CN114462533A - Clustering object clustering method - Google Patents
Clustering object clustering method Download PDFInfo
- Publication number
- CN114462533A CN114462533A CN202210117476.8A CN202210117476A CN114462533A CN 114462533 A CN114462533 A CN 114462533A CN 202210117476 A CN202210117476 A CN 202210117476A CN 114462533 A CN114462533 A CN 114462533A
- Authority
- CN
- China
- Prior art keywords
- clustering
- frame
- frames
- box
- minimum
- Prior art date
- Legal status (The legal status 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 status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 19
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000007418 data mining Methods 0.000 abstract description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Image Analysis (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a clustering method for clustered objects, which relates to the technical field of data mining and solves the technical problem that the existing clustering algorithm is not flexible enough. Initializing a cluster central point randomly; the clustering quantity does not need to be specified, and the use is more flexible; and distance degrees are introduced in the clustering process, and the distance of the relative distance of the object is reasonably analyzed.
Description
Technical Field
The application relates to the technical field of data mining, in particular to a clustering method for clustered objects.
Background
In some clustering tasks, attention needs to be paid to the dynamic of clustered targets and the correlation of similar targets, such as putting, pursuing, car accidents, etc., in which objects are not present by example, but a plurality of objects are present at the same time and have certain relation. When analyzing behaviors, attributes and states of these objects having a certain relation, the objects need to be combined according to a certain rule.
When the Kmeans clustering algorithm is used for solving the problem, the coordinates of the central points of all objects are generally clustered, the width and the height of the objects cannot be considered, and the clustering box at least needs to surround all the objects of the class, so that the maximum probability of surrounding the clustering box can contain the objects of other classes, and real clustering is not achieved.
Disclosure of Invention
The application provides a clustering object clustering method, which aims to make the clustering method more flexible, avoid initializing a clustering central point, introduce a distance degree in the clustering process and reasonably analyze the distance of relative distances of objects.
The technical purpose of the application is realized by the following technical scheme:
a clustered object clustering method, comprising:
step 1: acquiring a minimum bounding box set A of n objects, wherein A is { a ═ a }1,a2...,anGiving a search radius J at the same time; wherein n is>1;
Step 2: calculating the central point of each smallest enclosing frame in A as c1,c2,...,cn;
And step 3: traversing and calculating the straight-line distance between the central point of each minimum surrounding frame and the central points of the rest smallest surrounding frames in the step A, and acquiring the minimum value of the straight-line distances between the central points of each minimum surrounding frame and the rest smallest surrounding frames, wherein the minimum value is d1,d2,…,dnDividing the minimum value by the maximum value of the width and the height of the current minimum bounding box to obtain a distance b, wherein the calculation formula of b is as follows:
and 4, step 4: sorting the minimum bounding boxes in the A from small to large according to the corresponding b values to obtain a box set M, and setting M to { M ═ M1,m2,…,mn};
And 5: establishing a clustering frame empty set Q;
step 6: establish an empty set L and join m1While M rejects M1(ii) a Go through M, first with M1Center point mc1As a central point K, taking J as a search radius, finding all frames in the range, adding L, and then removing all frames meeting the conditions from M;
And 7: traversing the L, calculating the average value of the center points of all frames in the L, and updating the center point K; then repeating the step 6 until L is not updated; traversing the boundary coordinates of all frames in the L to obtain the minimum value and the maximum value of the boundary coordinates, wherein the minimum value and the maximum value are x respectivelymin、ymin、xmax、ymaxIf so, the coordinate is the boundary coordinate of the clustering frame q;
and 8: and (3) calculating the area ratio iou of the rest frame and the clustering frame q in the M in a traversing way, wherein the area ratio iou is expressed as:
when frame mnWhen the area ratio iou is not less than 0.8, the frame m is setnAdding the obtained data into a clustering box q, and updating the boundary range of the clustering box q to completely surround mnThen put the frame mnRemoving the M, emptying the set L, and adding all the cluster frames into a set Q, wherein the finally obtained cluster frames are the final current clustering result;
wherein the inter represents a frame mnArea intersection with the clustering box q, areanA representation frame mnThe area of (d);
and step 9: traversing the rest M, repeating the steps 6 to 8 until the M is empty, obtaining all clustering frames, and updating a set Q ═ Q1,q2,…,qj},j≤n。
The beneficial effect of this application lies in: according to the clustering object clustering method, more reasonable distance degree is introduced to judge the relative distance between objects, the dense center is used as an initial clustering center, surrounding adjacent objects are gradually merged, and the boundaries of clustering frames are updated. Initializing a cluster central point randomly; the clustering quantity does not need to be specified, and the use is more flexible; and distance degrees are introduced in the clustering process, and the distance of the relative distance of the object is reasonably analyzed.
Drawings
FIG. 1 is a flow chart of a method described herein;
FIG. 2 is a schematic diagram of the clustering effect of clustering by the method of the present application.
Detailed Description
The technical solution of the present application will be described in detail below with reference to the accompanying drawings.
Fig. 1 is a flow chart of a method according to the present application, as shown in fig. 1, the method comprising:
step 1: acquiring a minimum bounding box set A of n objects, wherein A is { a ═ a }1,a2...,anGiving a search radius J at the same time; wherein n is>1。
Step 2: calculating the central point of each smallest enclosing frame in A as c1,c2,...,cn。
And step 3: traversing and calculating the straight-line distance between the central point of each minimum surrounding frame and the central points of the rest smallest surrounding frames in the step A, and acquiring the minimum value of the straight-line distances between the central points of each minimum surrounding frame and the rest smallest surrounding frames, wherein the minimum value is d1,d2,…,dnDividing the minimum value by the maximum value of the width and the height of the current minimum bounding box to obtain a distance b, wherein the calculation formula of b is as follows:
and 4, step 4: sorting the minimum bounding boxes in the A from small to large according to the corresponding b values to obtain a box set M, and setting M to { M ═ M1,m2,…,mn}。
And 5: and establishing a clustering frame empty set Q.
Step 6: establish an empty set L and join m1While M rejects M1(ii) a Go through M, first with M1Center point mc1And taking J as a search radius as a central point K, finding all frames in the range, adding L, and then removing all frames meeting the conditions from M.
And 7: traversing the L, calculating the average value of the center points of all frames in the L, and updating the center point K; then repeating the step 6 until L is not updated; traversing the boundary coordinates of all frames in L to obtain the minimum value and the maximum value of the boundary coordinates, wherein the minimum value and the maximum value are x respectivelymin、ymin、xmax、ymaxAnd then the coordinate is the boundary coordinate of the clustering box q.
And 8: and (3) calculating the area ratio iou of the rest frame and the clustering frame q in the M in a traversing way, wherein the area ratio iou is expressed as:
when the frame mnWhen the area ratio iou is not less than 0.8, the frame m is setnAdding the obtained data into a clustering box q, and updating the boundary range of the clustering box q to completely surround mnThen put the frame mnRemoving the M, finally emptying the set L, and adding all the clustering frames into a set Q, wherein the finally obtained clustering frame is the final current clustering result;
wherein the inter represents a frame mnArea intersection with the clustering box q, areanA representation frame mnThe area of (a).
And step 9: traversing the rest M, repeating the steps 6 to 8 until the M is empty, obtaining all clustering frames, and updating a set Q ═ Q1,q2,…,qj},j≤n。
The distance degree attribute of each clustering box Q in the clustering box set Q finally obtained in the step 9 is different m in the step 61The distance degree of (c). The distance degree represents the relative distance between objects. When a plurality of frames exist in one clustered frame, the distance degree of the clustered frame is the minimum value of the plurality of frames, namely the distance degree of m1, and the use is that if 10 frames are clustered, only 3 frames are needed, and 3 clustered frames are selected according to the distance degree.
And selecting the clustering frames according to the sequence, wherein the sequence of the clustering frames is the sequence of the relation distances among the objects from near to far, selecting the frames which are most likely to have aggregation and contact states according to the sequence, and selecting the first frames by self-definition and then accessing the subsequent algorithm processing or business.
Fig. 2 is a schematic diagram of a clustering effect of clustering by the method of the present application, where (a) in fig. 2 is an object frame before clustering, and (b) in fig. 2 is an object frame after clustering and converted into a square, and a number at the center of the frame indicates a distance degree of the frame. Converting the clustering frame into a square object frame, specifically comprising: keeping the central point of the clustering frame unchanged, and enabling the side length of the square to be equal to the maximum value of the width and the height of the clustering frame.
The foregoing is an exemplary embodiment of the method described herein, and the scope of protection is defined by the claims and their equivalents.
Claims (2)
1. A clustering method for clustered objects, comprising:
step 1: acquiring a minimum bounding box set A of n objects, wherein A is { a ═ a }1,a2...,anGiving a search radius J at the same time; wherein n is>1;
Step 2: calculating the central point of each smallest enclosing frame in A as c1,c2,...,cn;
And step 3: traversing and calculating the straight-line distance between the central point of each minimum surrounding frame and the central points of the rest smallest surrounding frames in the step A, and acquiring the minimum value of the straight-line distances between the central points of each minimum surrounding frame and the rest smallest surrounding frames, wherein the minimum value is d1,d2,…,dnDividing the minimum value by the maximum value of the width and the height of the current minimum bounding box to obtain a distance b, wherein the calculation formula of b is as follows:
and 4, step 4: sorting the minimum bounding boxes in the A from small to large according to the corresponding b values to obtain a box set M, and setting M to { M ═ M1,m2,…,mn};
And 5: establishing a clustering frame empty set Q;
step 6: establish an empty set L and join m1While M rejects M1(ii) a Go through M, first with M1Center point mc1Taking J as a search radius as a central point K, finding out all frames in the range, adding L, and then removing all frames meeting the conditions from M;
and 7: traversing the L, calculating the average value of the center points of all frames in the L, and updating the center point K;then repeating the step 6 until L is not updated; traversing the boundary coordinates of all frames in the L to obtain the minimum value and the maximum value of the boundary coordinates, wherein the minimum value and the maximum value are x respectivelymin、ymin、xmax、ymaxIf so, the coordinate is the boundary coordinate of the clustering frame q;
and 8: and (3) calculating the area ratio iou of the rest frame and the clustering frame q in the M in a traversing way, wherein the area ratio iou is expressed as:
when the frame mnWhen the area ratio iou is not less than 0.8, the frame m is setnAdding the obtained data into a clustering box q, and updating the boundary range of the clustering box q to completely surround mnThen put the frame mnRemoving the M, emptying the set L, and adding all the cluster frames into a set Q, wherein the finally obtained cluster frames are the final current clustering result;
wherein the inter represents a frame mnArea intersection with the clustering box q, areanA representation frame mnThe area of (d);
and step 9: traversing the rest M, repeating the steps 6 to 8 until the M is empty, obtaining all clustering frames, and updating a set Q ═ Q1,q2,…,qj},j≤n。
2. The clustering method according to claim 1, wherein converting the cluster boxes in the cluster box set Q into squares comprises: keeping the central point of the clustering frame unchanged, wherein the side length of the square is equal to the maximum value of the width and the height of the clustering frame.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210117476.8A CN114462533B (en) | 2022-02-08 | 2022-02-08 | Clustered object clustering method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210117476.8A CN114462533B (en) | 2022-02-08 | 2022-02-08 | Clustered object clustering method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN114462533A true CN114462533A (en) | 2022-05-10 |
CN114462533B CN114462533B (en) | 2024-06-25 |
Family
ID=81411926
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210117476.8A Active CN114462533B (en) | 2022-02-08 | 2022-02-08 | Clustered object clustering method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114462533B (en) |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20130297141A1 (en) * | 2012-05-04 | 2013-11-07 | Chungbuk National University Industry-Academic Cooperation Foundation | Apparatus and method for monitoring abnormal state of vehicle using clustering technique |
CN104143009A (en) * | 2014-08-22 | 2014-11-12 | 河海大学 | Competition and cooperation clustering method based on maximum clearance segmentation of dynamic bounding box |
WO2018224634A1 (en) * | 2017-06-08 | 2018-12-13 | Renault S.A.S | Method and system for identifying at least one moving object |
CN110909792A (en) * | 2019-11-21 | 2020-03-24 | 安徽大学 | Clustering analysis method based on improved K-means algorithm and new clustering effectiveness index |
-
2022
- 2022-02-08 CN CN202210117476.8A patent/CN114462533B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20130297141A1 (en) * | 2012-05-04 | 2013-11-07 | Chungbuk National University Industry-Academic Cooperation Foundation | Apparatus and method for monitoring abnormal state of vehicle using clustering technique |
CN104143009A (en) * | 2014-08-22 | 2014-11-12 | 河海大学 | Competition and cooperation clustering method based on maximum clearance segmentation of dynamic bounding box |
WO2018224634A1 (en) * | 2017-06-08 | 2018-12-13 | Renault S.A.S | Method and system for identifying at least one moving object |
CN110909792A (en) * | 2019-11-21 | 2020-03-24 | 安徽大学 | Clustering analysis method based on improved K-means algorithm and new clustering effectiveness index |
Non-Patent Citations (1)
Title |
---|
张晓滨;母玉雪;: "改进的方差优化初始中心的K-medoids算法", 计算机技术与发展, no. 07, 10 July 2020 (2020-07-10) * |
Also Published As
Publication number | Publication date |
---|---|
CN114462533B (en) | 2024-06-25 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2014174932A1 (en) | Image processing device, program, and image processing method | |
CN114926699B (en) | Indoor three-dimensional point cloud semantic classification method, device, medium and terminal | |
CN111640089A (en) | Defect detection method and device based on feature map center point | |
CN110188763B (en) | Image significance detection method based on improved graph model | |
WO2022077863A1 (en) | Visual positioning method, and method for training related model, related apparatus, and device | |
CN108764726B (en) | Method and device for making decision on request according to rules | |
CN111460234A (en) | Graph query method and device, electronic equipment and computer readable storage medium | |
CN111583274A (en) | Image segmentation method and device, computer-readable storage medium and electronic equipment | |
CN112634457B (en) | Point cloud simplification method based on local entropy of Hausdorff distance and average projection distance | |
CN113436223B (en) | Point cloud data segmentation method and device, computer equipment and storage medium | |
CN111522968A (en) | Knowledge graph fusion method and device | |
CN111738040A (en) | Deceleration strip identification method and system | |
CN110706238A (en) | Method and device for segmenting point cloud data, storage medium and electronic equipment | |
CN114359632A (en) | Point cloud target classification method based on improved PointNet + + neural network | |
CN113674425B (en) | Point cloud sampling method, device, equipment and computer readable storage medium | |
CN115100099A (en) | Point cloud data processing method, device, equipment and medium | |
CN114417095A (en) | Data set partitioning method and device | |
CN113887630A (en) | Image classification method and device, electronic equipment and storage medium | |
CN112906652A (en) | Face image recognition method and device, electronic equipment and storage medium | |
CN115686379B (en) | Method and system for optimizing management of hollow white data area in flash memory | |
CN114462533A (en) | Clustering object clustering method | |
Kharinov et al. | Object detection in color image | |
CN114511862B (en) | Form identification method and device and electronic equipment | |
CN116363416A (en) | Image de-duplication method and device, electronic equipment and storage medium | |
CN116109649A (en) | 3D point cloud instance segmentation method based on semantic error correction |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
CB02 | Change of applicant information |
Country or region after: China Address after: 568 Longmian Avenue, Gaoxinyuan, Jiangning District, Nanjing City, Jiangsu Province Applicant after: Xiaoshi Technology (Jiangsu) Co.,Ltd. Address before: 568 Longmian Avenue, Gaoxinyuan, Jiangning District, Nanjing City, Jiangsu Province Applicant before: NANJING ZHENSHI INTELLIGENT TECHNOLOGY Co.,Ltd. Country or region before: China |
|
CB02 | Change of applicant information | ||
GR01 | Patent grant | ||
GR01 | Patent grant |