CN117056547B - Big data classification method and system based on image recognition - Google Patents
Big data classification method and system based on image recognition Download PDFInfo
- Publication number
- CN117056547B CN117056547B CN202311322173.0A CN202311322173A CN117056547B CN 117056547 B CN117056547 B CN 117056547B CN 202311322173 A CN202311322173 A CN 202311322173A CN 117056547 B CN117056547 B CN 117056547B
- Authority
- CN
- China
- Prior art keywords
- image
- feature
- missing
- classification
- images
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 25
- 238000012360 testing method Methods 0.000 claims abstract description 116
- 238000002372 labelling Methods 0.000 claims abstract description 52
- 238000012549 training Methods 0.000 claims abstract description 30
- 239000000758 substrate Substances 0.000 claims description 25
- 238000004364 calculation method Methods 0.000 claims description 15
- 238000005516 engineering process Methods 0.000 claims description 13
- 238000003708 edge detection Methods 0.000 claims description 10
- 239000003086 colorant Substances 0.000 claims description 4
- 230000000694 effects Effects 0.000 description 12
- 238000004458 analytical method Methods 0.000 description 5
- 238000012545 processing Methods 0.000 description 5
- 230000007547 defect Effects 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 238000012217 deletion Methods 0.000 description 3
- 230000037430 deletion Effects 0.000 description 3
- 238000012098 association analyses Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000008520 organization Effects 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 238000013441 quality evaluation Methods 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000010191 image analysis Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 230000008439 repair process Effects 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/55—Clustering; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/51—Indexing; Data structures therefor; Storage structures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/583—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Databases & Information Systems (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Library & Information Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a big data classification method and a system based on image recognition, which are characterized in that a high-definition camera is used for collecting images to be recognized of a target area image according to different time nodes in a time sharing way, and the images are divided into characteristic areas; training a test model, and obtaining a test result through the test model; performing feature labeling on the feature region according to the test result, calculating an image quality value through the feature labeling, and splicing to obtain a first new image according to the image quality value; calculating the relevance, obtaining a first relevance, comparing the first relevance with a relevance threshold value, setting the same group according to the comparison result, and further setting a total image index title; and setting a classification path according to the index title of the total image, classifying the image according to the classification path, establishing a multidimensional database, setting a classification priority, and carrying out priority classification according to the classification priority. More accurate and efficient image recognition and classification is achieved.
Description
Technical Field
The invention provides a big data classification method and a system based on image recognition, relates to the technical field of big data classification, and particularly relates to the technical field of big data classification of image recognition.
Background
In the current image recognition technology field, the development of image recognition and data classification technology is rapid, but there are some problems that in the large data classification process of image recognition, articles are usually blocked, so that part of features of images cannot be captured, the large data classification cannot be successfully completed, the images cannot be automatically grouped according to the image correlation, and whether the images have defects or not is difficult to automatically judge.
Disclosure of Invention
The invention provides a big data classification method and a system based on image recognition, which are used for solving the problems that in the big data classification process of the image recognition, articles are blocked, partial characteristics of the image cannot be shot, the big data classification cannot be successfully completed, the images cannot be automatically grouped according to the image correlation, and whether the images have defects or not is difficult to automatically judge:
the invention provides a big data classification method and a system based on image recognition, wherein the method comprises the following steps:
s1, acquiring images to be identified of target area images by a high-definition camera according to different time nodes in a time sharing manner, and dividing characteristic areas of the images to be identified; training a characteristic edge test model, and obtaining a test result through the test model;
S2, carrying out feature labeling on the feature region according to the test result, calculating an image quality value through the feature labeling, and splicing to obtain a first new image according to the image quality value;
s3, calculating the relevance of the first new image and the new images of other target areas, obtaining a first relevance, comparing the first relevance with a relevance threshold value, setting the same group according to a comparison result, and further setting a total image index title;
s4, setting a classification path according to the index title of the total image, classifying the image according to the classification path, establishing a multidimensional database, setting a classification priority, and carrying out priority classification according to the classification priority.
Further, the high-definition camera acquires images to be identified of the target area images in a time-sharing mode according to different time nodes, and the images to be identified are divided into characteristic areas; training a feature edge test model, obtaining a test result through the test model, including:
setting a plurality of time nodes, and acquiring target area images at the same position and in the same direction through a high-definition camera at each time node to obtain a plurality of images to be identified;
Respectively carrying out edge division on each image to be identified through an edge detection algorithm, and obtaining a plurality of characteristic areas by each image to be identified through the edge division;
article information of a target area is collected through a big data technology, the article information is preprocessed, a training set is obtained, a characteristic edge test model is trained through the training set, the characteristic area is tested through the characteristic edge test model, and a test result is obtained, wherein the test result comprises a missing result and an un-missing result.
Further, the feature labeling of the feature area according to the test result includes:
performing feature labeling on the feature region according to the test result, wherein the feature labeling is divided into a missing labeling and an non-missing labeling;
when the test result is a missing result, carrying out missing marking on the characteristic region, and when the test result is a non-missing result, carrying out non-missing marking on the characteristic region;
and setting exclusive marks for each characteristic region, wherein the characteristic regions of each image to be identified in the same target region are in one-to-one correspondence, and the exclusive marks are also in one-to-one correspondence.
Further, the calculating the image quality value through the feature labeling includes:
the calculation formula of the image quality value is as follows:
；
wherein Z is an image quality value, b 1 For the number of regions not to be missing, E 1 For the total number of characteristic areas, Q 1 Maximum sharpness of the same image, Q 2 Is the minimum sharpness of the same picture.
Further, the stitching to obtain a first new image according to the image quality value includes:
calculating the image quality value of each image, and ranking the image quality values from large to small to obtain an image quality value ranking;
extracting the maximum image quality value ranked first, taking an image corresponding to the maximum image quality value as an image substrate, and obtaining a missing mark of the image substrate;
collecting exclusive marks of the missing marks, and collecting feature marks of feature areas corresponding to the exclusive marks from images corresponding to the second rank of the image quality values;
when the feature labels are not missing labels, replacing the feature areas corresponding to the missing labels of the image substrate with the feature areas corresponding to the feature labels;
and when the feature labels are missing labels, continuously collecting the feature labels of the feature areas corresponding to the exclusive labels of the next rank until all the feature areas corresponding to the missing labels of the image substrate are replaced, so as to obtain a first new image.
Further, the calculating the relevance between the first new image and the new image of other target areas, obtaining a first relevance, comparing the first relevance with a relevance threshold, setting the same group according to the comparison result, and further setting a total image index title, including:
extracting information of the first new image through a big data technology, and setting information keywords as index titles of the first new image;
the new image of other target areas is called a second new image, the relevance between the first new image and the second new image is calculated, the relevance is called a first relevance, and a relevance threshold value is set;
when the first relevance is greater than the relevance threshold value and the second relevance among the second new images is smaller than the relevance threshold value, setting the first new image and the second new image corresponding to the first relevance into the same group; extracting the keywords of the same group, namely a total keyword, and setting a total image index title.
Further, the calculating the association between the first new image and the second new image, which is called as a first association, includes:
the first relevance formula is:
；
Wherein G is the association, A 1 For the number of feature areas of the first new image, A 2 For the number of feature areas of the second new image, P 1 For the number of texture categories, P, of the first new image 2 For the number of texture categories of the second new image, Y 1 For the number of color categories of the first new image, Y 2 The number of colors for the second new image; w (W) 1 、W 2 And W is 3 As a weight parameter, W 1 0.3 of the total weight, W 2 0.4 of the total weight, W 3 Accounting for 0.3 of the total weight.
Further, the setting a classification path according to the overall image index title, classifying the image according to the classification path, and establishing a multidimensional database includes:
setting a classification path through the total image index title, and classifying the image corresponding to the image index title through the classification path; each classification path is independently used for an image corresponding to a total image index title;
a multidimensional database is established through a plurality of classification paths, and classified images are stored.
Further, the setting of the classification priority, and the performing of the priority classification according to the classification priority, includes:
judging the number of images of each total index title, setting a classification priority for the total index titles, and classifying the images corresponding to the important or more total index titles according to the classification priority.
Further, the system comprises:
the characteristic test module is used for acquiring images to be identified of the target area images in a time-sharing mode through the high-definition camera according to different time nodes and dividing characteristic areas of the images to be identified; training a characteristic edge test model, and obtaining a test result through the test model;
the image splicing module is used for carrying out feature labeling on the feature areas according to the test result, calculating an image quality value according to the feature labeling, and splicing to obtain a first new image according to the image quality value;
the association calculation module is used for calculating the association between the first new image and the new image of other target areas, obtaining a first association, comparing the first association with an association threshold value, setting the same group according to a comparison result, and further setting a total image index title;
and the classification module is used for setting a classification path according to the total image index title, classifying the images according to the classification path, establishing a multidimensional database, setting a classification priority, and carrying out priority classification according to the classification priority.
The invention has the beneficial effects that:
the invention provides a big data classification method and a big data classification system based on image recognition. This may help extract important features of the target region. And testing on the image to be identified by training the characteristic edge test model, so as to obtain a test result. These test results may help determine feature edges in the image, further assisting in the recognition and annotation process. And carrying out feature labeling on the feature region according to the test result, and calculating the quality value of the image through the feature labeling. This quality value can evaluate the sharpness, texture, etc. characteristics of the image, providing a reference for subsequent processing. And splicing the images to be identified according to the image quality value, and generating a first new image. This may improve the quality and integrity of the image by selecting better quality feature areas or combining multiple image portions. And calculating the relevance between the first new image and the new images of other target areas, and comparing according to a relevance threshold value. The grouping of images is set according to the comparison result, and the total image index header is set. Thus, effective classification path setting can be performed according to the relevance and classification characteristics of each image. And establishing a multidimensional database according to the total image index title, and setting classification priority. And according to the classification priority, the images are classified preferentially, so that more efficient image recognition and classification are realized. By comprehensively applying image acquisition, feature extraction, test model training, image quality evaluation, relevance calculation, classification path setting and multidimensional database establishment, more accurate and efficient image recognition and classification are realized, and better technical effects are provided.
Drawings
Fig. 1 is a schematic diagram of a big data classification method based on image recognition.
Detailed Description
The preferred embodiments of the present invention will be described below with reference to the accompanying drawings, it being understood that the preferred embodiments described herein are for illustration and explanation of the present invention only, and are not intended to limit the present invention.
The invention provides a big data classification method and a system based on image recognition, wherein the method comprises the following steps:
s1, acquiring images to be identified of target area images by a high-definition camera according to different time nodes in a time sharing manner, and dividing characteristic areas of the images to be identified; training a characteristic edge test model, and obtaining a test result through the test model;
s2, carrying out feature labeling on the feature region according to the test result, calculating an image quality value through the feature labeling, and splicing to obtain a first new image according to the image quality value;
s3, calculating the relevance of the first new image and the new images of other target areas, obtaining a first relevance, comparing the first relevance with a relevance threshold value, setting the same group according to a comparison result, and further setting a total image index title;
S4, setting a classification path according to the index title of the total image, classifying the image according to the classification path, establishing a multidimensional database, setting a classification priority, and carrying out priority classification according to the classification priority.
The working principle of the technical scheme is as follows: firstly, acquiring images to be identified of target area images by a high-definition camera according to different time nodes in a time sharing manner, and dividing characteristic areas of the images to be identified; training a characteristic edge test model, and obtaining a test result through the test model; then, carrying out feature labeling on the feature region according to the test result, calculating an image quality value through the feature labeling, and splicing to obtain a first new image according to the image quality value; furthermore, calculating the relevance of the first new image and the new images of other target areas, obtaining a first relevance, comparing the first relevance with a relevance threshold value, setting the same group according to the comparison result, and further setting a total image index title; and finally, setting a classification path according to the index title of the total image, classifying the image according to the classification path, establishing a multidimensional database, setting a classification priority, and carrying out priority classification according to the classification priority.
The technical effects of the technical scheme are as follows: and acquiring an image of the target area by using a high-definition camera, and dividing the characteristic area of the image to be identified. This may help extract important features of the target region. And testing on the image to be identified by training the characteristic edge test model, so as to obtain a test result. These test results may help determine feature edges in the image, further assisting in the recognition and annotation process. And carrying out feature labeling on the feature region according to the test result, and calculating the quality value of the image through the feature labeling. This quality value can evaluate the sharpness, texture, etc. characteristics of the image, providing a reference for subsequent processing. And splicing the images to be identified according to the image quality value, and generating a first new image. This may improve the quality and integrity of the image by selecting better quality feature areas or combining multiple image portions. And calculating the relevance between the first new image and the new images of other target areas, and comparing according to a relevance threshold value. The grouping of images is set according to the comparison result, and the total image index header is set. Thus, effective classification path setting can be performed according to the relevance and classification characteristics of each image. And establishing a multidimensional database according to the total image index title, and setting classification priority. And according to the classification priority, the images are classified preferentially, so that more efficient image recognition and classification are realized. By comprehensively applying image acquisition, feature extraction, test model training, image quality evaluation, relevance calculation, classification path setting and multidimensional database establishment, more accurate and efficient image recognition and classification are realized, and better technical effects are provided.
According to one embodiment of the invention, the high-definition camera is used for collecting images to be identified of the target area images according to different time nodes in a time sharing manner, and the images to be identified are divided into characteristic areas; training a feature edge test model, obtaining a test result through the test model, including:
setting a plurality of time nodes, and acquiring target area images at the same position and in the same direction through a high-definition camera at each time node to obtain a plurality of images to be identified;
respectively carrying out edge division on each image to be identified through an edge detection algorithm, and obtaining a plurality of characteristic areas by each image to be identified through the edge division;
article information of a target area is collected through a big data technology, the article information is preprocessed, a training set is obtained, a characteristic edge test model is trained through the training set, the characteristic area is tested through the characteristic edge test model, and a test result is obtained, wherein the test result comprises a missing result and an un-missing result.
The working principle of the technical scheme is as follows: setting a plurality of time nodes, and acquiring target area images at the same position and in the same direction through a high-definition camera at each time node to obtain a plurality of images to be identified; the time nodes may be set by hour or by day. Respectively carrying out edge division on each image to be identified through an edge detection algorithm, and obtaining a plurality of characteristic areas by each image to be identified through the edge division; the edge detection algorithm comprises Sobel or Canny. Article information of a target area is collected through a big data technology, the article information is preprocessed, a training set is obtained, a characteristic edge test model is trained through the training set, the characteristic area is tested through the characteristic edge test model, and a test result is obtained, wherein the test result comprises a missing result and an un-missing result. The edge of the characteristic region is complete and is a non-missing result, and the edge of the characteristic region is incomplete and is a missing result.
The technical effects of the technical scheme are as follows: and acquiring the target area image by using the high-definition camera by arranging a plurality of time nodes at the same position and in the same direction. In this way, a plurality of images to be identified can be obtained, and the setting of the time node can be determined on an hourly or daily basis according to the requirements. And carrying out edge division on each image to be identified by adopting a common edge detection algorithm, such as a Sobel or Canny algorithm. This allows each image to be divided into a plurality of feature areas for subsequent processing and analysis. And collecting and preprocessing the article information of the target area by utilizing a big data technology. Thus, a training set can be established for training the feature edge test model, and a basis is provided for subsequent image analysis. The feature edge test model is trained using the training set. And then, testing the characteristic region by using the model to obtain a test result. The test results include missing results and absent results, as judged by comparing the edge integrity of the feature areas. The method can realize multiple collection and edge division of the target area image, and judge whether the characteristic area is missing or not by utilizing the characteristic edge test model. The technical scheme can improve the recognition precision and accuracy of the target area image, and is more accurate in evaluating the edge integrity.
The method combines the acquisition of the high-definition camera, the edge detection algorithm, the big data preprocessing and the characteristic edge test model, can effectively divide the characteristic region and judge missing and non-missing results, and improves the accuracy and reliability of image recognition.
According to one embodiment of the present invention, the feature labeling of the feature area according to the test result includes:
performing feature labeling on the feature region according to the test result, wherein the feature labeling is divided into a missing labeling and an non-missing labeling;
when the test result is a missing result, carrying out missing marking on the characteristic region, and when the test result is a non-missing result, carrying out non-missing marking on the characteristic region;
and setting exclusive marks for each characteristic region, wherein the characteristic regions of each image to be identified in the same target region are in one-to-one correspondence, and the exclusive marks are also in one-to-one correspondence.
The working principle of the technical scheme is as follows: performing feature labeling on the feature region according to the test result, wherein the feature labeling is divided into a missing labeling and an non-missing labeling;
when the test result is a missing result, carrying out missing marking on the characteristic region, and when the test result is a non-missing result, carrying out non-missing marking on the characteristic region;
And setting exclusive marks for each characteristic region, wherein the characteristic regions of each image to be identified in the same target region are in one-to-one correspondence, and the exclusive marks are also in one-to-one correspondence. The multiple images to be identified of the same target area image are collected at the same position and in the same direction, and only the collection time is different, so that the image without object shielding can be collected under the condition that the position of the target object image to be collected is unchanged, the object in the image can be conveniently subjected to the correspondence of the characteristic areas, all the characteristic areas without shielding are found out, and the image is spliced into a new image without shielding.
The technical effects of the technical scheme are as follows: and marking the characteristics of the characteristic areas according to the test result. If the test result is a missing result, carrying out missing marking on the characteristic region; and if the test result is an indeterminate result, marking the feature region without loss. Thus, whether each characteristic region has a defect or not can be judged according to the labeling result. A dedicated label is provided for each feature region. And distributing the exclusive labels to the characteristic areas of each image to be identified in the same target area in a one-to-one correspondence manner. Therefore, the characteristic areas with the same positions and directions can be ensured to have the same exclusive marks, and the matching and analysis of the characteristic areas are convenient to follow. The images to be identified are collected at the same position and in the same direction, and the collection time is different. Such a design is intended to maintain consistency of the position of the target object image and to ensure that images are acquired without object occlusion. In this way, the object in the image can be conveniently subjected to the correspondence of the characteristic region, the characteristic region which is not blocked can be found out, and the new image which is not blocked at all can be spliced again. The technical scheme can realize the deletion and non-deletion marking of the characteristic areas, set a special mark for each characteristic area, and acquire the non-shielding image through multiple acquisitions of the same position and direction. In this way, whether the characteristic area is missing or not can be accurately identified, and a reliable data basis is provided for subsequent object identification and analysis.
In one embodiment of the present invention, the calculating the image quality value by the feature labeling includes:
the calculation formula of the image quality value is as follows:
；
wherein Z is an image quality value, b 1 For the number of regions not to be missing, E 1 For the total number of characteristic areas, Q 1 Maximum sharpness of the same image, Q 2 Is the minimum sharpness of the same picture.
The working principle of the technical scheme is as follows: by performing the number of regions without missing labels, the number of total feature regions, the maximum sharpness of the same image and the minimum sharpness of the same image.
The technical effects of the technical scheme are as follows: the regions are divided using an edge detection algorithm, and then identified and marked by a target detection or instance segmentation algorithm. This can result in the total number of divided areas of one picture. By further analyzing the region marking results, the number of missing regions is identified and calculated. The missing region may refer to a portion of the image that lacks a specific feature, or a region that cannot be accurately identified due to occlusion or damage. And carrying out definition evaluation on each region by using an image definition measurement algorithm to obtain a definition value. Common image sharpness evaluation methods include techniques based on edge detection, texture analysis, or fourier transforms. The maximum value and the minimum value are found from the sharpness values of all the regions, and the ratio between them is calculated. This ratio can be used as an indicator of the image quality value to measure the relative degree of difference between the most clear and least clear regions in the image; the total number of areas, the number of area deletions and the ratio of maximum sharpness to minimum sharpness of a picture can be obtained, and these data can be used as a basis for evaluating the image quality. For example, a smaller number of missing regions and a larger maximum sharpness to minimum sharpness ratio can generally be regarded as an indicator of higher image quality.
According to one embodiment of the present invention, the stitching to obtain the first new image according to the image quality value includes:
calculating the image quality value of each image, and ranking the image quality values from large to small to obtain an image quality value ranking;
extracting the maximum image quality value ranked first, taking an image corresponding to the maximum image quality value as an image substrate, and obtaining a missing mark of the image substrate;
collecting exclusive marks of the missing marks, and collecting feature marks of feature areas corresponding to the exclusive marks from images corresponding to the second rank of the image quality values;
when the feature labels are not missing labels, replacing the feature areas corresponding to the missing labels of the image substrate with the feature areas corresponding to the feature labels;
and when the feature labels are missing labels, continuously collecting the feature labels of the feature areas corresponding to the exclusive labels of the next rank until all the feature areas corresponding to the missing labels of the image substrate are replaced, so as to obtain a first new image.
The working principle of the technical scheme is as follows: calculating the image quality value of each image, and ranking the image quality values from large to small to obtain an image quality value ranking; extracting the maximum image quality value ranked first, taking an image corresponding to the maximum image quality value as an image substrate, and obtaining a missing mark of the image substrate; collecting exclusive marks of the missing marks, and collecting feature marks of feature areas corresponding to the exclusive marks from images corresponding to the second rank of the image quality values; when the feature labels are not missing labels, replacing the feature areas corresponding to the missing labels of the image substrate with the feature areas corresponding to the feature labels; instead, feature regions of the same exclusive designation are replaced. And when the feature labels are missing labels, continuously collecting the feature labels of the feature areas corresponding to the exclusive labels of the next rank until all the feature areas corresponding to the missing labels of the image substrate are replaced, so as to obtain a first new image. The spliced new image is formed by carrying out non-shielding arrangement on each characteristic region or object, each characteristic region after replacement is subjected to non-shielding arrangement, and the image can be enlarged.
The technical effects of the technical scheme are as follows: for a set of images, the image quality value for each image is calculated and ranked from large to small in value. And extracting the image serving as the image base from the image with the highest rank, and acquiring the missing labels in the image. The missing mark refers to a mark for marking a missing region, and can be obtained by region identification and marking. Starting from the second ranked image, feature annotations of feature regions corresponding to missing annotations in the image base are gathered. For each collected feature annotation, a feature region corresponding to the missing annotation based on the image substrate is examined. If the feature region corresponding to the feature label is not missing, replacing the feature region with the feature region corresponding to the feature label in the image substrate, wherein the replaced region needs to have the same exclusive label. If the image substrate is provided with the feature areas corresponding to the missing marks, the feature marks corresponding to the exclusive marks of the next rank are continuously collected, and replacement operation is carried out until the feature areas corresponding to all the missing marks of the image substrate are replaced. When all the feature areas corresponding to the missing marks are replaced, the obtained image is a first new image, and the defect of feature missing is overcome by replacement operation based on the image with the highest image quality value in the original image set; the highest quality image is selected from the original image set as a substrate by the strategies of ranking the image quality values and replacing the characteristic areas, and the quality of the image is improved by the characteristic areas in other images. This can help to improve overall image quality, repair missing areas in the base image, and create new images with higher quality. The rate of image area replacement is mentioned.
In one embodiment of the present invention, the calculating the relevance between the first new image and the new image of the other target area, obtaining a first relevance, comparing the first relevance with a relevance threshold, setting the same group according to the comparison result, and further setting a total image index title, includes:
extracting information of the first new image through a big data technology, and setting information keywords as index titles of the first new image;
the new image of other target areas is called a second new image, the relevance between the first new image and the second new image is calculated, the relevance is called a first relevance, and a relevance threshold value is set;
when the first relevance is greater than the relevance threshold value and the second relevance among the second new images is smaller than the relevance threshold value, setting the first new image and the second new image corresponding to the first relevance into the same group; extracting the keywords of the same group, namely a total keyword, and setting a total image index title.
The working principle of the technical scheme is as follows: extracting information of the first new image through a big data technology, and setting information keywords as index titles of the first new image;
The new image of other target areas is called a second new image, the relevance between the first new image and the second new image is calculated, the relevance is called a first relevance, and a relevance threshold value is set; the relevance threshold is set to 0.7 or set by the user himself.
When the first relevance is greater than the relevance threshold value and the second relevance among the second new images is smaller than the relevance threshold value, setting the first new image and the second new image corresponding to the first relevance into the same group; extracting the keywords of the same group, namely a total keyword, and setting a total image index title. The keywords are selected by the user. The second relevance is the relevance between every two second new images, and the comparison is to ensure that the relevance between the second new images participating in calculating the first relevance and the new images is larger than a relevance threshold value, and the relevance between the second new images and other second new images is smaller than the relevance threshold value, so that the second new images and the new images are set into the same group. The first new image may be considered a reference for ease of understanding.
The technical effects of the technical scheme are as follows: and analyzing and processing the first new image by utilizing a big data technology, extracting key information and characteristics in the first new image, and taking the key information and the characteristics as index titles of the images. These information keywords may be keywords that describe image content, features, or other related attributes to provide an efficient index of image content. For each second new image, the association between it and the first new image is calculated. An association threshold, such as 0.7, is set for determining whether the association between the first new image and each of the second new images meets the desired requirements. The threshold may be set according to specific needs or may be specified by the user at his own discretion to control the accuracy and reliability of the association. When the association of a first new image with a certain second new image is greater than an association threshold and the association between the second new image and other second new images is less than the threshold, the images are set to the same group. This means that they have a higher similarity under certain correlation conditions and can be considered as related images. And the first relevance with the first new image is higher than the relevance threshold value, and the second relevance with other second new images with relevance smaller than the relevance threshold value can be separated into a group with the first relevance, so that the problem of disorder of the same image in different groups caused by disorder of the groups is avoided. For images in the same group, their keywords are extracted and combined into a total keyword. These keywords may be features, content, attributes, etc. common to the group images, which are used to describe and index the entire group. The overall image index header may be set according to the overall keywords to better describe and identify the content and characteristics of the entire group. Key information can be extracted and related image composition groups can be organized according to the relevance and similarity between images, so that better indexing and management of image sets can be realized.
In one embodiment of the present invention, the calculating the association between the first new image and the second new image, which is referred to as a first association, includes:
the first relevance formula is:
；
wherein G is the association, A 1 For the number of feature areas of the first new image, A 2 For the number of feature areas of the second new image, P 1 For the number of texture categories, P, of the first new image 2 For the number of texture categories of the second new image, Y 1 For the number of color categories of the first new image, Y 2 The number of colors for the second new image; w (W) 1 、W 2 And W is 3 As a weight parameter, W 1 0.3 of the total weight, W 2 0.4 of the total weight, W 3 Accounting for 0.3 of the total weight.
The working principle of the technical scheme is as follows: the number of the characteristic areas of the first new image, the number of the characteristic areas of the second new image, the number of the texture types of the first new image, the number of the texture types of the second new image, the number of the color types of the first new image, the number of the color types of the second new image, W 1 、W 2 And W is 3 As a weight parameter, W 1 0.3 of the total weight, W 2 0.4 of the total weight, W 3 The value range of G accounting for 0.3 of the total weight is as follows: 1>G>0, calculating the relevance.
The technical effects of the technical scheme are as follows: the relevance calculation formula comprehensively considers the differences of the number of the characteristic areas, the number of the texture types and the number of the color types, and performs weighted summation according to the weight parameters. By calculation, a correlation value between 0 and 1 can be derived, with higher values indicating that the two images are more similar in terms of features, texture and color, and lower values indicating that the two images differ more in these respects. The technical scheme can be used for evaluating and comparing the image relevance. By adjusting the weight parameters, the degree of contribution of features, textures and colors to relevance can be adjusted according to specific requirements and concerns. Accurate calculation and judgment of image relevance can be achieved.
In one embodiment of the present invention, the setting a classification path according to the total image index header, classifying the image according to the classification path, and creating the multidimensional database includes:
setting a classification path through the total image index title, and classifying the image corresponding to the image index title through the classification path; each classification path is independently used for an image corresponding to a total image index title;
a multidimensional database is established through a plurality of classification paths, and classified images are stored.
In one embodiment of the present invention, the setting a classification priority, and performing a priority classification according to the classification priority, includes:
judging the number of images of each total index title, setting a classification priority for the total index titles, and classifying the images corresponding to the important or more total index titles according to the classification priority. Judging the number of images of each total index title, setting a classification priority for the total index titles, and classifying the images corresponding to the important or more total index titles according to the classification priority.
The working principle of the technical scheme is as follows: setting a classification path through the total image index title, and classifying the image corresponding to the image index title through the classification path; each classification path is independently used for an image corresponding to a total image index title; a multidimensional database is established through a plurality of classification paths, and classified images are stored. Judging the number of images of each total index title, setting a classification priority for the total index titles, and classifying the images corresponding to the important or more total index titles according to the classification priority. The priority setting can be set by the user according to the specific situation.
The technical effects of the technical scheme are as follows: and setting a classification path through the total image index title, and classifying the images corresponding to the image index title by using the classification path. Each classification path is used independently for an image classification corresponding to a total image index title. By creating multiple classification paths, a multidimensional database can be constructed to store classified images. First, the number of images corresponding to each total index title needs to be determined. And then sets a classification priority for each total index title according to factors such as the number of images or importance. The priority determines the image classification sequence, so that images corresponding to the total index titles with high importance or a large number of images can be classified preferentially, and the images can be classified according to the total index titles to which the images belong, so that similar or related images are classified under the same path. By creating multiple classification paths and storing the classified images, a multidimensional database can be constructed. Thus, the images can be conveniently searched, inquired and managed. By setting the classification priority, the images corresponding to the total index titles with high importance or large number of images can be classified preferentially, and the classification efficiency and accuracy are improved. The user can set the classification priority according to specific conditions so as to adapt to different application scenes and requirements. The technical scheme can help to realize automation and high efficiency of image classification and management, so that organization, retrieval and analysis of images become more convenient and reliable.
In one embodiment of the invention, the system comprises:
the characteristic test module is used for acquiring images to be identified of the target area images in a time-sharing mode through the high-definition camera according to different time nodes and dividing characteristic areas of the images to be identified; training a characteristic edge test model, and obtaining a test result through the test model;
the image splicing module is used for carrying out feature labeling on the feature areas according to the test result, calculating an image quality value according to the feature labeling, and splicing to obtain a first new image according to the image quality value;
the association calculation module is used for calculating the association between the first new image and the new image of other target areas, obtaining a first association, comparing the first association with an association threshold value, setting the same group according to a comparison result, and further setting a total image index title;
and the classification module is used for setting a classification path according to the total image index title, classifying the images according to the classification path, establishing a multidimensional database, setting a classification priority, and carrying out priority classification according to the classification priority.
The working principle of the technical scheme is as follows: the characteristic test module is used for acquiring images to be identified of the target area images in a time-sharing mode through the high-definition camera according to different time nodes and dividing characteristic areas of the images to be identified; training a characteristic edge test model, and obtaining a test result through the test model; the image stitching module is used for carrying out feature labeling on the feature areas according to the test result, calculating an image quality value according to the feature labeling, and stitching to obtain a first new image according to the image quality value; the association calculation module is used for calculating the association between the first new image and the new image of other target areas, obtaining a first association, comparing the first association with an association threshold value, setting the same group according to a comparison result, and further setting a total image index title; the classification module is used for setting a classification path according to the index title of the total image, classifying the image according to the classification path, establishing a multidimensional database, setting a classification priority, and carrying out priority classification according to the classification priority.
The technical effects of the technical scheme are as follows: and acquiring images to be identified of the target area images by using a high-definition camera according to different time nodes in a time sharing way, and dividing characteristic areas of the images to be identified. This may help extract key features in the image. By training the characteristic edge test model, the characteristic region can be tested and a test result can be obtained. This model may help identify and detect particular edge features in the image. And carrying out feature labeling on the feature region according to the feature test result, and calculating an image quality value. And performing image stitching according to the image quality value to obtain a first new image. This module can help to improve image quality and sharpness. And calculating the relevance between the first new image and the new images of other target areas, and obtaining the first relevance. The first relevance is compared with a relevance threshold value, and the same group and the total image index title are set according to the comparison result. This may help to categorize related images into the same grouping in order to better organize and manage image data. A classification path is set according to the overall image index header, and the images are classified by the classification path. At the same time, a multidimensional database is built to store the classified images. By setting the classification priority, it is possible to classify preferentially the images corresponding to the total index titles having high importance or a large number of images. This module may provide efficient image classification and retrieval functions. Through the feature test module and the trained feature edge test model, feature edges in the image can be accurately extracted and tested. The image quality can be improved by the image stitching module based on the feature labels and the calculation of the image quality value. Through the association calculation module, the association and similarity between the images can be calculated, so that the organization and association analysis of the images are realized. Through the classification module and the multidimensional database, the images can be classified, stored and retrieved rapidly, and the management efficiency of the image data is improved. The technical scheme can help to improve the automation level of image processing and management, improve the image recognition and classification effects and provide efficient image retrieval and association analysis functions.
It will be apparent to those skilled in the art that various modifications and variations can be made to the present invention without departing from the spirit or scope of the invention. Thus, it is intended that the present invention also include such modifications and alterations insofar as they come within the scope of the appended claims or the equivalents thereof.
Claims (6)
1. A big data classification method based on image recognition, the method comprising:
s1, acquiring images to be identified of target area images by a high-definition camera according to different time nodes in a time sharing manner, and dividing characteristic areas of the images to be identified; training a characteristic edge test model, and obtaining a test result through the test model;
the method comprises the steps that a high-definition camera collects images to be identified of target area images in a time-sharing mode according to different time nodes, and feature areas of the images to be identified are divided; training a feature edge test model, and obtaining test results through the test model comprises:
setting a plurality of time nodes, and acquiring target area images at the same position and in the same direction through a high-definition camera at each time node to obtain a plurality of images to be identified;
Respectively carrying out edge division on each image to be identified through an edge detection algorithm, and obtaining a plurality of characteristic areas by each image to be identified through the edge division;
article information of a target area is collected through a big data technology, the article information is preprocessed, a training set is obtained, a characteristic edge test model is trained through the training set, the characteristic area is tested through the characteristic edge test model, and a test result is obtained, wherein the test result comprises a missing result and an un-missing result;
s2, carrying out feature labeling on the feature region according to the test result, calculating an image quality value through the feature labeling, and splicing to obtain a first new image according to the image quality value;
the feature labeling of the feature area according to the test result comprises the following steps:
performing feature labeling on the feature region according to the test result, wherein the feature labeling is divided into a missing labeling and an non-missing labeling;
when the test result is a missing result, carrying out missing marking on the characteristic region, and when the test result is a non-missing result, carrying out non-missing marking on the characteristic region;
Setting exclusive marks for each characteristic region, wherein the characteristic regions of each image to be identified in the same target region are in one-to-one correspondence, and the exclusive marks are also in one-to-one correspondence;
wherein the calculating the image quality value by the feature labeling includes:
the calculation formula of the image quality value is as follows:
；
wherein Z is an image quality value, b 1 For the number of regions not to be missing, E 1 For the total number of characteristic areas, Q 1 Maximum sharpness of the same image, Q 2 Minimum definition of the same picture;
the step of obtaining a first new image by stitching according to the image quality value comprises the following steps:
calculating the image quality value of each image, and ranking the image quality values from large to small to obtain an image quality value ranking;
extracting the maximum image quality value ranked first, taking an image corresponding to the maximum image quality value as an image substrate, and obtaining a missing mark of the image substrate;
collecting exclusive marks of the missing marks, and collecting feature marks of feature areas corresponding to the exclusive marks from images corresponding to the second rank of the image quality values;
when the feature labels are not missing labels, replacing the feature areas corresponding to the missing labels of the image substrate with the feature areas corresponding to the feature labels;
When the feature labels are missing labels, continuing to collect feature labels of the feature areas corresponding to the exclusive labels of the next rank until all feature areas corresponding to the missing labels of the image substrate are replaced, so as to obtain a first new image;
s3, calculating the relevance of the first new image and the new images of other target areas, obtaining a first relevance, comparing the first relevance with a relevance threshold value, setting the same group according to a comparison result, and further setting a total image index title;
s4, setting a classification path according to the index title of the total image, classifying the image according to the classification path, establishing a multidimensional database, setting a classification priority, and carrying out priority classification according to the classification priority.
2. The method for classifying big data based on image recognition according to claim 1, wherein the calculating the relevance of the first new image to the new image of the other target area, obtaining a first relevance, comparing the first relevance with a relevance threshold, setting the same group according to the comparison result, and further setting a total image index title, includes:
extracting information of the first new image through a big data technology, and setting information keywords as index titles of the first new image;
The new image of other target areas is called a second new image, the relevance between the first new image and the second new image is calculated, the relevance is called a first relevance, and a relevance threshold value is set;
when the first relevance is greater than the relevance threshold value and the second relevance among the second new images is smaller than the relevance threshold value, setting the first new image and the second new image corresponding to the first relevance into the same group; extracting the keywords of the same group, namely a total keyword, and setting a total image index title.
3. The method of claim 2, wherein the calculating the association between the first new image and the second new image, the association being referred to as a first association, comprises:
the first relevance formula is:
；
wherein G is the association, A 1 For the number of feature areas of the first new image, A 2 For the number of feature areas of the second new image, P 1 For the number of texture categories, P, of the first new image 2 For the number of texture categories of the second new image, Y 1 For the number of color categories of the first new image, Y 2 The number of colors for the second new image; w (W) 1 、W 2 And W is 3 As a weight parameter, W 1 0.3 of the total weight, W 2 0.4 of the total weight, W 3 Accounting for 0.3 of the total weight.
4. The method of claim 1, wherein the setting a classification path according to the overall image index header, classifying images according to the classification path, and creating a multidimensional database comprises:
setting a classification path through the total image index title, and classifying the image corresponding to the image index title through the classification path; each classification path is independently used for an image corresponding to a total image index title;
a multidimensional database is established through a plurality of classification paths, and classified images are stored.
5. The method for classifying big data based on image recognition according to claim 1, wherein said setting a classification priority, and performing a priority classification based on said classification priority, comprises:
judging the number of images of each total index title, setting a classification priority for the total index titles, and classifying the images corresponding to the important or more total index titles according to the classification priority.
6. A big data classification system based on image recognition, the system comprising:
The characteristic test module is used for acquiring images to be identified of the target area images in a time-sharing mode through the high-definition camera according to different time nodes and dividing characteristic areas of the images to be identified; training a characteristic edge test model, and obtaining a test result through the test model;
wherein, the feature test module includes:
setting a plurality of time nodes, and acquiring target area images at the same position and in the same direction through a high-definition camera at each time node to obtain a plurality of images to be identified;
respectively carrying out edge division on each image to be identified through an edge detection algorithm, and obtaining a plurality of characteristic areas by each image to be identified through the edge division;
article information of a target area is collected through a big data technology, the article information is preprocessed, a training set is obtained, a characteristic edge test model is trained through the training set, the characteristic area is tested through the characteristic edge test model, and a test result is obtained, wherein the test result comprises a missing result and an un-missing result;
the image splicing module is used for carrying out feature labeling on the feature areas according to the test result, calculating an image quality value according to the feature labeling, and splicing to obtain a first new image according to the image quality value;
The feature labeling of the feature region according to the test result includes:
performing feature labeling on the feature region according to the test result, wherein the feature labeling is divided into a missing labeling and an non-missing labeling;
when the test result is a missing result, carrying out missing marking on the characteristic region, and when the test result is a non-missing result, carrying out non-missing marking on the characteristic region;
setting exclusive marks for each characteristic region, wherein the characteristic regions of each image to be identified in the same target region are in one-to-one correspondence, and the exclusive marks are also in one-to-one correspondence;
wherein the calculating the image quality value by the feature labeling includes:
the calculation formula of the image quality value is as follows:
；
wherein Z is an image quality value, b 1 For the number of regions not to be missing, E 1 For the total number of characteristic areas, Q 1 Maximum sharpness of the same image, Q 2 Minimum definition of the same picture;
the step of obtaining a first new image by stitching according to the image quality value comprises the following steps:
calculating the image quality value of each image, and ranking the image quality values from large to small to obtain an image quality value ranking;
Extracting the maximum image quality value ranked first, taking an image corresponding to the maximum image quality value as an image substrate, and obtaining a missing mark of the image substrate;
collecting exclusive marks of the missing marks, and collecting feature marks of feature areas corresponding to the exclusive marks from images corresponding to the second rank of the image quality values;
when the feature labels are not missing labels, replacing the feature areas corresponding to the missing labels of the image substrate with the feature areas corresponding to the feature labels;
when the feature labels are missing labels, continuing to collect feature labels of the feature areas corresponding to the exclusive labels of the next rank until all feature areas corresponding to the missing labels of the image substrate are replaced, so as to obtain a first new image;
the association calculation module is used for calculating the association between the first new image and the new image of other target areas, obtaining a first association, comparing the first association with an association threshold value, setting the same group according to a comparison result, and further setting a total image index title;
and the classification module is used for setting a classification path according to the total image index title, classifying the images according to the classification path, establishing a multidimensional database, setting a classification priority, and carrying out priority classification according to the classification priority.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311322173.0A CN117056547B (en) | 2023-10-13 | 2023-10-13 | Big data classification method and system based on image recognition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311322173.0A CN117056547B (en) | 2023-10-13 | 2023-10-13 | Big data classification method and system based on image recognition |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117056547A CN117056547A (en) | 2023-11-14 |
| CN117056547B true CN117056547B (en) | 2024-01-26 |
Family
ID=88657627
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311322173.0A Active CN117056547B (en) | 2023-10-13 | 2023-10-13 | Big data classification method and system based on image recognition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117056547B (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106204435A (en) * | 2016-06-27 | 2016-12-07 | 北京小米移动软件有限公司 | Image processing method and device |
| CN108875834A (en) * | 2018-06-22 | 2018-11-23 | 北京达佳互联信息技术有限公司 | Image clustering method, device, computer equipment and storage medium |
| CN108875797A (en) * | 2018-05-29 | 2018-11-23 | 腾讯科技（深圳）有限公司 | A kind of method of determining image similarity, photograph album management method and relevant device |
| WO2020215952A1 (en) * | 2019-04-23 | 2020-10-29 | 北京京东尚科信息技术有限公司 | Object recognition method and system |
| CN114238744A (en) * | 2021-12-21 | 2022-03-25 | 支付宝(杭州)信息技术有限公司 | Data processing method, device and equipment |
| CN114387651A (en) * | 2022-01-12 | 2022-04-22 | 北京百度网讯科技有限公司 | Face recognition method, device, equipment and storage medium |
| CN114638980A (en) * | 2022-03-04 | 2022-06-17 | 支付宝(杭州)信息技术有限公司 | Dish type identification processing method and device |
-
2023
- 2023-10-13 CN CN202311322173.0A patent/CN117056547B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106204435A (en) * | 2016-06-27 | 2016-12-07 | 北京小米移动软件有限公司 | Image processing method and device |
| CN108875797A (en) * | 2018-05-29 | 2018-11-23 | 腾讯科技（深圳）有限公司 | A kind of method of determining image similarity, photograph album management method and relevant device |
| CN108875834A (en) * | 2018-06-22 | 2018-11-23 | 北京达佳互联信息技术有限公司 | Image clustering method, device, computer equipment and storage medium |
| WO2020215952A1 (en) * | 2019-04-23 | 2020-10-29 | 北京京东尚科信息技术有限公司 | Object recognition method and system |
| CN114238744A (en) * | 2021-12-21 | 2022-03-25 | 支付宝(杭州)信息技术有限公司 | Data processing method, device and equipment |
| CN114387651A (en) * | 2022-01-12 | 2022-04-22 | 北京百度网讯科技有限公司 | Face recognition method, device, equipment and storage medium |
| CN114638980A (en) * | 2022-03-04 | 2022-06-17 | 支付宝(杭州)信息技术有限公司 | Dish type identification processing method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117056547A (en) | 2023-11-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Scharr et al. | Leaf segmentation in plant phenotyping: a collation study | |
| CN104487843B (en) | Image processing apparatus, program, image processing method, computer-readable medium and image processing system | |
| CN110717534B (en) | Target classification and positioning method based on network supervision | |
| CN102880729A (en) | Figure image retrieval method and device based on human face detection and recognition | |
| US8170343B2 (en) | Method and system for searching images with figures and recording medium storing metadata of image | |
| CN108830332A (en) | A kind of vision vehicle checking method and system | |
| US7283659B1 (en) | Apparatus and methods for searching through and analyzing defect images and wafer maps | |
| CN111325103A (en) | Cell labeling system and method | |
| CN108090232A (en) | Utilize the semantic event detection of cross-domain knowledge | |
| CN101073078A (en) | Multi-tiered image clustering by event | |
| CN110533654A (en) | The method for detecting abnormality and device of components | |
| CN115271003B (en) | Abnormal data analysis method and system for automatic environment monitoring equipment | |
| CN112700408A (en) | Model training method, image quality evaluation method and device | |
| CN111242899B (en) | Image-based flaw detection method and computer-readable storage medium | |
| CN107316053A (en) | A kind of cloth image Rapid matching search method | |
| CN112633382A (en) | Mutual-neighbor-based few-sample image classification method and system | |
| CN113313149B (en) | Dish identification method based on attention mechanism and metric learning | |
| CN117115148B (en) | Chip surface defect intelligent identification method based on 5G technology | |
| Deulkar et al. | An automated tomato quality grading using clustering based support vector machine | |
| CN117809124B (en) | Medical image association calling method and system based on multi-feature fusion | |
| CN110309737A (en) | A kind of information processing method applied to cigarette sales counter, apparatus and system | |
| CN117056547B (en) | Big data classification method and system based on image recognition | |
| Montalvo et al. | A novel threshold to identify plant textures in agricultural images by Otsu and Principal Component Analysis | |
| CN106326927A (en) | Shoeprint new class detection method | |
| CN116524263A (en) | Semi-automatic labeling method for fine-grained images |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |