A fast graphic image recognition method based on cloud computing

Abstract

The present invention discloses a method for rapid recognition of graphic images based on cloud computing, which relates to the field of image recognition technology and solves the technical problem that the existing technology is unable to rapidly recognize graphic images due to limitations on manpower, material resources and recognition efficiency during image recognition. The present invention obtains a large amount of basic image data through a data service platform connected to a data acquisition module, constructs an image feature library by combining tree classification technology and basic image features corresponding to the basic image data, and associates the basic image data with the corresponding terminal branches in the image feature library. By constructing a complete and orderly image feature library, a foundation is laid for the rapid recognition of subsequent graphic images. The present invention searches and locates the target features in an orderly manner in the image feature library to determine the terminal branches. The basic image data associated with the image to be recognized and the terminal branches are compared one by one to obtain recognition results. The data processing amount during large-scale image recognition is reduced, and the efficiency of graphic image recognition is improved.

Description

Technical Field

[0001] The present invention belongs to the field of image recognition, and relates to a cloud computing-based rapid graphic image recognition technology, and specifically to a cloud computing-based rapid graphic image recognition method. Background Art

[0002] Image recognition technology uses computers to process, analyze, and interpret images to identify various patterns of objects. Image recognition technology addresses the question of how computers can automatically process large amounts of physical information instead of humans, resolving problems that humans struggle to identify or that require excessive resources to identify, thereby reducing labor intensity.

[0003] When performing image recognition, the existing technology first collects and processes the target image, inputs the target image into a trained artificial intelligence model, and obtains the output recognition result. However, the training of the artificial intelligence model requires manual annotation of a large amount of sample data, which is time-consuming and labor-intensive, and the image recognition accuracy is not ideal; or the target image is compared with the standard image in the image library to identify the target image. This method has high requirements for the integrity of the image library, but is limited by the comparison efficiency and cannot quickly recognize; therefore, a cloud computing-based graphic image rapid recognition method is urgently needed. Summary of the Invention

[0004] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes a method for rapid graphic image recognition based on cloud computing, which is used to solve the technical problem that the prior art is unable to quickly recognize graphic images due to limitations in manpower, material resources and recognition efficiency during image recognition.

[0005] To achieve the above objectives, the first aspect of the present invention provides a method for rapid graphic image recognition based on cloud computing, comprising:

[0006] Obtain basic image data, perform image preprocessing on the basic image data, and then send it to the central analysis module; wherein the image preprocessing includes image segmentation, image denoising or grayscale conversion, and the central analysis module is built on the cloud server;

[0007] Identify basic image features of basic image data; construct an image feature tree based on the basic image features and associate the basic image data; wherein the basic image features include attribute features and graphic features, and the graphic features include point, line, and surface features;

[0008] The image to be identified is preprocessed, and then its basic image features are extracted and marked as target features; the target features are searched in the image feature tree to determine whether there are similar images; if so, similar images are extracted and compared with the image to be identified; if not, the image feature tree is updated according to the target features, and the image with the highest similarity is fed back to the smart terminal.

[0009] Preferably, identifying the basic image data and obtaining basic image features includes:

[0010] Receive basic image data and extract attribute features of the basic image data; wherein the attribute features include application fields and color features, and the color features include color components and the proportion of each color component;

[0011] The point features, line features and surface features of the basic image data are extracted in sequence to obtain graphic features; the attribute features and graphic features are integrated to generate basic image features.

[0012] Preferably, constructing an image feature tree based on the basic image features includes:

[0013] Divide the basic image features into several levels of features according to the hierarchical relationship, extract the correlation between the features of each level, and build an image feature tree by combining the tree classification technology;

[0014] The basic image data is associated with the end branch of the corresponding image feature tree and updated and stored regularly.

[0015] Preferably, performing image preprocessing on the image to be identified to obtain target features, and extracting similar images in combination with the image feature tree, includes:

[0016] Extract the application fields of the target features and locate the first-level branches of the image feature tree according to the application fields;

[0017] Extracting color features from target features, locating at least one secondary branch based on the primary branch according to the color features; and determining the final branch according to the graphic features, extracting basic image data associated with the final branch as a similar image.

[0018] Preferably, locating the secondary branches according to the color feature includes:

[0019] The color features in the target features are extracted, and the secondary branches are preliminarily matched according to the color components. Then, a fine match is performed according to the proportion of each color cost to locate the secondary branches that meet the requirements.

[0020] Preferably, determining the last branch according to the graphic features includes:

[0021] Compare the graphic features in the target feature with the graphic features in the secondary branches respectively, and obtain the point feature coefficient DTX, line feature coefficient XTX and surface feature coefficient MTX in turn;

[0022] The branch location coefficient ZDX is determined by the formula ZDX = α × DTX × XTX × MTX. The last branch is determined by judging whether the branch location coefficient is greater than the branch location threshold. α is a proportional coefficient set based on experience.

[0023] Preferably, comparing and analyzing the image to be identified with similar images includes:

[0024] Compare the target features of the image to be identified with the basic image features of similar images to obtain similarity;

[0025] When the similarities are all less than the similarity threshold, the image feature tree is updated according to the target feature; when the similarity with only one basic image data is greater than the similarity threshold, the basic image data is used as the recognition result; when the similarity with at least two basic image data is greater than the similarity threshold, the one with the greatest similarity is used as the recognition result; wherein the similarity threshold is set based on experience.

[0026] A cloud computing-based graphic image rapid recognition system is used to implement a cloud computing-based graphic image rapid recognition method, including a central analysis module, an intelligent terminal connected to it, and several data acquisition modules;

[0027] The data acquisition module obtains basic image data through the data service platform connected thereto;

[0028] The central analysis module constructs an image feature tree based on the basic image data collected by the data collection modules, and performs recognition analysis on the image to be recognized based on the image feature tree;

[0029] The intelligent terminal is used to collect images to be recognized and display recognition results.

[0030] Preferably, the central analysis module is in communication and / or electrically connected with the smart terminal and the plurality of data acquisition modules respectively; wherein the smart terminal includes a smart phone or a computer;

[0031] The data acquisition module is in communication with the data service platform; and after authorization, the data acquisition module obtains basic image data of the corresponding field through the data service platform.

[0032] A device for rapid recognition of graphic images based on cloud computing includes a storage medium and a processor. The storage medium stores operation instructions, and the processor executes the operation instructions to implement a method for rapid recognition of graphic images based on cloud computing.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] 1. The present invention obtains a large amount of basic image data through a data service platform connected to a data acquisition module, and constructs an image feature library by combining tree classification technology and basic image features corresponding to the basic image data, and associates the basic image data with the corresponding terminal branches in the image feature library; by constructing a complete and orderly image feature library, it lays the foundation for the rapid recognition of subsequent graphic images.

[0035] 2. The present invention extracts the target features of the image to be identified through a central analysis module, and locates the target features in an orderly manner in the image feature library to determine the terminal branches; the basic image data associated with the image to be identified and the terminal branches are compared one by one to obtain the recognition results; the data processing volume during large-scale image recognition is reduced, and the efficiency of graphic image recognition is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0037] Figure 1 It is a schematic diagram of the working steps of the present invention;

[0038] Figure 2 Schematic diagram of the system principle of the present invention. DETAILED DESCRIPTION

[0039] The technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0040] See also Figure 1The first embodiment of the present invention provides a method for rapid graphic image recognition based on cloud computing, which obtains basic image data, performs image preprocessing on the basic image data, and then sends it to a central analysis module; identifies basic image features of the basic image data; constructs an image feature tree based on the basic image features and associates the basic image data; performs image preprocessing on the image to be recognized, and then extracts its basic image features and marks them as target features; searches for the target features in the image feature tree to determine whether there are similar images; if so, extracts similar images and compares them with the image to be recognized; if not, updates the image feature tree based on the target features, and feeds back the image with the highest similarity to the smart terminal.

[0041] The recognition of graphic images is to identify the pattern of the image to be recognized, and determine the specific meaning of the pattern based on the self-built image library for user reference. If the accuracy and efficiency of graphic image recognition are to be improved, a complete image library and an efficient retrieval method are required. The present invention obtains basic image data through a data service platform, and conducts targeted analysis of the basic image data to construct an image feature tree; theoretically, the image feature tree constructed by the present invention can meet the needs of rapid recognition of graphic images in all fields when the data service platform is large enough and complete enough. After the image library is constructed, the next difficulty lies in how to perform fast and effective retrieval and return. The present invention divides the basic image data into several levels based on the basic image features. When comparing, the matching degree of the basic image features is used to continuously narrow the scope until a similar image is located (the similar image in the present invention is an image that has a high degree of the same basic image features as the image to be recognized after retrieval).

[0042] Image preprocessing includes image segmentation, image denoising, or grayscale conversion. The central analysis module is built on a cloud server. Image preprocessing is to ensure the accuracy of the image feature library and the matching accuracy of the image to be identified. It does not necessarily require that all basic image data have the same size and resolution, but at least the resolution meets the requirements and all basic image features are preserved.

[0043] The present invention identifies basic image data and obtains basic image features, including: receiving basic image data and extracting attribute features of the basic image data; extracting point features, line features and surface features of the basic image data in sequence to obtain graphic features; and integrating attribute features and graphic features to generate basic image features.

[0044] Attribute features are mainly surface features of the basic image data, specifically the application fields and color features corresponding to the basic image data; for example, the no-parking sign belongs to the field of traffic management, and its color features include red and blue (taking only one no-parking sign as an example). Point features, line features, and surface features in graphic features are sequentially included, that is, line features are extensions of point features, and surface features are extensions of line features. Each basic image feature should include attribute features and graphic features. The lack of a certain feature will affect the accuracy and efficiency of recognition. Basic image features can express the uniqueness of basic image data as much as possible and help to construct the image feature tree.

[0045] In an optional embodiment, an image feature tree is constructed based on basic image features, including: dividing the basic image features into several levels of features according to a hierarchical relationship, extracting the correlation relationship between the features of each level, and constructing an image feature tree in combination with tree classification technology; associating the basic image data with the end branches of the corresponding image feature tree, and updating and storing them regularly.

[0046] The image feature library is built based on basic image features. The final branch of the image feature tree corresponds to at least one piece of basic image data. This basic image data is then associated with the final branch to facilitate retrieval and extraction of similar images. If the features corresponding to the final branch are limited to a range, the final branch may correspond to multiple pieces of basic image data.

[0047] The division of basic image features is actually a hierarchical division from large to small, that is, constructing an image feature tree based on application domain, color features, point features, line features, and surface features. It should be noted that many images to be recognized are black and white, so color features can be omitted when constructing the image feature tree.

[0048] For application areas, the official economic industry classification table should be used as much as possible to establish each application area. Of course, application areas can also be determined based on the collected user consensus. Each application area is a first-level branch, which means that each application area has several second-level branches.

[0049] Color features (if not ignored) mainly include color components, that is, the colors they are composed of and the proportions of these color components. Considering the need to calculate the proportions of color components, the three primary colors can be used as color components, namely red, green, and blue. In other words, random combinations of these three colors can produce six secondary branches (red, green, blue, red-green, red-blue, green-blue, and red-green-blue). The ratio between red, green, and blue in an image is used as part of the color feature.

[0050] Point features mainly refer to the distribution characteristics of other color points in the basic image data except the background color, such as the distribution range relative to the central point of the basic image data and the distribution ratio of each distribution range. Specifically, the basic image data is evenly divided into twelve areas based on the center point (similar to a clock), and the proportion of other color points in each area except the background color is.

[0051] Line features mainly refer to the regular image features in each area, such as the number of straight lines, arcs, and intersections. Surface features analyze whether the lines can be combined into regular shapes, such as circles, triangles, letters, and Chinese characters.

[0052] In the subsequent similarity calculation process, the closer the distribution ranges of point features and the distribution ratios of each distribution range, the higher the similarity. The closer the number of lines, arcs, and intersections, the higher the similarity. The similarity is also higher if similar images are matched in surface features. Similarity calculation is carried out through mining using artificial intelligence models.

[0053] In the present invention, image preprocessing is performed on the image to be identified to obtain target features, and similar images are extracted in combination with the image feature tree, including: extracting the application field in the target feature, and locating the first-level branches of the image feature tree according to the application field; extracting the color feature in the target feature, and locating at least one second-level branch based on the first-level branch according to the color feature; and determining the last branch according to the graphic feature, and extracting the basic image data associated with the last branch as a similar image.

[0054] When sending an image to be identified through a smart terminal, you can enter its application field. If the application field is not entered, it will be intelligently identified through the central analysis module. If the application field is not identified, it will directly enter the identification of the next basic image feature.

[0055] By extracting the color features of the target features, specifically which of red, green, and blue are included, we can initially locate the secondary branches. Then, we can further locate them based on the ratio of red, green, and blue. For example, if the ratio is 1:1:1, it will fall into the branches corresponding to 1:1:0.5-1 in the image feature library, while 1:1:9.3 will fall into the branches corresponding to 1:1:9-9.5 in the image feature library. It should be understood that the secondary branches are located based on the color components, while the tertiary branches below the secondary branches are located based on the ratio of the color components.

[0056] In an optional embodiment, the last branch is determined based on the graphic features, including: comparing the graphic features in the target features with the graphic features in the secondary branches respectively, and obtaining the point feature coefficient DTX, the line feature coefficient XTX and the surface feature coefficient MTX in turn; determining the branch positioning coefficient ZDX by the formula ZDX = α × DTX × XTX × MTX; and determining the last branch by judging whether the branch positioning coefficient is greater than the branch positioning threshold.

[0057] The point feature coefficient, line feature coefficient, and surface feature coefficient are essentially the similarity between the target feature and the graphic features below the second-level branches (sometimes the third-level branches) in the image feature library. They can be mined and obtained through artificial intelligence models with nonlinear fitting capabilities. They can also be compared sequentially to determine the similarity. For example, if the distribution area and distribution ratio of the point feature are exactly the same, the number of straight lines and the number of intersection points in the line feature are exactly the same, and the triangles in the surface feature are successfully matched, then the similarity can be understood as 100%. Specifically, reasonable weights can be set for point features, line features, and surface features to calculate the branch positioning coefficient.

[0058] When the branch location coefficient meets the set branch location threshold, it means that the final branch has been located, and the basic image data associated with the final branch is highly likely to completely match the image to be identified. It is important to note that the final branch does not necessarily have to be one, but can be multiple. Appropriately setting the branch location threshold can reduce the number of final branches located.

[0059] In the present invention, the image to be identified is compared and analyzed with similar images, including: comparing the target features of the image to be identified and the basic image features of the similar images to obtain similarity; when the similarities are both less than the similarity threshold, updating the image feature tree according to the target features; when the similarity with only one basic image data is greater than the similarity threshold, the basic image data is used as the recognition result; when the similarity with at least two basic image data is greater than the similarity threshold, the one with the greatest similarity is used as the recognition result.

[0060] Finally, the image to be identified is compared one by one with the basic image data associated with the located end branches. Once the comparison is successful, similar images that match the image to be identified are identified, and relevant descriptions and applications of the similar images are extracted and fed back to the smart terminal, thus achieving rapid recognition of the graphic image.

[0061] See also Figure 2The second aspect of the present invention provides a cloud computing-based graphic image rapid recognition system for implementing a cloud computing-based graphic image rapid recognition method, including a central analysis module, and an intelligent terminal and several data acquisition modules connected thereto; the data acquisition module obtains basic image data through a data service platform connected thereto; the central analysis module constructs an image feature tree based on the basic image data collected by several data acquisition modules, and performs recognition analysis on the image to be recognized based on the image feature tree; the intelligent terminal is used to collect the image to be recognized and display the recognition results.

[0062] In the present invention, the central analysis module communicates and / or is electrically connected with the smart terminal and several data acquisition modules respectively; wherein, the smart terminal includes a smart phone or a computer; the data acquisition module communicates and is connected with the data service platform; and after authorization, the data acquisition module obtains basic image data of the corresponding field through the data service platform.

[0063] The third aspect of the present invention provides a device for rapid recognition of graphic images based on cloud computing, including a storage medium and a processor, wherein the storage medium stores operation instructions, and the processor executes the operation instructions to implement a method for rapid recognition of graphic images based on cloud computing.

[0064] Some of the data in the above formula are calculated by removing the dimensions and taking their numerical values. The formula is a formula that is closest to the actual situation obtained by software simulation of a large amount of collected data; the preset parameters and preset thresholds in the formula are set by technical personnel in this field according to actual conditions or obtained through simulation of a large amount of data.

[0065] Working principle of the present invention:

[0066] The data acquisition module obtains basic image data through the data service platform connected to it, performs image preprocessing on the basic image data, and then sends it to the central analysis module.

[0067] The central analysis module receives basic image data collected by several data collection modules, identifies basic image features of the basic image data, constructs an image feature tree based on the basic image features, and associates the basic image data.

[0068] The central analysis module extracts the target features corresponding to the image to be identified; searches the target features in the image feature tree to determine whether there are similar images; if so, extracts similar images and compares them with the image to be identified; if not, updates the image feature tree based on the target features and feeds back the image with the highest similarity to the smart terminal.

[0069] The above embodiments are only used to illustrate the technical method of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A method for rapid recognition of graphic images based on cloud computing, characterized in that: include: Obtain basic image data, perform image preprocessing on the basic image data, and then send it to the central analysis module; wherein the image preprocessing includes image segmentation, image denoising or grayscale conversion, and the central analysis module is built on the cloud server; Identify basic image features of basic image data; construct an image feature tree based on the basic image features and associate the basic image data; wherein the basic image features include attribute features and graphic features, and the graphic features include point, line, and surface features; Identifying the basic image data and obtaining basic image features includes: Receive basic image data and extract attribute features of the basic image data; wherein the attribute features include application fields and color features, and the color features include color components and the proportion of each color component; Extract point features, line features, and surface features of the basic image data in sequence to obtain graphic features; integrate attribute features and graphic features to generate basic image features; Perform image preprocessing on the image to be identified to obtain target features, and extract similar images based on the image feature tree, including: Extracting application fields from the target features and locating a first-level branch of the image feature tree based on the application fields; extracting color features from the target features and locating at least one second-level branch based on the first-level branch based on the color features; and determining a final branch based on the graphic features and extracting basic image data associated with the final branch as a similar image; Locating the secondary branches according to the color feature includes: Extract the color features from the target features, perform preliminary matching of secondary branches based on the color components, and then perform fine matching based on the proportion of each color cost to locate the secondary branches that meet the requirements; Determining the end branch according to the graphic feature includes: Compare the graphic features in the target feature with the graphic features in the secondary branches respectively, and obtain the point feature coefficient DTX, line feature coefficient XTX and surface feature coefficient MTX in turn; The branch positioning coefficient ZDX is determined by the formula ZDX=α×DTX×XTX×MTX. The last branch is determined by judging whether the branch positioning coefficient is greater than the branch positioning threshold. α is a proportional coefficient set based on experience. The image to be identified is preprocessed, and then its basic image features are extracted and marked as target features; the target features are searched in the image feature tree to determine whether there are similar images; if so, similar images are extracted and compared with the image to be identified; if not, the image feature tree is updated according to the target features.

2. The method for rapid graphic image recognition based on cloud computing according to claim 1, characterized in that: Constructing an image feature tree based on the basic image features includes: Divide the basic image features into several levels of features according to the hierarchical relationship, extract the correlation between the features of each level, and build an image feature tree by combining the tree classification technology; The basic image data is associated with the end branch of the corresponding image feature tree and updated and stored regularly.

3. The method for rapid graphic image recognition based on cloud computing according to claim 1, characterized in that: Comparing and analyzing the image to be identified with similar images includes: Compare the target features of the image to be identified with the basic image features of similar images to obtain similarity; When the similarities are all less than the similarity threshold, the image feature tree is updated according to the target feature; When the similarity of a basic image data is greater than the similarity threshold, the basic image data is used as the recognition result; when the similarity with at least two basic image data is greater than the similarity threshold, the one with the greatest similarity is used as the recognition result; wherein the similarity threshold is set based on experience.

4. A cloud computing-based graphic image rapid recognition system, used to implement the cloud computing-based graphic image rapid recognition method according to any one of claims 1 to 3, characterized in that: It includes a central analysis module, an intelligent terminal connected to it, and several data acquisition modules; The data acquisition module obtains basic image data through the data service platform connected to it; the central analysis module constructs an image feature tree based on the basic image data collected by the data acquisition modules, and performs recognition analysis on the image to be recognized based on the image feature tree; The intelligent terminal is used to collect images to be recognized and display recognition results.

5. The cloud computing-based graphic image rapid recognition system according to claim 4, characterized in that: The central analysis module is in communication and / or electrically connected with the smart terminal and the plurality of data acquisition modules respectively; wherein the smart terminal includes a smart phone or a computer; The data acquisition module is in communication with the data service platform; and after authorization, the data acquisition module obtains basic image data of the corresponding field through the data service platform.

6. A graphic image rapid recognition device based on cloud computing, characterized in that: It includes a storage medium and a processor, the storage medium stores operation instructions, and the processor executes the operation instructions to implement a cloud computing-based graphic image rapid recognition method as described in any one of claims 1 to 3.

Images