CN112100419A - Single weather image identification method and system based on image retrieval - Google Patents
Single weather image identification method and system based on image retrieval Download PDFInfo
- Publication number
- CN112100419A CN112100419A CN202010967561.4A CN202010967561A CN112100419A CN 112100419 A CN112100419 A CN 112100419A CN 202010967561 A CN202010967561 A CN 202010967561A CN 112100419 A CN112100419 A CN 112100419A
- Authority
- CN
- China
- Prior art keywords
- image
- feature
- vector
- training
- feedback information
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 48
- 239000013598 vector Substances 0.000 claims abstract description 150
- 238000012549 training Methods 0.000 claims abstract description 144
- 238000000605 extraction Methods 0.000 claims abstract description 36
- 230000008569 process Effects 0.000 claims abstract description 18
- 238000012790 confirmation Methods 0.000 claims description 89
- 238000012937 correction Methods 0.000 claims description 56
- 238000012360 testing method Methods 0.000 claims description 26
- 238000005259 measurement Methods 0.000 claims description 22
- 238000007781 pre-processing Methods 0.000 claims description 20
- 238000010606 normalization Methods 0.000 claims description 10
- 238000004422 calculation algorithm Methods 0.000 claims description 9
- 238000012795 verification Methods 0.000 claims description 8
- 238000004364 calculation method Methods 0.000 claims description 7
- 230000015572 biosynthetic process Effects 0.000 claims description 6
- 238000006243 chemical reaction Methods 0.000 claims description 6
- 238000011478 gradient descent method Methods 0.000 claims description 6
- 230000006870 function Effects 0.000 description 22
- 238000013527 convolutional neural network Methods 0.000 description 8
- 230000009286 beneficial effect Effects 0.000 description 6
- 238000001514 detection method Methods 0.000 description 6
- 238000012545 processing Methods 0.000 description 6
- 230000002159 abnormal effect Effects 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 239000000463 material Substances 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 208000037170 Delayed Emergence from Anesthesia Diseases 0.000 description 2
- 238000013499 data model Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 238000012216 screening Methods 0.000 description 2
- 206010039203 Road traffic accident Diseases 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
Images
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
- G06F16/53—Querying
-
- 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
- G06F16/56—Information retrieval; Database structures therefor; File system structures therefor of still image data having vectorial format
-
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24147—Distances to closest patterns, e.g. nearest neighbour classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biophysics (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Library & Information Science (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a single weather image identification method and a single weather image identification system based on image retrieval, wherein the method comprises the following steps: step S1, acquiring a training image, and training a preset initial model based on the training image to obtain a retrieval model; storing first vector features of a training image to a feature repository during a training process; step S2, inputting the single search image into a retrieval model, and performing feature extraction to obtain a second vector feature; and retrieving a corresponding third vector feature from the feature repository based on the second vector feature, and outputting the weather description of the training image corresponding to the third vector feature as a recognition result. The invention relates to a single weather image identification method based on image retrieval, which is based on a feature repository established during model training, realizes the quick identification of a single weather image and improves the identification rate of the single image.
Description
Technical Field
The invention relates to the technical field of image processing and pattern recognition, in particular to a single weather image recognition method and a single weather image recognition system based on image retrieval.
Background
At present, the weather identification under the natural scene plays an important role in the field of intelligent transportation, and the traffic transportation efficiency can be effectively improved due to the improvement of the accuracy of the weather identification. The single image recognition means that the input of the model only has one image and the recognition is completely carried out based on the image content. The traditional weather image identification method is to classify by using some prior information in the image, combine the characteristics of different information domains into multi-dimensional characteristics and send the multi-dimensional characteristics into a support vector machine for classification judgment. Unlike the traditional identification method, the deep learning-based weather identification has better identification effect, such as the identification method based on image semantic segmentation and image classification. Since the continuous progress of the technology promotes the gradual improvement of the image recognition effect, the image recognition has become a hot problem in recent years.
Image recognition methods based on image retrieval have proven to be very effective methods, and are widely used because image retrieval can effectively calculate the similarity between the features of a query graph and the features of images in a repository, and further obtain an output result according to the size of the similarity. However, in the application field of weather identification in natural scenes at present, an image identification method based on image retrieval does not relate to the application field, and the problem that the existing identification method is low in identification rate is always to be solved urgently.
Disclosure of Invention
One of the purposes of the invention is to provide a single weather image identification method based on image retrieval, which is based on a feature repository established during model training, realizes the rapid identification of a single weather image and improves the identification rate of the single image.
The embodiment of the invention provides a single weather image identification method based on image retrieval, which comprises the following steps:
step S1, acquiring a training image, and training a preset initial model based on the training image to obtain a retrieval model; storing first vector features of a training image to a feature repository during a training process;
step S2, inputting the single search image into a retrieval model, and performing feature extraction to obtain a second vector feature; and retrieving a corresponding third vector feature from the feature repository based on the second vector feature, and outputting the weather description of the training image corresponding to the third vector feature as a recognition result.
Preferably, in step S1, a training image is obtained, and a preset initial model is trained based on the training image to obtain a retrieval model; storing first vector features of a training image to a feature repository during a training process; the method comprises the following steps:
step S11: preprocessing a training image;
step S12: performing feature extraction on a training image by using a residual error network structure ResNet50, performing array conversion on a training label list, calculating final loss by combining measurement loss and a cross entropy loss function, and optimizing network parameters by a batch random gradient descent method to minimize a loss function;
step S13, in each training iteration, a first vector feature formation queue extracted from the current batch-size image is introduced into a feature repository, and when the feature repository is full, the corresponding old vector features in the feature repository are removed;
step S14: and during iteration, pairing the first vector features of the current batch-size image and all third vector features in the feature repository to form a large number of sample pairs, calculating the Similarity of the sample pairs by using a measurement Loss function Multi-Similarity Loss, and calculating the classification Loss of the current batch of images by using a cross entropy Loss function.
Preferably, step S2, inputting the single search image into the search model, and performing feature extraction to obtain a second vector feature; retrieving a corresponding third vector feature from the feature repository based on the second vector feature, and outputting a weather description of the training image corresponding to the third vector feature as a recognition result, including:
sequentially extracting test images in the test data set, performing feature extraction on the images after image normalization preprocessing, and adding fourth vector features of the test images into a feature repository;
preprocessing a single image to be inquired, inputting the single image into a retrieval model, and performing feature extraction to obtain a second vector feature for retrieving the single image;
and performing feature comparison and search on the second vector features in a feature storage library, calculating and inquiring K-class distances between the second vector features and third vector features in the feature storage library by utilizing a K-nearest neighbor algorithm idea, and determining the class label of the single image based on the K-class distances.
Preferably, storing the first vector feature of the training image to the feature repository during the training process includes:
acquiring first identification information of a training image, wherein the first identification information comprises: one or more of shooting position information, shooting time information and shooting equipment number information are combined;
grouping the feature storage banks based on grouping information corresponding to the first identification information to obtain a plurality of feature storage groups;
storing the first vector features into corresponding feature storage groups in a classified manner based on first identification information of the training images;
retrieving a corresponding third vector feature from the feature repository based on the second vector feature, including:
acquiring second identification information of a single searched image, wherein the second identification information comprises: one or more of shooting position information, shooting time information and shooting equipment number information are combined;
calling a corresponding feature storage group based on the second identification information;
and searching in the feature storage group based on the second vector, and determining a corresponding third vector.
Preferably, the single weather image identification method based on image retrieval further includes:
acquiring feedback information for the recognition output result;
when the feedback information is that the recognition result is wrong, acquiring a correct result input by a user;
storing the single search image corresponding to the recognition output result and the correct result into a correction database in an associated manner;
when the data volume in the correction database reaches a preset condition, extracting a single search image in the correction database, adding the single search image into the training image, and retraining the retrieval model;
wherein the preset conditions include:
correcting the data amount in the database to reach a preset first amount;
and/or the presence of a gas in the gas,
correcting the similar data in the database to a preset second number;
and/or the presence of a gas in the gas,
the storage time interval of the similar data in the correction database is smaller than the preset time interval;
and/or the presence of a gas in the gas,
classifying the data based on the similarity between the data in the correction database, and when the number of classification items reaches a preset third number;
when the feedback information for the identification output result is obtained, the validity of the feedback information needs to be verified, when the verification is passed, the feedback information is obtained, otherwise, the feedback information is not obtained; the verified conditions include:
the number of source users of the same feedback information is larger than the preset number of users;
and/or the presence of a gas in the gas,
when a single user is used as a source user to feed back feedback information, the confirmation degree of the feedback information is greater than a preset first confirmation value; the confirmation degree calculation formula is as follows:
wherein Q isiThe confirmation degree of the feedback information when the feedback information is fed back for the ith user; a isiFeedback confirmation values of the i users are pre-distributed; bjAgreeing values of agreeing users for the jth approval feedback information; beta is ajAgreeing weight values selected when agreeing with the feedback information for the jth agreeing user; m is the number of approved users;
and/or the presence of a gas in the gas,
when all source users of the same feedback information feed back the feedback information, the sum of the confirmation degrees of the feedback information is greater than a preset second confirmation value;
after the retrieval model is retrained, the feedback confirmation value of the source user of the feedback information is corrected, and the correction formula is as follows:
wherein a' is a feedback confirmation value of the modified source user; a is a feedback confirmation value of the source user before correction; c is a correction coefficient, is related to the use condition of the feedback information, is verified as feedback error when the feedback information is used, c is-1, and is verified as feedback correct when the feedback information is used, c is 1; the amplitude value is a preset corrected amplitude value;0correcting amplitude values for the preset corrected amplitude values; a ismaxThe feedback confirmation value is a preset maximum feedback confirmation value; a isminThe feedback confirmation value is a preset minimum feedback confirmation value.
The invention also provides a single weather image identification system based on image retrieval, which comprises the following components:
the model training module is used for acquiring a training image, and training a preset initial model based on the training image to acquire a retrieval model;
the characteristic repository establishing module is used for storing the first vector characteristics of the training image into the characteristic repository in the training process;
the retrieval module is used for inputting the single search image into a retrieval model, and performing feature extraction to obtain a second vector feature; retrieving a corresponding third vector feature from the feature store based on the second vector feature,
and the output module is used for outputting the weather description of the training image corresponding to the third vector characteristic as a recognition result.
Preferably, the model training module performs the following operations:
preprocessing a training image;
performing feature extraction on a training image by using a residual error network structure ResNet50, performing array conversion on a training label list, calculating final loss by combining measurement loss and a cross entropy loss function, and optimizing network parameters by a batch random gradient descent method to minimize a loss function;
in each training iteration, a first vector feature formation queue extracted from a current batch-size image is introduced into a feature repository, and when the feature repository is full, corresponding old vector features in the feature repository are removed;
and during iteration, pairing the first vector features of the current batch-size image and all third vector features in the feature repository to form a large number of sample pairs, calculating the Similarity of the sample pairs by using a measurement Loss function Multi-Similarity Loss, and calculating the classification Loss of the current batch of images by using a cross entropy Loss function.
Preferably, the retrieval module performs the following operations:
sequentially extracting test images in the test data set, performing feature extraction on the images after image normalization preprocessing, and adding fourth vector features of the test images into a feature repository;
preprocessing a single image to be inquired, inputting the single image into a retrieval model, and performing feature extraction to obtain a second vector feature for retrieving the single image;
and performing feature comparison and search on the second vector features in a feature storage library, calculating and inquiring K-class distances between the second vector features and third vector features in the feature storage library by utilizing a K-nearest neighbor algorithm idea, and determining the class label of the single image based on the K-class distances.
Preferably, storing the first vector feature of the training image to the feature repository during the training process includes:
acquiring first identification information of a training image, wherein the first identification information comprises: one or more of shooting position information, shooting time information and shooting equipment number information are combined;
grouping the feature storage banks based on grouping information corresponding to the first identification information to obtain a plurality of feature storage groups;
storing the first vector features into corresponding feature storage groups in a classified manner based on first identification information of the training images;
retrieving a corresponding third vector feature from the feature repository based on the second vector feature, including:
acquiring second identification information of a single searched image, wherein the second identification information comprises: one or more of shooting position information, shooting time information and shooting equipment number information are combined;
calling a corresponding feature storage group based on the second identification information;
and searching in the feature storage group based on the second vector, and determining a corresponding third vector.
Preferably, the single weather image recognition system based on image retrieval further includes: an update module that performs the following operations:
acquiring feedback information for the recognition output result;
when the feedback information is that the recognition result is wrong, acquiring a correct result input by a user;
storing the single search image corresponding to the recognition output result and the correct result into a correction database in an associated manner;
when the data volume in the correction database reaches a preset condition, extracting a single search image in the correction database, adding the single search image into the training image, and retraining the retrieval model;
wherein the preset conditions include:
correcting the data amount in the database to reach a preset first amount;
and/or the presence of a gas in the gas,
correcting the similar data in the database to a preset second number;
and/or the presence of a gas in the gas,
the storage time interval of the similar data in the correction database is smaller than the preset time interval;
and/or the presence of a gas in the gas,
classifying the data based on the similarity between the data in the correction database, and when the number of classification items reaches a preset third number;
when the feedback information for the identification output result is obtained, the validity of the feedback information needs to be verified, when the verification is passed, the feedback information is obtained, otherwise, the feedback information is not obtained; the verified conditions include:
the number of source users of the same feedback information is larger than the preset number of users;
and/or the presence of a gas in the gas,
when a single user is used as a source user to feed back feedback information, the confirmation degree of the feedback information is greater than a preset first confirmation value; the confirmation degree calculation formula is as follows:
wherein Q isiThe confirmation degree of the feedback information when the feedback information is fed back for the ith user; a isiFeedback confirmation values of the i users are pre-distributed; bjAgreeing values of agreeing users for the jth approval feedback information; beta is ajAgreeing weight values selected when agreeing with the feedback information for the jth agreeing user; m is the number of approved users;
and/or the presence of a gas in the gas,
when all source users of the same feedback information feed back the feedback information, the sum of the confirmation degrees of the feedback information is greater than a preset second confirmation value;
after the retrieval model is retrained, the feedback confirmation value of the source user of the feedback information is corrected, and the correction formula is as follows:
wherein a' is a feedback confirmation value of the modified source user; a is a feedback confirmation value of the source user before correction; c is a correction coefficient, is related to the use condition of the feedback information, is verified as feedback error when the feedback information is used, c is-1, and is verified as feedback correct when the feedback information is used, c is 1; the amplitude value is a preset corrected amplitude value;0correcting amplitude values for the preset corrected amplitude values; a ismaxThe feedback confirmation value is a preset maximum feedback confirmation value; a isminThe feedback confirmation value is a preset minimum feedback confirmation value.
Compared with the prior art, the invention has the advantages that: firstly, the data set data image used by the invention is automatically collected and manufactured, focuses on the traffic scene under the natural scene, not only can identify the weather image, but also can identify the natural disaster image and the traffic accident image of the road, and has good application prospect. Secondly, an image retrieval method is innovatively adopted for the weather image recognition task, and the accuracy of weather image recognition is effectively improved.
Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.
The technical solution of the present invention is further described in detail by the accompanying drawings and embodiments.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the principles of the invention and not to limit the invention. In the drawings:
FIG. 1 is a schematic diagram of a single weather image identification method based on image retrieval according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of the creation of a search model;
fig. 3 is a schematic diagram of a single image search process.
Detailed Description
The preferred embodiments of the present invention will be described in conjunction with the accompanying drawings, and it will be understood that they are described herein for the purpose of illustration and explanation and not limitation.
The embodiment of the invention provides a single weather image identification method based on image retrieval, which comprises the following steps of:
step S1, acquiring a training image, and training a preset initial model based on the training image to obtain a retrieval model; storing first vector features of a training image to a feature repository during a training process;
step S2, inputting the single search image into a retrieval model, and performing feature extraction to obtain a second vector feature; and retrieving a corresponding third vector feature from the feature repository based on the second vector feature, and outputting the weather description of the training image corresponding to the third vector feature as a recognition result.
The working principle and the beneficial effects of the technical scheme are as follows:
as shown in fig. 2, a CNN network is used to perform feature extraction on a training image to obtain vector features representing the image; the CNN network utilizes a ResNet network architecture, and enables the network to move to a deeper level under the condition of not greatly increasing the calculated amount through a residual block structure; introducing a feature storage mechanism Memory Bank to the extracted image vector features, adding new features extracted from the current mini-batch samples into a queue every iteration, and removing the old features; in the training process, effectively performing metric learning by using an MS Loss metric function, and training to obtain a retrieval model;
as shown in fig. 3, a single search image is input to the search model for feature extraction; performing feature extraction under test image line concentration, and performing feature storage by using a Memory Bank; and carrying out feature comparison search on the feature vectors of the single searched image in a Memory Bank, and classifying by using a K neighbor algorithm to obtain the category of the searched image so as to realize the identification of the weather image.
Further, the weather training image is subjected to pixel normalization processing, is randomly cut and then is uniformly scaled to 227 multiplied by 227 image blocks, is randomly subjected to horizontal overturning, and then is input into a convolutional neural network;
the measurement Loss function adopts Multi-Similarity Loss to effectively calculate the Similarity of the current batch-size characteristic and all the characteristics in a characteristic Memory Bank, calculate the measurement Loss and improve a retrieval network, unite the Similarity measurement and the cross entropy Loss to supervise and train the convolutional neural network, the Similarity measurement during the test is consistent with the Similarity measurement during the training, and the test performance is more stable for the image with larger difference between the training and the test and is more suitable for the retrieval scene.
In one embodiment, step S1 includes obtaining a training image, training a preset initial model based on the training image to obtain a retrieval model; storing first vector features of a training image to a feature repository during a training process; the method comprises the following steps:
step S11: preprocessing a training image; preprocessing the image data of the training set before training, firstly, establishing indexes for the images and the categories to which the images belong, and forming a label list; secondly, performing normalization processing on the weather images, unifying the sizes of the images into 256 multiplied by 256, randomly selecting image blocks with the size of 227 multiplied by 227 from the 256 multiplied by 256 images during each training, and inputting the image blocks into a network for training after randomly performing horizontal turning;
step S12: and (3) performing feature extraction on the training image by using a residual network structure ResNet50, performing array conversion on a training label list, calculating final loss by combining measurement loss and a cross entropy loss function, and optimizing network parameters by a batch random gradient descent method to minimize the loss function. The ResNet-50 has smaller calculation amount and is more suitable for real-time scene application. And training and fine-tuning the ResNet-50 convolutional neural network by using the weather image based on the measurement loss and the cross entropy loss so as to be suitable for a weather retrieval scene. And a feature embedding layer is added on the ResNet-50 structure, so that the low-dimensional feature is obtained directly based on measurement loss training without performing dimension reduction processing on the feature.
Step S13, in each training iteration, a first vector feature formation queue extracted from the current batch-size image is introduced into a feature repository, and when the feature repository is full, the corresponding old vector features in the feature repository are removed; thus ensuring that all feature information in Memory Bank (feature repository) is kept up to date.
Step S14: and during iteration, pairing the first vector features of the current batch-size image and all third vector features in the feature repository to form a large number of sample pairs, calculating the Similarity of the sample pairs by using a measurement Loss function Multi-Similarity Loss, and calculating the classification Loss of the current batch of images by using a cross entropy Loss function. More efficient sample training is achieved.
In one embodiment, step S2, inputting the single search image into the search model, and performing feature extraction to obtain a second vector feature; retrieving a corresponding third vector feature from the feature repository based on the second vector feature, and outputting a weather description of the training image corresponding to the third vector feature as a recognition result, including:
sequentially extracting test images in the test data set, performing feature extraction on the images after image normalization preprocessing, and adding fourth vector features of the test images into a feature repository; and (4) sequentially extracting the test images in the test data set, performing feature extraction on the images after image normalization processing, and forming a feature repository Memory Bank in the step S1 to form dynamic storage of feature vectors.
Preprocessing a single image to be inquired, inputting the single image into a retrieval model, and performing feature extraction to obtain a second vector feature for retrieving the single image; and (4) preprocessing a single image to be inquired, inputting the preprocessed single image into the retrieval model obtained by training in the step S1, and performing feature extraction on the retrieved image to obtain a feature vector of the retrieved image.
And performing feature comparison and search on the second vector features in a feature storage library, calculating and inquiring K-class distances between the second vector features and third vector features in the feature storage library by utilizing a K-nearest neighbor algorithm idea, and determining the class label of the single image based on the K-class distances. Feature comparison and search are carried out on the feature vectors of the single weather image in a feature repository Memory Bank, K-type distances of the query feature vectors are calculated by utilizing the thought of a K-nearest neighbor algorithm, classification output is realized after sorting, a category label to which the query image belongs is obtained, and identification of the weather image is realized.
Further, a loss function of the training retrieval network adopts cross entropy cross Encopy loss and Multi-Similarity loss; in the Multi-Similarity loss, the weight parameter SCALE _ POS is set to be 2.0, and SCALE _ NEG is 40.0;
further, the batch-size at training is set to 55, the batch-size at testing is set to 128, and the number of work processes is set to 8.
In one embodiment, storing a first vector feature of a training image to a feature repository during a training process includes:
acquiring first identification information of a training image, wherein the first identification information comprises: one or more of shooting position information, shooting time information and shooting equipment number information are combined;
grouping the feature storage banks based on grouping information corresponding to the first identification information to obtain a plurality of feature storage groups;
storing the first vector features into corresponding feature storage groups in a classified manner based on first identification information of the training images;
retrieving a corresponding third vector feature from the feature repository based on the second vector feature, including:
acquiring second identification information of a single searched image, wherein the second identification information comprises: one or more of shooting position information, shooting time information and shooting equipment number information are combined;
calling a corresponding feature storage group based on the second identification information;
and searching in the feature storage group based on the second vector, and determining a corresponding third vector.
The working principle and the beneficial effects of the technical scheme are as follows:
weather conditions possibly occurring in the same season in the same place are much less than the total weather conditions, so that the weather conditions can be identified by using representative places [ shooting position information ] and representative seasons [ shooting time information ], and the feature repository is grouped based on the first identification information and the second identification information, so that the data comparison amount during single image detection is reduced, and the detection speed is improved; the number information of the shooting equipment is also used as identification information, and the shooting equipment is mainly installed and cannot move generally, so that the weather condition shot by the equipment is less than the total weather condition.
In one embodiment, the single weather image identification method based on image retrieval further comprises:
acquiring feedback information for the recognition output result;
when the feedback information is that the recognition result is wrong, acquiring a correct result input by a user;
storing the single search image corresponding to the recognition output result and the correct result into a correction database in an associated manner;
when the data volume in the correction database reaches a preset condition, extracting a single search image in the correction database, adding the single search image into the training image, and retraining the retrieval model;
wherein the preset conditions include:
correcting the data amount in the database to reach a preset first amount;
and/or the presence of a gas in the gas,
correcting the similar data in the database to a preset second number;
and/or the presence of a gas in the gas,
the storage time interval of the similar data in the correction database is smaller than the preset time interval;
and/or the presence of a gas in the gas,
classifying the data based on the similarity between the data in the correction database, and when the number of classification items reaches a preset third number;
when the feedback information for the identification output result is obtained, the validity of the feedback information needs to be verified, when the verification is passed, the feedback information is obtained, otherwise, the feedback information is not obtained; the verified conditions include:
the number of source users of the same feedback information is larger than the preset number of users;
and/or the presence of a gas in the gas,
when a single user is used as a source user to feed back feedback information, the confirmation degree of the feedback information is greater than a preset first confirmation value; the confirmation degree calculation formula is as follows:
wherein Q isiThe confirmation degree of the feedback information when the feedback information is fed back for the ith user; a isiFeedback confirmation values of the i users are pre-distributed; bjAgreeing values of agreeing users for the jth approval feedback information; beta is ajAgreeing weight values selected when agreeing with the feedback information for the jth agreeing user; m is the number of approved users;
and/or the presence of a gas in the gas,
when all source users of the same feedback information feed back the feedback information, the sum of the confirmation degrees of the feedback information is greater than a preset second confirmation value;
after the retrieval model is retrained, the feedback confirmation value of the source user of the feedback information is corrected, and the correction formula is as follows:
wherein a' is a feedback confirmation value of the modified source user; a is a feedback confirmation value of the source user before correction; c is a correction coefficient, is related to the use condition of the feedback information, is verified as feedback error when the feedback information is used, c is-1, and is verified as feedback correct when the feedback information is used, c is 1; the amplitude value is a preset corrected amplitude value;0correcting amplitude values for the preset corrected amplitude values; a ismaxThe feedback confirmation value is a preset maximum feedback confirmation value; a isminThe feedback confirmation value is a preset minimum feedback confirmation value.
The working principle and the beneficial effects of the technical scheme are as follows:
acquiring an updated material (a single search image with identification errors) by collecting and analyzing feedback information of an identification output result, collecting the material into a correction database, and calling data to retrain a retrieval model when the data in the correction database meets a preset condition so as to update a data model; the updating time is determined by adopting a preset condition, so that the condition that the stability of the identification system is influenced due to large fluctuation of a detection model caused by frequent updating is prevented; in the feedback information collection link, the feedback of the feedback information needs to be verified, the feedback information is screened, and the interfered feedback information is eliminated. When a user feeds back, the confirmation degree of the feedback information is introduced to identify the feedback information, the confirmation degree is mainly determined by the fed-back user and the approver approving the feedback information, and the determining method enhances the identification effect of the confirmation degree on the feedback information and can reflect that the feedback information is effective; when the feedback information is verified in use, the feedback information is judged to be abnormal data or abnormal feedback by a manager when the verification way is a training model; the accuracy of the next feedback information screening is improved by correcting the feedback confirmation value of the source user.
The invention also provides a single weather image identification system based on image retrieval, which comprises the following components:
the model training module is used for acquiring a training image, and training a preset initial model based on the training image to acquire a retrieval model;
the characteristic repository establishing module is used for storing the first vector characteristics of the training image into the characteristic repository in the training process;
the retrieval module is used for inputting the single search image into a retrieval model, and performing feature extraction to obtain a second vector feature; retrieving a corresponding third vector feature from the feature store based on the second vector feature,
and the output module is used for outputting the weather description of the training image corresponding to the third vector characteristic as a recognition result.
The working principle and the beneficial effects of the technical scheme are as follows:
as shown in fig. 2, a CNN network is used to perform feature extraction on a training image to obtain vector features representing the image; the CNN network utilizes a ResNet network architecture, and enables the network to move to a deeper level under the condition of not greatly increasing the calculated amount through a residual block structure; introducing a feature storage mechanism Memory Bank to the extracted image vector features, adding new features extracted by the current mini-batch into a queue every iteration, and removing old features; in the training process, effectively training the model by combining the MS Loss measurement Loss function and the cross entropy Loss function, and training to obtain a retrieval model;
as shown in fig. 3, a single search image is input to the search model for feature extraction; performing feature extraction under test image line concentration, and performing feature storage by using a Memory Bank; and carrying out feature comparison search on the feature vectors of the single searched image in a Memory Bank, and classifying by using a K neighbor algorithm to obtain the category of the searched image so as to realize the identification of the weather image.
Further, the weather training image is subjected to pixel normalization processing, is randomly cut and then is uniformly scaled to 227 multiplied by 227 image blocks, is randomly subjected to horizontal overturning, and then is input into a convolutional neural network;
the measurement Loss function adopts Multi-Similarity Loss to effectively calculate the Similarity of the characteristics of the current batch-size and all the characteristics in the characteristic repository Memory Bank, calculate the measurement Loss and improve the retrieval network.
In one embodiment, the model training module performs the following operations:
preprocessing a training image;
performing feature extraction on a training image by using a residual error network structure ResNet50, performing array conversion on a training label list, calculating final loss by combining measurement loss and a cross entropy loss function, and optimizing network parameters by a batch random gradient descent method to minimize a loss function;
in each training iteration, a first vector feature formation queue extracted from a current batch-size image is introduced into a feature repository, and when the feature repository is full, corresponding old vector features in the feature repository are removed;
and during iteration, pairing the first vector features of the current batch-size image and all third vector features in the feature repository to form a large number of sample pairs, calculating the Similarity of the sample pairs by using a measurement Loss function Multi-Similarity Loss, and calculating the classification Loss of the current batch of images by using a cross entropy Loss function.
In one embodiment, the retrieval module performs the following operations:
sequentially extracting test images in the test data set, performing feature extraction on the images after image normalization preprocessing, and adding fourth vector features of the test images into a feature repository;
preprocessing a single image to be inquired, inputting the single image into a retrieval model, and performing feature extraction to obtain a second vector feature for retrieving the single image;
and performing feature comparison and search on the second vector features in a feature storage library, calculating and inquiring K-class distances between the second vector features and third vector features in the feature storage library by utilizing a K-nearest neighbor algorithm idea, and determining the class label of the single image based on the K-class distances.
In one embodiment, storing a first vector feature of a training image to a feature repository during a training process includes:
acquiring first identification information of a training image, wherein the first identification information comprises: one or more of shooting position information, shooting time information and shooting equipment number information are combined;
grouping the feature storage banks based on grouping information corresponding to the first identification information to obtain a plurality of feature storage groups;
storing the first vector features into corresponding feature storage groups in a classified manner based on first identification information of the training images;
retrieving a corresponding third vector feature from the feature repository based on the second vector feature, including:
acquiring second identification information of a single searched image, wherein the second identification information comprises: one or more of shooting position information, shooting time information and shooting equipment number information are combined;
calling a corresponding feature storage group based on the second identification information;
and searching in the feature storage group based on the second vector, and determining a corresponding third vector.
The working principle and the beneficial effects of the technical scheme are as follows:
weather conditions possibly occurring in the same season in the same place are much less than the total weather conditions, so that the weather conditions can be identified by using representative places [ shooting position information ] and representative seasons [ shooting time information ], and the feature repository is grouped based on the first identification information and the second identification information, so that the data comparison amount during single image detection is reduced, and the detection speed is improved; the number information of the shooting equipment is also used as identification information, and the shooting equipment is mainly installed and cannot move generally, so that the weather condition shot by the equipment is less than the total weather condition.
In one embodiment, the single weather image recognition system based on image retrieval further comprises: an update module that performs the following operations:
acquiring feedback information for the recognition output result;
when the feedback information is that the recognition result is wrong, acquiring a correct result input by a user;
storing the single search image corresponding to the recognition output result and the correct result into a correction database in an associated manner;
when the data volume in the correction database reaches a preset condition, extracting a single search image in the correction database, adding the single search image into the training image, and retraining the retrieval model;
wherein the preset conditions include:
correcting the data amount in the database to reach a preset first amount;
and/or the presence of a gas in the gas,
correcting the similar data in the database to a preset second number;
and/or the presence of a gas in the gas,
the storage time interval of the similar data in the correction database is smaller than the preset time interval;
and/or the presence of a gas in the gas,
classifying the data based on the similarity between the data in the correction database, and when the number of classification items reaches a preset third number;
when the feedback information for the identification output result is obtained, the validity of the feedback information needs to be verified, when the verification is passed, the feedback information is obtained, otherwise, the feedback information is not obtained; the verified conditions include:
the number of source users of the same feedback information is larger than the preset number of users;
and/or the presence of a gas in the gas,
when a single user is used as a source user to feed back feedback information, the confirmation degree of the feedback information is greater than a preset first confirmation value; the confirmation degree calculation formula is as follows:
wherein Q isiThe confirmation degree of the feedback information when the feedback information is fed back for the ith user; a isiFeedback confirmation values of the i users are pre-distributed; bjAgreeing values of agreeing users for the jth approval feedback information; beta is ajAgreeing weight values selected when agreeing with the feedback information for the jth agreeing user; m is the number of approved users;
and/or the presence of a gas in the gas,
when all source users of the same feedback information feed back the feedback information, the sum of the confirmation degrees of the feedback information is greater than a preset second confirmation value;
after the retrieval model is retrained, the feedback confirmation value of the source user of the feedback information is corrected, and the correction formula is as follows:
wherein a' is a feedback confirmation value of the modified source user; a is correctionFeedback confirmation value of previous source user; c is a correction coefficient, is related to the use condition of the feedback information, is verified as feedback error when the feedback information is used, c is-1, and is verified as feedback correct when the feedback information is used, c is 1; the amplitude value is a preset corrected amplitude value;0correcting amplitude values for the preset corrected amplitude values; a ismaxThe feedback confirmation value is a preset maximum feedback confirmation value; a isminThe feedback confirmation value is a preset minimum feedback confirmation value.
The working principle and the beneficial effects of the technical scheme are as follows:
acquiring an updated material (a single search image with identification errors) by collecting and analyzing feedback information of an identification output result, collecting the material into a correction database, and calling data to retrain a retrieval model when the data in the correction database meets a preset condition so as to update a data model; the updating time is determined by adopting a preset condition, so that the condition that the stability of the identification system is influenced due to large fluctuation of a detection model caused by frequent updating is prevented; in the feedback information collection link, the feedback of the feedback information needs to be verified, the feedback information is screened, and the interfered feedback information is eliminated. When a user feeds back, the confirmation degree of the feedback information is introduced to identify the feedback information, the confirmation degree is mainly determined by the fed-back user and the approver approving the feedback information, and the determining method enhances the identification effect of the confirmation degree on the feedback information and can reflect that the feedback information is effective; when the feedback information is verified in use, the feedback information is judged to be abnormal data or abnormal feedback by a manager when the verification way is a training model; the accuracy of the next feedback information screening is improved by correcting the feedback confirmation value of the source user.
It will be apparent to those skilled in the art that various changes and modifications may be made in the present invention without departing from the spirit and scope of the invention. Thus, if such modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include such modifications and variations.
Claims (10)
1. A single weather image identification method based on image retrieval is characterized by comprising the following steps:
step S1, acquiring a training image, and training a preset initial model based on the training image to obtain a retrieval model; storing a first vector feature of the training image to a feature repository during a training process;
step S2, inputting a single search image into the search model, and performing feature extraction to obtain a second vector feature; and retrieving a corresponding third vector feature from the feature repository based on the second vector feature, and outputting the weather description of the training image corresponding to the third vector feature as a recognition result.
2. The method for recognizing a single weather image based on image retrieval as claimed in claim 1, wherein the step S1 comprises obtaining a training image, training a preset initial model based on the training image to obtain a retrieval model; storing a first vector feature of the training image to a feature repository during a training process; the method comprises the following steps:
step S11: preprocessing the training image;
step S12: performing feature extraction on the training image by using a residual error network structure ResNet50, performing array conversion on a training label list, calculating final loss by combining measurement loss and a cross entropy loss function, and optimizing network parameters by a batch random gradient descent method to minimize a loss function;
step S13, in each training iteration, a first vector feature formation queue extracted from the current batch-size image is introduced into a feature repository, and when the feature repository is full, the corresponding old vector features in the feature repository are removed;
step S14: and during iteration, matching the first vector features of the current batch-size image and all third vector features in the feature repository to form a large number of sample pairs, calculating the Similarity of the sample pairs by using a measurement Loss function Multi-Similarity Loss, and calculating the classification Loss of the current batch of images by using a cross entropy Loss function.
3. The method for recognizing the single weather image based on the image retrieval as claimed in claim 1, wherein the step S2 is to input the single search image into the retrieval model for feature extraction to obtain the second vector feature; retrieving a corresponding third vector feature from the feature repository based on the second vector feature, and outputting a weather description of the training image corresponding to the third vector feature as a recognition result, including:
sequentially extracting test images in the test data set, performing feature extraction on the images after image normalization preprocessing, and adding fourth vector features of the test images into the feature repository;
preprocessing a single image to be inquired, inputting the single image into the retrieval model, and performing feature extraction to obtain a second vector feature for retrieving the single image;
and performing feature comparison and search on the second vector features in the feature storage library, calculating and inquiring K-class distances between the second vector features and third vector features in the feature storage library by utilizing a K-nearest neighbor algorithm idea, and determining the class label of the single image based on the K-class distances.
4. The image-retrieval-based single weather image recognition method of claim 1, wherein storing the first vector features of the training image to a feature repository during training comprises:
acquiring first identification information of the training image, wherein the first identification information comprises: one or more of shooting position information, shooting time information and shooting equipment number information are combined;
grouping the feature storage libraries based on grouping information corresponding to the first identification information to obtain a plurality of feature storage groups;
storing the first vector feature into the corresponding feature storage group based on the first identification information of the training image in a classified manner;
retrieving, by the processor, a corresponding third vector feature from the feature repository based on the second vector feature, including:
acquiring second identification information of the single search image, wherein the second identification information comprises: one or more of shooting position information, shooting time information and shooting equipment number information are combined;
calling the corresponding feature storage group based on the second identification information;
and searching in the feature storage group based on the second vector, and determining the corresponding third vector.
5. The image-retrieval-based single weather image recognition method of claim 1, further comprising:
acquiring feedback information of the recognition output result;
when the feedback information is that the identification result is wrong, acquiring a correct result input by a user;
storing the single search image corresponding to the recognition output result and the correct result in a correction database in a correlation manner;
when the data volume in the correction database reaches a preset condition, extracting the single search image in the correction database, adding the single search image into the training image, and retraining the retrieval model;
wherein the preset conditions include:
the data volume in the correction database reaches a preset first quantity;
and/or the presence of a gas in the gas,
the similar data in the correction database reach a preset second quantity;
and/or the presence of a gas in the gas,
the storage time interval of the similar data in the correction database is smaller than a preset time interval;
and/or the presence of a gas in the gas,
classifying the data based on the similarity between the data in the correction database, when the number of classification items reaches a preset third number;
when feedback information for the identification output result is obtained, the validity of the feedback information needs to be verified, when the verification is passed, the feedback information is obtained, otherwise, the feedback information is not obtained; the verified conditions include:
the number of source users of the same feedback information is larger than the preset number of users;
and/or the presence of a gas in the gas,
when a single user is used as a source user to feed back the feedback information, the confirmation degree of the feedback information is greater than a preset first confirmation value; the confirmation degree calculation formula is as follows:
wherein Q isiConfirming degree of the feedback information when the feedback information is fed back for the ith user; a isiFeedback confirmation values of the i users are pre-distributed; bjAn approval value for the jth approved user who approves the feedback information; beta is ajAgreeing weight values selected by the jth agreeing user when agreeing with the feedback information; m is the number of approved users;
and/or the presence of a gas in the gas,
when all source users of the same feedback information feed back the feedback information, the sum of the confirmation degrees of the feedback information is greater than a preset second confirmation value;
after the retrieval model is retrained, feedback confirmation value correction is carried out on the source user of the feedback information, and a correction formula is as follows:
wherein a' is a feedback confirmation value of the modified source user; a is a feedback confirmation value of the source user before correction; c is a correction coefficient, and is related to the use condition of the feedback information when the feedback information is usedC is verified as a feedback error, c is-1, when the feedback information is verified as a correct feedback when in use, c is 1; the amplitude value is a preset corrected amplitude value;0correcting amplitude values for the preset corrected amplitude values; a ismaxThe feedback confirmation value is a preset maximum feedback confirmation value; a isminThe feedback confirmation value is a preset minimum feedback confirmation value.
6. A single weather image identification system based on image retrieval is characterized by comprising:
the model training module is used for acquiring a training image, and training a preset initial model based on the training image to acquire a retrieval model;
the characteristic repository establishing module is used for storing the first vector characteristic of the training image into a characteristic repository in the training process;
the retrieval module is used for inputting a single search image into the retrieval model, and performing feature extraction to obtain a second vector feature; retrieving a corresponding third vector feature from the feature store based on the second vector feature,
and the output module is used for outputting the weather description of the training image corresponding to the third vector feature as a recognition result.
7. The image-retrieval-based single weather image recognition system of claim 6, wherein the model training module performs the following operations:
preprocessing the training image;
performing feature extraction on the training image by using a residual error network structure ResNet50, performing array conversion on a training label list, calculating final loss by combining measurement loss and a cross entropy loss function, and optimizing network parameters by a batch random gradient descent method to minimize a loss function;
in each training iteration, a first vector feature formation queue extracted from a current batch-size image is introduced into a feature repository, and when the feature repository is full, corresponding old vector features in the feature repository are removed;
and during iteration, matching the first vector features of the current batch-size image and all third vector features in the feature repository to form a large number of sample pairs, calculating the Similarity of the sample pairs by using a measurement Loss function Multi-Similarity Loss, and calculating the classification Loss of the current batch of images by using a cross entropy Loss function.
8. The image-retrieval-based single weather image recognition system of claim 6, wherein the retrieval module performs the following operations:
sequentially extracting test images in the test data set, performing feature extraction on the images after image normalization preprocessing, and adding fourth vector features of the test images into the feature repository;
preprocessing a single image to be inquired, inputting the single image into the retrieval model, and performing feature extraction to obtain a second vector feature for retrieving the single image;
and performing feature comparison and search on the second vector features in the feature storage library, calculating and inquiring K-class distances between the second vector features and third vector features in the feature storage library by utilizing a K-nearest neighbor algorithm idea, and determining the class label of the single image based on the K-class distances.
9. The image-retrieval-based single weather image recognition system of claim 6, wherein storing the first vector features of the training image to a feature repository during training comprises:
acquiring first identification information of the training image, wherein the first identification information comprises: one or more of shooting position information, shooting time information and shooting equipment number information are combined;
grouping the feature storage libraries based on grouping information corresponding to the first identification information to obtain a plurality of feature storage groups;
storing the first vector feature into the corresponding feature storage group based on the first identification information of the training image in a classified manner;
retrieving, by the processor, a corresponding third vector feature from the feature repository based on the second vector feature, including:
acquiring second identification information of the single search image, wherein the second identification information comprises: one or more of shooting position information, shooting time information and shooting equipment number information are combined;
calling the corresponding feature storage group based on the second identification information;
and searching in the feature storage group based on the second vector, and determining the corresponding third vector.
10. The image-retrieval-based single weather image recognition system of claim 6, further comprising: an update module that performs the following operations:
acquiring feedback information of the recognition output result;
when the feedback information is that the identification result is wrong, acquiring a correct result input by a user;
storing the single search image corresponding to the recognition output result and the correct result in a correction database in a correlation manner;
when the data volume in the correction database reaches a preset condition, extracting the single search image in the correction database, adding the single search image into the training image, and retraining the retrieval model;
wherein the preset conditions include:
the data volume in the correction database reaches a preset first quantity;
and/or the presence of a gas in the gas,
the similar data in the correction database reach a preset second quantity;
and/or the presence of a gas in the gas,
the storage time interval of the similar data in the correction database is smaller than a preset time interval;
and/or the presence of a gas in the gas,
classifying the data based on the similarity between the data in the correction database, when the number of classification items reaches a preset third number;
when feedback information for the identification output result is obtained, the validity of the feedback information needs to be verified, when the verification is passed, the feedback information is obtained, otherwise, the feedback information is not obtained; the verified conditions include:
the number of source users of the same feedback information is larger than the preset number of users;
and/or the presence of a gas in the gas,
when a single user is used as a source user to feed back the feedback information, the confirmation degree of the feedback information is greater than a preset first confirmation value; the confirmation degree calculation formula is as follows:
wherein Q isiConfirming degree of the feedback information when the feedback information is fed back for the ith user; a isiFeedback confirmation values of the i users are pre-distributed; bjAn approval value for the jth approved user who approves the feedback information; beta is ajAgreeing weight values selected by the jth agreeing user when agreeing with the feedback information; m is the number of approved users;
and/or the presence of a gas in the gas,
when all source users of the same feedback information feed back the feedback information, the sum of the confirmation degrees of the feedback information is greater than a preset second confirmation value;
after the retrieval model is retrained, feedback confirmation value correction is carried out on the source user of the feedback information, and a correction formula is as follows:
wherein a' is a feedback confirmation value of the modified source user; a is the source user before correctionFeeding back an acknowledgement value; c is a correction coefficient, is related to the use condition of the feedback information, when the feedback information is verified to be in a feedback error in use, c is-1, when the feedback information is verified to be in a correct feedback in use, c is 1; the amplitude value is a preset corrected amplitude value;0correcting amplitude values for the preset corrected amplitude values; a ismaxThe feedback confirmation value is a preset maximum feedback confirmation value; a isminThe feedback confirmation value is a preset minimum feedback confirmation value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010967561.4A CN112100419B (en) | 2020-09-15 | 2020-09-15 | Single weather image recognition method and system based on image retrieval |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010967561.4A CN112100419B (en) | 2020-09-15 | 2020-09-15 | Single weather image recognition method and system based on image retrieval |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112100419A true CN112100419A (en) | 2020-12-18 |
| CN112100419B CN112100419B (en) | 2024-02-06 |
Family
ID=73759769
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010967561.4A Active CN112100419B (en) | 2020-09-15 | 2020-09-15 | Single weather image recognition method and system based on image retrieval |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112100419B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115331062A (en) * | 2022-08-29 | 2022-11-11 | 北京达佳互联信息技术有限公司 | Image recognition method and device, electronic equipment and computer-readable storage medium |
| WO2024038941A1 (en) * | 2022-08-17 | 2024-02-22 | 쿠팡 주식회사 | Data providing method and device |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104598916A (en) * | 2014-09-11 | 2015-05-06 | 单勇 | Establishment method of train recognition system and train recognition method |
| CN106776675A (en) * | 2016-11-01 | 2017-05-31 | 上海斐讯数据通信技术有限公司 | A kind of plants identification method and system based on mobile terminal |
| CN107169148A (en) * | 2017-06-21 | 2017-09-15 | 北京百度网讯科技有限公司 | Image search method, device, equipment and storage medium |
| CN109118779A (en) * | 2018-10-12 | 2019-01-01 | 东软集团股份有限公司 | Break in traffic rules and regulations information identifying method, equipment and readable storage medium storing program for executing |
| CN109740627A (en) * | 2018-11-27 | 2019-05-10 | 南京邮电大学 | A kind of insect image identification identifying system and its method based on parallel-convolution neural network |
| CN110046738A (en) * | 2019-01-28 | 2019-07-23 | 南京林业大学 | A kind of forest fire prediction technique based on artificial intelligence perceptron model |
| CN110197258A (en) * | 2019-05-29 | 2019-09-03 | 北京市商汤科技开发有限公司 | Neural network searching method, image processing method and device, equipment and medium |
| US20200027459A1 (en) * | 2019-09-09 | 2020-01-23 | Lg Electronics Inc. | Artificial intelligence apparatus and method for recognizing speech of user |
| US20200051180A1 (en) * | 2017-08-28 | 2020-02-13 | The Climate Corporation | Crop disease recognition and yield estimation |
| US20200184278A1 (en) * | 2014-03-18 | 2020-06-11 | Z Advanced Computing, Inc. | System and Method for Extremely Efficient Image and Pattern Recognition and Artificial Intelligence Platform |
-
2020
- 2020-09-15 CN CN202010967561.4A patent/CN112100419B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20200184278A1 (en) * | 2014-03-18 | 2020-06-11 | Z Advanced Computing, Inc. | System and Method for Extremely Efficient Image and Pattern Recognition and Artificial Intelligence Platform |
| CN104598916A (en) * | 2014-09-11 | 2015-05-06 | 单勇 | Establishment method of train recognition system and train recognition method |
| CN106776675A (en) * | 2016-11-01 | 2017-05-31 | 上海斐讯数据通信技术有限公司 | A kind of plants identification method and system based on mobile terminal |
| CN107169148A (en) * | 2017-06-21 | 2017-09-15 | 北京百度网讯科技有限公司 | Image search method, device, equipment and storage medium |
| US20200051180A1 (en) * | 2017-08-28 | 2020-02-13 | The Climate Corporation | Crop disease recognition and yield estimation |
| CN109118779A (en) * | 2018-10-12 | 2019-01-01 | 东软集团股份有限公司 | Break in traffic rules and regulations information identifying method, equipment and readable storage medium storing program for executing |
| CN109740627A (en) * | 2018-11-27 | 2019-05-10 | 南京邮电大学 | A kind of insect image identification identifying system and its method based on parallel-convolution neural network |
| CN110046738A (en) * | 2019-01-28 | 2019-07-23 | 南京林业大学 | A kind of forest fire prediction technique based on artificial intelligence perceptron model |
| CN110197258A (en) * | 2019-05-29 | 2019-09-03 | 北京市商汤科技开发有限公司 | Neural network searching method, image processing method and device, equipment and medium |
| US20200027459A1 (en) * | 2019-09-09 | 2020-01-23 | Lg Electronics Inc. | Artificial intelligence apparatus and method for recognizing speech of user |
Non-Patent Citations (4)
| Title |
|---|
| MARIA TZELEPI 等: "Relevance Feedback in Deep Convolutional Neural Networks for Content Based Image Retrieval", 《SETN \'16: PROCEEDINGS OF THE 9TH HELLENIC CONFERENCE ON ARTIFICIAL INTELLIGENCE》, pages 1 * |
| XUN WANG 等: "Cross-Batch Memory for Embedding Learning", 《PROCEEDINGS OF THE IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR)》, pages 6388 - 6397 * |
| 刘凯雄: "基于交通视频的目标车辆检索方法及应用研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》, no. 05, pages 034 - 225 * |
| 岳清清: "深度学习在岩石薄片图像检索中的应用研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》, no. 08, pages 138 - 1155 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2024038941A1 (en) * | 2022-08-17 | 2024-02-22 | 쿠팡 주식회사 | Data providing method and device |
| CN115331062A (en) * | 2022-08-29 | 2022-11-11 | 北京达佳互联信息技术有限公司 | Image recognition method and device, electronic equipment and computer-readable storage medium |
| CN115331062B (en) * | 2022-08-29 | 2023-08-08 | 北京达佳互联信息技术有限公司 | Image recognition method, image recognition device, electronic device and computer-readable storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112100419B (en) | 2024-02-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113516012B (en) | Pedestrian re-identification method and system based on multi-level feature fusion | |
| US7362892B2 (en) | Self-optimizing classifier | |
| CN112990335B (en) | Intelligent recognition self-learning training method and system for power grid unmanned aerial vehicle inspection image defects | |
| CN110851645B (en) | Image retrieval method based on similarity maintenance under deep metric learning | |
| CN111832608B (en) | Iron spectrum image multi-abrasive particle identification method based on single-stage detection model yolov3 | |
| CN112633382B (en) | Method and system for classifying few sample images based on mutual neighbor | |
| CN109063649B (en) | Pedestrian re-identification method based on twin pedestrian alignment residual error network | |
| CN110213222A (en) | Network inbreak detection method based on machine learning | |
| CN112417381B (en) | Method and device for rapidly positioning infringement image applied to image copyright protection | |
| CN109871885A (en) | A kind of plants identification method based on deep learning and Plant Taxonomy | |
| CN112100419B (en) | Single weather image recognition method and system based on image retrieval | |
| CN110516098A (en) | Image labeling method based on convolutional neural networks and binary coding feature | |
| CN112950445B (en) | Compensation-based detection feature selection method in image steganalysis | |
| CN111368867B (en) | File classifying method and system and computer readable storage medium | |
| CN114723994B (en) | Hyperspectral image classification method based on dual classifier antagonism enhancement network | |
| CN111104398A (en) | Detection method and elimination method for approximate repeated record of intelligent ship | |
| CN110825642B (en) | Software code line-level defect detection method based on deep learning | |
| CN115292532A (en) | Remote sensing image domain adaptive retrieval method based on pseudo label consistency learning | |
| CN114842371A (en) | Unsupervised video anomaly detection method | |
| CN112818859B (en) | Deep hash-based multi-level retrieval pedestrian re-identification method | |
| CN113705215A (en) | Meta-learning-based large-scale multi-label text classification method | |
| CN110298399B (en) | Freeman chain code and moment feature fusion-based pumping well fault diagnosis method | |
| Liu et al. | Container-code recognition system based on computer vision and deep neural networks | |
| CN112084353A (en) | Bag-of-words model method for rapid landmark-convolution feature matching | |
| CN110750672B (en) | Image retrieval method based on deep measurement learning and structure distribution learning loss |
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 |