CN111178409B - Image matching and recognition system based on big data matrix stability analysis - Google Patents
Image matching and recognition system based on big data matrix stability analysis Download PDFInfo
- Publication number
- CN111178409B CN111178409B CN201911321300.9A CN201911321300A CN111178409B CN 111178409 B CN111178409 B CN 111178409B CN 201911321300 A CN201911321300 A CN 201911321300A CN 111178409 B CN111178409 B CN 111178409B
- Authority
- CN
- China
- Prior art keywords
- image
- sub
- initial
- characteristic data
- input image
- 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
- 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
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/583—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Library & Information Science (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Databases & Information Systems (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Image Analysis (AREA)
Abstract
The invention provides an image matching and identifying system and method based on big data matrix stability analysis and a computer readable storage medium. According to the technical scheme, input original images are not preprocessed, but are firstly identified before being inquired, one or a group of sub-images containing the most key elements of the original images are identified from the original images, and then image matching search is carried out on the basis of the sub-images. The method for identifying the key elements with the most number comprises similarity judgment and characteristic matrix stability judgment, so that the input query retrieval elements can contain the key information of the original input image, the data processing amount can be greatly reduced, and the image retrieval speed is improved on the premise of not losing the main information of the image. The scheme of the invention can be carried out through an automatic computer image recognition program and a matching program without manual intervention and prior experience.
Description
Technical Field
The invention belongs to the technical field of image recognition, and particularly relates to an image matching and recognition system and method based on big data matrix stability analysis and a computer readable storage medium.
Background
The development of mobile internet, smart phones and social networks brings huge amounts of picture information, and the uploading amount of images of an Instagram per day is about 6000 million; the daily picture delivery amount of WhatsApp is 5 hundred million; the WeChat friend circle in China is also driven by picture sharing. Pictures which are not limited by regions and languages gradually replace complicated and delicate characters and become main media for transferring word meaning. The reason why the pictures become the main media for internet information exchange is mainly two reasons: firstly, from the habit of reading information by a user, compared with characters, the picture can provide more vivid, easily understood, interesting and artistic information for the user; secondly, from the view of picture sources, the smart phone brings convenient shooting and screen capturing means for people, and helps people to acquire and record information by pictures more quickly.
But with pictures becoming the main information carrier in the internet, difficulties arise. When the information is recorded by characters, the required content can be easily found through keyword search and can be edited at will, and when the information is recorded by pictures, the content in the pictures cannot be retrieved, so that the efficiency of finding the key content from the pictures is influenced. The picture brings a fast information recording and sharing mode to us, but reduces the information retrieval efficiency. Under such circumstances, the image recognition technology of the computer is very important.
Image recognition is a technique in which a computer processes, analyzes, and understands images to recognize various different patterns of objects and objects. The identification process comprises image preprocessing, image segmentation, feature extraction and judgment matching. In brief, image recognition is how a computer reads the contents of a picture like a person. By means of the image recognition technology, the information can be acquired more quickly through picture searching, a new mode of interacting with the external world can be generated, and the external world can operate more intelligently.
Picture search is different from text search. The prior art picture search function is generally to provide a source image, pre-process it, and input it into a search engine (e.g., a picture matching system) to obtain at least one target image meeting specific requirements as an output. In the process, complicated preprocessing needs to be carried out on the source image, a large amount of characteristic information is extracted, and then semantic-based retrieval is carried out according to the processed image characteristic information.
The retrieval system disclosed by the Chinese patent with the publication number of CN103927387B and the related method and device thereof extract the characteristics of a sample image or a preprocessed sample image, and the extracted characteristic data comprises the position information, the scale, the direction and the characteristic description information of each characteristic point in an image area; classifying the feature description information of the sample image by using a classifier to find out an optimal classification result, wherein each classification corresponds to a classification index after classification; performing dimension reduction processing on the feature description information by combining the classification index value of the classification to which each feature description information belongs, taking the result data after dimension reduction as a label corresponding to the feature point, wherein each feature point corresponds to one label data; sequentially storing sample image content data into a retrieval database in a unit of sample image index, the content data of one sample image comprising: sample image index values, the number of characteristic points, and position information, scale, direction and characteristic description information of each characteristic point; and sequentially storing the label data into a retrieval database according to the classification index values in each classification by taking the label as a unit, wherein each label data corresponds to one classification index value, and each classification index value corresponds to a group of similar label data sets. Before the classifying is carried out on the feature description information of the sample image by using the classifier, the method further comprises the following steps: judging whether a classifier exists; if yes, classifying the feature description information of the sample image according to the existing classifier; if not, training a data set formed by the feature description information of all the sample images to generate a classifier. Preferably, training a data set composed of the feature description information to generate a classifier specifically includes: and generating a plurality of clustering centers by adopting a K-means clustering algorithm, and classifying the description data according to the distribution condition of the clustering centers by using a neighbor method. Preferably, the dimensionality reduction of the feature description information is to generate a dimensionality reduction matrix by adopting a Principal Component Analysis (PCA) method.
The chinese patent application with the application number of cn201811608994.x provides an image retrieval method and a related product, wherein an image retrieval device can compare each face image in a first face image set with each face image in a second face image set to obtain a plurality of comparison values, the comparison values can be understood as the similarity between each face image in the first face image and each face image in the second face image, if the comparison values are larger, the greater the similarity is, the greater the probability of representing the same face image is, the maximum comparison value can be selected from the plurality of comparison values, and the face image corresponding to the maximum comparison value is taken as a target object, so that the target object can be confirmed, and the accuracy of identifying the target object is improved;
chinese patent application No. CN201810903853.4 proposes an image retrieval method and system based on image segmentation and fuzzy pattern recognition, which performs image segmentation, sequentially performs similarity matching on each image region at the same position corresponding to a query image and each retrieved image, and comprehensively considers the similarity between each image region to measure the similarity between the query image and each retrieved image, thereby further enhancing the contrast between the images; in the image feature extraction process, the color feature and the texture feature of the image are comprehensively considered, so that the representativeness of the image feature is better than that when the color feature or the texture feature is considered independently; in the process of extracting the color features and the texture features of the image, a fuzzy mathematical algorithm is introduced, so that the representativeness of the color features and the texture features to the image can be further improved; the similarity between the query image and each image to be searched is comprehensively measured by utilizing the similarity between the k adjacent images of the query image and the k adjacent images of each image to be searched, so that the performance of the image retrieval system can be further improved; the neighbor number k is set as a dynamic parameter through a function, so that the adaptability of the image retrieval system to different query images can be further improved; and the performance of the image retrieval system can be further improved by carrying out information feedback through the satisfaction degree of the user.
The Chinese invention patent application with the application number of CN201810296979.X provides an image retrieval method and an image storage method, wherein a node of a service searching cluster extracts characteristics of an input image, the image and the characteristic data of the image are stored separately, an original image is stored in a database, and the characteristic data is stored in each node, so that each node can be matched with the locally stored image characteristic data independently during image retrieval.
However, in the above image identification or retrieval scheme in the prior art, the image to be queried or retrieved needs to be preprocessed mostly, including links such as image automation processing and manual labeling, and for each image, as many identification features as possible need to be extracted, and for a single image, the calculation amount is still barely acceptable; however, if the search engine is applied to a large amount of large-scale images, the above process will bring a large amount of computational complexity at the same time, so that the search efficiency is reduced.
Disclosure of Invention
In order to solve the technical problems, the invention provides an image matching and recognition system and method based on big data matrix stability analysis and a computer readable storage medium. According to the technical scheme, input original images are not preprocessed, but are firstly identified before being inquired, one or a group of sub-images containing the most key elements of the original images are identified from the original images, and then image matching search is carried out on the basis of the sub-images. The method for identifying the key elements with the most number comprises similarity judgment and characteristic matrix stability judgment, so that the input query retrieval elements can contain the key information of the original input image, the data processing amount can be greatly reduced, and the image retrieval speed is improved on the premise of not losing the main information of the image. The scheme of the invention can be carried out through an automatic computer image recognition program and a matching program without manual intervention and prior experience.
In particular, in the first aspect of the invention, a large data matrix stability analysis-based identification system is provided, which comprises an image segmentation module, an image feature extraction module, a feature data matrixing engine, a feature data stability judgment component and an identification feature output interface,
the image segmentation component is used for picking up at least one sub-image of an initial input image, and the size of the sub-image is smaller than that of the initial input image;
the feature extraction module is used for extracting a plurality of sub-image feature points aiming at the sub-images;
the characteristic data matrixing engine is used for forming a sub-image characteristic data matrix based on the extracted plurality of sub-image characteristic points;
the characteristic data stability judging component is used for judging the stability of a characteristic data matrix and outputting the stability on the identification characteristic output interface;
as one of the key technologies for embodying the innovation point of the present invention, the characteristic data stability determining component is configured to determine the stability of the characteristic data matrix, and specifically includes:
based on the plurality of sub-image feature points extracted aiming at the sub-images, restoring at least one comparison image by adopting an image reconstruction technology;
determining a similarity of the comparison image and the initial input image,
if the similarity is smaller than a preset threshold value, extracting a plurality of initial image feature points from an initial input image by using the feature extraction module;
the characteristic data matrixing engine is used for forming an initial image characteristic data matrix based on the extracted plurality of initial image characteristic points;
obtaining a combination matrix of the sub-image characteristic data matrix and the initial image characteristic data matrix,
if the trace of the combined matrix is larger than a preset value, outputting an identification result on the identification feature output interface;
if the similarity is greater than a predetermined threshold, updating the initial input image to at least one currently picked-up sub-image; the sub-image characteristic data matrix and the initial image characteristic data matrix are N-order matrixes with equal size, and N is a positive integer larger than 1.
It should be noted that, in the above recognition process, the size ratio of the sub-image to the initial input image is greater than a preset ratio value.
Preferably, the difference operation may be performed on the sub-image feature data matrix and the initial image feature data matrix to obtain the combination matrix.
Additionally, the initial input image is updated to the currently picked up at least one sub-image if the trace of the combined matrix is smaller than a predetermined value.
In a second aspect of the present invention, there is provided an image recognition method, which adopts the above-mentioned recognition system based on big data matrix stability analysis to perform a pattern recognition process and output at least one initial input image as a source input image (a retrieval image to be queried) of a subsequent image matching or retrieval system, the method comprising the following steps:
step S501: acquiring an original input image, and taking the original input image as an initial input image;
step S502: picking up at least one sub-image from the initial input image based on the set preset proportion value;
step S503: extracting a plurality of sub-image feature points of the sub-images;
step S504: reconstructing at least one comparison image based on the extracted plurality of image feature points;
step S505: judging the similarity between the comparison image and the initial input image;
step S506: judging whether the similarity is larger than a preset threshold value or not;
if yes, taking at least one currently picked sub-image as an initial input image, and returning to the step S502;
if not, go to step S507;
step S507: extracting a plurality of initial input image feature points of the initial input image;
step S508: respectively obtaining a sub-image characteristic data matrix and an initial image characteristic data matrix based on the image characteristic points extracted in the steps S503 and S507;
step S509: obtaining a combined matrix based on the sub-image characteristic data matrix and the initial image characteristic data matrix;
step S510: determining whether a trace of the combined matrix is less than a predetermined value,
if yes, taking at least one currently picked sub-image as an initial input image, and returning to the step S502;
otherwise, outputting the initial input image on the identification feature output interface.
In step S509, the method performs a difference operation on the sub-image characteristic data matrix and the initial image characteristic data matrix in combination with the preset ratio value to obtain the combined matrix.
In a third aspect of the present invention, an image matching system is provided, the image matching system is used for retrieving a matching image, for example, inputting an image to be inquired, and the image matching system outputs at least one target image meeting a predetermined condition.
Different from the prior art, the image matching system provided by the invention processes the image to be inquired through the image recognition system provided by the invention after the image to be inquired is input, so as to obtain at least one initial input image, and then inquires.
Preferably, if a plurality of initial input images are output, an initial input image having the greatest similarity to the original input image is taken as the input query image.
In a fourth aspect of the present invention, an image matching method is further provided, where the method queries, for a source input image, at least one target image whose matching degree meets a predetermined requirement in a massive image library, and specifically includes: and identifying the source input image by using the image identification method, outputting at least one initial input image, taking the initial input image as an input query image, and querying at least one target image with matching degree meeting a preset requirement in a massive image library.
The above methods of the present invention can be implemented automatically in the form of a computer program of instructions, and the execution process does not require human intervention, therefore, the present invention further provides a computer-readable storage medium having a computer-executable program of instructions stored thereon, which is executed by a processor and a memory, for implementing the aforementioned image recognition method and image matching method.
Further advantages of the invention will be apparent in the detailed description section in conjunction with the drawings attached hereto.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without creative efforts.
FIG. 1 is an overall architecture diagram of an image recognition system according to an embodiment of the present invention
FIG. 2 is a flow chart of an image recognition method according to an embodiment of the invention
FIG. 3 is a flow chart of an image recognition method according to another embodiment of the present invention
FIG. 4 is an overall architecture diagram of an image matching system of one embodiment of the present invention
FIG. 5 is a flow chart of an image matching method according to an embodiment of the invention
FIG. 6 is a graph comparing image matching accuracy of the present invention with that of the prior art
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention. The invention is further described with reference to the following drawings and detailed description:
fig. 1 is a diagram illustrating an overall architecture of an image recognition system according to an embodiment of the present invention.
In fig. 1, an identification system based on big data matrix stability analysis is provided, where the identification system includes an image segmentation module, an image feature extraction module, a feature data matrixing engine, a feature data stability determination component, and an identification feature output interface.
In this embodiment, the image segmentation component is configured to pick up at least one sub-image of the initial input image, the sub-image having a smaller size than the initial input image.
Preferably, the size ratio of the sub-image to the initial input image is greater than a preset ratio value.
The preset proportional value may be set in the system in advance, for example, to 90% or 0.95. It should be noted that, in order to enable the technical solution of the present invention to obtain a better recognition effect and enable the recognition result to be used for subsequent matching, the preset ratio value should be relatively high, for example, 90% or more.
As an example, assuming the initial input image size is X X Y, the sub-image size may be a b, where 1 > a/X > 0.9, 1 > b/Y > 0.9
In fig. 1, the feature extraction module is configured to extract a plurality of sub-image feature points for the sub-image.
There are many common image feature extraction methods in the field for extracting a plurality of image feature points for an image, and the present invention does not limit this, and as an illustrative example, a feature value, such as an RGB three-channel value, or an average value thereof, may be extracted for each pixel point of an image; edge recognition can be carried out on the image, and the characteristic value of each edge sub-picture is extracted; or, the picture is transformed, for example, a characteristic value is calculated for each p pixel points;
on the basis, the characteristic data matrixing engine is used for forming a sub-image characteristic data matrix based on the extracted plurality of sub-image characteristic points;
referring to the foregoing example, the feature data matrix may be a Z-order matrix, Z being less than X and less than Y;
as a key technical means for embodying the innovation point of the invention, the recognition system carries out image recognition in the following way:
the characteristic data stability judging component is used for judging the stability of a characteristic data matrix and outputting the stability on the identification characteristic output interface;
specifically, the feature data stability determining component is configured to determine the stability of the feature data matrix, and includes the following steps:
based on the plurality of sub-image feature points extracted aiming at the sub-images, restoring at least one comparison image by adopting an image reconstruction technology;
determining a similarity of the comparison image and the initial input image,
if the similarity is smaller than a preset threshold value, extracting a plurality of initial image feature points from an initial input image by using the feature extraction module;
the characteristic data matrixing engine is used for forming an initial image characteristic data matrix based on the extracted plurality of initial image characteristic points;
obtaining a combination matrix of the sub-image characteristic data matrix and the initial image characteristic data matrix,
if the trace of the combined matrix is larger than a preset value, outputting an identification result on the identification feature output interface;
updating the initial input image to at least one currently picked up sub-image if the trace of the combined matrix is less than a predetermined value;
if the similarity is greater than a predetermined threshold, updating the initial input image to at least one currently picked-up sub-image;
the subimage characteristic data matrix and the initial image characteristic data matrix are N-order matrixes with equal size, N is a positive integer larger than 1, and N is less than or equal to Z;
in this example, the sub-image feature data matrix is Pnext, and the initial image feature data matrix is Pcurrent;
the combined matrix may be Pcurrent-Pcurrent, i.e., a difference matrix of the two.
The image recognition method performed by the above recognition system of the present embodiment may further refer to the flow described in fig. 2-3.
In fig. 2, the method comprises an iterative loop consisting of steps S501-S510, in particular:
step S501: acquiring an original input image, and taking the original input image as an initial input image;
step S502: picking up at least one sub-image from the initial input image based on the set preset proportion value;
step S503: extracting a plurality of sub-image feature points of the sub-images;
step S504: reconstructing at least one comparison image based on the extracted plurality of image feature points;
step S505: judging the similarity between the comparison image and the initial input image;
step S506: judging whether the similarity is larger than a preset threshold value or not;
if yes, taking at least one currently picked sub-image as an initial input image, and returning to the step S502;
if not, go to step S507;
as one of the core concepts of the present invention, if the similarity of the sub-image and the initial input image is greater than the threshold, it means that the currently picked up sub-image can still be reduced again; therefore, the initial input image is updated, and the above process is continued until the picked-up sub-image can not be reduced;
step S507: extracting a plurality of initial input image feature points of the initial input image;
step S508: respectively obtaining a sub-image characteristic data matrix and an initial image characteristic data matrix based on the image characteristic points extracted in the steps S503 and S507;
step S509: obtaining a combined matrix based on the sub-image characteristic data matrix and the initial image characteristic data matrix;
step S510: determining whether a trace of the combined matrix is less than a predetermined value,
if yes, taking at least one currently picked sub-image as an initial input image, and returning to the step S502;
otherwise, outputting the initial input image on the identification feature output interface.
Since the combination matrix is the difference between the sub-image characteristic data matrix and the initial image characteristic data matrix, if the two retained information are similar, the trace of the combination matrix will be smaller than the predetermined value, which means that the sub-image can be further reduced.
As a further preferred option, the difference operation is performed on the sub-image characteristic data matrix and the initial image characteristic data matrix in combination with the preset ratio value to obtain the combined matrix.
For example, the sub-image feature data matrix is Pnext, the initial image feature data matrix is Pcurrent, the preset ratio value is 0.95,
the combined matrix may be Pnext-0.95 Pcurrent.
Through the above process, a sub-image that retains key information of an original image but has a minimum size may be output, and in an implementation, if a plurality of initial input images are output, the initial input image having a maximum similarity to the original input image is taken as an input query image.
The steps included in the method of fig. 2 may be performed in selected portions to provide greater efficiency. The key technical means is to use the trace of the combined matrix to carry out the cyclic judgment. Therefore, referring to fig. 3, a flowchart of an image recognition method according to another embodiment is also provided, and the method of fig. 3 includes the following steps:
picking up at least one sub-image from an initial input image according to the initial input image, extracting a plurality of sub-image characteristic points of the sub-image, and obtaining a sub-image characteristic data matrix
Extracting a plurality of initial input image characteristic points of an initial input image aiming at the initial input image to obtain an initial image characteristic data matrix;
the two steps can be executed in parallel without sequence, thereby improving the calculation efficiency;
on the basis, obtaining a combined matrix, then judging whether the trace of the combined matrix is smaller than a preset value, if so, taking at least one currently picked sub-image as an initial input image, and returning to the processing step aiming at the initial input image;
otherwise, outputting the initial input image on the identification feature output interface.
Fig. 4 is an overall architecture diagram of an image matching system of one embodiment of the present invention.
The image matching is used for retrieving a matching image, for example, an image to be inquired is input, and the image matching system outputs at least one target image meeting a preset condition.
Different from the prior art, the image matching system provided by the invention processes the image to be inquired through the image recognition system provided by the invention after the image to be inquired is input, so as to obtain at least one initial input image, and then inquires.
The image recognition and matching process described in fig. 1-4 does not require preprocessing operations such as noise reduction and smoothing, which are similar to those in the prior art, on the picture itself, nor labeling, including manual labeling or automatic labeling.
Preferably, if a plurality of initial input images are output, an initial input image having the greatest similarity to the original input image is taken as the input query image.
Referring next to fig. 5, fig. 5 is a flowchart of an image matching method according to an embodiment of the invention.
The method shown in fig. 5 is a method for searching at least one target image, of which the matching degree meets a predetermined requirement, in a massive image library for one source input image, and specifically includes: and identifying the source input image by using the image identification method, outputting at least one initial input image, taking the initial input image as an input query image, and querying at least one target image with matching degree meeting a preset requirement in a massive image library.
Fig. 6 is a graph comparing the image matching accuracy of the present invention with the prior art.
In fig. 6, a schematic illustration of the difference between the processing time and the accuracy used in the image recognition and matching process of different orders of magnitude in the solution of the present invention and two common prior art techniques is compared.
One type of prior art is a solution that includes a pre-processing process and another type of prior art is a solution that includes a manual annotation process.
It can be seen that when the number of pictures is extremely small, the recognition accuracy rate of the manual labeling method is the highest, and the processing time can be accepted; however, as the number of pictures increases significantly, manual labeling is hardly possible (e.g., the processing time is not shown to 10 in fig. 6)5Thereafter), and the accuracy decreases drastically; although the scheme including the preprocessing process can be automatically realized, the processing time of the scheme also increases exponentially with the increase of the picture order, the accuracy also decreases, and the scheme cannot be applied to image recognition and matching at a big data level;
in contrast, the scheme of the invention has longer processing time and lower accuracy when the number of pictures is less, but has better effects on both the processing time and the accuracy along with the increase of the order of magnitude of the pictures; according to the matrix stability criterion based on the big data picture characteristics, the larger the number is according to the matrix properties, the higher the stability accuracy embodied by matrix traces is, so that the accuracy is gradually and stably guaranteed along with the increase of the number of pictures; in terms of processing time, the technical scheme of the invention has no complex preprocessing process and labeling process and only performs feature extraction and matrix trace calculation, so that the calculated amount is also obviously reduced, and the processing time tends to be stable when the number of pictures is sharply increased.
Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that changes, modifications, substitutions and alterations can be made in these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.
Claims (9)
1. A recognition system based on big data matrix stability analysis comprises an image segmentation module, an image feature extraction module, a feature data matrixing engine, a feature data stability judgment component and a recognition feature output interface;
the image segmentation component is used for picking up at least one sub-image of an initial input image, and the size of the sub-image is smaller than that of the initial input image;
the feature extraction module is used for extracting a plurality of sub-image feature points aiming at the sub-images;
the characteristic data matrixing engine is used for forming a sub-image characteristic data matrix based on the extracted plurality of sub-image characteristic points;
the characteristic data stability judging component is used for judging the stability of a characteristic data matrix and outputting the stability on the identification characteristic output interface;
the method is characterized in that:
the characteristic data stability judging component is configured to judge stability of the characteristic data matrix, and specifically includes:
based on the plurality of sub-image feature points extracted aiming at the sub-images, restoring at least one comparison image by adopting an image reconstruction technology;
determining a similarity of the comparison image and the initial input image,
if the similarity is smaller than a preset threshold value, extracting a plurality of initial image feature points from an initial input image by using the feature extraction module;
the characteristic data matrixing engine is used for forming an initial image characteristic data matrix based on the extracted plurality of initial image characteristic points;
obtaining a combination matrix of the sub-image characteristic data matrix and the initial image characteristic data matrix;
if the trace of the combined matrix is larger than a preset value, outputting an identification result on the identification feature output interface; updating the initial input image to at least one currently picked up sub-image if the trace of the combined matrix is less than a predetermined value;
if the similarity is greater than a predetermined threshold, updating the initial input image to at least one currently picked-up sub-image; the sub-image characteristic data matrix and the initial image characteristic data matrix are N-order matrixes with equal size, and N is a positive integer larger than 1.
2. The identification system of claim 1, wherein:
and the size ratio of the sub-image to the initial input image is larger than a preset ratio value.
3. The identification system of claim 1, wherein:
and performing difference operation on the sub-image characteristic data matrix and the initial image characteristic data matrix to obtain the combined matrix.
4. The identification system of claim 3, wherein:
if the trace of the combination matrix is smaller than a predetermined value, the initial input image is updated to the currently picked up at least one sub-image.
5. A method for image recognition using feature matrix stability and image similarity, the method being based on the recognition system of any one of claims 1-4,
the method is characterized in that:
the method comprises the following steps:
step S501: acquiring an original input image, and taking the original input image as an initial input image;
step S502: picking up at least one sub-image from the initial input image based on the set preset proportion value;
step S503: extracting a plurality of sub-image feature points of the sub-images;
step S504: reconstructing at least one comparison image based on the extracted plurality of image feature points;
step S505: judging the similarity between the comparison image and the initial input image;
step S506: judging whether the similarity is larger than a preset threshold value or not;
if yes, taking at least one currently picked sub-image as an initial input image, and returning to the step S502;
if not, go to step S507;
step S507: extracting a plurality of initial input image feature points of the initial input image;
step S508: respectively obtaining a sub-image characteristic data matrix and an initial image characteristic data matrix based on the image characteristic points extracted in the steps S503 and S507;
step S509: obtaining a combined matrix based on the sub-image characteristic data matrix and the initial image characteristic data matrix;
step S510: determining whether a trace of the combined matrix is less than a predetermined value,
if yes, taking at least one currently picked sub-image as an initial input image, and returning to the step S502;
otherwise, outputting the initial input image on the identification feature output interface.
6. The method of claim 5, wherein:
step S509 further includes:
and performing difference operation on the sub-image characteristic data matrix and the initial image characteristic data matrix by combining the preset proportion value to obtain the combined matrix.
7. An image matching system connected to the recognition system of any one of claims 1-4, characterized by:
an image to be subjected to image matching is used as an original input image, at least one initial input image is output after the image is processed by the identification system of any one of claims 1 to 4, and the initial input image is used as an input query image and is input into the image matching system for matching query.
8. The image matching system of claim 7, wherein:
and if a plurality of initial input images are output, taking the initial input image with the maximum similarity with the original input image as an input query image.
9. An image matching method, which aims at a source input image, queries at least one target image with matching degree meeting a preset requirement in a massive image library,
the method is characterized in that:
identifying the source input image by using the method of any one of claims 5 to 6, outputting at least one initial input image, using the initial input image as an input query image, and querying at least one target image with matching degree meeting a predetermined requirement in a massive image library.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911321300.9A CN111178409B (en) | 2019-12-19 | 2019-12-19 | Image matching and recognition system based on big data matrix stability analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911321300.9A CN111178409B (en) | 2019-12-19 | 2019-12-19 | Image matching and recognition system based on big data matrix stability analysis |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111178409A CN111178409A (en) | 2020-05-19 |
| CN111178409B true CN111178409B (en) | 2021-11-16 |
Family
ID=70652077
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911321300.9A Active CN111178409B (en) | 2019-12-19 | 2019-12-19 | Image matching and recognition system based on big data matrix stability analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111178409B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111881191B (en) * | 2020-08-05 | 2021-06-11 | 留洋汇(厦门)金融技术服务有限公司 | Client portrait key feature mining system and method under mobile internet |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101251889A (en) * | 2007-12-25 | 2008-08-27 | 哈尔滨工业大学 | Personal identification method and near-infrared image forming apparatus based on palm vena and palm print |
| WO2012030872A1 (en) * | 2010-09-02 | 2012-03-08 | Edge3 Technologies Inc. | Method and apparatus for confusion learning |
| CN103207910A (en) * | 2013-04-08 | 2013-07-17 | 河南大学 | Image retrieval method based on hierarchical features and genetic programming relevance feedback |
| CN104091350A (en) * | 2014-06-20 | 2014-10-08 | 华南理工大学 | Object tracking method achieved through movement fuzzy information |
| CN105427296A (en) * | 2015-11-11 | 2016-03-23 | 北京航空航天大学 | Ultrasonic image low-rank analysis based thyroid lesion image identification method |
| CN105447100A (en) * | 2015-11-11 | 2016-03-30 | 宁波大学 | Cloud image retrieval method based on shape feature |
-
2019
- 2019-12-19 CN CN201911321300.9A patent/CN111178409B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101251889A (en) * | 2007-12-25 | 2008-08-27 | 哈尔滨工业大学 | Personal identification method and near-infrared image forming apparatus based on palm vena and palm print |
| WO2012030872A1 (en) * | 2010-09-02 | 2012-03-08 | Edge3 Technologies Inc. | Method and apparatus for confusion learning |
| CN103207910A (en) * | 2013-04-08 | 2013-07-17 | 河南大学 | Image retrieval method based on hierarchical features and genetic programming relevance feedback |
| CN104091350A (en) * | 2014-06-20 | 2014-10-08 | 华南理工大学 | Object tracking method achieved through movement fuzzy information |
| CN105427296A (en) * | 2015-11-11 | 2016-03-23 | 北京航空航天大学 | Ultrasonic image low-rank analysis based thyroid lesion image identification method |
| CN105447100A (en) * | 2015-11-11 | 2016-03-30 | 宁波大学 | Cloud image retrieval method based on shape feature |
Non-Patent Citations (2)
| Title |
|---|
| Orbit segmentation by surface reconstruction with vertex screening;Tai-Chiu Hsung;《2014 19th International Conference on Digital Signal Processing》;20140918;全文 * |
| 高分辨率遥感影像典型地物目标的特征选择及其稳定性研究;季金胜;《中国优秀硕士学位论文全文数据库》;20160215;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111178409A (en) | 2020-05-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US9430719B2 (en) | System and method for providing objectified image renderings using recognition information from images | |
| US10621755B1 (en) | Image file compression using dummy data for non-salient portions of images | |
| US7809192B2 (en) | System and method for recognizing objects from images and identifying relevancy amongst images and information | |
| US7809722B2 (en) | System and method for enabling search and retrieval from image files based on recognized information | |
| US7519200B2 (en) | System and method for enabling the use of captured images through recognition | |
| US20170024384A1 (en) | System and method for analyzing and searching imagery | |
| US20200334486A1 (en) | System and a method for semantic level image retrieval | |
| CN112738556B (en) | Video processing method and device | |
| CN112347284B (en) | Combined trademark image retrieval method | |
| US7277584B2 (en) | Form recognition system, form recognition method, program and storage medium | |
| Choi et al. | Automatic face annotation in personal photo collections using context-based unsupervised clustering and face information fusion | |
| AU2018202767A1 (en) | Data structure and algorithm for tag less search and svg retrieval | |
| CN113657087B (en) | Information matching method and device | |
| CN109241299B (en) | Multimedia resource searching method, device, storage medium and equipment | |
| CN108875828B (en) | Rapid matching method and system for similar images | |
| Singh et al. | Content-based image retrieval based on supervised learning and statistical-based moments | |
| Kalaiarasi et al. | Clustering of near duplicate images using bundled features | |
| CN109697240A (en) | A kind of image search method and device based on feature | |
| US20230072445A1 (en) | Self-supervised video representation learning by exploring spatiotemporal continuity | |
| CN111178409B (en) | Image matching and recognition system based on big data matrix stability analysis | |
| US11869127B2 (en) | Image manipulation method and apparatus | |
| Elakkiya et al. | Interactive real time fuzzy class level gesture similarity measure based sign language recognition using artificial neural networks | |
| JP2015097036A (en) | Recommended image presentation apparatus and program | |
| Maihami et al. | Color Features and Color Spaces Applications to the Automatic Image Annotation | |
| Zeng et al. | Robust and efficient visual tracking under illumination changes based on maximum color difference histogram and min-max-ratio metric |
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 |