the overview of the method and system in accordance to the present invention
  • FIG 2 illustrates the tasks of one of the main components of the present invention
  • FIG 3 illustrates the tasks of another main component of the present invention
  • FIG 4 illustrates the tasks of yet another main component of the present invention
  • FIG 5 illustrates the tasks involved in starting up the system of the present invention
  • FIG 6 illustrates the start-up process in accordance with the present invention
  • FIG 7 and FIG 8 illustrate the classification process in accordance with the present invention
  • FIG 9 illustrates the tasks involved for re-hashing the configuration process
  • FIG 10 illustrates the flowchart for the re-hashing process in accordance to the present invention.
  • the present invention provides an effective and simple method and system for use in image classification and recognition, whereby the present invention is configured in a manner such that it supports the incorporation of multiple classifiers dynamically for producing a combined image classification output or result.
  • FIG 1 shows a flowchart containing components and steps involved in accordance with the present invention.
  • the main components of the present invention comprise of at least one means configured for classification engine handler (100), at least one means configured for classification result analyzer handler (200) and at least one dynamic module loader (300) .
  • Each of these components are interconnected with several other sub-components such as a digester means (400), classification configuration mediator means (500) result analyzer (600) and result merger means (700) so as to operate based on the preferred embodiment of the present invention.
  • the engine handler (100) as portrayed in FIG 2 is regarded as the chief component within the method and system of the present invention, as its responsibilities comprises, but not limiting to, determining the suitable classifier to invoke and determining whether further classification is necessary based on the received image.
  • the engine handler (100) may perform several steps to accomplish the aforementioned tasks, these steps may include but not limiting to, initiating classification of image request, downloading said image, consulting the mediator for the classifier engine to invoke in the event that the image downloading is successful, detecting the presence of classifiers, invoking the selected classifier ( s ) and processing the result prior to forwarding said result to the next component .
  • the result analyzer (600) in accordance to the present invention is configured to analyze results from one or more classifiers (60, 70) within the system so as to assist in determining the best approach to process the results, apart from merging the classification results obtained from one or more classifiers to provide a uniformed output to calling component.
  • the result analyzer handler (200) may perform several steps during operation, said steps may include but not limiting to; initiating classification . result processing, delegating the result to an analyzer or analyzing sub-component within the system, invoking or instructing analyzer plugin based on ID, upon successful plugin the analyzer (600) proceeds to analyzing the results and will allow continuous receipt of results until prompted to merge the classification results.
  • the dynamic module loader (300) as seen in FIG 4 is responsible for determining whether a particular module needs to be loaded, and if required, injecting the loaded module into the system to be reviewed by the engine handler (100) and the result analyzer handler (200) .
  • the dynamic loader (300) in accordance to the present invention may perform several steps apart from the above during operation, said steps may include, but not limiting to, initiating load engine request, loading the requested engine, detecting engine (s) for plugin directory, and injecting said engine into memory. These steps may be repeated in the event that any form of error occurs during operation .
  • the operative mode of the system and method of the present invention is shown in FIG 5.
  • the system is initialized and process is delegated to a digester (400) .
  • the digester (400) reads the configuration file which may be in the form of XML file or other suitable format from which the result from said step is then forwarded to the classification engine handler (100).
  • the classification engine handler (100) Upon prompted the classification engine handler (100) fires up each of the classifiers by means of the dynamic module loader (300).
  • the dynamic module loader (300) determines whether the classifier engine plug in requested is available within the memory of the system. In the event that the classifier fails to start up, the module loader (300) then proceeds to detect the designated plugin folder and attempt to inject the same into memory prior to re-starting it up.
  • the configuration is forwarded to the classification result handler (200) for further processing.
  • the classification result analyzer handler (200) then fires up each of the analyzers (600) within the system by means of the dynamic module loader (300).
  • the dynamic module (300) is configured to determine whether the prompted classifier analyzer plugin requested is available for the purpose. If otherwise, the loader (300) therefore proceeds to search the designated plugin folder and attempt to inject into memory before starting it up. It is understood by a person skilled in the art that startup process in operating the present invention may further include other sub-component steps or sub-steps to improve the efficiency however to attain a similar result or object with the steps disclosed herein.
  • the first step is to receive image input and delegating the classification task to the classification engine handler (100).
  • the engine handler (100) then proceeds to consult another component within the system , known as the classification configuration mediator (500) to aid in determining the most suitable classifier (60, 70) to invoke.
  • the mediator (500) then proceeds to process the request from said engine handler (100) then aids to determine the suitable or correct classifier ( s ) (60, 70) to invoke and generate or provide form of classifier ID to the engine handler (100).
  • the engine handler (100) Upon receipt of such ID the engine handler (100) invokes the respective classifier ( s ) (60, 70). The invoked engine plugin therefore returns the classification result to the engine handler (100) . It is preferred that the steps of detecting the classifiers which are correct and suitable for the image input are repeated until the classification is thoroughly and accurately performed that no further classification is required.
  • the result generated by the engine handler (100) is then forwarded to the classification result analyzer handler (200) for further processing. Based on said result the analyzer (200) therefore invokes the suitable analyzers (600) to analyze all obtained results.
  • the selected analyzer ' (s) (600) then invokes the result plugin based on the ID of the results.
  • the invoked result plugin then returns the analyzed result to the result analyzer (600) whereby at this point the result analyzer (600) may determine as to whether further analyzing is necessary or otherwise.
  • Definite result (s) are then forwarded to the result analyzer handler (200) .
  • the handler (200) then proceeds to pass all the results to a result merger (700) which is a component configured to merge all obtained result (s).
  • FIG 9 shows the preferred paths the method and system of the present invention may follow in the event that a re-hash configuration is requested.
  • the essential steps are as shown in FIG 10. It is apparent that these steps may vary however to achieve a similar primary object of the present invention.
  • the first step which may occur for a re-hash is a generic image classifier (30) sends a request to the digester (400) for a re-hash.
  • the digester (400) reads from the updated configuration file and proceeds to return the new configuration to the generic image classifier (30).
  • the generic image classifier (30) then proceeds to pass the new configuration to the classification engine handler (100) for startup, whereby at this stage the engine handler (100) shuts down all previous classifier engines starts up the newly received classifier engine by means of the dynamic module loader (300) .
  • the dynamic loader (300) determines as to whether the classifier plugin engine requested is available in the memory. If it is available, the module loader (300) starts it up and in the event that the plugin engine fails to start up the module loader proceeds to search in the designated plugin folder and then tries to inject it into memory before restarting it up.
  • the generic image classifier (30) then proceeds to pass the new configuration to the classification image handler (100) for analyzer startup.
  • the classification result analyzer (600) then shuts down all previous analyzer plugins and starts up each of the new analyzers by means of the dynamic module loader (300) .
  • the dynamic module loader (300) determines whether the classifier analyzer plugin requested is available in memory. In the event that such plugin is not available in the memory, the module loader (300) performs a search and attempt to inject the searched plugin into memory before re-starting it up.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Library & Information Science (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Databases & Information Systems (AREA)
  • Software Systems (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Image Analysis (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

La présente invention concerne un procédé et un système de classification d'images. Les principaux composants en sont: au moins un organe configuré en pilote de moteur de classification (100); au moins un organe servant de pilote d'analyseur de résultats (200); au moins un organe configuré en chargeur dynamique de modules (300); au moins un organe configuré en classificateur d'images, servant à associer aux classes qui conviennent l'image fournie en entrée (30); au moins un organe servant à analyser les résultats (600); au moins un organe servant à effectuer des fusions, et ainsi à produire un résultat générique (700). La sortie obtenue par le procédé et le système de la présente invention est une sortie de classification fusionnée se basant sur les résultats de classification provenant d'une pluralité de classificateurs.
PCT/MY2010/000265 2010-05-11 2010-11-10 Système et procédé de classification d'images WO2011142651A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
MYPI2010002192 2010-05-11
MYPI2010002192A MY147282A (en) 2010-05-11 2010-05-11 System and method for use in image classification

Publications (1)

Publication Number Publication Date
WO2011142651A1 true WO2011142651A1 (fr) 2011-11-17

Family

ID=44914557

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/MY2010/000265 WO2011142651A1 (fr) 2010-05-11 2010-11-10 Système et procédé de classification d'images

Country Status (2)

Country Link
MY (1) MY147282A (fr)
WO (1) WO2011142651A1 (fr)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107341190A (zh) * 2017-06-09 2017-11-10 努比亚技术有限公司 图片筛选方法、终端及计算机可读存储介质
CN109902178A (zh) * 2019-02-28 2019-06-18 云孚科技(北京)有限公司 一种多级文本分类方法及系统
CN110249304A (zh) * 2017-01-19 2019-09-17 三星电子株式会社 电子设备的视觉智能管理
CN111813843A (zh) * 2019-04-12 2020-10-23 阿里巴巴集团控股有限公司 一种数据处理方法、装置及平台
US10909371B2 (en) 2017-01-19 2021-02-02 Samsung Electronics Co., Ltd. System and method for contextual driven intelligence

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20050286772A1 (en) * 2004-06-24 2005-12-29 Lockheed Martin Corporation Multiple classifier system with voting arbitration
KR20070011970A (ko) * 2005-07-22 2007-01-25 삼성에스디에스 주식회사 이미지 처리 시스템 및 그 방법
US20070092133A1 (en) * 2005-10-26 2007-04-26 Huitao Luo Pre-normalization data classification

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20050286772A1 (en) * 2004-06-24 2005-12-29 Lockheed Martin Corporation Multiple classifier system with voting arbitration
KR20070011970A (ko) * 2005-07-22 2007-01-25 삼성에스디에스 주식회사 이미지 처리 시스템 및 그 방법
US20070092133A1 (en) * 2005-10-26 2007-04-26 Huitao Luo Pre-normalization data classification

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110249304A (zh) * 2017-01-19 2019-09-17 三星电子株式会社 电子设备的视觉智能管理
EP3559804A4 (fr) * 2017-01-19 2019-11-27 Samsung Electronics Co., Ltd. Gestion de l'intelligence de vision destinée à des dispositifs électroniques
US10902262B2 (en) 2017-01-19 2021-01-26 Samsung Electronics Co., Ltd. Vision intelligence management for electronic devices
US10909371B2 (en) 2017-01-19 2021-02-02 Samsung Electronics Co., Ltd. System and method for contextual driven intelligence
CN110249304B (zh) * 2017-01-19 2023-05-23 三星电子株式会社 电子设备的视觉智能管理
CN107341190A (zh) * 2017-06-09 2017-11-10 努比亚技术有限公司 图片筛选方法、终端及计算机可读存储介质
CN109902178A (zh) * 2019-02-28 2019-06-18 云孚科技(北京)有限公司 一种多级文本分类方法及系统
CN111813843A (zh) * 2019-04-12 2020-10-23 阿里巴巴集团控股有限公司 一种数据处理方法、装置及平台
CN111813843B (zh) * 2019-04-12 2024-06-11 阿里巴巴集团控股有限公司 一种数据处理方法、装置及平台

Also Published As

Publication number Publication date
MY147282A (en) 2012-11-30

Similar Documents

Publication Publication Date Title
US11599348B2 (en) Container image building using shared resources
US10657264B2 (en) Techniques for correlating vulnerabilities across an evolving codebase
RU2654146C1 (ru) Система и способ обнаружения вредоносных файлов с использованием элементов статического анализа
Duan et al. Detective: Automatically identify and analyze malware processes in forensic scenarios via DLLs
RU2614557C2 (ru) Система и способ обнаружения вредоносных файлов на мобильных устройствах
JP5963008B2 (ja) コンピュータシステムの分析方法および装置
WO2011142651A1 (fr) Système et procédé de classification d'images
WO2007103984A2 (fr) entrée d'images multiples pour les systèmes et méthodes de reconnaissance optique de caractères
US9355298B2 (en) Fingerprint authentication apparatus, method, and recording medium
US11521078B2 (en) Leveraging entity relations to discover answers using a knowledge graph
RU2634177C1 (ru) Система и способ обнаружения нежелательного программного обеспечения
US10055583B2 (en) Method and apparatus for processing file
Garn et al. Eris: A tool for combinatorial testing of the Linux system call interface
RU2756186C2 (ru) Система и способ категоризации .NET приложений
KR102367859B1 (ko) 특징 벡터를 이용하여 데이터를 분류하는 장치 및 방법
US7367056B1 (en) Countering malicious code infections to computer files that have been infected more than once
US20160267271A1 (en) Detection of malicious software behavior using signature-based static analysis
US8312137B1 (en) Live experiment framework
US8689327B2 (en) Method for characterization of a computer program part
CN111752570A (zh) 一种编译方法、装置、终端及计算机可读存储介质
Mokhov et al. File type analysis using signal processing techniques and machine learning vs. file unix utility for forensic analysis
CN114547628B (zh) 漏洞检测方法及装置
KR102566928B1 (ko) 사용자 의도를 판별하는 딥러닝 모델의 강화 학습을 수행하기 위한 훈련 세트를 생성하는 전자 장치 및 그 동작 방법
CN115544496A (zh) 基于可信执行环境的无服务器计算方法、装置及设备
Liu et al. Correlating ui contexts with sensitive api calls: Dynamic semantic extraction and analysis

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 10851475

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 10851475

Country of ref document: EP

Kind code of ref document: A1