CN116597286A - Image recognition self-adaptive learning method and system based on deep learning - Google Patents
Image recognition self-adaptive learning method and system based on deep learning Download PDFInfo
- Publication number
- CN116597286A CN116597286A CN202310870633.7A CN202310870633A CN116597286A CN 116597286 A CN116597286 A CN 116597286A CN 202310870633 A CN202310870633 A CN 202310870633A CN 116597286 A CN116597286 A CN 116597286A
- Authority
- CN
- China
- Prior art keywords
- image
- image processing
- agcn
- recognition
- model
- 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 34
- 238000013135 deep learning Methods 0.000 title claims abstract description 27
- 238000012545 processing Methods 0.000 claims abstract description 256
- 230000009466 transformation Effects 0.000 claims abstract description 14
- 238000007781 pre-processing Methods 0.000 claims abstract description 11
- 238000012549 training Methods 0.000 claims description 48
- 230000003044 adaptive effect Effects 0.000 claims description 42
- 238000012360 testing method Methods 0.000 claims description 38
- 238000011176 pooling Methods 0.000 claims description 16
- 238000001514 detection method Methods 0.000 claims description 5
- 230000006870 function Effects 0.000 claims description 4
- 238000010276 construction Methods 0.000 claims description 3
- 230000008929 regeneration Effects 0.000 claims description 2
- 238000011069 regeneration method Methods 0.000 claims description 2
- 230000008901 benefit Effects 0.000 description 6
- 238000010586 diagram Methods 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- 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
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- 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
- G06N3/0985—Hyperparameter optimisation; Meta-learning; Learning-to-learn
-
- 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
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Multimedia (AREA)
- Medical Informatics (AREA)
- Databases & Information Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to the field of image recognition, in particular to an image recognition self-adaptive learning method and system based on deep learning. The self-adaptive recognition of the images can be realized aiming at the images with multiple scenes and multiple labels, and the recognition result is accurate. Image preprocessing is carried out on the image data set by utilizing an OpenCV image processing model through acquiring the image data set in the database, so as to obtain an image data set to be trained and an image trying data set to be tested; establishing an initial AGCN image processing recognition model based on an LMS self-adaptive recognition algorithm, and explicitly calculating the mean value difference of the image processing recognition domains of the first AGCN image processing recognition model and the second AGCN image processing recognition model based on multi-core transformation to obtain a target AGCN image processing recognition model; inputting the real-time image data set into a target AGCN image processing recognition model for image self-adaptive recognition; judging whether the result of the image self-adaptive recognition is correct, if not, inputting the real-time image data set into a first AGCN image processing recognition model for retraining.
Description
Technical Field
The invention relates to the field of image recognition, in particular to an image recognition self-adaptive learning method and system based on deep learning.
Background
With the development of multimedia technology and integrated circuits, advances in image technology are being promoted, and people use cameras to obtain rich image data. The pictures and the video images change the life of people, bring visual effects to people, and simultaneously, people dig image information service. The image contains rich data information, and computer vision is to mine out useful data for people in complex information. The computer vision research subjects are numerous, and the detection of moving objects in image sequences is a popular research field. Images in adaptive recognition or acquired in real time often have certain recognition errors when acquired and adaptively recognized. The self-adaptive identification of the image is difficult to realize when facing the images with multiple scenes and multiple labels; the existing image recognition technology has the problem of overfitting, and is easy to be interfered by image noise.
Disclosure of Invention
The invention aims to solve the problems and designs an image recognition self-adaptive learning method and system based on deep learning.
The technical scheme for achieving the purpose is that in the image recognition self-adaptive learning method based on deep learning, the image recognition self-adaptive learning method comprises the following steps:
acquiring an image data set in a database, and performing image preprocessing on the image data set by using an OpenCV image processing model to obtain an image data set to be trained and an image trying data set to be tested;
establishing an initial AGCN image processing identification model based on an LMS self-adaptive identification algorithm, wherein the initial AGCN image processing identification model comprises a self-adaptive layer, a convolution layer, a pooling layer, a full-connection layer and an output layer;
inputting the image data set to be trained into the initial AGCN image processing recognition model for training to obtain a first AGCN image processing recognition model;
acquiring the to-be-detected image trying data set, and testing the first AGCN image processing identification model by utilizing the to-be-detected image trying data set to obtain a second AGCN image processing identification model;
explicitly calculating the mean difference of the image processing recognition domains of the first AGCN image processing recognition model and the second AGCN image processing recognition model based on multi-core transformation to obtain a target image classifier; optimizing the target image classifier by using label-free image data to obtain a target AGCN image processing recognition model;
acquiring a real-time image data set, and inputting the real-time image data set into the target AGCN image processing recognition model for image self-adaptive recognition;
and judging whether the result of the image self-adaptive recognition is correct, if not, inputting the real-time image data set into the first AGCN image processing recognition model for retraining.
Further, in the above image recognition adaptive learning method, the acquiring the image dataset in the database, performing image preprocessing on the image dataset by using an OpenCV image processing model to obtain an image dataset to be trained and an attempt image dataset to be tested, includes:
acquiring an image data set in a database, wherein the image data set is image data of different categories, and at least comprises 2000 pictures;
scaling, clipping and normalizing the image data set based on an OpenCV image processing model; obtaining an initial image dataset;
and classifying the initial image data set to obtain an image data set to be trained and an image data set to be tested.
Further, in the image recognition adaptive learning method, the creating an initial AGCN image processing recognition model based on the LMS adaptive recognition algorithm, where the initial AGCN image processing recognition model includes an adaptive layer, a convolution layer, a pooling layer, a full connection layer, and an output layer, and includes:
constructing category image data by using an LMS self-adaptive recognition algorithm;
modeling the correlation of the category image data by utilizing a multi-scale graph convolution network to obtain a convolution image processing identification model;
assigning node weights to adjacent matrixes in the convolution image processing recognition model to obtain a continuous image processing recognition model;
inputting the classifier of the category image data into the continuous image processing recognition model to obtain an initial AGCN image processing recognition model; the initial AGCN image processing identification model comprises an adaptive layer, a convolution layer, a pooling layer, a full connection layer and an output layer.
Further, in the above image recognition adaptive learning method, the inputting the image dataset to be trained into the initial AGCN image processing recognition model for training, to obtain a first AGCN image processing recognition model includes:
the image data set to be trained at least comprises a group of training images for training the initial AGCN image processing identification model;
the training image at least comprises size image information, angle image information, environment image information, character image information and animal image information; the image data set to be trained at least comprises 1000 training images;
acquiring a first super parameter in the initial AGCN image processing identification model; the super parameters comprise loss function parameters, optimizer parameters and learning rate parameters;
inputting the image data set to be trained and the first super parameter into the initial AGCN image processing identification model for training;
and adjusting the first super parameter according to a training result to obtain a second super parameter, and adjusting the initial AGCN image processing recognition model based on the second super parameter to obtain a first AGCN image processing recognition model.
Further, in the above image recognition adaptive learning method, the acquiring the to-be-detected image dataset, and testing the first AGCN image processing recognition model with the to-be-detected image dataset to obtain a second AGCN image processing recognition model includes:
the to-be-tested image attempting data set at least comprises a group of image testing information for testing the first AGCN image processing identification model;
the image test information comprises category label image information, confidence image information and frame image information;
inputting the data set of the to-be-tested image to the first AGCN image processing identification model for testing;
the first AGCN image processing recognition model must detect each trained target object in the image test information during testing, and retrains if detection is missed or repeated;
and adjusting the first AGCN image processing recognition model according to the training result to obtain a second AGCN image processing recognition model.
Further, in the image recognition self-adaptive learning method, the average difference of the image processing recognition domains of the first AGCN image processing recognition model and the second AGCN image processing recognition model is explicitly calculated based on multi-core transformation, so as to obtain a target image classifier; optimizing the target image classifier by using the label-free image data to obtain a target AGCN image processing identification model, wherein the target AGCN image processing identification model comprises the following steps:
acquiring the first AGCN image processing identification model and the second AGCN image processing identification model;
using a pre-trained ResNet network as a basic network to carry out self-adaptive training on the first AGCN image processing recognition model to obtain a first image processing recognition domain;
using a pre-trained ResNet network as a basic network to carry out self-adaptive training on the second AGCN image processing recognition model to obtain a second image processing recognition domain;
based on a regeneration kernel Hilbert space, explicitly calculating the mean difference of the first image processing identification domain and the second image processing identification domain by utilizing multi-core transformation to obtain a target image classifier;
and optimizing the target image classifier by using the label-free image data to obtain a target AGCN image processing identification model.
Further, in the above image recognition adaptive learning method, the acquiring a real-time image dataset, inputting the real-time image dataset into the target AGCN image processing recognition model for image adaptive recognition includes:
acquiring a real-time image data set, wherein the real-time image data set at least comprises a group of image information to be identified;
the image information to be identified comprises environment image information, animal and plant image information, building image information, face image information and daily necessities image information which are acquired by an image acquisition device in real time;
and inputting the real-time image data set into the target AGCN image processing recognition model to perform one or more image self-adaption recognition.
Further, in the above image recognition adaptive learning method, the determining whether the result of the image adaptive recognition is correct, if not, inputting the real-time image dataset into the first AGCN image processing recognition model for retraining includes:
judging whether the result of the image self-adaptive recognition is correct or not;
if not, inputting the real-time image data set into the first AGCN image processing recognition model for retraining;
if yes, outputting the image self-adaption identification result to a server or a display terminal, and inputting the image self-adaption identification result into a database for storage.
The technical scheme of the invention for achieving the purpose is that in the image recognition self-adaptive learning system based on deep learning, the image recognition self-adaptive learning system comprises:
the data acquisition module is used for acquiring an image data set in the database, and carrying out image preprocessing on the image data set by using an OpenCV image processing model to obtain an image data set to be trained and an image trying data set to be tested;
the model generation module is used for establishing an initial AGCN image processing identification model based on an LMS self-adaptive identification algorithm, wherein the initial AGCN image processing identification model comprises a self-adaptive layer, a convolution layer, a pooling layer, a full-connection layer and an output layer;
the model training module is used for inputting the image data set to be trained into the initial AGCN image processing recognition model for training to obtain a first AGCN image processing recognition model;
the model test module is used for acquiring the to-be-tested image trying data set, and testing the first AGCN image processing identification model by utilizing the to-be-tested image trying data set to obtain a second AGCN image processing identification model;
the model adjustment module is used for explicitly calculating the mean value difference of the image processing recognition domains of the first AGCN image processing recognition model and the second AGCN image processing recognition model based on multi-core transformation to obtain a target image classifier; optimizing the target image classifier by using label-free image data to obtain a target AGCN image processing recognition model;
the self-adaptive recognition module is used for acquiring a real-time image data set, and inputting the real-time image data set into the target AGCN image processing recognition model for image self-adaptive recognition;
and the identification judging module is used for judging whether the result of the image self-adaptive identification is correct, if not, inputting the real-time image data set into the first AGCN image processing identification model for retraining.
Further, in the image recognition adaptive learning system based on deep learning, the model generating module includes the following modules:
the construction submodule is used for constructing category image data by using an LMS self-adaptive recognition algorithm;
the modeling module is used for modeling the correlation of the category image data by utilizing a multi-scale graph convolution network to obtain a convolution image processing identification model;
the assigning submodule is used for assigning node weights to adjacent matrixes in the convolution image processing recognition model to obtain a continuous image processing recognition model;
the obtaining submodule is used for inputting the classifier of the category image data into the continuous image processing recognition model to obtain an initial AGCN image processing recognition model; the initial AGCN image processing identification model comprises an adaptive layer, a convolution layer, a pooling layer, a full connection layer and an output layer.
The method has the advantages that the image data set in the database is obtained, and the image data set is subjected to image preprocessing by using an OpenCV image processing model to obtain an image data set to be trained and an image trying data set to be tested; establishing an initial AGCN image processing identification model based on an LMS self-adaptive identification algorithm, wherein the initial AGCN image processing identification model comprises a self-adaptive layer, a convolution layer, a pooling layer, a full-connection layer and an output layer; inputting the image data set to be trained into the initial AGCN image processing recognition model for training to obtain a first AGCN image processing recognition model; acquiring the to-be-detected image trying data set, and testing the first AGCN image processing identification model by utilizing the to-be-detected image trying data set to obtain a second AGCN image processing identification model; explicitly calculating the mean difference of the image processing recognition domains of the first AGCN image processing recognition model and the second AGCN image processing recognition model based on multi-core transformation to obtain a target image classifier; optimizing the target image classifier by using label-free image data to obtain a target AGCN image processing recognition model; acquiring a real-time image data set, and inputting the real-time image data set into the target AGCN image processing recognition model for image self-adaptive recognition; and judging whether the result of the image self-adaptive recognition is correct, if not, inputting the real-time image data set into the first AGCN image processing recognition model for retraining. 1. The self-adaptive recognition of the images can be realized aiming at the images with multiple scenes and multiple labels, and the recognition result is accurate; 2. the problem of excessive fitting of the conventional image recognition is solved, and the accuracy of the image recognition is improved; 3. the extracted image is more stable, and the detected image is recognized more completely; 4. more pixel information can be obtained, and the anti-interference capability is stronger.
Drawings
Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The drawings are only for purposes of illustrating the preferred embodiments and are not to be construed as limiting the invention.
FIG. 1 is a schematic diagram of a first embodiment of an image recognition adaptive learning method based on deep learning in an embodiment of the present invention;
FIG. 2 is a schematic diagram of a second embodiment of an image recognition adaptive learning method based on deep learning in an embodiment of the present invention;
FIG. 3 is a schematic diagram of a third embodiment of an image recognition adaptive learning method based on deep learning according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of a first embodiment of an image recognition adaptive learning system based on deep learning in an embodiment of the present invention;
fig. 5 is a schematic diagram of a second embodiment of an image recognition adaptive learning system based on deep learning in an embodiment of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present invention more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
As used herein, the singular forms "a", "an", "the" and "the" are intended to include the plural forms as well, unless expressly stated otherwise, as understood by those skilled in the art. It will be further understood that the terms "comprises" and/or "comprising," when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.
The invention is specifically described below with reference to the accompanying drawings, as shown in fig. 1, an image recognition self-adaptive learning method based on deep learning, the image recognition self-adaptive learning method includes the following steps:
step 101, acquiring an image dataset in a database, and performing image preprocessing on the image dataset by using an OpenCV image processing model to obtain an image dataset to be trained and an image dataset to be tested;
specifically, in this embodiment, an image dataset in a database is obtained, where the image dataset is image data of different types, and the image dataset at least includes 2000 pictures; scaling, cutting and normalizing the image data set based on the OpenCV image processing model; obtaining an initial image dataset; and classifying the initial image data set to obtain an image data set to be trained and an image data set to be tested.
102, establishing an initial AGCN image processing identification model based on an LMS self-adaptive identification algorithm, wherein the initial AGCN image processing identification model comprises a self-adaptive layer, a convolution layer, a pooling layer, a full-connection layer and an output layer;
specifically, in this embodiment, the LMS adaptive recognition algorithm is used to construct the category image data; modeling the relevance of the category image data by using a multi-scale graph convolution network to obtain a convolution image processing identification model; the method comprises the steps of providing node weights for adjacent matrixes in a convolution image processing recognition model to obtain a continuous image processing recognition model, inputting a classifier of category image data into the continuous image processing recognition model to obtain an initial AGCN image processing recognition model; the initial AGCN image processing identification model comprises an adaptive layer, a convolution layer, a pooling layer, a full connection layer and an output layer.
Step 103, inputting an image data set to be trained into an initial AGCN image processing recognition model for training to obtain a first AGCN image processing recognition model;
specifically, the image dataset to be trained in this embodiment includes at least one set of training images for training the initial AGCN image processing recognition model; the training image at least comprises size image information, angle image information, environment image information, character image information and animal image information; the image data set to be trained at least comprises 1000 training images; acquiring a first super parameter in an initial AGCN image processing identification model; the super parameters comprise loss function parameters, optimizer parameters and learning rate parameters; inputting the image data set to be trained and the first super parameter into an initial AGCN image processing identification model for training; and adjusting the first super-parameters according to the training result to obtain second super-parameters, and adjusting the initial AGCN image processing recognition model based on the second super-parameters to obtain a first AGCN image processing recognition model.
104, acquiring an image dataset to be tested, and testing the first AGCN image processing identification model by utilizing the image dataset to be tested to obtain a second AGCN image processing identification model;
specifically, in this embodiment, an image dataset to be tested is obtained; at least comprises a group of image test information for testing the first AGCN image processing identification model; the image test information comprises category label image information, confidence image information and frame image information; inputting the data set of the to-be-tested image to a first AGCN image processing identification model for testing; the first AGCN image processing recognition model must detect each trained target object in the image test information during testing, and retrains if detection is missed or repeated; and adjusting the first AGCN image processing recognition model according to the training result to obtain a second AGCN image processing recognition model.
Step 105, explicitly calculating the mean value difference of the image processing recognition domains of the first AGCN image processing recognition model and the second AGCN image processing recognition model based on multi-core transformation to obtain a target image classifier; optimizing a target image classifier by using the label-free image data to obtain a target AGCN image processing identification model;
specifically, in this embodiment, a first AGCN image processing recognition model and a second AGCN image processing recognition model are obtained; using a pre-trained ResNet network as a basic network to carry out self-adaptive training on a first AGCN image processing recognition model to obtain a first image processing recognition domain; using a pre-trained ResNet network as a basic network to carry out self-adaptive training on a second AGCN image processing recognition model to obtain a second image processing recognition domain; based on the regenerated kernel Hilbert space, explicitly calculating the mean difference of the first image processing identification domain and the second image processing identification domain by utilizing multi-core transformation to obtain a target image classifier; and optimizing the target image classifier by using the label-free image data to obtain a target AGCN image processing recognition model.
Step 106, acquiring a real-time image dataset, and inputting the real-time image dataset into a target AGCN image processing recognition model for image self-adaptive recognition;
specifically, in this embodiment, a real-time image dataset is obtained, where the real-time image dataset includes at least a set of image information to be identified; the image information to be identified comprises environment image information, animal and plant image information, building image information, face image information and daily necessities image information which are acquired by the image acquisition device in real time; and inputting the real-time image data set into a target AGCN image processing identification model to carry out one or more image self-adaption identifications.
And 107, judging whether the result of the image self-adaptive recognition is correct, if not, inputting the real-time image data set into a first AGCN image processing recognition model for retraining.
Specifically, in this embodiment, whether the result of the image adaptive recognition is correct is determined; if not, inputting the real-time image data set into a first AGCN image processing recognition model for retraining; if yes, outputting an image self-adaption identification result to a server or a display terminal, and inputting the image self-adaption identification result into a database for storage.
The method has the advantages that the image data set in the database is obtained, and the image data set is subjected to image preprocessing by using an OpenCV image processing model to obtain an image data set to be trained and an image trying data set to be tested; establishing an initial AGCN image processing identification model based on an LMS self-adaptive identification algorithm, wherein the initial AGCN image processing identification model comprises a self-adaptive layer, a convolution layer, a pooling layer, a full-connection layer and an output layer; inputting an image data set to be trained into an initial AGCN image processing recognition model for training to obtain a first AGCN image processing recognition model; acquiring an image data set to be tested, and testing the first AGCN image processing identification model by utilizing the image data set to be tested to obtain a second AGCN image processing identification model; explicitly calculating the mean difference of the image processing recognition domains of the first AGCN image processing recognition model and the second AGCN image processing recognition model based on multi-core transformation to obtain a target image classifier; optimizing a target image classifier by using the label-free image data to obtain a target AGCN image processing identification model; acquiring a real-time image data set, and inputting the real-time image data set into a target AGCN image processing recognition model to carry out image self-adaptive recognition; judging whether the result of the image self-adaptive recognition is correct, if not, inputting the real-time image data set into a first AGCN image processing recognition model for retraining. 1. The self-adaptive recognition of the images can be realized aiming at the images with multiple scenes and multiple labels, and the recognition result is accurate; 2. the problem of excessive fitting of the conventional image recognition is solved, and the accuracy of the image recognition is improved; 3. the extracted image is more stable, and the detected image is recognized more completely; 4. more pixel information can be obtained, and the anti-interference capability is stronger.
In this embodiment, referring to fig. 2, in a second embodiment of an image recognition adaptive learning method based on deep learning in the embodiment of the present invention, inputting an image dataset to be trained into an initial AGCN image processing recognition model for training, obtaining a first AGCN image processing recognition model includes the following steps:
step 201, the image data set to be trained at least comprises a group of training images for training an initial AGCN image processing identification model;
step 202, training images at least comprise size image information, angle image information, environment image information, character image information and animal image information; the image data set to be trained at least comprises 1000 training images;
step 203, acquiring a first super parameter in an initial AGCN image processing identification model; the super parameters comprise loss function parameters, optimizer parameters and learning rate parameters;
step 204, inputting the image data set to be trained and the first super parameter into an initial AGCN image processing recognition model for training;
step 205, adjusting the first super parameter according to the training result to obtain a second super parameter, and adjusting the initial AGCN image processing recognition model based on the second super parameter to obtain a first AGCN image processing recognition model.
In this embodiment, referring to fig. 3, in a third embodiment of an image recognition adaptive learning method based on deep learning in an embodiment of the present invention, an image dataset to be tested is obtained, and an image dataset to be tested is used to test a first AGCN image processing recognition model, so as to obtain a second AGCN image processing recognition model, which includes the following steps:
step 301, acquiring an image dataset to be tested; at least comprises a group of image test information for testing the first AGCN image processing identification model;
step 302, the image test information comprises category label image information, confidence coefficient image information and frame image information;
step 303, inputting a data set of an attempted image to be tested into a first AGCN image processing identification model for testing;
step 304, the first AGCN image processing recognition model must detect each trained target object in the image test information during the test, and retrains if the detection is missed or repeated;
and 305, adjusting the first AGCN image processing recognition model according to the training result to obtain a second AGCN image processing recognition model.
The description of the image recognition adaptive learning method based on deep learning provided by the embodiment of the present invention is given above, and the description of the image recognition adaptive learning system based on deep learning of the embodiment of the present invention is given below, referring to fig. 4, and one embodiment of the image recognition adaptive learning system in the embodiment of the present invention includes:
401. the data acquisition module is used for acquiring an image data set in the database, and performing image preprocessing on the image data set by using an OpenCV image processing model to obtain an image data set to be trained and an image trying data set to be tested;
402. the model generation module is used for establishing an initial AGCN image processing identification model based on an LMS self-adaptive identification algorithm, wherein the initial AGCN image processing identification model comprises a self-adaptive layer, a convolution layer, a pooling layer, a full-connection layer and an output layer;
403. the model training module is used for inputting the image data set to be trained into the initial AGCN image processing recognition model for training to obtain a first AGCN image processing recognition model;
404. the model test module is used for acquiring an image data set to be tested, and testing the first AGCN image processing identification model by utilizing the image data set to be tested to obtain a second AGCN image processing identification model;
405. the model adjustment module is used for explicitly calculating the mean value difference of the image processing recognition domains of the first AGCN image processing recognition model and the second AGCN image processing recognition model based on multi-core transformation to obtain a target image classifier; optimizing a target image classifier by using the label-free image data to obtain a target AGCN image processing identification model;
406. the self-adaptive recognition module is used for acquiring a real-time image data set, inputting the real-time image data set into the target AGCN image processing recognition model for image self-adaptive recognition;
407. and the recognition judging module is used for judging whether the result of the image self-adaptive recognition is correct, if not, inputting the real-time image data set into the first AGCN image processing recognition model for retraining.
In this embodiment, referring to fig. 5, in a second embodiment of an image recognition adaptive learning system based on deep learning according to an embodiment of the present invention, a model generating module includes:
501. the construction submodule is used for constructing category image data by using an LMS self-adaptive recognition algorithm;
502. the modeling module is used for modeling the relevance of the category image data by utilizing a multi-scale graph convolution network to obtain a convolution image processing identification model;
503. the assigning submodule is used for assigning node weights to adjacent matrixes in the convolution image processing recognition model to obtain a continuous image processing recognition model;
504. the obtaining sub-module is used for inputting the classifier of the category image data into the continuous image processing recognition model to obtain an initial AGCN image processing recognition model; the initial AGCN image processing identification model comprises an adaptive layer, a convolution layer, a pooling layer, a full connection layer and an output layer.
The foregoing has shown and described the basic principles, principal features and advantages of the invention. It will be understood by those skilled in the art that the present invention is not limited to the above-described embodiments, and that the above-described embodiments and descriptions are only preferred embodiments of the present invention, and are not intended to limit the invention, and that various changes and modifications may be made therein without departing from the spirit and scope of the invention as claimed. The scope of the invention is defined by the appended claims and equivalents thereof.
Claims (10)
1. The image recognition self-adaptive learning method based on deep learning is characterized by comprising the following steps of:
acquiring an image data set in a database, and performing image preprocessing on the image data set by using an OpenCV image processing model to obtain an image data set to be trained and an image trying data set to be tested;
establishing an initial AGCN image processing identification model based on an LMS self-adaptive identification algorithm, wherein the initial AGCN image processing identification model comprises a self-adaptive layer, a convolution layer, a pooling layer, a full-connection layer and an output layer;
inputting the image data set to be trained into the initial AGCN image processing recognition model for training to obtain a first AGCN image processing recognition model;
acquiring the to-be-detected image trying data set, and testing the first AGCN image processing identification model by utilizing the to-be-detected image trying data set to obtain a second AGCN image processing identification model;
explicitly calculating the mean difference of the image processing recognition domains of the first AGCN image processing recognition model and the second AGCN image processing recognition model based on multi-core transformation to obtain a target image classifier; optimizing the target image classifier by using label-free image data to obtain a target AGCN image processing recognition model;
acquiring a real-time image data set, and inputting the real-time image data set into the target AGCN image processing recognition model for image self-adaptive recognition;
and judging whether the result of the image self-adaptive recognition is correct, if not, inputting the real-time image data set into the first AGCN image processing recognition model for retraining.
2. The image recognition adaptive learning method based on deep learning as claimed in claim 1, wherein the obtaining the image dataset in the database, performing image preprocessing on the image dataset by using an OpenCV image processing model, to obtain an image dataset to be trained and an attempt image dataset to be tested, includes:
acquiring an image data set in a database, wherein the image data set is image data of different categories, and at least comprises 2000 pictures;
scaling, clipping and normalizing the image data set based on an OpenCV image processing model; obtaining an initial image dataset;
and classifying the initial image data set to obtain an image data set to be trained and an image data set to be tested.
3. The image recognition adaptive learning method based on deep learning as claimed in claim 1, wherein the image recognition adaptive learning method based on the LMS builds an initial AGCN image processing recognition model, the initial AGCN image processing recognition model including an adaptive layer, a convolution layer, a pooling layer, a full connection layer and an output layer, and includes:
constructing category image data by using an LMS self-adaptive recognition algorithm;
modeling the correlation of the category image data by utilizing a multi-scale graph convolution network to obtain a convolution image processing identification model;
assigning node weights to adjacent matrixes in the convolution image processing recognition model to obtain a continuous image processing recognition model;
inputting the classifier of the category image data into the continuous image processing recognition model to obtain an initial AGCN image processing recognition model; the initial AGCN image processing identification model comprises an adaptive layer, a convolution layer, a pooling layer, a full connection layer and an output layer.
4. The image recognition adaptive learning method based on deep learning of claim 1, wherein inputting the image dataset to be trained into the initial AGCN image processing recognition model for training to obtain a first AGCN image processing recognition model comprises:
the image data set to be trained at least comprises a group of training images for training the initial AGCN image processing identification model;
the training image at least comprises size image information, angle image information, environment image information, character image information and animal image information; the image data set to be trained at least comprises 1000 training images;
acquiring a first super parameter in the initial AGCN image processing identification model; the super parameters comprise loss function parameters, optimizer parameters and learning rate parameters;
inputting the image data set to be trained and the first super parameter into the initial AGCN image processing identification model for training;
and adjusting the first super parameter according to a training result to obtain a second super parameter, and adjusting the initial AGCN image processing recognition model based on the second super parameter to obtain a first AGCN image processing recognition model.
5. The method for adaptive learning of image recognition based on deep learning of claim 1, wherein the obtaining the dataset of the image to be tested, and the testing the first AGCN image processing recognition model with the dataset of the image to be tested, to obtain the second AGCN image processing recognition model, comprises:
the to-be-tested image attempting data set at least comprises a group of image testing information for testing the first AGCN image processing identification model;
the image test information comprises category label image information, confidence image information and frame image information;
inputting the data set of the to-be-tested image to the first AGCN image processing identification model for testing;
the first AGCN image processing recognition model must detect each trained target object in the image test information during testing, and retrains if detection is missed or repeated;
and adjusting the first AGCN image processing recognition model according to the training result to obtain a second AGCN image processing recognition model.
6. The image recognition adaptive learning method based on deep learning according to claim 1, wherein the multi-core transformation explicitly calculates a mean difference of image processing recognition domains of the first AGCN image processing recognition model and the second AGCN image processing recognition model to obtain a target image classifier; optimizing the target image classifier by using the label-free image data to obtain a target AGCN image processing identification model, wherein the target AGCN image processing identification model comprises the following steps:
acquiring the first AGCN image processing identification model and the second AGCN image processing identification model;
using a pre-trained ResNet network as a basic network to carry out self-adaptive training on the first AGCN image processing recognition model to obtain a first image processing recognition domain;
using a pre-trained ResNet network as a basic network to carry out self-adaptive training on the second AGCN image processing recognition model to obtain a second image processing recognition domain;
based on a regeneration kernel Hilbert space, explicitly calculating the mean difference of the first image processing identification domain and the second image processing identification domain by utilizing multi-core transformation to obtain a target image classifier;
and optimizing the target image classifier by using the label-free image data to obtain a target AGCN image processing identification model.
7. The image recognition adaptive learning method based on deep learning according to claim 1, wherein the acquiring a real-time image dataset, inputting the real-time image dataset into the target AGCN image processing recognition model for image adaptive recognition, comprises:
acquiring a real-time image data set, wherein the real-time image data set at least comprises a group of image information to be identified;
the image information to be identified comprises environment image information, animal and plant image information, building image information, face image information and daily necessities image information which are acquired by an image acquisition device in real time;
and inputting the real-time image data set into the target AGCN image processing recognition model to perform one or more image self-adaption recognition.
8. The method according to claim 1, wherein the determining whether the result of the image adaptive recognition is correct, if not, inputting the real-time image dataset into the first AGCN image processing recognition model for retraining, includes:
judging whether the result of the image self-adaptive recognition is correct or not;
if not, inputting the real-time image data set into the first AGCN image processing recognition model for retraining;
if yes, outputting the image self-adaption identification result to a server or a display terminal, and inputting the image self-adaption identification result into a database for storage.
9. An image recognition self-adaptive learning system based on deep learning is characterized by comprising the following modules
The data acquisition module is used for acquiring an image data set in the database, and carrying out image preprocessing on the image data set by using an OpenCV image processing model to obtain an image data set to be trained and an image trying data set to be tested;
the model generation module is used for establishing an initial AGCN image processing identification model based on an LMS self-adaptive identification algorithm, wherein the initial AGCN image processing identification model comprises a self-adaptive layer, a convolution layer, a pooling layer, a full-connection layer and an output layer;
the model training module is used for inputting the image data set to be trained into the initial AGCN image processing recognition model for training to obtain a first AGCN image processing recognition model;
the model test module is used for acquiring the to-be-tested image trying data set, and testing the first AGCN image processing identification model by utilizing the to-be-tested image trying data set to obtain a second AGCN image processing identification model;
the model adjustment module is used for explicitly calculating the mean value difference of the image processing recognition domains of the first AGCN image processing recognition model and the second AGCN image processing recognition model based on multi-core transformation to obtain a target image classifier; optimizing the target image classifier by using label-free image data to obtain a target AGCN image processing recognition model;
the self-adaptive recognition module is used for acquiring a real-time image data set, and inputting the real-time image data set into the target AGCN image processing recognition model for image self-adaptive recognition;
and the identification judging module is used for judging whether the result of the image self-adaptive identification is correct, if not, inputting the real-time image data set into the first AGCN image processing identification model for retraining.
10. The image recognition adaptive learning system of claim 9 wherein the model generation module comprises:
the construction submodule is used for constructing category image data by using an LMS self-adaptive recognition algorithm;
the modeling module is used for modeling the correlation of the category image data by utilizing a multi-scale graph convolution network to obtain a convolution image processing identification model;
the assigning submodule is used for assigning node weights to adjacent matrixes in the convolution image processing recognition model to obtain a continuous image processing recognition model;
the obtaining submodule is used for inputting the classifier of the category image data into the continuous image processing recognition model to obtain an initial AGCN image processing recognition model; the initial AGCN image processing identification model comprises an adaptive layer, a convolution layer, a pooling layer, a full connection layer and an output layer.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310870633.7A CN116597286B (en) | 2023-07-17 | 2023-07-17 | Image recognition self-adaptive learning method and system based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310870633.7A CN116597286B (en) | 2023-07-17 | 2023-07-17 | Image recognition self-adaptive learning method and system based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116597286A true CN116597286A (en) | 2023-08-15 |
| CN116597286B CN116597286B (en) | 2023-09-15 |
Family
ID=87604823
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310870633.7A Active CN116597286B (en) | 2023-07-17 | 2023-07-17 | Image recognition self-adaptive learning method and system based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116597286B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109657552A (en) * | 2018-11-16 | 2019-04-19 | 北京邮电大学 | The vehicle type recognition device being cold-started across scene and method are realized based on transfer learning |
| CN110781921A (en) * | 2019-09-25 | 2020-02-11 | 浙江农林大学 | Depth residual error network and transfer learning-based muscarinic image identification method and device |
| CN113408662A (en) * | 2021-07-19 | 2021-09-17 | 北京百度网讯科技有限公司 | Image recognition method and device, and training method and device of image recognition model |
| CN113837217A (en) * | 2021-07-02 | 2021-12-24 | 中国空间技术研究院 | Passive non-visual field image identification method and device based on deep learning |
| CN114821178A (en) * | 2022-05-05 | 2022-07-29 | 中国科学院水生生物研究所 | Processing method of modular image recognition and classification system based on deep learning |
| CN116434252A (en) * | 2023-04-10 | 2023-07-14 | 京东方科技集团股份有限公司 | Training of image recognition model and image recognition method, device, medium and equipment |
-
2023
- 2023-07-17 CN CN202310870633.7A patent/CN116597286B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109657552A (en) * | 2018-11-16 | 2019-04-19 | 北京邮电大学 | The vehicle type recognition device being cold-started across scene and method are realized based on transfer learning |
| CN110781921A (en) * | 2019-09-25 | 2020-02-11 | 浙江农林大学 | Depth residual error network and transfer learning-based muscarinic image identification method and device |
| CN113837217A (en) * | 2021-07-02 | 2021-12-24 | 中国空间技术研究院 | Passive non-visual field image identification method and device based on deep learning |
| CN113408662A (en) * | 2021-07-19 | 2021-09-17 | 北京百度网讯科技有限公司 | Image recognition method and device, and training method and device of image recognition model |
| CN114821178A (en) * | 2022-05-05 | 2022-07-29 | 中国科学院水生生物研究所 | Processing method of modular image recognition and classification system based on deep learning |
| CN116434252A (en) * | 2023-04-10 | 2023-07-14 | 京东方科技集团股份有限公司 | Training of image recognition model and image recognition method, device, medium and equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116597286B (en) | 2023-09-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109919031B (en) | Human behavior recognition method based on deep neural network | |
| CN110348376B (en) | Pedestrian real-time detection method based on neural network | |
| CN110796057A (en) | Pedestrian re-identification method and device and computer equipment | |
| CN111738054B (en) | Behavior anomaly detection method based on space-time self-encoder network and space-time CNN | |
| CN108564673A (en) | A kind of check class attendance method and system based on Global Face identification | |
| CN112347964B (en) | Behavior detection method and device based on graph network | |
| CN110458235B (en) | Motion posture similarity comparison method in video | |
| CN112861575A (en) | Pedestrian structuring method, device, equipment and storage medium | |
| CN111814611A (en) | Multi-scale face age estimation method and system embedded with high-order information | |
| CN112507778B (en) | Loop detection method of improved bag-of-words model based on line characteristics | |
| CN116977937A (en) | Pedestrian re-identification method and system | |
| CN113780145A (en) | Sperm morphology detection method, sperm morphology detection device, computer equipment and storage medium | |
| CN116030396A (en) | Accurate segmentation method for video structured extraction | |
| CN113870254A (en) | Target object detection method and device, electronic equipment and storage medium | |
| CN113313179A (en) | Noise image classification method based on l2p norm robust least square method | |
| CN116597286B (en) | Image recognition self-adaptive learning method and system based on deep learning | |
| CN112488072A (en) | Method, system and equipment for acquiring face sample set | |
| CN114708645A (en) | Object identification device and object identification method | |
| CN113076860A (en) | Bird detection system under field scene | |
| CN115862119B (en) | Attention mechanism-based face age estimation method and device | |
| CN115019367A (en) | Genetic disease face recognition device and method | |
| CN114155411A (en) | Intelligent detection and identification method for small and weak targets | |
| CN111523598A (en) | Image recognition method based on neural network and visual analysis | |
| CN113850301B (en) | Training data acquisition method and device, model training method and device | |
| CN115761561A (en) | Framework behavior recognition-based violation video detection method |
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 |