CN115063617B - Advertisement image recognition method and advertisement image recognition system - Google Patents
Advertisement image recognition method and advertisement image recognition system Download PDFInfo
- Publication number
- CN115063617B CN115063617B CN202210980468.6A CN202210980468A CN115063617B CN 115063617 B CN115063617 B CN 115063617B CN 202210980468 A CN202210980468 A CN 202210980468A CN 115063617 B CN115063617 B CN 115063617B
- Authority
- CN
- China
- Prior art keywords
- advertisement
- image
- target
- key points
- key
- 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.)
- Active
Links
Images
Classifications
-
- 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
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/751—Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0242—Determining effectiveness of advertisements
-
- 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
- G06V10/70—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
Abstract
The invention relates to the technical field of image data processing, in particular to an advertisement image identification method and an advertisement image identification system, wherein the method is a method for identifying by using electronic equipment, the advertisement image identification system is an artificial intelligence system in the production field, the identification of an advertisement image is completed, and key points and multi-dimensional principal component characteristics of an advertisement material image and an advertisement template image are firstly obtained; classifying the key points according to the difference of the main component characteristics of the key points of all dimensions to obtain key point categories, and matching the key point categories of the advertisement material images and the advertisement template images to obtain key point category pairs; calculating the distance between the key points based on the key point category pairs and the principal component characteristics, and matching the key points to obtain a plurality of groups of key point pairs; and matching the advertisement material image and the advertisement template image by the key point pairs and the gray difference of each pixel point of the image. According to the method, the key points are matched through principal component analysis, so that the accuracy and the speed of key point matching are improved, and the purpose of fast image matching is achieved.
Description
Technical Field
The invention relates to the technical field of image data processing, in particular to an advertisement image identification method and an advertisement image identification system.
Background
With the rapid development of advertisement media, more and more advertisement schemes are put on the advertising board of the bus station, the advertisement is a big means for attracting consumers of products, advertisement operators need to timely and accurately feed back advertisement putting results to customers, usually, the feedback of the putting results is to collect actually put advertisement images, and the advertisers can correctly identify and rearrange the advertisement schemes through the feedback results.
At present, the common method for acquiring the feedback image of the advertisement image is manually acquired and transmitted back by a worker, the efficiency is low, a large amount of manpower, material resources and financial resources are consumed, and the workload of the worker is greatly increased along with the increase of the advertisement delivery quantity and the shortening of the delivery period.
Disclosure of Invention
In order to solve the above technical problems, an object of the present invention is to provide an advertisement image recognition method and an advertisement image recognition system, wherein the adopted technical scheme is as follows:
in a first aspect, an embodiment of the present invention provides an advertisement image identification method, including the following steps:
acquiring an advertisement material image and acquiring an advertisement template image;
acquiring key points and corresponding multi-dimensional principal component characteristics of an advertisement material image and an advertisement template image;
selecting any dimension as a target dimension based on the advertisement material image or the advertisement template image, calculating a mean value of principal component characteristics of all key points in the target dimension, and classifying the key points according to the difference between the mean value and the principal component characteristics corresponding to the key points in the target dimension to obtain key point categories; matching the key point categories of the advertisement material images and the key point categories of the advertisement template images to obtain a plurality of key point category pairs;
calculating the distance between key points by taking the characteristic value of the principal component characteristic of each dimensionality of the key points as weight based on a group of key point category pairs, and matching the key points according to the distance to obtain a plurality of groups of key point pairs; selecting any advertisement material image as a target material image, and obtaining an advertisement template image corresponding to the target material image according to key points in the target material image and the key point pairs, wherein each advertisement template image corresponds to a plurality of advertisement material images;
selecting any advertisement template image as a target template image, and obtaining an advertisement material image matched with the target template image according to the gray value difference of each pixel point of the target template image and the corresponding multiple advertisement material images.
Preferably, the acquiring key points and corresponding multi-dimensional principal component features of the advertisement material image and the advertisement template image includes:
extracting key points and feature descriptors of the key points in the advertisement material image and the advertisement template image through a scale invariant feature transformation algorithm;
calculating a covariance matrix of the feature descriptor to obtain an eigenvalue of the covariance matrix and a corresponding eigenvector; and obtaining the multi-dimensional principal component characteristics of the key points according to the characteristic descriptors and the corresponding characteristic vectors.
Preferably, the classifying the key points according to the difference between the mean value and the principal component characteristics corresponding to the key points under the target dimension to obtain a key point category includes:
selecting any key point as a target key point; when the principal component characteristics of the target key points under the target dimension are larger than or equal to the mean value, marking 1; when the principal component characteristics of the target key points under the target dimension are smaller than the mean value, marking 0; each dimension of the target key point corresponds to a mark, and a binary number is obtained according to the mark corresponding to each dimension;
and classifying the key points in the advertisement material images with the same binary number into one class, and classifying the key points in the advertisement template images with the same binary number into one class to obtain a plurality of key point categories.
Preferably, the calculating the distance between the key points, and matching the key points according to the distance to obtain a plurality of groups of key point pairs includes:
the calculation formula of the distance is as follows:
wherein the content of the first and second substances,
is the distance;
is the first of the key points of the advertisement template image
Principal component characteristics of the dimension;
being key points of an image of advertising material
Principal component characteristics of the dimension;
is characterized by principal component
A characteristic value of (d);
is the dimensionality of the keypoints.
Preferably, the obtaining of the advertisement template image corresponding to the target material image according to the key points in the target material image and the key point pairs includes:
acquiring the number of key points of key point pairs in each advertisement template image and the key points in the target material image based on the key point pairs corresponding to the key points in the target material image; and the advertisement template image corresponding to the maximum number of the key points is the advertisement template image corresponding to the target material image.
Preferably, the obtaining of the advertisement material image matched with the target template image according to the gray value difference of each pixel point of the target template image and the corresponding advertisement material images includes:
matching the target template image with the corresponding multiple advertisement material images according to the gray value difference to obtain corresponding matching degrees; the advertisement material image corresponding to the maximum matching degree is the advertisement material image matched with the target template image;
the calculation formula of the matching degree is as follows:
wherein the content of the first and second substances,
is the degree of matching;
is a natural constant;
the number of pixel points of each line of the advertisement template image is set;
the number of pixel points in each row of the advertisement template image is set;
for advertising template image
Go to the first
Gray values of the column pixel points;
as an image of the advertising material
Go to the first
The gray values of the column pixels.
In a second aspect, an embodiment of the present invention provides an advertisement image recognition system, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the advertisement image recognition method when executing the computer program.
The embodiment of the invention at least has the following beneficial effects:
the embodiment of the invention utilizes an image data processing technology, the advertisement image identification method is a method for identifying by applying electronic equipment, the advertisement image identification system is a generation field artificial intelligence system and an artificial intelligence optimization operation system to complete the identification of the advertisement image, and firstly, key points and corresponding multi-dimensional principal component characteristics of an advertisement material image and an advertisement template image are obtained; based on the advertisement material image or the advertisement template image, selecting any dimensionality as a target dimensionality, calculating the mean value of the principal component characteristics of all key points in the target dimensionality, classifying the key points according to the difference between the mean value and the principal component characteristics corresponding to the key points in the target dimensionality to obtain key point categories, namely classifying the key points according to the similarity of the projection sizes of the principal components in the dimensionality to obtain the key point categories, so that the matching accuracy of the key points is improved, and the matching accuracy of the advertisement material image and the advertisement template image is further improved. Matching the key point types of the advertisement material image and the advertisement template image to obtain a plurality of key point type pairs; based on a group of key point category pairs, the feature values of the principal component features of all dimensions of the key points are used as weights, the distances between the key points are calculated, the key points are matched according to the distances to obtain a plurality of groups of key point pairs, and the weights of all the dimension features are not considered when the key points are normally matched through the distances. Selecting any advertisement material image as a target material image, and obtaining an advertisement template image corresponding to the target material image according to key points and key point pairs in the target material image; selecting any advertisement template image as a target template image, obtaining an advertisement material image matched with the target template image according to the gray value difference of each pixel point of the target template image and the corresponding multiple advertisement material images, and mutually matching the advertisement material image and the advertisement template image, so that the matching accuracy is improved, and a high-quality feedback image is obtained. According to the method, the key points are matched by combining a scale invariant feature transformation algorithm and principal component analysis, so that the accuracy and the speed of matching the key points are improved, and a high-quality feedback image matched with an advertisement template image, namely an advertisement material image matched with the high-quality feedback image, is quickly obtained.
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 description of the embodiments or 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 flowchart of a method for identifying an advertisement image according to an embodiment of the present invention;
FIG. 2 is a flowchart illustrating steps for obtaining key points and corresponding multi-dimensional principal component features of an advertisement template image according to an embodiment of the present invention;
fig. 3 is a flowchart of steps for classifying the keypoints in the advertisement material image and the advertisement template image, and matching different keypoint categories to obtain a plurality of keypoint category pairs according to an embodiment of the present invention.
Detailed Description
To further illustrate the technical means and effects of the present invention adopted to achieve the predetermined objects, the following detailed description of the embodiments, structures, features and effects of the method and system for recognizing advertisement images according to the present invention will be made 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 embodiment of the invention provides an advertisement image identification method and a specific implementation method of an advertisement image identification system. The invention relates to a method for identifying by using electronic equipment, which is used for collecting a bus station billboard image of a bus station staying in a way of each bus by using a vehicle event data recorder on each bus and aiming at solving the problems of high cost and low efficiency caused by artificial feedback of an advertisement image.
The following describes the specific schemes of the advertisement image recognition method and the advertisement image recognition system provided by the present invention in detail with reference to the accompanying drawings.
Referring to fig. 1, a flowchart illustrating steps of an advertisement image identification method according to an embodiment of the present invention is provided, where the method includes the following steps:
and step S100, acquiring an advertisement material image and acquiring an advertisement template image.
The number and types of advertisements delivered by an advertiser in a certain time are fixed, so that the advertisement images delivered in a specified time period are used as the advertisement template images.
The feedback image of the advertisement template image, namely the feedback result, required by the invention is intercepted from the video stored by the bus running recorder. When the bus is close to the station, the passenger flow is large, the bus station advertising board can be shielded, and the quality of the advertising board image collected by the bus driving recorder is further influenced, so that one frame of image before the vehicle is started is intercepted as an advertising material image, and because the personnel needing to take the bus get on the bus before the vehicle is started, the influence on the quality of the advertising board image is minimum.
For each bus station, a plurality of buses pass through the same bus station for many times every day, so that only one frame of image before the start of the bus is captured as the image of the billboard.
And removing the complex background of the billboard image through a semantic segmentation network. The training set used by the semantic segmentation network is a billboard image acquired by a bus driving recorder in a history acquisition process, the corresponding label labeling process of the training set is a single-channel semantic label, the pixel point of the corresponding position pixel belonging to the billboard is labeled as 1, and the pixel point of the corresponding position pixel belonging to the background of the non-billboard is labeled as 0. The loss function of the semantic segmentation network is a cross entropy loss function. And obtaining a mask image through a semantic segmentation network, and multiplying the mask image and the billboard image to obtain an image, namely the image of the advertising material which is free of the complex background and only contains the billboard.
And step S200, acquiring key points and corresponding multi-dimensional principal component characteristics of the advertisement material image and the advertisement template image.
Due to the influences of shooting angles, complex environments, weather illumination, shelters and the like, collected advertisement material images cannot be directly matched with advertisement template images, feature point detection and matching are not influenced by the factors, and accuracy is high. In the embodiment of the invention, the key points are obtained through a Scale Invariant Feature Transform (SIFT) algorithm, the algorithm matches the key points through Euclidean distance, the weights of all the features are 1, and the difference of the features is not considered, so the embodiment of the invention weights the features through principal component analysis.
Referring to fig. 2, the steps of obtaining key points of an advertisement template image and corresponding multi-dimensional principal component features specifically include:
step S201, extracting key points and feature descriptors of the key points in the advertisement material images and the advertisement template images through a Scale Invariant Feature Transform (SIFT) algorithm.
The steps of obtaining key points and feature descriptors based on the advertisement template image are as follows: converting the advertisement template image into a gray-scale image; and generating a Gaussian difference pyramid and finishing the construction of the scale space. The preliminary investigation of key points is realized through the detection of spatial extreme points; removing noise points, realizing the accurate positioning of stable key points, and recording the key points corresponding to the advertisement template image as first key points; and calculating 4 multiplied by 4 partitioned gradient direction histograms around the first key point, counting gradient amplitudes of 8 directions in each histogram, and acquiring a feature descriptor of the first key point, wherein the dimension of the feature descriptor is 128 dimensions.
And based on the advertisement material image, acquiring a second key point and a corresponding feature descriptor in the advertisement material image according to the same method.
Step S202, calculating a covariance matrix of the feature descriptors to obtain an eigenvalue of the covariance matrix and a corresponding eigenvector; and obtaining the multi-dimensional principal component characteristics of the key points according to the characteristic descriptors and the corresponding characteristic vectors.
Based on the advertisement template image, calculating a covariance matrix of a feature descriptor of the first key point, further determining a feature value of the covariance matrix, sequencing the feature values from small to large to obtain a feature vector corresponding to the feature value, standardizing the feature vector, obtaining principal component features according to the feature vector and the feature descriptor, and recording the principal component features as multi-dimensional first principal component features corresponding to the first key point.
And multiplying the feature vector of the feature descriptor corresponding to the first key point of the acquired advertisement template image and the feature descriptor of the second key point of the advertisement material image to obtain the multidimensional second principal component feature corresponding to the second key point.
Step S300, selecting any dimension as a target dimension based on an advertisement material image or an advertisement template image, calculating the mean value of the principal component characteristics of all key points in the target dimension, and classifying the key points according to the difference between the mean value and the principal component characteristics corresponding to the key points in the target dimension to obtain key point categories; and matching the key point categories of the advertisement material images and the key point categories of the advertisement template images to obtain a plurality of key point category pairs.
Referring to fig. 3, the steps of classifying the key points in the advertisement material image and the advertisement template image, and matching different key point classes to obtain a plurality of key point class pairs include:
step S301, based on the advertisement material image or the advertisement template image, selecting any dimension as a target dimension, and calculating the mean value of the principal component characteristics of all key points in the target dimension.
For any key point, the characteristic values of the principal component characteristics of each dimension are arranged from large to small to obtain a characteristic value sequence, and the accumulated contribution rate of the front p-dimension principal component characteristics is calculated according to the characteristic values
Wherein
Is as follows
Characterised by the principal component of the dimensionEigenvalues, as cumulative contribution rates
Then, the corresponding p value is obtained.
And selecting any dimensionality as a target dimensionality for each dimensionality of the key points with the p-dimensional principal component characteristics, and calculating the mean value of the principal component characteristics of all the key points under the target dimensionality.
And step S302, classifying the key points according to the difference between the mean value and the principal component characteristics corresponding to the key points under the target dimensionality to obtain key point categories.
And comparing the main component characteristics corresponding to the key points under the mean value and the target dimension. Selecting any key point as a target key point, and marking 1 when the principal component characteristics of the target key point under the target dimension are greater than or equal to the mean value corresponding to the target dimension; when the principal component characteristics of the target key points under the target dimension are smaller than the mean value corresponding to the target dimension, marking 0; and repeatedly comparing the principal component characteristics of the key point a in each dimension with the corresponding average value, wherein each dimension of the target key point corresponds to one mark, and obtaining a binary number according to the mark corresponding to each dimension to obtain a p-bit binary number. It should be noted that the binary number of p bits is obtained because one key point corresponds to p-dimensional principal component features, and each dimension corresponds to one mark.
And classifying the key points in the advertisement material images with the same corresponding binary number into one class, and classifying the key points in the advertisement template images with the same corresponding binary number into one class to obtain a plurality of key point categories.
Step S303, the key point categories of the advertisement material images and the key point categories of the advertisement template images are paired to obtain a plurality of key point category pairs.
Binary numbers corresponding to the key points in each key point category in the advertisement material image or the advertisement template image are the same.
And matching the key point categories with the same binary number in the advertisement material image and the advertisement template image to obtain a plurality of key point category pairs. If the binary number of the key point category a of the advertisement material image is 101, the binary number of the key point category b in the advertisement template image is also 101, and the key point category a and the key point category b are a pair of key point category pairs.
Step S400, based on a group of key point category pairs, calculating the distance between key points by taking the characteristic value of the principal component characteristic of each dimensionality of the key points as weight, and matching the key points according to the distance to obtain a plurality of groups of key point pairs; selecting any advertisement material image as a target material image, and obtaining an advertisement template image corresponding to the target material image according to key points and key point pairs in the target material image, wherein each advertisement template image corresponds to a plurality of advertisement material images.
In the embodiment of the invention, the key points are obtained through a Scale Invariant Feature Transform (SIFT) algorithm, the algorithm matches the key points through Euclidean distance, the weights of all the features are 1, and the difference of the features is not considered, so the embodiment of the invention weights the features through principal component analysis, further improves the accuracy of key point matching, and achieves the aim of accurately matching the advertisement template images and the advertisement material images.
And for any second key point d in the advertisement material image, finding a first key point matched with the second key point d in the corresponding key point category in the advertisement template image. Specifically, the method comprises the following steps: based on a group of key point category pairs, calculating the distance between key points by taking the characteristic value of the principal component characteristic of each dimensionality of the key points as weight, and matching the key points according to the distance to obtain a plurality of groups of key point pairs.
The distance
The calculation formula of (c):
wherein the content of the first and second substances,
as advertising templatesFirst of key points of image
Principal component characteristics of the dimension;
being key points of an image of advertising material
Principal component features of the dimension;
is characterized by principal component
A characteristic value of (d);
is the dimension of the keypoint.
The more similar the principal component features of each dimension between two keypoints, the smaller the distance between the two keypoints.
And for a second key point d in the advertisement material image, calculating distances corresponding to a plurality of first key points in the same key point category pair to which the second key point d belongs, and acquiring a first key point e corresponding to the minimum distance as a matching result, wherein the first key point e and the second key point d corresponding to the minimum distance are a group of key point pairs if matching is successful.
And selecting any advertisement material image as a target material image, and acquiring the number of key points of key point pairs formed by key points in each advertisement template image and key points in the target material image based on the key point pairs corresponding to the key points in the target material image. The advertisement template image corresponding to the maximum number of the key points is the advertisement template image corresponding to the target material image. Each advertisement template image corresponds to a plurality of advertisement material images.
And S500, selecting any advertisement template image as a target template image, and obtaining an advertisement material image matched with the target template image according to the gray value difference of each pixel point of the target template image and the corresponding multiple advertisement material images.
Selecting any advertisement template image as a target template image, matching the target template image with the advertisement material images according to the gray value difference of each pixel point of the target template image and the corresponding multiple advertisement material images to obtain the corresponding matching degree, and using the advertisement material image corresponding to the maximum matching degree as the advertisement material image matched with the target template image.
The degree of matching
The calculation formula of (2) is as follows:
wherein, the first and the second end of the pipe are connected with each other,
is a natural constant;
the number of pixel points of each line of the advertisement template image is set;
the number of pixel points in each row of the advertisement template image is set;
for advertising template image
Go to the first
Gray values of the column pixel points;
as an image of the advertising material
Go to the first
The gray values of the column pixels. It should be noted that the size of the advertisement template image is the same as that of the advertisement material image, and the number of pixels in each row and each column is the same.
The smaller the gray difference of each pixel point in the target template image and the advertisement material image is, the larger the matching degree of the corresponding target template image and the advertisement material image is.
And the advertisement material image corresponding to the maximum matching degree is a feedback image of the target template image.
And obtaining feedback images of the advertisement template images, and feeding back the delivery results of the advertisement template images according to the advertiser requirement in time and quantity.
In summary, the embodiment of the present invention utilizes an image data processing technology, the advertisement image identification method is a method for identifying by using an electronic device, and the advertisement image identification system is a generation field artificial intelligence system and an artificial intelligence optimization operation system to complete identification of an advertisement image, and first obtains key points and corresponding multidimensional principal component characteristics of an advertisement material image and an advertisement template image; selecting any dimension as a target dimension based on an advertisement material image or an advertisement template image, calculating the mean value of the principal component characteristics of all key points in the target dimension, and classifying the key points according to the difference between the mean value and the principal component characteristics corresponding to the key points in the target dimension to obtain key point categories; matching the key point types of the advertisement material image and the advertisement template image to obtain a plurality of key point type pairs; calculating the distance between the key points by taking the characteristic value of the principal component characteristic of each dimensionality of the key points as weight based on a group of key point category pairs, and matching the key points according to the distance to obtain a plurality of groups of key point pairs; selecting any advertisement material image as a target material image, and obtaining an advertisement template image corresponding to the target material image according to key points and key point pairs in the target material image; and selecting any advertisement template image as a target template image, and obtaining an advertisement material image matched with the target template image according to the gray value difference of each pixel point of the target template image and the corresponding multiple advertisement material images. According to the method, the key points are matched by combining a scale invariant feature transformation algorithm and principal component analysis, so that the accuracy and speed of key point matching are improved, and a high-quality feedback image matched with the advertisement template image is quickly obtained.
The embodiment of the present invention further provides an advertisement image recognition system, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the computer program. Since the advertisement image recognition method is described in detail above, it will not be described again.
It should be noted that: the precedence order of the above embodiments of the present invention is only for description, and does not represent the merits of the embodiments. And specific embodiments thereof 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 addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing may also be possible or may be advantageous.
All 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 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. An advertising image recognition method, characterized in that the method comprises the steps of:
acquiring an advertisement material image and acquiring an advertisement template image;
acquiring key points and corresponding multi-dimensional principal component characteristics of an advertisement material image and an advertisement template image;
selecting any dimension as a target dimension based on the advertisement material image or the advertisement template image, calculating the mean value of the principal component characteristics of all key points in the target dimension, and classifying the key points according to the difference between the mean value and the principal component characteristics corresponding to the key points in the target dimension to obtain key point categories; matching the key point categories of the advertisement material images and the key point categories of the advertisement template images to obtain a plurality of key point category pairs;
calculating the distance between key points by taking the characteristic value of the principal component characteristic of each dimensionality of the key points as weight based on a group of key point category pairs, and matching the key points according to the distance to obtain a plurality of groups of key point pairs; selecting any advertisement material image as a target material image, and obtaining an advertisement template image corresponding to the target material image according to key points in the target material image and the key point pairs, wherein each advertisement template image corresponds to a plurality of advertisement material images;
selecting any advertisement template image as a target template image, and obtaining an advertisement material image matched with the target template image according to the gray value difference of each pixel point of the target template image and the corresponding multiple advertisement material images.
2. The method for identifying advertisement images according to claim 1, wherein the acquiring key points and corresponding multi-dimensional principal component features of the advertisement material images and the advertisement template images comprises:
extracting key points and feature descriptors of the key points in the advertisement material image and the advertisement template image through a scale invariant feature transformation algorithm;
calculating a covariance matrix of the feature descriptor to obtain an eigenvalue of the covariance matrix and a corresponding eigenvector; and obtaining the multi-dimensional principal component characteristics of the key points according to the characteristic descriptors and the corresponding characteristic vectors.
3. The method of claim 1, wherein the classifying the key points according to the difference between the average value and the principal component feature corresponding to each key point in the target dimension to obtain a key point category comprises:
selecting any key point as a target key point; when the principal component characteristics of the target key points under the target dimension are larger than or equal to the mean value, marking the target dimension as 1; when the principal component characteristics of the target key points under the target dimension are smaller than the mean value, marking the target dimension as 0; each dimension of the target key point corresponds to a mark, and a binary number is obtained according to the mark corresponding to each dimension;
and classifying the key points in the advertisement material images with the same binary number into one class, and classifying the key points in the advertisement template images with the same binary number into one class to obtain a plurality of key point categories.
4. The method of claim 1, wherein the calculating distances between the key points and matching the key points according to the distances to obtain a plurality of key point pairs comprises:
the calculation formula of the distance is as follows:
wherein the content of the first and second substances,
is the distance;
as key points of the advertisement template image
Principal component characteristics of the dimension;
being key points of an image of advertising material
Principal component characteristics of the dimension;
is characterized by principal component
A characteristic value of (d);
is the dimensionality of the keypoints.
5. The method for identifying an advertisement image according to claim 1, wherein the obtaining an advertisement template image corresponding to a target material image according to a key point in the target material image and the key point pair comprises:
acquiring the number of key points of key point pairs in each advertisement template image and the key points in the target material image based on the key point pairs corresponding to the key points in the target material image; and the advertisement template image corresponding to the maximum number of the key points is the advertisement template image corresponding to the target material image.
6. The method for identifying an advertisement image according to claim 1, wherein obtaining an advertisement material image matching the target template image according to a gray value difference between pixels of the target template image and corresponding advertisement material images comprises:
matching the target template image with the corresponding multiple advertisement material images according to the gray value difference to obtain corresponding matching degrees; the advertisement material image corresponding to the maximum matching degree is the advertisement material image matched with the target template image;
the calculation formula of the matching degree is as follows:
wherein the content of the first and second substances,
is the degree of matching;
is a natural constant;
the number of pixel points of each line of the advertisement template image is set;
the number of pixel points in each row of the advertisement template image is set;
for advertising template image
Go to the first
Gray values of the column pixel points;
as an image of the advertising material
Go to the first
The gray values of the column pixels.
7. An advertisement image recognition system comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor when executing the computer program implements the steps of the method of any one of claims 1~6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210980468.6A CN115063617B (en) | 2022-08-16 | 2022-08-16 | Advertisement image recognition method and advertisement image recognition system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210980468.6A CN115063617B (en) | 2022-08-16 | 2022-08-16 | Advertisement image recognition method and advertisement image recognition system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115063617A CN115063617A (en) | 2022-09-16 |
| CN115063617B true CN115063617B (en) | 2022-12-13 |
Family
ID=83208488
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210980468.6A Active CN115063617B (en) | 2022-08-16 | 2022-08-16 | Advertisement image recognition method and advertisement image recognition system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115063617B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116563400B (en) * | 2023-07-12 | 2023-09-05 | 南通原力云信息技术有限公司 | Small program image information compression processing method |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114492681A (en) * | 2022-04-19 | 2022-05-13 | 南通人民彩印有限公司 | Method for identifying color printing packaging pattern by using electronic equipment and computer vision system thereof |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103729654A (en) * | 2014-01-22 | 2014-04-16 | 青岛新比特电子科技有限公司 | Image matching retrieval system on account of improving Scale Invariant Feature Transform (SIFT) algorithm |
| US20220092334A1 (en) * | 2020-09-22 | 2022-03-24 | Apple Inc. | Contextual Matching |
-
2022
- 2022-08-16 CN CN202210980468.6A patent/CN115063617B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114492681A (en) * | 2022-04-19 | 2022-05-13 | 南通人民彩印有限公司 | Method for identifying color printing packaging pattern by using electronic equipment and computer vision system thereof |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115063617A (en) | 2022-09-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP2808827B1 (en) | System and method for OCR output verification | |
| CN107085585B (en) | Accurate tag relevance prediction for image search | |
| WO2018137358A1 (en) | Deep metric learning-based accurate target retrieval method | |
| US8452096B2 (en) | Identifying descriptor for person or object in an image | |
| US10664719B2 (en) | Accurate tag relevance prediction for image search | |
| US8165410B2 (en) | Bags of visual context-dependent words for generic visual categorization | |
| Fehérvári et al. | Scalable logo recognition using proxies | |
| Yu et al. | Propagative hough voting for human activity recognition | |
| CN111814584A (en) | Vehicle weight identification method under multi-view-angle environment based on multi-center measurement loss | |
| CN111339975B (en) | Target detection, identification and tracking method based on central scale prediction and twin neural network | |
| CN111241931A (en) | Aerial unmanned aerial vehicle target identification and tracking method based on YOLOv3 | |
| CN102663401B (en) | Image characteristic extracting and describing method | |
| CN105930873B (en) | A kind of walking across mode matching method certainly based on subspace | |
| CN115063617B (en) | Advertisement image recognition method and advertisement image recognition system | |
| WO2024060684A1 (en) | Model training method, image processing method, device, and storage medium | |
| CN115115825B (en) | Method, device, computer equipment and storage medium for detecting object in image | |
| CN111898418A (en) | Human body abnormal behavior detection method based on T-TINY-YOLO network | |
| CN114596546A (en) | Vehicle weight recognition method and device, computer and readable storage medium | |
| CN113592030B (en) | Image retrieval method and system based on complex value singular spectrum analysis | |
| CN114663899A (en) | Financial bill processing method, device, equipment and medium | |
| CN113780335A (en) | Small sample commodity image classification method, device, equipment and storage medium | |
| CN113743359A (en) | Vehicle weight recognition method, model training method and related device | |
| CN112214639A (en) | Video screening method, video screening device and terminal equipment | |
| CN113362096A (en) | Frame advertisement image matching method based on deep learning | |
| Zhou et al. | A novel part-based model for fine-grained vehicle recognition |
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 |