Image feature intelligent identification method and system based on multi-source big data

Abstract

The invention relates to the technical field of image recognition, in particular to an intelligent image feature recognition method and system based on multi-source big data, and the method comprises the steps: receiving an intelligent recognition instruction, obtaining a multi-source image set, extracting multi-source images, calculating image definition and image information amount, and carrying out the collection to obtain an image definition set and an image information amount set; screening a definition screening set and an information amount screening set, matching a target image set, extracting a first target image, obtaining a first image coordinate system, performing coordinate conversion on the target image, obtaining a unified target image set, setting a target feature image, extracting the unified target image, calculating comprehensive feature similarity, and summarizing the comprehensive feature similarity; and screening a high-similarity set, matching a target matching image set, and completing image feature intelligent identification based on the target matching image set. According to the invention, the precision of image feature intelligent identification can be improved.

Description

Technical Field

[0001] The present invention relates to the field of image recognition technology, and in particular to a method and system for intelligent recognition of image features based on multi-source big data. Background Art

[0002] With the continuous advancement of image processing technology, intelligent recognition of image features has become an important research direction in the field of modern computer vision, especially in complex environments. Intelligent recognition of image features plays a vital role in realizing automated target detection, accurate recognition and autonomous decision-making. Through in-depth analysis and feature extraction of multi-source images, image features can be accurately identified in changing and complex environments.

[0003] Traditional image feature recognition methods mainly rely on manually designed feature extraction algorithms, such as edge detection methods. These methods will reduce the accuracy of intelligent recognition of image features when there are many interference factors. Therefore, how to improve the accuracy of intelligent recognition of image features is an important issue that needs to be solved urgently. Summary of the invention

[0004] The present invention provides a method and system for intelligent recognition of image features based on multi-source big data, the main purpose of which is to improve the accuracy of intelligent recognition of image features.

[0005] To achieve the above object, the present invention provides an image feature intelligent recognition method based on multi-source big data, comprising:

[0006] Receiving an intelligent recognition instruction, acquiring a multi-source image set based on the intelligent recognition instruction, extracting multi-source images from the multi-source image set in sequence, and performing definition calculation and information volume calculation on the extracted multi-source images to obtain image definition and image information volume, respectively summarizing the image definition and image information volume to obtain an image definition set and an image information volume set;

[0007] Preliminarily screening the image clarity set and the image information set by using the preset clarity threshold and information threshold, respectively, to obtain a clarity screening set and an information screening set, wherein the clarity of the images in the clarity screening set is greater than the clarity threshold, and the information amount of the images in the information screening set is greater than the information threshold;

[0008] Matching a target image set according to the definition screening set and the information volume screening set, wherein the image definition of all target images in the target image set is greater than a definition threshold and the image information volume of the target images is greater than an information volume threshold;

[0009] Randomly extracting a first target image based on the target image set, obtaining a first image coordinate system of the first target image, and performing a coordinate transformation operation on all target images in the target image set based on the first image coordinate system and the first target image to obtain a unified target image set;

[0010] Setting a target feature image, extracting a unified target image in sequence based on the unified target image set, calculating a comprehensive feature similarity based on the unified target image and the target feature image, and summarizing the comprehensive feature similarities to obtain a similarity set;

[0011] The similarity set is finally screened according to a preset similarity threshold to obtain a high similarity set, and the multi-source image set is matched based on the high similarity set to obtain a target matching image set, and the image feature intelligent recognition is completed based on the target matching image set.

[0012] Optionally, the performing definition calculation and information volume calculation on the extracted multi-source images to obtain image definition and image information volume includes:

[0013] Obtain the number of pixel rows, the number of pixel columns and the image grayscale value of the multi-source image, and set the pixel horizontal index and the pixel column index;

[0014] The horizontal change rate and the vertical change rate are calculated based on the image grayscale value, and the image clarity is calculated based on the number of pixel rows, the number of pixel columns, the pixel horizontal index, the pixel column index, the horizontal change rate and the vertical change rate:

[0015]

[0016] Wherein, Q refers to the image definition, W refers to the number of pixel rows, E refers to the number of pixel columns, ω refers to the pixel horizontal index, e refers to the pixel column index, α(ω, e) refers to the horizontal change rate of the pixel point with the pixel horizontal index ω and the pixel column index e, and β(ω, e) refers to the vertical change rate of the pixel point with the pixel horizontal index ω and the pixel column index e;

[0017] Calculate the pixel ratio of multi-source images and calculate the image information based on the pixel ratio.

[0018] Optionally, the calculating the pixel ratio of the multi-source images and calculating the image information amount based on the pixel ratio includes:

[0019] Identify the grayscale values ​​in the multi-source image to obtain a recognition grayscale value set, wherein the recognition grayscale value set includes a plurality of recognition grayscale values, extract recognition grayscale values ​​from the recognition grayscale value set in sequence, and perform the following operations on the extracted recognition grayscale values:

[0020] The number of grayscale values ​​identical to the extracted identification grayscale value is counted in the multi-source image to obtain the identification number, the grayscale value index is set according to the identification grayscale value, and the pixel ratio is calculated based on the number of pixel rows, the identification number, the number of pixel columns and the grayscale value index:

[0021]

[0022] Among them, R μ Refers to the pixel ratio when the gray value index is μ, μ refers to the gray value index, t μ Refers to the number of recognitions when the gray value index is μ;

[0023] The adjustment parameter and the grayscale weight parameter are set according to the grayscale value index, and the image information amount is calculated based on the pixel ratio, grayscale value index, adjustment parameter and grayscale weight parameter:

[0024]

[0025] Among them, T refers to the amount of image information, δ refers to the adjustment parameter, ∈ μ Refers to the grayscale weight parameter when the grayscale value index is μ.

[0026] Optionally, the calculating of the comprehensive feature similarity based on the unified target image and the target feature image includes:

[0027] Calculating color feature similarity, shape feature similarity and texture feature similarity based on the unified target image and the target feature image;

[0028] The comprehensive feature similarity is calculated based on the color feature similarity, shape feature similarity and texture feature similarity:

[0029] G=ρ1×H+ρ2×J+ρ3×K

[0030] Among them, G refers to the comprehensive feature similarity, ρ1 refers to the preset color feature weight parameter, H refers to the color feature similarity, ρ2 refers to the preset shape feature weight parameter, J refers to the shape feature similarity, ρ3 refers to the preset texture feature weight parameter, and K refers to the texture feature similarity.

[0031] Optionally, the calculating the color feature similarity, the shape feature similarity and the texture feature similarity based on the unified target image and the target feature image includes:

[0032] Converting the unified target image into an HSV target image, and converting the target feature image into an HSV feature image;

[0033] Divide the HSV target image and the HSV feature image based on the preset horizontal and vertical division values ​​to obtain a target image block set and a feature image block set;

[0034] Extracting target image blocks in sequence according to the target image block set;

[0035] Constructing a target hue histogram of the extracted target image block, performing feature recognition on the target hue histogram to obtain a hue target vector, summarizing the hue target vectors to form a hue target vector set, and splicing the hue target vectors in the hue target vector set to obtain a target hue feature vector;

[0036] Obtaining a characteristic hue feature vector based on a characteristic image block set;

[0037] Calculate the color feature similarity based on the target hue feature vector and the feature hue feature vector;

[0038] Respectively performing object recognition on the unified target image and the target feature image to obtain a recognized object target image and a recognized object feature image;

[0039] Generate a target bounding box according to the target image of the identified object, generate a feature bounding box according to the feature image of the identified object, and calculate the similarity of the appearance features based on the target bounding box and the feature bounding box;

[0040] A target texture feature vector of the unified target image and a feature texture feature vector of the target feature image are obtained, and texture feature similarity is calculated based on the target texture feature vector and the feature texture feature vector.

[0041] Optionally, calculating the color feature similarity according to the target hue feature vector and the feature hue feature vector includes:

[0042] Set the adjustment fixed value, and calculate the color feature similarity based on the target hue feature vector, the feature hue feature vector and the adjustment fixed value:

[0043]

[0044] Among them, ||*|| refers to calculating the Euclidean distance, Z1 refers to the target hue feature vector, Z2 refers to the feature hue feature vector, Δ refers to dot multiplication, Refers to adjusting a fixed value.

[0045] Optionally, the calculating the similarity of the appearance features based on the target bounding box and the feature bounding box includes:

[0046] Calculate the boundary overlap value based on the target bounding box and the feature bounding box, and obtain the target geometric center coordinates of the target bounding box and the feature geometric center coordinates of the feature bounding box respectively;

[0047] Set the boundary weight parameter, and calculate the shape feature similarity based on the boundary overlap value, the target geometric center coordinates, the feature geometric center coordinates and the boundary weight parameter:

[0048]

[0049] Among them, θ refers to the boundary weight parameter, φ(F1, F2) refers to the boundary overlap value between the target bounding box F1 and the feature bounding box F2, V1 refers to the target geometric center coordinates, and V2 refers to the feature geometric center coordinates.

[0050] Optionally, the calculating of texture feature similarity based on the target texture feature vector and the feature texture feature vector comprises:

[0051] Calculate the texture feature similarity based on the target texture feature vector and the feature texture feature vector:

[0052]

[0053] Among them, υ refers to the preset gradient weight parameter, ε(C1) refers to the vector norm of the target texture feature vector C1, ε(C2) refers to the vector norm of the characteristic texture feature vector C2, C1 refers to the target texture feature vector, C2 refers to the characteristic texture feature vector, ξ(C1) refers to the rate of change of the grayscale value of the target texture feature vector C1, and ξ(C2) refers to the rate of change of the grayscale value of the characteristic texture feature vector C2.

[0054] Optionally, the final screening of the similarity set according to a preset similarity threshold to obtain a high similarity set includes:

[0055] Extracting comprehensive feature similarities from the similarity set in sequence, and performing the following operations on the extracted comprehensive feature similarities:

[0056] Determining whether the extracted comprehensive feature similarity is not less than the similarity threshold;

[0057] If it is confirmed that the extracted comprehensive feature similarity is not less than the similarity threshold, the extracted comprehensive feature similarity is confirmed as high similarity;

[0058] The high similarities are collected to obtain a high similarity set.

[0059] To achieve the above object, the present invention also provides an image feature intelligent recognition system based on multi-source big data, comprising:

[0060] A preliminary screening module is used to receive an intelligent recognition instruction, obtain a multi-source image set based on the intelligent recognition instruction, extract multi-source images from the multi-source image set in sequence, and perform clarity calculation and information volume calculation on the extracted multi-source images to obtain image clarity and image information volume, respectively summarize the image clarity and image information volume to obtain an image clarity set and an image information volume set; perform preliminary screening on the image clarity set and the image information volume set respectively using a preset clarity threshold and information volume threshold to obtain a clarity screening set and an information volume screening set, wherein the clarity of the images in the clarity screening set are all greater than the clarity threshold, and the information volume of the images in the information volume screening set are all greater than the information volume threshold;

[0061] A target matching module, used for matching a target image set according to the definition screening set and the information volume screening set, wherein the image definition of all target images in the target image set is greater than a definition threshold and the image information volume of the target images is greater than an information volume threshold;

[0062] A coordinate conversion module is used to randomly extract a first target image based on the target image set, obtain a first image coordinate system of the first target image, and perform a coordinate conversion operation on all target images in the target image set based on the first image coordinate system and the first target image to obtain a unified target image set;

[0063] The final screening module is used to set the target feature image, extract the unified target image in sequence based on the unified target image set, calculate the comprehensive feature similarity based on the unified target image and the target feature image, summarize the comprehensive feature similarity to obtain a similarity set; perform a final screening on the similarity set according to a preset similarity threshold to obtain a high similarity set, match the multi-source image set based on the high similarity set to obtain a target matching image set, and complete image feature intelligent recognition based on the target matching image set.

[0064] In order to solve the above problem, the present invention further provides an electronic device, the electronic device comprising:

[0065] A memory storing at least one instruction;

[0066] The processor executes the instructions stored in the memory to implement the above-mentioned image feature intelligent recognition method based on multi-source big data.

[0067] In order to solve the above problems, the present invention also provides a computer-readable storage medium, in which at least one instruction is stored. The at least one instruction is executed by a processor in an electronic device to implement the above-mentioned image feature intelligent recognition method based on multi-source big data.

[0068] The present invention solves the problems described in the background technology. First, an intelligent recognition instruction is received, and a multi-source image set is obtained based on the intelligent recognition instruction. The multi-source image set is obtained from different sources, so that the data source is more diverse, and can cover more diverse scenes and objects, providing a reliable data basis for subsequent image feature recognition; secondly, multi-source images are extracted in sequence based on the multi-source image set, and the image clarity and image information volume of the extracted multi-source images are calculated. Through the calculation of the image clarity and the image information volume, the image clarity and image information volume of the multi-source images can be evaluated; then, the image clarity set and the image information volume set are preliminarily screened by using the clarity threshold and the information volume threshold, respectively, to obtain the clarity screening set and the information volume screening set, and the image clarity below the clarity threshold and the image information volume below the information volume threshold are filtered out by the clarity threshold and the information volume threshold, thereby reducing the amount of unnecessary image data, alleviating the burden of subsequent processing, and improving the efficiency of image feature recognition; then, according to the clarity screening set and the information volume screening set, The target image set is selected and matched, and the target image set is matched on the basis of the clarity screening set and the information screening set, so as to improve the accuracy of image feature recognition; further, based on the first image coordinate system, coordinate conversion operation is performed on all target images in the target image set to obtain a unified target image set, and the spatial position and direction of all target images in the target image set are kept consistent through the coordinate conversion operation, so as to reduce the complexity of calculation, make the calculation of comprehensive feature similarity more accurate, and avoid the error caused by coordinate difference; then, the comprehensive feature similarity is calculated, and the comprehensive feature similarity combines the color feature similarity, the shape feature similarity and the texture feature similarity, and performs image feature recognition from three dimensions of color feature similarity, shape feature similarity and texture feature similarity, so as to improve the accuracy of image feature recognition; finally, the similarity set is finally screened according to the similarity threshold to obtain a high similarity set, and the high similarity is screened out by setting the similarity threshold to ensure that the target matching image set finally obtained matches the target feature image. Therefore, the present invention can improve the accuracy of image feature recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 A schematic diagram of a process flow of an image feature intelligent recognition method based on multi-source big data provided by an embodiment of the present invention;

[0070] Figure 2 A functional module diagram of an image feature intelligent recognition system based on multi-source big data provided by an embodiment of the present invention;

[0071] Figure 3 A schematic diagram of the structure of an electronic device for implementing the method for intelligently identifying image features based on multi-source big data provided by one embodiment of the present invention.

[0072] Description of reference numerals:

[0073] 1. Electronic device; 10. Processor; 11. Memory; 12. Bus.

[0074] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0075] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0076] The embodiment of the present application provides an image feature intelligent recognition method based on multi-source big data. The execution subject of the image feature intelligent recognition method based on multi-source big data includes but is not limited to at least one of the electronic devices such as a server and a terminal that can be configured to execute the method provided by the embodiment of the present application. In other words, the image feature intelligent recognition method based on multi-source big data can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc.

[0077] Reference Figure 1 FIG. 1 is a flow chart of an intelligent image feature recognition method based on multi-source big data provided by an embodiment of the present invention. In this embodiment, the intelligent image feature recognition method based on multi-source big data includes:

[0078] S1. Receive an intelligent recognition instruction, obtain a multi-source image set based on the intelligent recognition instruction, extract multi-source images from the multi-source image set in sequence, and perform clarity calculation and information volume calculation on the extracted multi-source images to obtain image clarity and image information volume, summarize the image clarity and image information volume respectively, and obtain an image clarity set and an image information volume set.

[0079] Explainable, intelligent recognition instructions refer to instructions initiated by humans for intelligent recognition of image features of multi-source big data, multi-source image set refers to a set formed by multi-source images, multi-source images refer to images from different sources, and multi-source refers to different sources of images, such as satellite images, medical images, etc. Image clarity refers to the clarity of multi-source images. The higher the image clarity, the higher the clarity of the multi-source image. Image information refers to the information volume of multi-source images. The information volume reflects the distribution of grayscale values ​​in multi-source images. The more uniform the distribution of grayscale values ​​in multi-source images, the greater the image information volume. For example, a natural landscape image contains trees, leaves, sunlight and shadows, and the natural landscape image contains all grayscale values ​​from black to white. The grayscale value distribution is uniform, so the natural landscape image has a large amount of information, where the grayscale value of black is 0 and the grayscale value of white is 255. The more uneven the distribution of grayscale values ​​in a multi-source image, the smaller the image information volume. For example, an image with black text on a white background only has grayscale values ​​of white and black, and the grayscale value distribution is uneven. The image with black text on a white background has a small amount of information. The image clarity set refers to the set formed by the image clarity of all multi-source images in the multi-source image set, and the image information set refers to the set formed by the image information of all multi-source images in the multi-source image set.

[0080] In detail, the step of calculating the clarity and information volume of the extracted multi-source images to obtain the image clarity and image information volume includes:

[0081] Obtain the number of pixel rows, the number of pixel columns and the image grayscale value of the multi-source image, and set the pixel horizontal index and the pixel column index;

[0082] The horizontal change rate and the vertical change rate are calculated based on the image grayscale value, and the image clarity is calculated based on the number of pixel rows, the number of pixel columns, the pixel horizontal index, the pixel column index, the horizontal change rate and the vertical change rate:

[0083]

[0084] Wherein, Q refers to the image definition, W refers to the number of pixel rows, E refers to the number of pixel columns, ω refers to the pixel horizontal index, e refers to the pixel column index, α(ω, e) refers to the horizontal change rate of the pixel point with the pixel horizontal index ω and the pixel column index e, and β(ω, e) refers to the vertical change rate of the pixel point with the pixel horizontal index ω and the pixel column index e;

[0085] Calculate the pixel ratio of multi-source images and calculate the image information based on the pixel ratio.

[0086] Explainable, the number of pixel rows refers to the number of rows of pixels in the multi-source image, that is, the height of the multi-source image, the number of pixel columns refers to the number of columns of pixels in the multi-source image, that is, the width of the multi-source image. For example, for a 1920×1080 image, the number of pixel rows is 1080, and the number of pixel columns is 1920. The image grayscale value refers to the grayscale value of the pixel in the multi-source image. The pixel horizontal index refers to the parameter for traversing the number of pixel rows, and the pixel column index refers to the parameter for traversing the number of pixel columns. The horizontal change rate refers to the rate of change of the grayscale value in the horizontal direction. The larger the horizontal change rate, the faster the grayscale value changes in the horizontal direction. The horizontal direction refers to the direction from left to right in the multi-source image. The vertical change rate refers to the rate of change of the grayscale value in the vertical direction. The larger the vertical change rate, the faster the grayscale value changes in the vertical direction. The vertical direction refers to the direction from top to bottom in the multi-source image. For example, the grayscale values ​​in the horizontal direction are 1, 4, 2, and 9, then the horizontal change rate is The calculation method of the vertical change rate is the same as that of the horizontal change rate, which will not be repeated here. The pixel ratio refers to the ratio of the number of pixels with a gray value of μ to the total number of pixels in the multi-source image. The value range of μ is 0 to 255, and the total number of pixels refers to the number of pixels in the multi-source image.

[0087] In detail, the step of calculating the pixel ratio of the multi-source images and calculating the image information amount based on the pixel ratio includes:

[0088] Identify the grayscale values ​​in the multi-source image to obtain a recognition grayscale value set, wherein the recognition grayscale value set includes a plurality of recognition grayscale values, extract recognition grayscale values ​​from the recognition grayscale value set in sequence, and perform the following operations on the extracted recognition grayscale values:

[0089] The number of grayscale values ​​identical to the extracted identification grayscale value is counted in the multi-source image to obtain the identification number, the grayscale value index is set according to the identification grayscale value, and the pixel ratio is calculated based on the number of pixel rows, the identification number, the number of pixel columns and the grayscale value index:

[0090]

[0091] Among them, R μ Refers to the pixel ratio when the gray value index is μ, μ refers to the gray value index, t μ Refers to the number of recognitions when the gray value index is μ;

[0092] The adjustment parameter and the grayscale weight parameter are set according to the grayscale value index, and the image information amount is calculated based on the pixel ratio, grayscale value index, adjustment parameter and grayscale weight parameter:

[0093]

[0094] Among them, T refers to the amount of image information, δ refers to the adjustment parameter, ∈ μRefers to the grayscale weight parameter when the grayscale value index is μ.

[0095] Explainable, the identification gray value set refers to the set of identification gray values, the identification gray value refers to the gray value in the multi-source image, the identification number refers to the number of pixels with the same number of identification gray values ​​in the identification gray value set, the gray value index refers to the parameter of traversing gray values, and the value range is 0 to 255, and the adjustment parameter refers to a positive number. Optionally, set the adjustment parameter to 10 -6 It is used to avoid log20 errors and improve the stability of image information calculation. The grayscale weight parameter refers to the parameter that adjusts the influence of grayscale value on the image information. The larger the grayscale weight parameter, the greater the influence of grayscale value on the image information.

[0096] S2. Use the preset clarity threshold and information threshold to preliminarily screen the image clarity set and image information set respectively to obtain a clarity screening set and an information screening set, wherein the image clarity in the clarity screening set is greater than the clarity threshold, and the image information in the information screening set is greater than the information threshold.

[0097] It can be explained that the clarity threshold and the information amount threshold are pre-set parameters for screening multi-source images. Using the preset clarity threshold and information amount threshold to preliminarily screen the image clarity set and the image information amount set respectively refers to the operations of screening the image clarity greater than the clarity threshold and screening the image information amount greater than the information amount threshold. The clarity screening set refers to the set obtained by screening the image clarity using the clarity threshold, and the information amount screening set refers to the set obtained by screening the image information amount using the information amount threshold.

[0098] S3. Matching a target image set according to the clarity screening set and the information volume screening set, wherein the image clarity of all target images in the target image set is greater than a clarity threshold and the image information volume of the target images is greater than an information volume threshold.

[0099] Explainably, the target image set refers to a set obtained by matching the clarity filtering set and the information filtering set, and the image clarity of all target images in the target image set is greater than the clarity threshold and the image information of the target images is greater than the information threshold.

[0100] S4. Randomly extract a first target image based on the target image set, obtain a first image coordinate system of the first target image, and perform a coordinate transformation operation on all target images in the target image set based on the first image coordinate system and the first target image to obtain a unified target image set.

[0101] It can be explained that the first target image refers to an image randomly extracted from the target image set, the first image coordinate system refers to a coordinate system constructed based on the first target image, optionally, the first image coordinate system takes the upper left corner of the first target image as the origin, the positive direction of the horizontal axis of the first image coordinate system points to the right of the origin, and the positive direction of the vertical axis of the first image coordinate system points to the bottom of the origin, and performing a coordinate transformation operation on all target images in the target image set refers to transforming the coordinates of all target images in the target image set into the first image coordinate system according to an affine transformation, so that the coordinates of all target images in the target image set are unified into the same coordinate system. This conversion process is a prior art and will not be repeated here. The unified target image set refers to a set formed by unified target images, and the unified target image refers to an image obtained after transforming the coordinates of the target image into the first image coordinate system.

[0102] Exemplarily, the target image set includes three target images, a target image is randomly extracted from the target image set as the first target image, and a coordinate system is constructed in the first target image, wherein the coordinate system can be a camera coordinate system or an artificially set coordinate system, and all target images in the target image set except the extracted first target image are coordinate transformed, so that the coordinate systems corresponding to the target images after the coordinate transformation are all the coordinate systems constructed in the first target image, and the target images after the coordinate transformation and the first target image are summarized to obtain a unified target image set.

[0103] S5. Setting a target feature image, extracting a unified target image in sequence based on the unified target image set, calculating a comprehensive feature similarity based on the unified target image and the target feature image, and summarizing the comprehensive feature similarities to obtain a similarity set.

[0104] Explainable, target feature image refers to an image containing target features that we need to identify a specific target, specific target refers to a target that we pre-set to be identified, for example, a car, target feature refers to the feature on a specific target, for example, DIY prints on a car body, comprehensive feature similarity refers to the similarity that combines color feature similarity, shape feature similarity and texture feature similarity, similarity refers to the degree of similarity between two images, the greater the similarity, the more similar the two images are, color feature similarity refers to the degree of similarity between the two images in color, the higher the color feature similarity, the higher the degree of similarity between the two images in color, shape feature similarity refers to the degree of similarity between the target features in the two images in shape, the higher the shape feature similarity, the higher the degree of similarity between the target features in the two images in shape, shape refers to the outline of the target feature, texture feature similarity refers to the degree of similarity between the two images in texture features, similarity set refers to the set composed of comprehensive feature similarities.

[0105] In detail, the calculation of the comprehensive feature similarity based on the unified target image and the target feature image includes:

[0106] Calculating color feature similarity, shape feature similarity and texture feature similarity based on the unified target image and the target feature image;

[0107] The comprehensive feature similarity is calculated based on the color feature similarity, shape feature similarity and texture feature similarity:

[0108] G=ρ1×H+ρ2×J+ρ3×K

[0109] Among them, G refers to the comprehensive feature similarity, ρ1 refers to the preset color feature weight parameter, H refers to the color feature similarity, ρ2 refers to the preset shape feature weight parameter, J refers to the shape feature similarity, ρ3 refers to the preset texture feature weight parameter, and K refers to the texture feature similarity.

[0110] Explainable, the color feature weight parameter refers to the parameter for adjusting the influence of color feature similarity on comprehensive feature similarity. The larger the color feature weight parameter, the greater the influence of color feature similarity on comprehensive feature similarity. The shape feature weight parameter refers to the parameter for adjusting the influence of shape feature similarity on comprehensive feature similarity. The larger the shape feature weight parameter, the greater the influence of shape feature similarity on comprehensive feature similarity. The texture feature weight parameter refers to the parameter for adjusting the influence of texture feature similarity on comprehensive feature similarity. The larger the texture feature weight parameter, the greater the influence of texture feature similarity on comprehensive feature similarity.

[0111] For example, there is currently a pile of fallen leaves with similar colors and shapes but different textures, and it is necessary to identify fallen leaves with different textures from other fallen leaves in the pile. Therefore, a larger texture feature weight parameter can be set to increase the influence of texture feature similarity on comprehensive feature similarity. The effects that can be achieved by the color feature weight parameter and the shape feature weight parameter are the same as those of the texture feature weight parameter, and will not be repeated here.

[0112] In detail, the calculating of color feature similarity, shape feature similarity and texture feature similarity based on the unified target image and the target feature image includes:

[0113] Converting the unified target image into an HSV target image, and converting the target feature image into an HSV feature image;

[0114] Divide the HSV target image and the HSV feature image based on the preset horizontal and vertical division values ​​to obtain a target image block set and a feature image block set;

[0115] Extracting target image blocks in sequence according to the target image block set;

[0116] Constructing a target hue histogram of the extracted target image block, performing feature recognition on the target hue histogram to obtain a hue target vector, summarizing the hue target vectors to form a hue target vector set, and splicing the hue target vectors in the hue target vector set to obtain a target hue feature vector;

[0117] Obtaining a characteristic hue feature vector based on a characteristic image block set;

[0118] Calculate the color feature similarity based on the target hue feature vector and the feature hue feature vector;

[0119] Respectively performing object recognition on the unified target image and the target feature image to obtain a recognized object target image and a recognized object feature image;

[0120] Generate a target bounding box according to the target image of the identified object, generate a feature bounding box according to the feature image of the identified object, and calculate the similarity of the appearance features based on the target bounding box and the feature bounding box;

[0121] A target texture feature vector of the unified target image and a feature texture feature vector of the target feature image are obtained, and texture feature similarity is calculated based on the target texture feature vector and the feature texture feature vector.

[0122] Explainably, HSV target image refers to the image obtained after performing HSV conversion on the unified target image, HSV refers to the HSV color model, performing HSV conversion on the unified target image refers to using the HSV color space to represent the unified target image, HSV color space refers to a color model that describes color by three components, and the three components are hue, saturation and brightness, HSV feature image refers to the image obtained after performing HSV conversion on the target feature image, and the horizontal division value and the vertical division value refer to the parameters used to divide the HSV target image and the HSV feature image. For example, the HSV target image and the HSV feature image are divided into 5×4 images, wherein 5 refers to the horizontal division value and 4 refers to the vertical division value, and the HSV target image and the HSV feature image can be divided into 20 parts according to the horizontal division value and the vertical division value. The target image block set refers to a set composed of target image blocks, the feature image block set refers to a set composed of feature image blocks, the target image block refers to an image block obtained by dividing the HSV target image according to the horizontal division value and the vertical division value, the feature image block refers to an image block obtained by dividing the HSV feature image according to the horizontal division value and the vertical division value, the target hue histogram is used to represent the distribution of different hues in the target image block, and the construction process of the target hue histogram is: divide the hue range into ten equally spaced intervals, the hue range is [0, 1], obtain the hue values ​​of all pixels in the target image block, count the number of pixels in each interval, and construct the target hue histogram with the interval as the horizontal axis and the number of pixels in each interval as the vertical axis. The hue value refers to the specific numerical value of the hue, and different hue values ​​represent different colors. For example, when the hue value is 0, the color is red.For example, the information in the target hue histogram is: when the interval is 0-0.1, the number of pixels is 10, when the interval is 0.1-0.2, the number of pixels is 200, when the interval is 0.2-0.3, the number of pixels is 300, when the interval is 0.3-0.4, the number of pixels is 100, when the interval is 0.4-0.5, the number of pixels is 200, when the interval is 0.5-0.6, the number of pixels is 200, when the interval is 0.6-0.7, the number of pixels is 100, when the interval is 0.7-0.8, the number of pixels is 100, when the interval is 0.8-0.9, the number of pixels is 100, when the interval is 0.9-1.0, the number of pixels is 10, and different intervals are marked, and the interval 0 is -0.1 is identified as 1, the interval 0.1-0.2 is identified as 2, the interval 0.2-0.3 is identified as 3, the interval 0.3-0.4 is identified as 4, the interval 0.4-0.5 is identified as 5, the interval 0.5-0.6 is identified as 6, the interval 0.6-0.7 is identified as 7, the interval 0.7-0.8 is identified as 8, the interval 0.8-0.9 is identified as 9, and the interval 0.9-1.0 is identified as 10. The hue target vector is 10 vectors, each interval corresponds to a hue target vector, and the hue target vector corresponding to the interval 0-0.1 is: X1={1, 10}, the hue target vector set refers to the set composed of hue target vectors, and the target hue feature vector refers to the feature vector obtained by splicing all the hue target vectors in the hue target vector set, for example, X=[X1, X2,…, X]. 10 ], X refers to the target hue feature vector, X1 refers to the hue target vector corresponding to the interval 0-0.1, X2 refers to the hue target vector corresponding to the interval 0.1-0.2, X 10 Refers to the hue target vector corresponding to the interval 0.9-1.0. The method of obtaining the characteristic hue feature vector is the same as the method of obtaining the target hue feature vector, which will not be repeated here.

[0123] It can be understood that respectively performing object recognition on the unified target image and the target feature image means using existing technology to realize the recognition of the target to be detected in the image. Optionally, the unified target image and the target feature image are recognized by the YOLO target detection algorithm. The YOLO target detection algorithm is an existing technology and will not be repeated here. Recognizing the object target image refers to the image of the object obtained after the unified target image is recognized by the YOLO target detection algorithm. Optionally, recognizing the object feature image refers to the image of the object obtained after the target feature image is recognized by the YOLO target detection algorithm. For example, the target feature image is recognized by the YOLO target detection algorithm. After recognition, an image of a car is obtained. Generating a target bounding box according to the target image of the recognized object refers to the bounding box information of the target image of the recognized object output by the YOLO target detection algorithm, and then using OpenCV to draw the bounding box using the bounding box information. The method of generating a feature bounding box according to the feature image of the recognized object is consistent with the method of generating a target bounding box according to the target image of the recognized object, which will not be repeated here. The bounding box information refers to the geometric center of the bounding box, the width of the bounding box and the height of the bounding box. The bounding box refers to a minimum rectangular box that completely surrounds the target image of the recognized object and the feature image of the recognized object. Minimum refers to the smallest area of ​​the rectangular box. OpenCV is a prior art and will not be repeated here. The target bounding box is the smallest image area that can include the target image of the identified object, the feature bounding box refers to the smallest image area that can include the feature image of the identified object, the target texture feature vector refers to the vector representing the texture information of the unified target image, the feature texture feature vector refers to the vector representing the texture information of the target feature image, the texture information refers to the direction of change of the grayscale value in the unified target image and the target feature image, the spatial distribution of the grayscale value and the rate of change of the grayscale value, the direction of change of the grayscale value refers to the direction in which the grayscale value changes with the change of the spatial position. For example, in a face image, the grayscale value of the bridge of the nose area is higher than that of the cheek area, and the direction of change of the grayscale value is from a larger grayscale value to a smaller grayscale value. Therefore, in a face image, the direction of change of the grayscale value is from the bridge of the nose area to the cheek area.The unified target image and the target feature image are both regarded as two-dimensional planes. Optionally, the spatial position is represented by the two-dimensional coordinates of the two-dimensional coordinate system of the unified target image and the target feature image. The two-dimensional coordinate system refers to the upper left corner vertex of the unified target image as the origin. The positive direction of the horizontal axis of the two-dimensional coordinate system points to the right of the origin, and the positive direction of the vertical axis of the two-dimensional coordinate system points to the bottom of the origin. The two-dimensional coordinate refers to the coordinate expressing the position in the two-dimensional coordinate system. The spatial distribution of grayscale values ​​describes the distribution of grayscale values ​​in the two-dimensional plane. The distribution conditions include: uniform distribution and uneven distribution. Uniform distribution means that the standard deviation of grayscale values ​​is less than or equal to 10. For example, the image of the blue sky, the image of white paper, and the image of the blue sky The standard deviation of the grayscale values ​​of images such as images of white paper is less than or equal to 10. Uneven distribution means that the standard deviation of the grayscale value is greater than 10. For example, in images of cities, there are various types of objects in the city, such as cars, houses, etc. The standard deviation of the grayscale value is greater than 10. The rate of change of the grayscale value refers to the speed at which the grayscale value changes with the spatial position. The greater the rate of change of the grayscale value, the faster the grayscale value changes with the spatial position. For example, there are five grayscale values ​​of 30, 50, 70, 90 and 110 in the image. The five grayscale values ​​are arranged from left to right in the order of 30, 50, 70, 90 and 110. The rate of change of the grayscale value between 30 and 50 is calculated as. Calculate the rate of change of gray values ​​between gray values ​​50 and 70 as The change rate of the grayscale values ​​between the remaining grayscale values ​​is consistent with the change rate of the grayscale values ​​between the calculated grayscale values ​​30 and 50, which will not be repeated here. The target texture feature vector and the feature texture feature vector are both extracted through the local binary pattern, which is a prior art and will not be repeated here.

[0124] In detail, the color feature similarity is calculated according to the target hue feature vector and the feature hue feature vector, including:

[0125] Set the adjustment fixed value, and calculate the color feature similarity based on the target hue feature vector, the feature hue feature vector and the adjustment fixed value:

[0126]

[0127] Among them, ||*|| refers to calculating the Euclidean distance, Z1 refers to the target hue feature vector, Z2 refers to the feature hue feature vector, Δ refers to dot multiplication, Refers to adjusting a fixed value.

[0128] Interpretable, the adjustment fixed value refers to a positive value, optionally, the adjustment fixed value is set to 10 -6 , which is used to prevent the denominator from being 0, improves the stability of color feature similarity calculation. Calculating the Euclidean distance is an existing technology and will not be described in detail here.

[0129] In detail, the calculation of the shape feature similarity based on the target bounding box and the feature bounding box includes:

[0130] Calculate the boundary overlap value based on the target bounding box and the feature bounding box, and obtain the target geometric center coordinates of the target bounding box and the feature geometric center coordinates of the feature bounding box respectively;

[0131] Set the boundary weight parameter, and calculate the shape feature similarity based on the boundary overlap value, the target geometric center coordinates, the feature geometric center coordinates and the boundary weight parameter:

[0132]

[0133] Among them, θ refers to the boundary weight parameter, φ(F1, F2) refers to the boundary overlap value between the target bounding box F1 and the feature bounding box F2, V1 refers to the target geometric center coordinates, and V2 refers to the feature geometric center coordinates.

[0134] Explainably, the boundary overlap value refers to the ratio of the area of ​​the intersection of the target bounding box and the feature bounding box to the area of ​​the union of the target bounding box and the feature bounding box, the target geometric center coordinates refer to the geometric center of the target bounding box, the feature geometric center coordinates refer to the geometric center of the feature bounding box, and the boundary weight parameter refers to the parameter for adjusting the influence of the boundary overlap value on the similarity of the shape features. The larger the boundary weight parameter, the greater the influence of the boundary overlap value on the similarity of the shape features.

[0135] In detail, the calculating of texture feature similarity based on the target texture feature vector and the feature texture feature vector includes:

[0136] Calculate the texture feature similarity based on the target texture feature vector and the feature texture feature vector:

[0137]

[0138] Among them, v refers to the preset gradient weight parameter, ε(C1) refers to the vector norm of the target texture feature vector C1, ε(C2) refers to the vector norm of the characteristic texture feature vector C2, C1 refers to the target texture feature vector, C2 refers to the characteristic texture feature vector, ξ(C1) refers to the rate of change of the grayscale value of the target texture feature vector C1, and ξ(C2) refers to the rate of change of the grayscale value of the characteristic texture feature vector C2.

[0139] It can be explained that the rate of change of the grayscale value of the target texture feature vector C1 is obtained by unifying the texture information of the target image. The gradient weight parameter refers to a parameter for adjusting the influence of the gradient difference value on the texture feature similarity. The larger the gradient weight parameter, the greater the influence of the gradient difference value on the texture feature similarity. The gradient difference value refers to ||ξ(C1)-ξ(C2)|| in the above formula. The vector norm points to the Euclidean norm of the quantity, which is the prior art and will not be repeated here.

[0140] S6. Perform a final screening on the similarity set according to a preset similarity threshold to obtain a high similarity set, match the multi-source image set based on the high similarity set to obtain a target matching image set, and complete intelligent image feature recognition based on the target matching image set.

[0141] Explainably, the similarity threshold refers to the value used to filter the similarity set, the high similarity set refers to the set composed of high similarity, high similarity refers to the comprehensive feature similarity greater than the similarity threshold, the target matching image set refers to the set composed of target matching images, and the target matching image refers to the multi-source image corresponding to the high similarity in the high similarity set.

[0142] In detail, the similarity set is finally screened according to a preset similarity threshold to obtain a high similarity set, including:

[0143] Extracting comprehensive feature similarities from the similarity set in sequence, and performing the following operations on the extracted comprehensive feature similarities:

[0144] Determining whether the extracted comprehensive feature similarity is not less than the similarity threshold;

[0145] If it is confirmed that the extracted comprehensive feature similarity is not less than the similarity threshold, the extracted comprehensive feature similarity is confirmed as high similarity;

[0146] The high similarities are collected to obtain a high similarity set.

[0147] The present invention solves the problems described in the background technology. First, an intelligent recognition instruction is received, and a multi-source image set is obtained based on the intelligent recognition instruction. The multi-source image set is obtained from different sources, so that the data source is more diverse, and can cover more diverse scenes and objects, providing a reliable data basis for subsequent image feature recognition; secondly, multi-source images are extracted in sequence based on the multi-source image set, and the image clarity and image information volume of the extracted multi-source images are calculated. Through the calculation of the image clarity and the image information volume, the image clarity and image information volume of the multi-source images can be evaluated; then, the image clarity set and the image information volume set are preliminarily screened by using the clarity threshold and the information volume threshold, respectively, to obtain the clarity screening set and the information volume screening set, and the image clarity below the clarity threshold and the image information volume below the information volume threshold are filtered out by the clarity threshold and the information volume threshold, thereby reducing the amount of unnecessary image data, alleviating the burden of subsequent processing, and improving the efficiency of image feature recognition; then, according to the clarity screening set and the information volume screening set, The target image set is selected and matched, and the target image set is matched on the basis of the clarity screening set and the information screening set, so as to improve the accuracy of image feature recognition; further, based on the first image coordinate system, coordinate conversion operation is performed on all target images in the target image set to obtain a unified target image set, and the spatial position and direction of all target images in the target image set are kept consistent through the coordinate conversion operation, so as to reduce the complexity of calculation, make the calculation of comprehensive feature similarity more accurate, and avoid the error caused by coordinate difference; then, the comprehensive feature similarity is calculated, and the comprehensive feature similarity combines the color feature similarity, the shape feature similarity and the texture feature similarity, and performs image feature recognition from three dimensions of color feature similarity, shape feature similarity and texture feature similarity, so as to improve the accuracy of image feature recognition; finally, the similarity set is finally screened according to the similarity threshold to obtain a high similarity set, and the high similarity is screened out by setting the similarity threshold to ensure that the target matching image set finally obtained matches the target feature image. Therefore, the present invention can improve the accuracy of image feature recognition.

[0148] like Figure 2 , which is a functional module diagram of an image feature intelligent recognition system based on multi-source big data provided by one embodiment of the present invention.

[0149] The image feature intelligent recognition system 100 based on multi-source big data of the present invention can be installed in an electronic device. According to the functions implemented, the image feature intelligent recognition system 100 based on multi-source big data can include a preliminary screening module 101, a target matching module 102, a coordinate conversion module 103 and a final screening module 104. The module of the present invention can also be called a unit, which refers to a series of computer program segments that can be executed by an electronic device processor and can complete fixed functions, which are stored in the memory of the electronic device.

[0150] The preliminary screening module 101 is used to receive an intelligent recognition instruction, obtain a multi-source image set based on the intelligent recognition instruction, extract multi-source images from the multi-source image set in sequence, and perform clarity calculation and information volume calculation on the extracted multi-source images to obtain image clarity and image information volume, respectively summarize the image clarity and image information volume to obtain an image clarity set and an image information volume set; perform preliminary screening on the image clarity set and the image information volume set respectively using a preset clarity threshold and information volume threshold to obtain a clarity screening set and an information volume screening set, wherein the clarity of the images in the clarity screening set are all greater than the clarity threshold, and the information volume of the images in the information volume screening set are all greater than the information volume threshold;

[0151] The target matching module 102 is used to match the target image set according to the clarity screening set and the information volume screening set, wherein the image clarity of all target images in the target image set is greater than a clarity threshold and the image information volume of the target images is greater than an information volume threshold;

[0152] The coordinate conversion module 103 is used to randomly extract a first target image based on the target image set, obtain a first image coordinate system of the first target image, and perform a coordinate conversion operation on all target images in the target image set based on the first image coordinate system and the first target image to obtain a unified target image set;

[0153] The final screening module 104 is used to set a target feature image, extract a unified target image in sequence based on the unified target image set, calculate a comprehensive feature similarity based on the unified target image and the target feature image, summarize the comprehensive feature similarity to obtain a similarity set; perform a final screening on the similarity set according to a preset similarity threshold to obtain a high similarity set, match the multi-source image set based on the high similarity set to obtain a target matching image set, and complete intelligent recognition of image features based on the target matching image set.

[0154] In detail, each module in the image feature intelligent recognition system 100 based on multi-source big data in the embodiment of the present invention is used in the same manner as above. Figure 1 The same technical means are used as the method for intelligent recognition of image features based on multi-source big data described in , and can produce the same technical effects, so they will not be repeated here.

[0155] like Figure 3 , is a schematic diagram of the structure of an electronic device for implementing an image feature intelligent recognition method based on multi-source big data provided by an embodiment of the present invention.

[0156] The electronic device 1 may include a processor 10, a memory 11 and a bus 12, and may also include a computer program stored in the memory 11 and executable on the processor 10, such as an image feature intelligent recognition method program based on multi-source big data.

[0157] Among them, the memory 11 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, mobile hard disk, multimedia card, card-type memory (for example: SD or DX memory, etc.), magnetic memory, disk, optical disk, etc. The memory 11 can be an internal storage unit of the electronic device 1 in some embodiments, such as a mobile hard disk of the electronic device 1. The memory 11 can also be an external storage device of the electronic device 1 in other embodiments, such as a plug-in mobile hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device 1. Further, the memory 11 also includes an internal storage unit of the electronic device 1 and an external storage device. The memory 11 can not only be used to store application software and various types of data installed in the electronic device 1, such as the code of the image feature intelligent recognition method program based on multi-source big data, but also can be used to temporarily store data that has been output or is to be output.

[0158] The processor 10 may be composed of an integrated circuit in some embodiments, for example, a single packaged integrated circuit, or a plurality of packaged integrated circuits with the same or different functions, including one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and combinations of various control chips, etc. The processor 10 is the control core (Control Unit) of the electronic device, and uses various interfaces and lines to connect various components of the entire electronic device, and executes or executes programs or modules stored in the memory 11 (such as image feature intelligent recognition method programs based on multi-source big data, etc.), and calls data stored in the memory 11 to execute various functions of the electronic device 1 and process data.

[0159] The bus 12 may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus 12 may be divided into an address bus, a data bus, a control bus, etc. The bus 12 is configured to realize connection and communication between the memory 11 and at least one processor 10, etc.

[0160] Figure 3 Only an electronic device with components is shown, and those skilled in the art will understand that Figure 3 The structure shown does not constitute a limitation on the electronic device 1, and may include fewer or more components than shown in the figure, or combine certain components, or arrange the components differently.

[0161] For example, although not shown, the electronic device 1 may also include a power source (such as a battery) for supplying power to each component. Preferably, the power source may be logically connected to the at least one processor 10 through a power management device, so that the power management device can realize functions such as charging management, discharging management, and power consumption management. The power source may also include any components such as one or more DC or AC power sources, recharging devices, power failure detection circuits, power converters or inverters, power status indicators, etc. The electronic device 1 may also include a variety of sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be repeated here.

[0162] Furthermore, the electronic device 1 may also include a network interface. Optionally, the network interface may include a wired interface and / or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), which is generally used to establish a communication connection between the electronic device 1 and other electronic devices.

[0163] Optionally, the electronic device 1 may further include a user interface, which may be a display, an input unit (such as a keyboard), or a standard wired interface or a wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, and an OLED (Organic Light-Emitting Diode) touch device. The display may also be appropriately referred to as a display screen or a display unit, which is used to display information processed in the electronic device 1 and to display a visual user interface.

[0164] The image feature intelligent recognition method program based on multi-source big data stored in the memory 11 of the electronic device 1 is a combination of multiple instructions. When running in the processor 10, it can achieve:

[0165] Receiving an intelligent recognition instruction, acquiring a multi-source image set based on the intelligent recognition instruction, extracting multi-source images from the multi-source image set in sequence, and performing definition calculation and information volume calculation on the extracted multi-source images to obtain image definition and image information volume, respectively summarizing the image definition and image information volume to obtain an image definition set and an image information volume set;

[0166] Preliminarily screening the image clarity set and the image information set by using the preset clarity threshold and information threshold, respectively, to obtain a clarity screening set and an information screening set, wherein the clarity of the images in the clarity screening set is greater than the clarity threshold, and the information amount of the images in the information screening set is greater than the information threshold;

[0167] Matching a target image set according to the definition screening set and the information volume screening set, wherein the image definition of all target images in the target image set is greater than a definition threshold and the image information volume of the target images is greater than an information volume threshold;

[0168] Randomly extracting a first target image based on the target image set, obtaining a first image coordinate system of the first target image, and performing a coordinate transformation operation on all target images in the target image set based on the first image coordinate system and the first target image to obtain a unified target image set;

[0169] Setting a target feature image, extracting a unified target image in sequence based on the unified target image set, calculating a comprehensive feature similarity based on the unified target image and the target feature image, and summarizing the comprehensive feature similarities to obtain a similarity set;

[0170] The similarity set is finally screened according to a preset similarity threshold to obtain a high similarity set, and the multi-source image set is matched based on the high similarity set to obtain a target matching image set, and the image feature intelligent recognition is completed based on the target matching image set.

[0171] Specifically, the specific implementation method of the processor 10 for the above instructions can refer to Figures 1 to 3 The description of the relevant steps in the corresponding embodiments will not be repeated here.

[0172] Furthermore, if the module / unit integrated in the electronic device 1 is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. The computer-readable storage medium can be volatile or non-volatile. For example, the computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, and a read-only memory (ROM).

[0173] The present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor of an electronic device, the computer program can implement:

[0174] Receiving an intelligent recognition instruction, acquiring a multi-source image set based on the intelligent recognition instruction, extracting multi-source images from the multi-source image set in sequence, and performing definition calculation and information volume calculation on the extracted multi-source images to obtain image definition and image information volume, respectively summarizing the image definition and image information volume to obtain an image definition set and an image information volume set;

[0175] Preliminarily screening the image clarity set and the image information set by using the preset clarity threshold and information threshold, respectively, to obtain a clarity screening set and an information screening set, wherein the clarity of the images in the clarity screening set is greater than the clarity threshold, and the information amount of the images in the information screening set is greater than the information threshold;

[0176] Matching a target image set according to the definition screening set and the information volume screening set, wherein the image definition of all target images in the target image set is greater than a definition threshold and the image information volume of the target images is greater than an information volume threshold;

[0177] Randomly extracting a first target image based on the target image set, obtaining a first image coordinate system of the first target image, and performing a coordinate transformation operation on all target images in the target image set based on the first image coordinate system and the first target image to obtain a unified target image set;

[0178] Setting a target feature image, extracting a unified target image in sequence based on the unified target image set, calculating a comprehensive feature similarity based on the unified target image and the target feature image, and summarizing the comprehensive feature similarities to obtain a similarity set;

[0179] The similarity set is finally screened according to a preset similarity threshold to obtain a high similarity set, and the multi-source image set is matched based on the high similarity set to obtain a target matching image set, and the image feature intelligent recognition is completed based on the target matching image set.

[0180] In the several embodiments provided by the present invention, it should be understood that the disclosed devices, systems and methods can be implemented in other ways. For example, the system embodiments described above are only illustrative, and actual implementation may have other division methods.

[0181] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0182] In addition, each functional module in each embodiment of the present invention may be integrated into one processing unit, each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0183] It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0184] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. 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 solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. An intelligent image feature recognition method based on multi-source big data, characterized in that: The method comprises: Receiving an intelligent recognition instruction, acquiring a multi-source image set based on the intelligent recognition instruction, extracting multi-source images from the multi-source image set in sequence, and performing definition calculation and information volume calculation on the extracted multi-source images to obtain image definition and image information volume, respectively summarizing the image definition and image information volume to obtain an image definition set and an image information volume set; Preliminarily screening the image clarity set and the image information set by using the preset clarity threshold and information threshold, respectively, to obtain a clarity screening set and an information screening set, wherein the clarity of the images in the clarity screening set is greater than the clarity threshold, and the information amount of the images in the information screening set is greater than the information threshold; Matching a target image set according to the definition screening set and the information volume screening set, wherein the image definition of all target images in the target image set is greater than a definition threshold and the image information volume of the target images is greater than an information volume threshold; Randomly extracting a first target image based on the target image set, obtaining a first image coordinate system of the first target image, and performing a coordinate transformation operation on all target images in the target image set based on the first image coordinate system and the first target image to obtain a unified target image set; Setting a target feature image, extracting a unified target image in sequence based on the unified target image set, calculating a comprehensive feature similarity based on the unified target image and the target feature image, and summarizing the comprehensive feature similarities to obtain a similarity set; The similarity set is finally screened according to a preset similarity threshold to obtain a high similarity set, and the multi-source image set is matched based on the high similarity set to obtain a target matching image set, and the image feature intelligent recognition is completed based on the target matching image set.

2. The method for intelligent image feature recognition based on multi-source big data according to claim 1, characterized in that: The step of calculating the clarity and information volume of the extracted multi-source images to obtain the image clarity and image information volume includes: Obtain the number of pixel rows, the number of pixel columns and the image grayscale value of the multi-source image, and set the pixel horizontal index and the pixel column index; The horizontal change rate and the vertical change rate are calculated based on the image grayscale value, and the image clarity is calculated based on the number of pixel rows, the number of pixel columns, the pixel horizontal index, the pixel column index, the horizontal change rate and the vertical change rate: Wherein, Q refers to the image definition, W refers to the number of pixel rows, E refers to the number of pixel columns, ω refers to the pixel horizontal index, e refers to the pixel column index, α(ω, e) refers to the horizontal change rate of the pixel point with the pixel horizontal index ω and the pixel column index e, and β(ω, e) refers to the vertical change rate of the pixel point with the pixel horizontal index ω and the pixel column index e; Calculate the pixel ratio of multi-source images and calculate the image information based on the pixel ratio.

3. The method for intelligent image feature recognition based on multi-source big data as claimed in claim 2, characterized in that: The step of calculating the pixel ratio of the multi-source images and calculating the image information amount based on the pixel ratio includes: Identify the grayscale values ​​in the multi-source image to obtain a recognition grayscale value set, wherein the recognition grayscale value set includes a plurality of recognition grayscale values, extract recognition grayscale values ​​from the recognition grayscale value set in sequence, and perform the following operations on the extracted recognition grayscale values: The number of grayscale values ​​identical to the extracted identification grayscale value is counted in the multi-source image to obtain the identification number, the grayscale value index is set according to the identification grayscale value, and the pixel ratio is calculated based on the number of pixel rows, the identification number, the number of pixel columns and the grayscale value index: Among them, R μ Refers to the pixel ratio when the gray value index is μ, μ refers to the gray value index, t μ Refers to the number of recognitions when the gray value index is μ; The adjustment parameter and the grayscale weight parameter are set according to the grayscale value index, and the image information amount is calculated based on the pixel ratio, grayscale value index, adjustment parameter and grayscale weight parameter: Among them, T refers to the amount of image information, δ refers to the adjustment parameter, ∈ μ Refers to the grayscale weight parameter when the grayscale value index is μ.

4. The method for intelligent image feature recognition based on multi-source big data according to claim 3, characterized in that: The calculating of the comprehensive feature similarity based on the unified target image and the target feature image includes: Calculating color feature similarity, shape feature similarity and texture feature similarity based on the unified target image and the target feature image; The comprehensive feature similarity is calculated based on the color feature similarity, shape feature similarity and texture feature similarity: G=ρ1×H+ρ2×J+ρ3×K Among them, G refers to the comprehensive feature similarity, ρ1 refers to the preset color feature weight parameter, H refers to the color feature similarity, ρ2 refers to the preset shape feature weight parameter, J refers to the shape feature similarity, ρ3 refers to the preset texture feature weight parameter, and K refers to the texture feature similarity.

5. The method for intelligent recognition of image features based on multi-source big data according to claim 4, characterized in that: The calculating of color feature similarity, shape feature similarity and texture feature similarity based on the unified target image and the target feature image includes: Converting the unified target image into an HSV target image, and converting the target feature image into an HSV feature image; Divide the HSV target image and the HSV feature image based on the preset horizontal and vertical division values ​​to obtain a target image block set and a feature image block set; Extracting target image blocks in sequence according to the target image block set; Constructing a target hue histogram of the extracted target image block, performing feature recognition on the target hue histogram to obtain a hue target vector, summarizing the hue target vectors to form a hue target vector set, and splicing the hue target vectors in the hue target vector set to obtain a target hue feature vector; Obtaining a characteristic hue feature vector based on a characteristic image block set; Calculate the color feature similarity based on the target hue feature vector and the feature hue feature vector; Respectively performing object recognition on the unified target image and the target feature image to obtain a recognized object target image and a recognized object feature image; Generate a target bounding box according to the target image of the identified object, generate a feature bounding box according to the feature image of the identified object, and calculate the similarity of the appearance features based on the target bounding box and the feature bounding box; A target texture feature vector of the unified target image and a feature texture feature vector of the target feature image are obtained, and texture feature similarity is calculated based on the target texture feature vector and the feature texture feature vector.

6. The method for intelligent recognition of image features based on multi-source big data according to claim 5, characterized in that: The color feature similarity is calculated according to the target hue feature vector and the feature hue feature vector, including: Set the adjustment fixed value, and calculate the color feature similarity based on the target hue feature vector, the feature hue feature vector and the adjustment fixed value: Among them, ||*|| refers to calculating the Euclidean distance, Z1 refers to the target hue feature vector, Z2 refers to the feature hue feature vector, Δ refers to dot multiplication, Refers to adjusting a fixed value.

7. The method for intelligent recognition of image features based on multi-source big data according to claim 6, characterized in that: The calculating of the shape feature similarity based on the target bounding box and the feature bounding box includes: Calculate the boundary overlap value based on the target bounding box and the feature bounding box, and obtain the target geometric center coordinates of the target bounding box and the feature geometric center coordinates of the feature bounding box respectively; Set the boundary weight parameter, and calculate the shape feature similarity based on the boundary overlap value, the target geometric center coordinates, the feature geometric center coordinates and the boundary weight parameter: Among them, θ refers to the boundary weight parameter, φ(F1, F2) refers to the boundary overlap value between the target bounding box F1 and the feature bounding box F2, V1 refers to the target geometric center coordinates, and V2 refers to the feature geometric center coordinates.

8. The method for intelligent recognition of image features based on multi-source big data according to claim 7, characterized in that: The calculating of texture feature similarity based on the target texture feature vector and the feature texture feature vector comprises: Calculate the texture feature similarity based on the target texture feature vector and the feature texture feature vector: Among them, υ refers to the preset gradient weight parameter, ε(C1) refers to the vector norm of the target texture feature vector C1, ε(C2) refers to the vector norm of the characteristic texture feature vector C2, C1 refers to the target texture feature vector, C2 refers to the characteristic texture feature vector, ξ(C1) refers to the rate of change of the grayscale value of the target texture feature vector C1, and ξ(C2) refers to the rate of change of the grayscale value of the characteristic texture feature vector C2.

9. The method for intelligent image feature recognition based on multi-source big data according to claim 8, characterized in that: The similarity set is finally screened according to a preset similarity threshold to obtain a high similarity set, including: Extracting comprehensive feature similarities from the similarity set in sequence, and performing the following operations on the extracted comprehensive feature similarities: Determining whether the extracted comprehensive feature similarity is not less than the similarity threshold; If it is confirmed that the extracted comprehensive feature similarity is not less than the similarity threshold, the extracted comprehensive feature similarity is confirmed as high similarity; The high similarities are collected to obtain a high similarity set.

10. An image feature intelligent recognition system based on multi-source big data, characterized in that: The system comprises: A preliminary screening module is used to receive an intelligent recognition instruction, obtain a multi-source image set based on the intelligent recognition instruction, extract multi-source images from the multi-source image set in sequence, and perform clarity calculation and information volume calculation on the extracted multi-source images to obtain image clarity and image information volume, respectively summarize the image clarity and image information volume to obtain an image clarity set and an image information volume set; perform preliminary screening on the image clarity set and the image information volume set respectively using a preset clarity threshold and information volume threshold to obtain a clarity screening set and an information volume screening set, wherein the clarity of the images in the clarity screening set are all greater than the clarity threshold, and the information volume of the images in the information volume screening set are all greater than the information volume threshold; A target matching module, used for matching a target image set according to the definition screening set and the information volume screening set, wherein the image definition of all target images in the target image set is greater than a definition threshold and the image information volume of the target images is greater than an information volume threshold; A coordinate conversion module is used to randomly extract a first target image based on the target image set, obtain a first image coordinate system of the first target image, and perform a coordinate conversion operation on all target images in the target image set based on the first image coordinate system and the first target image to obtain a unified target image set; The final screening module is used to set the target feature image, extract the unified target image in sequence based on the unified target image set, calculate the comprehensive feature similarity based on the unified target image and the target feature image, summarize the comprehensive feature similarity to obtain a similarity set; perform a final screening on the similarity set according to a preset similarity threshold to obtain a high similarity set, match the multi-source image set based on the high similarity set to obtain a target matching image set, and complete image feature intelligent recognition based on the target matching image set.

Images