CN114637873A - Intelligent door and window recommendation method and system based on image similarity - Google Patents
Intelligent door and window recommendation method and system based on image similarity Download PDFInfo
- Publication number
- CN114637873A CN114637873A CN202210326723.5A CN202210326723A CN114637873A CN 114637873 A CN114637873 A CN 114637873A CN 202210326723 A CN202210326723 A CN 202210326723A CN 114637873 A CN114637873 A CN 114637873A
- Authority
- CN
- China
- Prior art keywords
- node
- door
- target
- layer
- window
- 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
Images
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
- G06F16/53—Querying
- G06F16/535—Filtering based on additional data, e.g. user or group profiles
-
- 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
- G06F16/55—Clustering; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Databases & Information Systems (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention relates to the field of artificial intelligence, in particular to a door and window intelligent recommendation method and system based on image similarity, wherein the method comprises the following steps: acquiring a plurality of door and window pictures based on the basic demand information of a user for doors and windows, and mapping each door and window picture into a two-dimensional characteristic point; determining a root node, and constructing a tree structure based on the position distribution information of the feature points in the two-dimensional coordinate system relative to the root node; acquiring a subtree of each child node directly connected with the root node in the tree structure; each node in the subtree is respectively used as a target node, the probability that a characteristic point corresponding to the target node is a clustering center is calculated based on the degree of the target node in the subtree and the number of nodes in a target layer where the target node is located and a layer adjacent to the target layer, and an initial clustering center corresponding to the subtree is selected; and clustering the feature points based on the initial clustering center, wherein after the clustering is finished, the door and window picture corresponding to the new clustering center is the door and window picture to be recommended. The invention can provide more accurate door and window recommendation information for the user.
Description
Technical Field
The invention relates to the field of artificial intelligence, in particular to a door and window intelligent recommendation method and system based on image similarity.
Background
The existing picture recommendation system based on picture features usually utilizes a DNN network model to calculate picture similarity, but the method needs a large amount of training data, labels are marked and other early-stage work, the early-stage training process is long, and the calculation amount is large.
Disclosure of Invention
In order to solve the technical problems, the invention aims to provide an intelligent door and window recommendation method and system based on image similarity, and the adopted technical scheme is as follows:
in a first aspect, an embodiment of the present invention provides an intelligent door and window recommendation method based on image similarity, including the following specific steps:
acquiring basic demand information of a user for doors and windows, acquiring a plurality of door and window pictures based on the basic demand information, and mapping each door and window picture into a two-dimensional characteristic point;
determining a root node, and constructing a tree structure based on the position distribution information of the feature points in the two-dimensional coordinate system relative to the root node; obtaining a reference point according to the mean value of the horizontal and vertical coordinates of all the feature points in the two-dimensional coordinate system, wherein the feature point closest to the reference point is a root node;
for each child node directly connected with the root node in the tree structure, acquiring a sub-tree corresponding to the child node; each node in the subtree is respectively used as a target node, and the probability that the characteristic point corresponding to the target node is a clustering center is calculated based on the degree of the target node in the subtree and the number of nodes in a target layer where the target node is located and a layer adjacent to the target layer; selecting an initial clustering center corresponding to the sub-tree based on the probability;
and clustering the feature points based on the initial clustering center, wherein after the clustering is finished, the door and window picture corresponding to the new clustering center is the door and window picture to be recommended.
Further, each door and window picture is mapped into two-dimensional characteristic points, specifically: and based on a self-coding network, reducing the dimension of the feature data of the door and window pictures to obtain the two-dimensional feature points.
Further, the construction of the tree structure specifically includes:
determining an initial search range by taking a root node as a center in a two-dimensional coordinate system, searching feature points in the initial search range according to a preset search rule, and updating a tree structure based on the searched feature points; expanding the search range, acquiring the searched newly added feature points, determining father nodes of the newly added feature points, and updating the tree structure; and continuously expanding the search range, and updating the tree structure after each expansion until the tree structure comprises all the feature points.
Further, the specific method for determining the father node of the newly added feature point is as follows:
acquiring newly added nodes in the tree structure after the previous tree structure is updated, and regarding each newly added feature point, if the position deviation between the newly added feature point and the newly added node with the minimum distance from the newly added feature point is less than or equal to a deviation threshold value, taking the newly added node with the minimum distance from the newly added feature point as a father node of the newly added feature point; otherwise, the root node is the parent node.
Further, calculating the probability that the feature point corresponding to the target node is the clustering center, specifically:
the ratio of the degree of the target node to the maximum value of the degrees of the nodes in the subtree to which the target node belongs is a first numerical value;
setting layer weights for a target layer and a target layer adjacent layer, wherein the layer weight is larger when the target layer adjacent layer is closer to the target layer, and the number of nodes in the target layer and the target layer adjacent layer is subjected to weighted summation and then normalization processing to obtain a second numerical value;
and obtaining the probability that the feature point corresponding to the target node is the clustering center based on the first numerical value and the second numerical value.
Further, the layer weights set for the target layer and the layer adjacent to the target layer conform to Gaussian distribution, and the layer weight of the target layer is greater than the layer weight of the layer adjacent to the target layer.
In a second aspect, another embodiment of the present invention provides an intelligent door and window recommendation system based on image similarity, which specifically includes a memory, a processor, and a computer program stored in the memory and executable on the processor, where the computer program, when executed by the processor, implements the steps of an intelligent door and window recommendation method based on image similarity.
The embodiment of the invention at least has the following beneficial effects: the method and the device recommend the door and window pictures meeting the requirements of the user based on the tree structure, have small calculation amount, and can provide more accurate door and window recommendation information for the user more quickly.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions and advantages of the prior art, the drawings used in the embodiments or the description of the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.
FIG. 1 is a distribution diagram of feature points in a two-dimensional coordinate system according to an embodiment of the present invention.
Fig. 2 is a schematic diagram of a tree structure corresponding to fig. 1.
Detailed Description
To further illustrate the technical means and effects of the present invention adopted to achieve the predetermined invention purpose, the following describes in detail the specific implementation, structure, features and effects of an intelligent door and window recommendation method and system based on image similarity according to the present invention with reference to the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" refers to not necessarily the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
The following application scenarios are taken as examples to illustrate the present invention:
the application scene is as follows: when the user is uncertain about the actual door and window that oneself required, only when having a rough requirement, can recommend the user with the best door and window picture of different aspect through screening the door and window picture of a large amount that accords with the user requirement, and then the supplementary user carries out door and window's selection.
One embodiment of the invention provides an intelligent door and window recommendation method based on image similarity, which comprises the following steps:
and step S1, acquiring basic demand information of the user for the door and window, acquiring a plurality of door and window pictures based on the basic demand information, and mapping each door and window picture into a two-dimensional characteristic point.
Specifically, the basic requirement information of the user for the door and window comprises information such as color, material, style and the like; and crawling a plurality of door and window pictures based on the basic requirement information of the user on the door and window.
Specifically, each door and window picture is mapped into two-dimensional characteristic points: based on a self-coding network, reducing the dimension of the feature data of the door and window pictures to obtain two-dimensional feature points, namely representing the door and window pictures by using two data; wherein, the middle hidden layer of the self-coding network comprises two neurons. It should be noted that when the crawled door and window pictures include interference factors such as backgrounds, the door and window pictures need to be intercepted, and the door and window pictures including only doors and windows are obtained; preferably, in the embodiment, a window and door mask is obtained based on the semantic segmentation network, the pixel value of the window and door in the window and door mask is 1, and the values of other pixels are 0, and the window and door are intercepted based on the window and door mask.
Step S2, determining a root node, and constructing a tree structure based on the position distribution information of the feature points in the two-dimensional coordinate system relative to the root node; the reference point is obtained according to the mean value of the horizontal and vertical coordinates of all the feature points in the two-dimensional coordinate system, and the feature point closest to the reference point is a root node.
Specifically, M door and window pictures are crawled together, two data corresponding to the door and window pictures are two-dimensional coordinates of corresponding characteristic points, correspondingly, the two-dimensional coordinate system comprises M characteristic points, the coordinates of the reference point are (X, Y),
(xi,11) Is the ith characteristic point in the two-dimensional coordinate system.
The construction of the tree structure is specifically as follows: determining an initial search range by taking a root node as a center in a two-dimensional coordinate system, searching feature points in the initial search range according to a preset search rule, and updating a tree structure based on the searched feature points; expanding the search range, acquiring the searched newly added feature points, determining father nodes of the newly added feature points, and updating the tree structure; and continuously expanding the search range, and updating the tree structure after each expansion until the tree structure comprises all the feature points.
The search range may be a rectangular search range or a circular search range, and preferably, the search range in the embodiment is a rectangular search range; the preset search rule is as follows: searching in the clockwise direction by taking the north-righting direction as the starting direction, wherein the first priority of the north-righting direction is the highest, and the first priority of the feature points is gradually reduced along the clockwise direction, namely the feature points searched firstly are added into the tree structure firstly; if there are multiple feature points in a certain direction, then the second priorities of the multiple feature points in the direction need to be determined according to the distance between each feature point and the root node, and the smaller the distance, the higher the second priority. It should be noted that the implementer may set the starting direction and the searching direction by himself, for example, the searching direction is the righteast direction and the searching direction is counterclockwise.
The specific method for determining the father node of the newly added feature point comprises the following steps: acquiring newly added nodes in the tree structure after the previous tree structure is updated, and regarding each newly added feature point, if the position deviation between the newly added feature point and the newly added node with the minimum distance from the newly added feature point is less than or equal to a deviation threshold value, taking the newly added node with the minimum distance from the newly added feature point as a father node of the newly added node; otherwise, the root node is the father node; specifically, the direction in which the root node points to the newly added node with the minimum distance from the newly added feature point is taken as a reference direction, whether the angle between the direction in which the newly added node with the minimum distance from the newly added feature point points to the newly added feature point and the reference direction is smaller than or equal to an angle threshold value or not is judged, the angle represents the position deviation between the newly added feature point and the newly added node with the minimum distance from the newly added feature point, and if the angle is smaller than or equal to the angle threshold value, the newly added node with the minimum distance from the newly added feature point is taken as a parent node of the newly added node; preferably, the angle threshold is 5 ° in an embodiment.
Assuming that a distribution diagram of feature points in a two-dimensional coordinate system is shown in fig. 1, the feature points marked with letters a in fig. 1 are root nodes, and a dashed rectangle box represents a search range, a tree structure corresponding to the feature points marked with letters a-k in fig. 1 is shown in fig. 2, specifically, a construction process of the tree structure in fig. 2 is as follows:
(i) the middle-most dotted rectangle in fig. 1 is an initial search range in which no feature point is searched; similarly, no feature point is searched after the search range is enlarged once.
(ii) After the search range is expanded twice, newly-added feature points a and b are discovered clockwise in sequence, the first priority of the point b is higher than that of the point a, therefore, the point b is selected as a first child node of the root node A, and the point a is selected as a second child node of the root node A.
(iii) After the search range is enlarged for three times, four newly-added feature points of c, d, e and f are discovered clockwise in sequence; newly added nodes in the tree structure after the previous tree structure is updated are points a and b; the newly-added node closest to the point c is a point b, and the angle between the bc direction and the Ab reference direction is less than or equal to an angle threshold value, so that the point b is a parent node of the point c; the newly added node closest to the point d is the point b, but the angle between the bd direction and the Ab reference direction is greater than the angle threshold, so the parent node of the point d is the root node a; the newly added nodes closest to the point f and the point e are all the point a, and the angles between the af and ae directions and the Aa reference direction are all smaller than or equal to the angle threshold, so that the point a is a parent node of the point f and the point e.
(iiii) after the search range is expanded four times, finding points g, j, h, i and k clockwise in turn; newly added nodes in the tree structure after the previous tree structure updating are points c, d, e and f; for the point g, the newly added node closest to the point g is a point c, and the angle between the cg direction and the Ac reference direction is less than or equal to the angle threshold, so that the point c is a father node of the point g; for point j, the newly added node closest to point j is point d, but the angle between the dj direction and the Ad reference direction is greater than the angle threshold, so the parent node of point j is the root node a; point h and point j are the same; for point i, the newly added node closest to point i is point d, and the angle between the di direction and the Ad reference direction is less than or equal to the angle threshold, so that point d is the parent node of point i; for point k, the upper node closest to point k is point d, but the angle between the dk direction and the Ad reference direction is greater than the angle threshold, so the parent node of point k is the root node a.
At this time, the tree structure includes all the feature points in the coordinate system, and the search for the feature points is stopped.
In order to recommend a more accurate door and window picture, the maximum number n of the finally recommended door and window pictures is preset in one embodiment; after the tree structure is obtained, if the number of each child node directly connected with the root node in the tree structure is greater than n, sub-trees are needed to be merged, and the sub-trees are sub-trees corresponding to the child nodes directly connected with the root node; specifically, the method for combining subtrees includes:
respectively calculating the distance variance between the characteristic points corresponding to the nodes in each subtree; any two subtrees form a subtree combination, the distance variance of the characteristic points corresponding to the nodes in the subtree combination is the subtree combination variance, the combination with the minimum variance change value compared with the distance variances corresponding to the two subtrees in the subtree combination is selected, and the two subtrees corresponding to the subtree combination are combined; and subtracting the distance variances corresponding to the two subtrees in the subtree combination from the subtree combination variance to obtain a variance change value. For example, two subtrees in a combination in which a variance change value of a subtree combination variance and a distance variance corresponding to the two subtrees in the subtree combination is the smallest are a first word tree and a second subtree, and root nodes corresponding to the two subtrees are ZS1 and ZS2, it is first determined according to a first priority, that is, a searched precedence order, which subtree's root node is merged into the other subtree, if the first priority of ZS1 is greater than the first priority of ZS2, the second subtree is merged into the first subtree, and specifically, ZS2 is a child node of ZS 1; determining the position of the ZS2 in the first subtree according to the ZS2 and the distance between the original subnode which is directly connected with the ZS1 in the first subtree and the ZS 1; for the child nodes on the second layer in the second subtree, because the first priority of the child nodes on the second layer is lower than the first priority of the child nodes which are directly connected with the ZS1 and originally in the first subtree, the child nodes on the second layer in the second subtree are added to the right side of the child node which is positioned on the rightmost side of the second layer in the first subtree according to the original sequence from left to right, and the child nodes in the rest layers in the second subtree keep the original parent nodes unchanged and are merged into the first subtree.
It should be noted that the merging of the subtrees according to the above method can keep the structure of the original subtree as much as possible, and the destructiveness to the original subtree is smaller.
Step S3, for each child node directly connected with the root node in the tree structure, obtaining the sub-tree corresponding to the child node; each node in the subtree is respectively used as a target node, and the probability that the characteristic point corresponding to the target node is a clustering center is calculated based on the degree of the target node in the subtree and the number of nodes in a target layer where the target node is located and a layer adjacent to the target layer; and selecting the initial clustering center corresponding to the sub-tree based on the probability.
Preferably, the calculation of the probability that the feature point corresponding to the target node is the cluster center specifically includes: the ratio of the degree of the target node to the maximum value of the degrees of the nodes in the subtree to which the target node belongs is a first numerical value; setting layer weights for a target layer and a target layer adjacent layer, wherein the layer weight is larger when the target layer adjacent layer is closer to the target layer, and the number of nodes in the target layer and the target layer adjacent layer is subjected to weighted summation and then normalization processing to obtain a second numerical value; wherein, the layer weight values set for the target layer and the target layer adjacent layer are in accordance with Gaussian distribution, and the layer weight value of the target layer is greater than the layer weight value of the target layer adjacent layer; and obtaining the probability that the characteristic point corresponding to the target node is the clustering center based on the first numerical value and the second numerical value, and specifically, obtaining the probability that the characteristic point corresponding to the target node is the clustering center by multiplying the first numerical value and the second numerical value.
The method for acquiring the second numerical value comprises the following steps: for each sub-tree, acquiring a node number sequence based on the number of nodes in each layer in the sub-tree; the template with the preset size is used for counting nodes in a target layer where target nodes are located in the sequence of the number of the nodesThe number is the center, the values in the template are the layer weights of the target layer and the layer adjacent to the target layer respectively, wherein the values in the template accord with Gaussian distribution; preferably, the size of the template is 1 × 3, three values are obtained, the three values are sequentially recorded from left to right as P1, P2 and P3, the value of P1 is the function value corresponding to the midpoint of the horizontal axis on the left half of the symmetry axis of the gaussian function, the value of P2 is the maximum value of the gaussian function, the value of P3 is the function value corresponding to the midpoint of the horizontal axis on the right half of the symmetry axis of the gaussian function, and the values of P1 and P3 are the same. Normalization of P1, P2, P3 gave P1 ', P2 ', P3 ', Pz ═ Pz/Σ3z＝1Pz, z is in the range of [1,3 ]](ii) a P1 ', P2 ', P3 ' are values in the size 1 × 3 template; specifically, P1 ' and P3 ' are layer weights of two target layer adjacent layers on both sides of the target layer, and P2 ' is a layer weight of the target layer. It should be noted that the gaussian function can be set by the implementer. And carrying out weighted summation according to the layer weight and the node number of the target layer and the adjacent layer of the target layer to obtain a sum value corresponding to the target layer, and normalizing the sum value, specifically, calculating the sum value corresponding to each layer in the subtree by using each numerical value in the node number sequence as the center through a 1-by-3 template, and comparing the sum value corresponding to the target layer with the maximum value of the sum value to carry out normalization to obtain a second numerical value corresponding to the target layer.
In order to obtain the accurate probability that the feature point is the clustering center, in one embodiment, the number of adjacent layers of the target layer is changed, that is, the size of the template is changed, a plurality of second values corresponding to the target layer are obtained, and the probability that the feature point corresponding to the target node is the clustering center is obtained based on the first value and the mean value of the plurality of second values; specifically, the size of the template may also be 1 × 5, 1 × 7, … …, 1 × V, and for each subtree, the number of layers corresponding to the subtree is divided by 10 to obtain V.
Therefore, the probability that the feature point corresponding to each node in each sub-tree is the clustering center can be obtained, the feature point corresponding to the node with the highest probability in each sub-tree is the initial clustering center, and how many initial clustering centers can be obtained by how many sub-trees are needed to be explained.
And step S4, clustering the characteristic points based on the initial clustering center, and after the clustering is finished, taking the door and window picture corresponding to the new clustering center as the door and window picture to be recommended.
Clustering characteristic points based on the initial clustering center, calculating the distance from each class object to the clustering center, re-dividing the classes, and calculating a standard measure function until the maximum iteration times is reached; and after stopping, obtaining a new clustering center, wherein the door and window picture corresponding to the new clustering center is the door and window picture to be recommended to the user.
Based on the same inventive concept as the method embodiment, an embodiment of the present invention provides an intelligent door and window recommendation system based on image similarity, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the intelligent door and window recommendation method based on image similarity.
It should be noted that: the sequence of the above embodiments of the present invention is only for description, and does not represent the advantages or disadvantages of the embodiments. And that specific embodiments have been described above. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims may be performed in a different order than in the embodiments and still achieve desirable results. In some embodiments, multitasking and parallel processing may also be possible or may be advantageous.
The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.
Claims (7)
1. The intelligent door and window recommendation method based on image similarity is characterized by comprising the following steps:
acquiring basic demand information of a user for doors and windows, acquiring a plurality of door and window pictures based on the basic demand information, and mapping each door and window picture into a two-dimensional characteristic point;
determining a root node, and constructing a tree structure based on the position distribution information of the feature points in the two-dimensional coordinate system relative to the root node; obtaining a reference point according to the mean value of the horizontal and vertical coordinates of all the feature points in the two-dimensional coordinate system, wherein the feature point closest to the reference point is a root node;
for each child node directly connected with the root node in the tree structure, acquiring a sub-tree corresponding to the child node; each node in the subtree is respectively used as a target node, and the probability that the characteristic point corresponding to the target node is a clustering center is calculated based on the degree of the target node in the subtree and the number of nodes in a target layer where the target node is located and a target layer adjacent layer; selecting an initial clustering center corresponding to the sub-tree based on the probability;
and clustering the feature points based on the initial clustering center, wherein after the clustering is finished, the door and window picture corresponding to the new clustering center is the door and window picture to be recommended.
2. The method of claim 1, wherein each window and door picture is mapped to two-dimensional feature points, specifically: and based on a self-coding network, reducing the dimension of the feature data of the door and window pictures to obtain the two-dimensional feature points.
3. The method according to claim 2, characterized in that the construction of the tree structure is in particular:
determining an initial search range by taking a root node as a center in a two-dimensional coordinate system, searching feature points in the initial search range according to a preset search rule, and updating a tree structure based on the searched feature points; expanding the search range, acquiring the searched newly added feature points, determining father nodes of the newly added feature points, and updating the tree structure; and continuously expanding the search range, and updating the tree structure after each expansion until the tree structure comprises all the feature points.
4. The method of claim 3, wherein the specific method for determining the parent node of the newly added feature point is as follows:
acquiring newly added nodes in the tree structure after the previous tree structure is updated, and regarding each newly added feature point, if the position deviation between the newly added feature point and the newly added node with the minimum distance from the newly added feature point is less than or equal to a deviation threshold value, taking the newly added node with the minimum distance from the newly added feature point as a father node of the newly added node; otherwise, the root node is the parent node.
5. The method according to claim 4, characterized by calculating the probability that the feature point corresponding to the target node is the cluster center, in particular:
the ratio of the degree of the target node to the maximum value of the degrees of the nodes in the subtree to which the target node belongs is a first numerical value;
setting layer weights for a target layer and a target layer adjacent layer, wherein the layer weight is larger when the target layer adjacent layer is closer to the target layer, and the number of nodes in the target layer and the target layer adjacent layer is subjected to weighted summation and then normalization processing to obtain a second numerical value;
and obtaining the probability that the feature point corresponding to the target node is the clustering center based on the first numerical value and the second numerical value.
6. The method of claim 5, wherein the layer weights set for the target layer and the target layer neighbor layer conform to a Gaussian distribution, the layer weight of the target layer being greater than the layer weight of the target layer neighbor layer.
7. An intelligent door and window recommendation system based on image similarity, comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the method according to any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210326723.5A CN114637873B (en) | 2022-03-30 | 2022-03-30 | Intelligent door and window recommendation method and system based on image similarity |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210326723.5A CN114637873B (en) | 2022-03-30 | 2022-03-30 | Intelligent door and window recommendation method and system based on image similarity |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114637873A true CN114637873A (en) | 2022-06-17 |
| CN114637873B CN114637873B (en) | 2022-12-23 |
Family
ID=81951167
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210326723.5A Active CN114637873B (en) | 2022-03-30 | 2022-03-30 | Intelligent door and window recommendation method and system based on image similarity |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114637873B (en) |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101299213A (en) * | 2008-06-17 | 2008-11-05 | 中国地质大学（武汉） | N-dimension clustering order recording tree space index method |
| CN105373804A (en) * | 2015-09-29 | 2016-03-02 | 北京深视科技有限公司 | A human body part positioning method and system based on multi-dimensional space quick clustering |
| CN107729348A (en) * | 2017-08-23 | 2018-02-23 | 浙江大学 | A kind of the approximate KNN search method and searching system of the high dimensional data based on radiation stretching, extension figure |
| CN107766897A (en) * | 2017-12-06 | 2018-03-06 | 上海师范大学 | A kind of clustering algorithm based on minimum spanning tree and initial cluster center |
| CN108427723A (en) * | 2018-02-09 | 2018-08-21 | 哈尔滨工业大学深圳研究生院 | A kind of author's recommendation method and system based on clustering algorithm and local sensing reconstructing model |
| JP2018190154A (en) * | 2017-05-02 | 2018-11-29 | 株式会社リコー | Recommendation system, device, and program |
| JP2019079319A (en) * | 2017-10-25 | 2019-05-23 | 株式会社リコー | Recommendation system, apparatus, method, and program |
| CN111221937A (en) * | 2020-01-10 | 2020-06-02 | 江苏大学 | Method for constructing theme R tree by dynamic K value clustering |
| CN111382299A (en) * | 2020-01-20 | 2020-07-07 | 江苏艾佳家居用品有限公司 | Method, device, computer equipment and storage medium for accelerating image retrieval |
| CN111382760A (en) * | 2018-12-28 | 2020-07-07 | Tcl集团股份有限公司 | Image category identification method and device and computer readable storage medium |
| CN112200874A (en) * | 2020-10-30 | 2021-01-08 | 中国科学院自动化研究所 | Multilayer scene reconstruction and rapid segmentation method, system and device in narrow space |
| CN113177999A (en) * | 2021-03-25 | 2021-07-27 | 杭州易现先进科技有限公司 | Visual three-dimensional reconstruction method, system, electronic device and storage medium |
| CN113255714A (en) * | 2020-02-11 | 2021-08-13 | 北京三星通信技术研究有限公司 | Image clustering method and device, electronic equipment and computer readable storage medium |
-
2022
- 2022-03-30 CN CN202210326723.5A patent/CN114637873B/en active Active
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101299213A (en) * | 2008-06-17 | 2008-11-05 | 中国地质大学（武汉） | N-dimension clustering order recording tree space index method |
| CN105373804A (en) * | 2015-09-29 | 2016-03-02 | 北京深视科技有限公司 | A human body part positioning method and system based on multi-dimensional space quick clustering |
| JP2018190154A (en) * | 2017-05-02 | 2018-11-29 | 株式会社リコー | Recommendation system, device, and program |
| CN107729348A (en) * | 2017-08-23 | 2018-02-23 | 浙江大学 | A kind of the approximate KNN search method and searching system of the high dimensional data based on radiation stretching, extension figure |
| JP2019079319A (en) * | 2017-10-25 | 2019-05-23 | 株式会社リコー | Recommendation system, apparatus, method, and program |
| CN107766897A (en) * | 2017-12-06 | 2018-03-06 | 上海师范大学 | A kind of clustering algorithm based on minimum spanning tree and initial cluster center |
| CN108427723A (en) * | 2018-02-09 | 2018-08-21 | 哈尔滨工业大学深圳研究生院 | A kind of author's recommendation method and system based on clustering algorithm and local sensing reconstructing model |
| CN111382760A (en) * | 2018-12-28 | 2020-07-07 | Tcl集团股份有限公司 | Image category identification method and device and computer readable storage medium |
| CN111221937A (en) * | 2020-01-10 | 2020-06-02 | 江苏大学 | Method for constructing theme R tree by dynamic K value clustering |
| CN111382299A (en) * | 2020-01-20 | 2020-07-07 | 江苏艾佳家居用品有限公司 | Method, device, computer equipment and storage medium for accelerating image retrieval |
| CN113255714A (en) * | 2020-02-11 | 2021-08-13 | 北京三星通信技术研究有限公司 | Image clustering method and device, electronic equipment and computer readable storage medium |
| CN112200874A (en) * | 2020-10-30 | 2021-01-08 | 中国科学院自动化研究所 | Multilayer scene reconstruction and rapid segmentation method, system and device in narrow space |
| CN113177999A (en) * | 2021-03-25 | 2021-07-27 | 杭州易现先进科技有限公司 | Visual three-dimensional reconstruction method, system, electronic device and storage medium |
Non-Patent Citations (3)
| Title |
|---|
| CHONG YANG等: "Continuous KNN Join Processing for Real-Time Recommendation", 《2014 IEEE INTERNATIONAL CONFERENCE ON DATA MINING》 * |
| 刘璐: "基于联合域聚类和稀疏表示的极化SAR图像分类", 《中国博士学位论文全文数据库 信息科技辑》 * |
| 王华等: "二维空间聚类的树ART2模型", 《计算机应用》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114637873B (en) | 2022-12-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110930454B (en) | Six-degree-of-freedom pose estimation algorithm based on boundary box outer key point positioning | |
| Unnikrishnan et al. | Toward objective evaluation of image segmentation algorithms | |
| CN109614935B (en) | Vehicle damage assessment method and device, storage medium and electronic equipment | |
| CN114926699B (en) | Indoor three-dimensional point cloud semantic classification method, device, medium and terminal | |
| CN110827398B (en) | Automatic semantic segmentation method for indoor three-dimensional point cloud based on deep neural network | |
| CN108984785B (en) | Historical data and increment-based fingerprint database updating method and device | |
| CN110084299B (en) | Target detection method and device based on multi-head fusion attention | |
| US10360672B2 (en) | Automated separation of binary overlapping trees | |
| US20110075927A1 (en) | Fast image segmentation using region merging with a k-nearest neighbor graph | |
| WO1999067727A1 (en) | Method and system for visualization of clusters and classifications | |
| Lee et al. | Perceptual organization of 3D surface points | |
| CN111291768B (en) | Image feature matching method and device, equipment and storage medium | |
| JP2004164596A (en) | Method and system for traversing n-dimensional bi-tree | |
| CN104899883A (en) | Indoor object cube detection method for depth image scene | |
| US11074274B2 (en) | Large scale social graph segmentation | |
| CN109993753B (en) | Method and device for segmenting urban functional area in remote sensing image | |
| CN110210431A (en) | A kind of point cloud classifications method based on cloud semantic tagger and optimization | |
| CN111091618A (en) | Combined indoor layout method and system based on house type data driving | |
| CN115731355A (en) | SuperPoint-NeRF-based three-dimensional building reconstruction method | |
| CN114399784A (en) | Automatic identification method and device based on CAD drawing | |
| Zhang et al. | A data-driven approach for adding facade details to textured lod2 citygml models | |
| CN114637873B (en) | Intelligent door and window recommendation method and system based on image similarity | |
| CN116109649A (en) | 3D point cloud instance segmentation method based on semantic error correction | |
| Dong et al. | An optimization-based approach to image binarization | |
| CN114494693A (en) | Method and device for performing semantic segmentation on image |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| PE01 | Entry into force of the registration of the contract for pledge of patent right | ||
| PE01 | Entry into force of the registration of the contract for pledge of patent right |
Denomination of invention: Intelligent recommendation method and system for doors and windows based on image similarity Effective date of registration: 20230130 Granted publication date: 20221223 Pledgee: China Construction Bank Fengxian sub branch Pledgor: XUZHOU DAGONG ELECTRONIC TECHNOLOGY Co.,Ltd. Registration number: Y2023980031809 |