WO2011140679A1 - 基于拓扑知觉组织理论的形状图像分类方法 - Google Patents
基于拓扑知觉组织理论的形状图像分类方法 Download PDFInfo
- Publication number
- WO2011140679A1 WO2011140679A1 PCT/CN2010/000684 CN2010000684W WO2011140679A1 WO 2011140679 A1 WO2011140679 A1 WO 2011140679A1 CN 2010000684 W CN2010000684 W CN 2010000684W WO 2011140679 A1 WO2011140679 A1 WO 2011140679A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- global
- local
- topological
- feature
- features
- Prior art date
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
- G06V20/58—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
- G06V20/582—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads of traffic signs
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2137—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on criteria of topology preservation, e.g. multidimensional scaling or self-organising maps
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/478—Contour-based spectral representations or scale-space representations, e.g. by Fourier analysis, wavelet analysis or curvature scale-space [CSS]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/7715—Feature extraction, e.g. by transforming the feature space, e.g. multi-dimensional scaling [MDS]; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
- G06V10/806—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level of extracted features
Definitions
- Shape image classification method based on topological perceptual organization theory
- the present invention relates to pattern recognition, and more particularly to object recognition based on cognitive psychology. Background technique
- Shape image recognition is a key technology in intelligent video surveillance and autopilot systems. It can help intelligent video surveillance systems classify objects of interest and help autopilot systems identify signs in feature scenes.
- SIFT scale-invariant feature transform algorithm
- Topology Perceptual Organization Theory proposed by Professor Chen Lin (Computational perceptual Organization theory, also known as TPO theory, argues that global features take precedence over local features, global features are more important than local features, and global features are determined by their topological invariance.
- the theory of topological perceptual organization has been proved to be in line with the working mechanism of biological vision systems by a series of neurophysiological, psychological and medical image analysis experiments. Here are two interesting experiments.
- topological perceptual organization theory has made a major breakthrough in cognitive visual psychology, the theory is based on a large number of psychological, physiological, and medical imaging experiments, without strict mathematical descriptions and computable models.
- An object of the present invention is to provide a method for efficiently extracting global features of a shape image, and fusing global features and local features to more accurately describe the shape image.
- a shape image classification method based on topological perceptual organization theory includes the following steps:
- the global feature and the local feature are merged, and the weight of the local feature in the fusion process is adjusted according to the matching degree of the global feature S5;
- the shape image S6 is classified according to the merged features.
- the invention is applicable to an intelligent visual monitoring system, which helps the monitoring system to classify the targets in the scene, so that the monitoring system can truly understand what is happening in the scene, and can adopt different security levels according to different target categories. Applicable to the automatic driving system, judging the type of traffic signs, so that the automatic driving system is more intelligent.
- Figure 1 is a system block diagram of a shape image recognition method based on topological perceptual organization theory
- Figure 2 is an example diagram: illustrating that the European space has a weak ability to express semantic features
- FIG. 1 Example diagram: It shows that the geodesic distance can better express the semantic features
- Figure 4 is an example diagram: illustrating the role of tolerance
- Figure 5 is an example diagram: illustrating global feature extraction, green solid line indicates geodesic distance, red dashed line indicates Euclidean distance, we use the ratio of geodesic distance to Euclidean distance to extract global features;
- Figure 6 is a schematic diagram of the database constructed by the present invention ;
- Figure 7 is an effect of the present invention on a database constructed by the present invention in which the same shapes represent the same topological shape, which are brought together. Different sizes are allowed in the same shape, representing different geometric information in the same topology;
- Figure 8 is the effect of the SIFT method on the database constructed by the present invention.
- the same shape (see Figure 7. for a detailed shape) represents the same topological shape, and they are very cluttered, indicating that SIFT features have a weak ability to distinguish between different topologies;
- Figure 9 is a prior art test of the bee to identify the shape image
- Figure 10 is an experimental diagram of a prior art similar to the prior art. detailed description
- Fig. 1 shows a flow chart of a shape image recognition method based on topological perceptual organization theory.
- Step S1 Extracting edge points of the shape image.
- the present invention employs the Canny algorithm to extract the edge points of the shape image.
- Step S2 Construct a topological space, and calculate the expression of the extracted edge points in the topological space.
- topological space Compared with the European space used by most computer vision algorithms, topological space not only has a strong cognitive basis, but also can be better explained in computer vision algorithms. Below we analyze.
- the human visual system does not use European space, but instead transforms the European space.
- two straight lines that are parallel in the real external space are two intersecting lines in the human eye (the intersections are called “disappearing points" in computer vision).
- the intersections are called "disappearing points" in computer vision.
- the two points that are very close together are actually far away when expressing the meaning of the "S" shape. More precisely, the distance between the two red forks is expressed by their shortest connection distance (measuring distance).
- d (i, j) represents the Euclidean distance between point i and point j.
- Step S3 Extract the global feature according to the expression of the edge point in the topological space. We use the ratio of d * (i, j) and d (i, j) as a vote to construct a statistical distribution histogram as a global feature of the shape image. The histogram is defined as follows: , )
- n is the number of edge points
- ⁇ and ⁇ are the upper and lower bounds of the kth square (bin) of the histogram.
- Figure 5 shows an example plot of the ratio of geodesic distance to Euclidean distance as a global feature. It is worth noting that this is a very simple global feature extraction method, which is equivalent to the distribution of statistical pixel brightness in European space. The reason why we use such a simple global feature extraction method is to highlight the effectiveness of our proposed topological space. In other words, if we use the simple method in the topological space to achieve better results than the higher-level method in the European space, it fully demonstrates that the improvement of the effect is caused by the topological space.
- Step S4 Extract local features according to the expression of the edge points in the Euclidean space. We use SIFT features to extract local features of the shape image.
- Step S5 merging the global feature and the local feature, and adjusting the weight of the local feature in the fusion process according to the matching degree of the global feature. In the fusion process, we first calculate the matching score of the global feature (that is, the reciprocal of the global feature histogram distance between the shape images) and normalize it, and then use the reciprocal of the normalized global feature matching score as the weight of the local feature. , so the distance between the final two shape images is calculated using the following formula:
- Step S6 The shape image is described based on the merged features. According to steps S1-S5, we can get the distance between any two shape images. Using these distances, we can get the classification results of these shape images by using the common clustering algorithm.
- Step S1 extracting edge points of the shape image by using the Canny algorithm
- Step S2 construct a topological space by using formula (1) and formula (2), and calculate the expression of the extracted edge point in the topological space;
- Step S3 using formula (3) and formula (4), extracting global features according to the expression of edge points in the topological space;
- Step S4 extracting the SIFT feature as a local feature according to the expression of the edge point in the Euclidean space
- Step S5 using the formula (5) to fuse the global feature and the local feature, that is, adjusting the weight of the local feature in the fusion process according to the matching degree of the global feature;
- Step S6 classifying the shape image according to the merged feature, specifically, in this embodiment
- Figure 7 and Figure 8 show the classification results of our method on these shape images, and give the comparison results of the SIFT method.
- the present invention proposes a shape image recognition method based on topological perceptual organization theory.
- the invention is easy to implement and stable in performance.
- the invention can improve the ability of the intelligent monitoring system to understand the monitoring scene, and can improve the adaptability of the automatic driving system to the environment.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Software Systems (AREA)
- Medical Informatics (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Image Analysis (AREA)
Description
Claims
Priority Applications (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US13/142,650 US8732172B2 (en) | 2010-05-13 | 2010-05-13 | Shape classification method based on the topological perceptual organization theory |
CN2010800037256A CN102511049B (zh) | 2010-05-13 | 2010-05-13 | 基于拓扑知觉组织理论的形状图像分类方法 |
PCT/CN2010/000684 WO2011140679A1 (zh) | 2010-05-13 | 2010-05-13 | 基于拓扑知觉组织理论的形状图像分类方法 |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/CN2010/000684 WO2011140679A1 (zh) | 2010-05-13 | 2010-05-13 | 基于拓扑知觉组织理论的形状图像分类方法 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2011140679A1 true WO2011140679A1 (zh) | 2011-11-17 |
Family
ID=44913821
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2010/000684 WO2011140679A1 (zh) | 2010-05-13 | 2010-05-13 | 基于拓扑知觉组织理论的形状图像分类方法 |
Country Status (3)
Country | Link |
---|---|
US (1) | US8732172B2 (zh) |
CN (1) | CN102511049B (zh) |
WO (1) | WO2011140679A1 (zh) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104111960A (zh) * | 2013-04-22 | 2014-10-22 | 阿里巴巴集团控股有限公司 | 一种页面的匹配方法和装置 |
CN111931873A (zh) * | 2020-09-28 | 2020-11-13 | 支付宝(杭州)信息技术有限公司 | 图像识别方法和装置 |
Families Citing this family (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9268765B1 (en) * | 2012-07-30 | 2016-02-23 | Weongozi Inc. | Systems, methods and computer program products for neurolinguistic text analysis |
US8942515B1 (en) * | 2012-10-26 | 2015-01-27 | Lida Huang | Method and apparatus for image retrieval |
CN103336945B (zh) * | 2013-06-10 | 2017-11-10 | 黑龙江大学 | 融合局部特征与全局特征的指静脉识别方法 |
CN106156798B (zh) * | 2016-07-25 | 2019-10-25 | 河海大学 | 基于环形空间金字塔和多核学习的场景图像分类方法 |
WO2018119406A1 (en) | 2016-12-22 | 2018-06-28 | Aestatix LLC | Image processing to determine center of balance in a digital image |
CN107766791A (zh) * | 2017-09-06 | 2018-03-06 | 北京大学 | 一种基于全局特征和粗粒度局部特征的行人再识别方法及装置 |
CN109543708B (zh) * | 2018-10-11 | 2021-05-25 | 上海交通大学 | 融合拓扑特征的面向图数据的模式识别方法 |
CN113392766A (zh) * | 2021-06-16 | 2021-09-14 | 哈尔滨理工大学 | 一种基于注意力机制的人脸表情识别方法 |
CN114170607A (zh) * | 2021-12-08 | 2022-03-11 | 成都理工大学 | 一种视觉拓扑认知模型的构建方法、手写汉字识别方法 |
CN117623735B (zh) * | 2023-12-01 | 2024-05-14 | 广东雅诚德实业有限公司 | 高强度抗污日用陶瓷的生产方法 |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2004086503A1 (en) * | 2003-03-27 | 2004-10-07 | Green Vision Systems Ltd. | Generating intra-particle crystallographic parameter maps and histograms of a chemically pure crystalline particulate substance |
CN101110100A (zh) * | 2006-07-17 | 2008-01-23 | 松下电器产业株式会社 | 检测图像的几何形状的方法和装置 |
CN101276418A (zh) * | 2008-04-18 | 2008-10-01 | 首都师范大学 | 一种基于显微图像的微生物识别***及方法 |
WO2010027476A1 (en) * | 2008-09-03 | 2010-03-11 | Rutgers, The State University Of New Jersey | System and method for accurate and rapid identification of diseased regions on biological images with applications to disease diagnosis and prognosis |
CN101702197A (zh) * | 2009-10-26 | 2010-05-05 | 宁波大学 | 一种道路交通标志的检测方法 |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP3103793B2 (ja) * | 1997-03-11 | 2000-10-30 | 株式会社モノリス | オブジェクトの構造グラフ生成装置、それを利用することの可能なデータ変換装置およびシステム |
CN100485710C (zh) * | 2006-12-22 | 2009-05-06 | 四川川大智胜软件股份有限公司 | 利用数字图像处理技术识别车辆类型的方法 |
-
2010
- 2010-05-13 CN CN2010800037256A patent/CN102511049B/zh active Active
- 2010-05-13 WO PCT/CN2010/000684 patent/WO2011140679A1/zh active Application Filing
- 2010-05-13 US US13/142,650 patent/US8732172B2/en not_active Expired - Fee Related
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2004086503A1 (en) * | 2003-03-27 | 2004-10-07 | Green Vision Systems Ltd. | Generating intra-particle crystallographic parameter maps and histograms of a chemically pure crystalline particulate substance |
CN101110100A (zh) * | 2006-07-17 | 2008-01-23 | 松下电器产业株式会社 | 检测图像的几何形状的方法和装置 |
CN101276418A (zh) * | 2008-04-18 | 2008-10-01 | 首都师范大学 | 一种基于显微图像的微生物识别***及方法 |
WO2010027476A1 (en) * | 2008-09-03 | 2010-03-11 | Rutgers, The State University Of New Jersey | System and method for accurate and rapid identification of diseased regions on biological images with applications to disease diagnosis and prognosis |
CN101702197A (zh) * | 2009-10-26 | 2010-05-05 | 宁波大学 | 一种道路交通标志的检测方法 |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104111960A (zh) * | 2013-04-22 | 2014-10-22 | 阿里巴巴集团控股有限公司 | 一种页面的匹配方法和装置 |
CN111931873A (zh) * | 2020-09-28 | 2020-11-13 | 支付宝(杭州)信息技术有限公司 | 图像识别方法和装置 |
CN111931873B (zh) * | 2020-09-28 | 2020-12-22 | 支付宝(杭州)信息技术有限公司 | 图像识别方法和装置 |
Also Published As
Publication number | Publication date |
---|---|
CN102511049B (zh) | 2013-07-17 |
CN102511049A (zh) | 2012-06-20 |
US8732172B2 (en) | 2014-05-20 |
US20130046762A1 (en) | 2013-02-21 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2011140679A1 (zh) | 基于拓扑知觉组织理论的形状图像分类方法 | |
WO2021022970A1 (zh) | 一种基于多层随机森林的零部件识别方法及*** | |
CN108334848B (zh) | 一种基于生成对抗网络的微小人脸识别方法 | |
WO2020151489A1 (zh) | 基于面部识别的活体检测的方法、电子设备和存储介质 | |
CN110188807B (zh) | 基于级联超分辨率网络与改进Faster R-CNN的隧道行人目标检测方法 | |
CN112932663B (zh) | 一种提高腹腔镜胆囊切除术安全性的智能辅助*** | |
WO2018188453A1 (zh) | 人脸区域的确定方法、存储介质、计算机设备 | |
WO2019227954A1 (zh) | 识别交通灯信号的方法、装置、可读介质及电子设备 | |
TWI484423B (zh) | 臉部識別裝置及方法 | |
CN113052029A (zh) | 基于动作识别的异常行为监管方法、装置及存储介质 | |
CN106295600A (zh) | 驾驶员状态实时检测方法和装置 | |
CN108596062B (zh) | 基于深度学习的人脸图片实时高光去除方法及装置 | |
CN110765833A (zh) | 一种基于深度学习的人群密度估计方法 | |
WO2021051547A1 (zh) | 暴力行为检测方法及*** | |
JP2005190400A (ja) | 顔画像検出方法及び顔画像検出システム並びに顔画像検出プログラム | |
WO2019119515A1 (zh) | 人脸分析、过滤方法、装置、嵌入式设备、介质和集成电路 | |
US20150278584A1 (en) | Object discriminating apparatus and method | |
TWI498857B (zh) | 瞌睡提醒裝置 | |
CN110909724A (zh) | 一种多目标图像的缩略图生成方法 | |
CN109670517A (zh) | 目标检测方法、装置、电子设备和目标检测模型 | |
CN111325107A (zh) | 检测模型训练方法、装置、电子设备和可读存储介质 | |
WO2020022329A1 (ja) | 物体検出認識装置、方法、及びプログラム | |
WO2015131710A1 (zh) | 人眼定位方法及装置 | |
CN113158971A (zh) | 一种事件检测模型训练方法及事件分类方法、*** | |
CN111582278B (zh) | 人像分割方法、装置及电子设备 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
WWE | Wipo information: entry into national phase |
Ref document number: 201080003725.6 Country of ref document: CN |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 10851186 Country of ref document: EP Kind code of ref document: A1 |
|
WWE | Wipo information: entry into national phase |
Ref document number: 13142650 Country of ref document: US |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 10851186 Country of ref document: EP Kind code of ref document: A1 |