CN111400572A - Content safety monitoring system and method for realizing image feature recognition based on convolutional neural network - Google Patents
Content safety monitoring system and method for realizing image feature recognition based on convolutional neural network Download PDFInfo
- Publication number
- CN111400572A CN111400572A CN202010133407.7A CN202010133407A CN111400572A CN 111400572 A CN111400572 A CN 111400572A CN 202010133407 A CN202010133407 A CN 202010133407A CN 111400572 A CN111400572 A CN 111400572A
- Authority
- CN
- China
- Prior art keywords
- image
- neural network
- features
- convolutional neural
- training
- 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.)
- Pending
Links
- 238000012544 monitoring process Methods 0.000 title claims abstract description 51
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 38
- 238000000034 method Methods 0.000 title claims abstract description 37
- 230000006870 function Effects 0.000 claims abstract description 22
- 238000013528 artificial neural network Methods 0.000 claims abstract description 18
- 238000005070 sampling Methods 0.000 claims abstract description 14
- 230000004913 activation Effects 0.000 claims abstract description 11
- 238000005516 engineering process Methods 0.000 claims abstract description 11
- 238000007781 pre-processing Methods 0.000 claims abstract description 11
- 239000011159 matrix material Substances 0.000 claims description 13
- 238000010276 construction Methods 0.000 claims description 8
- 238000004458 analytical method Methods 0.000 claims description 7
- 238000000605 extraction Methods 0.000 claims description 6
- 239000013598 vector Substances 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000007477 logistic regression Methods 0.000 claims description 3
- 230000006740 morphological transformation Effects 0.000 claims description 3
- 238000011176 pooling Methods 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 230000001902 propagating effect Effects 0.000 claims description 2
- 230000009193 crawling Effects 0.000 abstract 1
- 238000013135 deep learning Methods 0.000 description 3
- 230000007547 defect Effects 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 238000005192 partition Methods 0.000 description 2
- 239000003086 colorant Substances 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 239000003550 marker Substances 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 238000006467 substitution reaction 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
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/951—Indexing; Web crawling techniques
-
- 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
-
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Databases & Information Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Multimedia (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a content security monitoring system and a content security monitoring method for realizing image feature recognition based on a convolutional neural network, wherein the method comprises the steps of crawling images of a target UR L by using a crawler technology, preprocessing a crawled image to be detected and a training sample image, extracting features of the preprocessed image through a filter bank, performing machine training on sample features by using a neural network learning rule, constructing a risk recognition model according to a learning and training result, storing the extracted feature to be detected, performing convolution, sampling and classification through the risk recognition model, outputting the recognition result, monitoring an activation function threshold value, performing risk marking when the threshold value reaches a specified value, automatically early warning the system and feeding back the monitoring result to a user.
Description
Technical Field
The invention belongs to the technical field of network security, and particularly relates to a content security monitoring system and a content security monitoring method for realizing image feature identification based on a convolutional neural network.
Background
In the big data era of the internet, network information is increasingly abundant, illegal, pornographic, violent and other bad contents become problems to be solved urgently on the internet, and the legality, health and safety of information contents become hot problems in the network field.
At present, the problems of unavailable identification, incomplete searching, missed searching and the like exist in the image information security construction process. The traditional image content safety monitoring system mostly utilizes the bottom layer characteristics of the image to identify, for example, color and texture characteristics are global characteristics and cannot well reflect local characteristics; also for example, shape features are often affected by image segmentation effects; for another example, the spatial relationship features cannot accurately express scene information.
In order to overcome the defects of the traditional safety monitoring system and realize accurate description of local features, the existing developed mature technology is Scale-Invariant Feature Transform (SIFT) algorithm, and can detect key points in an image and accurately describe the local features. However, the technology has the following defects: the real-time performance is not high, and the characteristic points cannot be accurately extracted from the blurred, smooth-edge and round images. Therefore, the specified target cannot be completely and accurately identified in massive and complex network resources, so that the image information security construction work cannot be efficiently realized, which is a problem to be solved in the field.
Disclosure of Invention
The invention provides a content safety monitoring system and a content safety monitoring method for realizing image feature recognition based on a Convolutional Neural Network (CNN), wherein the system realizes artificial intelligent deep learning by using the Convolutional Neural Network (CNN), constructs a plurality of groups of sample feature models with different dimensions, recognizes specified image features, image original colors, character information and the like from massive images through efficient machine training, screens out target features for risk marking, and achieves the aim of safety monitoring of image contents. The technology combines the convolutional neural network technology with the image information safety monitoring technology, and realizes rapid, efficient and accurate website content safety monitoring through multi-dimensional and multi-level deep learning training.
The invention provides a content security monitoring method for realizing image feature identification based on a convolutional neural network, which has the following realization principle:
the method comprises the steps of utilizing a convolutional neural network to carry out primary extraction on features of a target source, multiplying the primary extracted features with a corresponding filter to obtain main features, carrying out partition sampling on the target source according to the main features, judging the target features through comparison with a sample model, assembling and classifying the partition target features through a weight matrix, outputting all the features of the target source, utilizing an activation function to carry out learning training on output and judging a threshold value, and carrying out risk marking on the target source reaching a specified threshold value.
The invention provides a content security monitoring method for realizing image feature identification based on a convolutional neural network, which comprises the following specific implementation steps of:
the system crawls images of a target UR L by using a crawler technology, preprocesses the crawled images to be detected and training sample images, extracts the characteristics of the preprocessed images by using a filter bank, performs machine training on the sample characteristics by using a neural network learning rule, constructs a risk identification model according to the learning training result, stores the extracted characteristics to be detected, performs convolution, sampling and classification by using the risk identification model, outputs the identification result, monitors an activation function threshold value, performs risk marking when the threshold value reaches a specified value, automatically warns and feeds the monitoring result back to a user.
Further, the neural network learning rule includes a gradient descent rule, a back propagation learning rule, a Delta (Wdrow-Holf) learning rule, wherein:
the gradient descent rule is a mathematical description of the method of reducing the difference between the actual output error and the desired output error;
the back propagation learning rule is divided into two stages, the first stage is forward propagation, input data is input into a network, the network calculates the output of each unit from front to back, the output of each unit is compared with the expected output, and errors are calculated; the second stage is back propagation, the error is recalculated from back to front and the weight is modified, and new data can be input after the two stages are finished;
delta learning rules reduce the error of the actual output of the system from the desired output by varying the connection weights between the units.
Further, constructing the risk identification model refers to: an image sample is taken from a characteristic sample library and input into a convolutional neural network, convolution and sampling are carried out after parameters are initialized, corresponding actual output is calculated through forward feedback transformation, the difference between the actual output and expected output is calculated, enhancement and logistic regression are carried out according to the back propagation of a learning method of minimizing errors, a weight matrix is updated and adjusted according to error feedback and weights, and finally expected results are output to construct a risk identification model.
Further, the convolution, sampling and classification by the risk identification model means: deconvolving the input image and the features by using a group of trainable filter banks to obtain a primary extracted feature map, converting a group of pixels in a feature image domain into a pixel unit through pooling to obtain a main feature mapping map, multiplying the vectorized mapping map by an optimized weight matrix for assembly, classifying through an activation function, and outputting an optimal result.
In addition, the invention also provides a content safety monitoring system for realizing image feature identification based on the convolutional neural network, which comprises the following modules:
the crawler module crawls image resources from the target UR L;
an image preprocessing module: preprocessing and segmenting the crawled image resources, and removing noise through image morphological transformation;
a model construction and training module; extracting and storing features from the preprocessed image resources, performing classification management on the features, establishing samples, establishing a neural network learning rule, performing machine training by using the learning rule, and establishing a risk identification model according to a training result;
the safety monitoring module: monitoring and analyzing the classified characteristics, and judging to obtain a risk level;
the early warning module: and feeding back the monitoring result with serious and significant risk level to the user.
Further, the model building and training module comprises a feature extraction submodule, a model trainer, a feature sample library and a rule library.
Further, the safety monitoring module comprises a safety information analysis submodule and a safety level judgment submodule.
Compared with the prior art, the content safety monitoring system and the method thereof for realizing image feature identification based on the convolutional neural network have the following advantages:
the method combines the convolutional neural network technology with the image content safety monitoring technology, adopts a multilayer neural network structure to safely identify the massive images, and is not limited to a single dimension; by using a filter bank, a sub-sampling and weight sharing deep learning algorithm with robustness, the specific local features of the multi-element mixed scene image can be safely identified, the fault tolerance is very high, and the problems of incomplete image identification, missed search and unavailable identification can be effectively solved; the system has the functions of automatic organization and autonomous learning, effectively weakens the dependence on the external environment, and has great superiority; the configuration can be completed quickly by having an application entry with a unified standard; the method has the advantages of massive sample library training, higher precision rate and higher safety, and effectively realizes the image information safety construction work.
Drawings
Fig. 1 is a flowchart of a content security monitoring method for implementing image feature recognition based on a convolutional neural network according to an embodiment.
Fig. 2 is a flowchart of a sample training and model building method based on a convolutional neural network according to an embodiment.
Fig. 3 is a flowchart of an image feature identification method based on a convolutional neural network according to an embodiment.
Fig. 4 is a schematic diagram of a content security monitoring system for implementing image feature recognition based on a convolutional neural network according to the second embodiment.
Detailed Description
The above description is only an overview of the technical solutions of the present invention, and the present invention can be implemented by looking up the content of the description in order to make the technical means of the present invention more clearly understood, and the following detailed description of the present invention is made in order to make the above and other objects, features, and advantages of the present invention more clearly understandable.
Example one
Referring to fig. 1, the content security monitoring method for implementing image feature recognition based on the convolutional neural network provided in this embodiment is only used for explaining the present invention, and is not used for limiting the scope of the present invention. The method specifically comprises the following steps:
s1, the system crawls images of the target UR L by using a crawler technology;
s2, preprocessing the crawled image to be detected and the training sample image;
s3, extracting the image characteristics after preprocessing through a filter bank;
s4, performing machine training on the extracted sample image features by using a neural network learning rule, and constructing a risk identification model according to a learning training result;
s5, storing the extracted image features to be detected, and performing convolution, sampling and classification on the image features to be detected through a risk identification model;
s6, outputting a recognition result and monitoring an activation function threshold;
s7, when the threshold value does not reach the designated value, continuing monitoring; when the threshold value reaches a specified value, carrying out risk marking;
and S8, the system automatically sends out an early warning prompt and feeds back the monitoring result to the front-end user.
The "activation function" in S6 may be a Sigmoid function, a Tanh function, a Re L U function, a L eaky Re L U function, or a Maxout function.
Wherein, the "risk marker" in S7 means: and marking the risk according to the threshold value of the activation function, wherein when the threshold value is 1, the 'no risk' is corresponded, and when the threshold value is 0, the 'at risk' is corresponded, and marking the risk.
Referring to fig. 2, the sample training and model building method based on the convolutional neural network provided in this embodiment is only for explaining the present invention, and is not intended to limit the scope of the present invention.
Wherein, S4 further includes the following steps:
s4.1, taking an image sample from the characteristic sample library and inputting the image sample into a convolutional neural network;
s4.2, calculating corresponding actual output;
s4.3, calculating the difference between the actual output and the expected output;
s4.4, reversely propagating according to a learning method of minimizing errors, and adjusting a weight matrix;
and S4.5, constructing a risk identification model according to the output result.
Wherein, the step of inputting the image sample into the convolutional neural network in S4.1 means: and initializing sample parameters, performing convolution and sampling, and performing forward feedback transformation and calculation.
Wherein, the said "back propagation" in S4.4 means: and (4) performing enhancement and logistic regression on the difference value between the actual output and the expected output, adjusting the weight matrix according to the feedback error and the updated weight, and finally outputting a result which is in line with the expected output.
Referring to fig. 3, the image feature identification method based on the convolutional neural network provided in the present embodiment is only used for explaining the present invention, and is not used to limit the scope of the present invention.
Wherein, S5 further includes the following steps:
s5.1, the input image passes through a filter bank WxObtaining a characteristic group xn, adding an offset bn to carry out convolution to obtain MxLayer, n represents the number of feature groups;
s5.2, to MxDown-sampling the layer characteristics to obtain Nx+1A layer;
s5.3, adding Nx+1The features of the layers are rasterized to form vectors, the vectors are input into a weight matrix of the fully-connected neural network for assembly and classification, an output feature group an is obtained, and the calculation formula is as follows:
where x1, x2, and x3 are the inputs of step S5.1, a1, a2, and a3 are the outputs of step S5.3, and b1, b2, and b3 are the offsets.
Wherein, the convolution process in S5.1 refers to: using a trainable set of filter banks WxDeconvoluting the input image and features, adding the bias bn to obtain the convolutional layer Mx。
Wherein, the process of down-sampling in S5.2 refers to: pooling a group of pixels in the feature image domain into a pixel unit to generate a feature map Nx+1。
Wherein, the step of inputting the weight matrix of the fully-connected neural network for assembling and classifying in the step S5.3 means that: map features to Nx+1And after vectorization, multiplying the vector by an optimized weight matrix for assembly, classifying through an activation function, and outputting an optimal result an.
Example two
Referring to fig. 4, the content security monitoring system for implementing image feature recognition based on the convolutional neural network is provided in this embodiment, which is only used for explaining the present invention, and is not used to limit the scope of the present invention. The system specifically comprises the following modules:
the crawler module crawls image resources from the target UR L;
an image preprocessing module: preprocessing, segmenting and carrying out image morphological transformation on the crawled image resources and training sample images to remove noise;
a model construction and training module: extracting and storing features from the preprocessed image resources, performing classification management on the features, establishing samples, establishing a neural network learning rule, performing machine training on the image features by using the learning rule, and establishing a risk identification model according to a learning training result;
the safety monitoring module: comparing and analyzing the classified characteristic result with a risk identification model, judging the risk level, and performing risk marking on a critical threshold;
the early warning module: and drawing a safety monitoring report according to the grade judgment result, and informing a front-end user through instant tools such as short messages, WeChats or mails.
Wherein the model building and training module further comprises the following:
a feature extraction submodule: performing primary extraction of features on the preprocessed image, sending the preliminarily extracted image features to be detected to a feature sample library for classification, and sending the sample image features to a model trainer for feature training and model construction;
a model trainer: calling a neural network learning rule from a rule base, performing feature training by using the learning rule, and constructing a risk identification model according to a training result so as to ensure the accuracy of identification of the features to be detected and the sample features in the feature sample base;
a characteristic sample library: storing and managing the extracted image features in a classified manner;
a rule base: establishing and managing neural network learning rules, including gradient descent rules, back propagation learning rules and Delta (Wdrow-Holf) learning rules.
Wherein, the safety monitoring module further comprises the following contents:
the safety information analysis submodule comprises: comparing and analyzing the classified image features in the feature sample library with the risk identification model, and sending an analysis result to the security level judgment submodule;
a safety level judgment submodule: and judging the safety grade of the analysis result, marking a risk when the grade reaches a critical threshold value, and informing the early warning module of the grade judgment result.
Wherein a critical threshold is marked as "at risk" when it is a specified value, and otherwise marked as "no risk".
The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present invention should be covered within the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (11)
1. A content security monitoring method for realizing image feature recognition based on a convolutional neural network is characterized by comprising the following steps: the method comprises the following steps:
s1, the system crawls images of the target UR L by using a crawler technology;
s2, preprocessing the crawled image to be detected and the training sample image;
s3, extracting the image characteristics after preprocessing through a filter bank;
s4, performing machine training on the extracted sample image features by using a neural network learning rule, and constructing a risk identification model according to a learning training result;
s5, storing the extracted image features to be detected, and performing convolution, sampling and classification on the image features to be detected through a risk identification model;
s6, outputting a recognition result and monitoring an activation function threshold;
s7, when the threshold value does not reach the designated value, continuing monitoring; when the threshold value reaches a specified value, carrying out risk marking;
and S8, the system automatically sends out an early warning prompt and feeds back the monitoring result to the front-end user.
2. The content security monitoring method for realizing image feature recognition based on the convolutional neural network as claimed in claim 1, characterized in that: the "neural network learning rule" in S4 includes: gradient descent rule, back propagation learning rule, Delta (Wdrow-Holf) learning rule, wherein:
the gradient descent rule is a mathematical description of the method of reducing the difference between the actual output error and the desired output error;
the back propagation learning rule is divided into two stages, the first stage is forward propagation, input data is input into a network, the network calculates the output of each unit from front to back, the output of each unit is compared with the expected output, and errors are calculated; the second stage is back propagation, the error is recalculated from back to front and the weight is modified, and new data can be input after the two stages are finished;
delta learning rules reduce the error of the actual output of the system from the desired output by varying the connection weights between the units.
3. The content security monitoring method for realizing image feature recognition based on the convolutional neural network as claimed in claim 1, characterized in that: the S4 further includes the steps of:
s4.1, taking an image sample from the characteristic sample library and inputting the image sample into a convolutional neural network;
s4.2, calculating corresponding actual output;
s4.3, calculating the difference between the actual output and the expected output;
s4.4, reversely propagating according to a learning method of minimizing errors, and adjusting a weight matrix;
and S4.5, constructing a risk identification model according to the output result.
4. The content security monitoring method based on the convolutional neural network for realizing the image feature recognition as claimed in claim 3, characterized in that: the step of inputting the image sample into the convolutional neural network in the step S4.1 means that: initializing sample parameters, performing convolution and sampling, and performing forward feedback transformation and calculation; the "back propagation" in S4.4 means that: and (4) performing enhancement and logistic regression on the difference value between the actual output and the expected output, adjusting the weight matrix according to the feedback error and the updated weight, and finally outputting a result which is in line with the expected output.
5. The content security monitoring method for realizing image feature recognition based on the convolutional neural network as claimed in claim 1, characterized in that: the S5 further includes the steps of:
s5.1, the input image passes through a filter bank WxObtaining a characteristic group xn, adding an offset bn to carry out convolution to obtain MxLayer, n represents the number of feature groups;
s5.2, to MxDown-sampling the layer characteristics to obtain Nx+1A layer;
s5.3, adding Nx+1The features of the layers are rasterized to form vectors, the vectors are input into a weight matrix of the fully-connected neural network for assembly and classification, an output feature group an is obtained, and the calculation formula is as follows:
where x1, x2, and x3 are the inputs of step S5.1, a1, a2, and a3 are the outputs of step S5.3, and b1, b2, and b3 are the offsets.
6. The content security monitoring method for realizing image feature recognition based on the convolutional neural network as claimed in claim 5, characterized in that: the convolution process in S5.1 refers to the use of a trainable bank of filters WxDeconvoluting the input image and features, adding the bias bn to obtain the convolutional layer Mx(ii) a The "down-sampling" process in S5.2 refers to converting a group of pixels in the feature image domain into a pixel unit through pooling, and generating a feature map Nx+1(ii) a The step of inputting the weight matrix of the fully-connected neural network for assembling and classifying in the step S5.3 is as follows: map features to Nx+1And after vectorization, multiplying the vector by an optimized weight matrix for assembly, classifying through an activation function, and outputting an optimal result an.
7. The method as claimed in claim 1, wherein the "activation function" in S6 is selected from the group consisting of Sigmoid function, Tanh function, Re L U function, L eakyRe L U function, and Maxout function.
8. A content safety monitoring system for realizing image feature recognition based on a convolutional neural network is characterized in that:
the system comprises the following modules:
the crawler module crawls image resources from the target UR L;
an image preprocessing module: preprocessing, segmenting and carrying out image morphological transformation on the crawled image resources and training sample images to remove noise;
a model construction and training module: extracting and storing features from the preprocessed image resources, performing classification management on the features, establishing samples, establishing a neural network learning rule, performing machine training on the image features by using the learning rule, and establishing a risk identification model according to a learning training result;
the safety monitoring module: comparing and analyzing the classified characteristic result with a risk identification model, judging the risk level, and performing risk marking on a critical threshold;
the early warning module: and drawing a safety monitoring report according to the grade judgment result, and informing a front-end user through a short message, a WeChat or an email.
9. The system for monitoring the content security based on the convolutional neural network for realizing the image feature recognition as claimed in claim 8, wherein: the model building and training module further comprises the following modules:
a feature extraction submodule: performing primary extraction of features on the preprocessed image, sending the preliminarily extracted image features to be detected to a feature sample library for classification, and sending the sample image features to a model trainer for feature training and model construction;
a model trainer: calling a neural network learning rule from a rule base, performing feature training by using the learning rule, and constructing a risk identification model according to a training result so as to ensure the accuracy of identification of the features to be detected and the sample features in the feature sample base;
a characteristic sample library: storing and managing the extracted image features in a classified manner;
a rule base: establishing and managing neural network learning rules, including gradient descent rules, back propagation learning rules and Delta (Wdrow-Holf) learning rules.
10. The system for monitoring the content security based on the convolutional neural network for realizing the image feature recognition as claimed in claim 8, wherein: the safety monitoring module further comprises the following modules:
the safety information analysis submodule comprises: comparing and analyzing the classified image features in the feature sample library with the risk identification model, and sending an analysis result to the security level judgment submodule;
a safety level judgment submodule: and judging the safety grade of the analysis result, marking a risk when the grade reaches a critical threshold value, and informing the early warning module of the grade judgment result.
11. The system for monitoring the content security based on the convolutional neural network for realizing the image feature recognition as claimed in claim 8, wherein: the critical threshold is marked as "at risk" for a given value, otherwise marked as "no risk".
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010133407.7A CN111400572A (en) | 2020-02-28 | 2020-02-28 | Content safety monitoring system and method for realizing image feature recognition based on convolutional neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010133407.7A CN111400572A (en) | 2020-02-28 | 2020-02-28 | Content safety monitoring system and method for realizing image feature recognition based on convolutional neural network |
Publications (1)
Publication Number | Publication Date |
---|---|
CN111400572A true CN111400572A (en) | 2020-07-10 |
Family
ID=71434112
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010133407.7A Pending CN111400572A (en) | 2020-02-28 | 2020-02-28 | Content safety monitoring system and method for realizing image feature recognition based on convolutional neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111400572A (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112418556A (en) * | 2020-12-10 | 2021-02-26 | 华东交通大学 | Gridding service system based on Internet sharing platform |
CN112560029A (en) * | 2020-12-25 | 2021-03-26 | 中国南方电网有限责任公司超高压输电公司 | Website content monitoring and automatic response protection method based on intelligent analysis technology |
CN112597828A (en) * | 2020-12-11 | 2021-04-02 | 京东数字科技控股股份有限公司 | Training method and device of webpage recognition model and webpage recognition method |
CN112906712A (en) * | 2021-03-02 | 2021-06-04 | 湖南金烽信息科技有限公司 | Neural network image preprocessing method based on light intensity analysis |
CN114148103A (en) * | 2021-11-22 | 2022-03-08 | 北京柏瑞安电子技术有限公司 | Solder paste printing mark identification method and system based on neural network learning |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20140300758A1 (en) * | 2013-04-04 | 2014-10-09 | Bao Tran | Video processing systems and methods |
CN105022835A (en) * | 2015-08-14 | 2015-11-04 | 武汉大学 | Public safety recognition method and system for crowd sensing big data |
CN107104978A (en) * | 2017-05-24 | 2017-08-29 | 赖洪昌 | A kind of network risks method for early warning based on deep learning |
CN107341518A (en) * | 2017-07-07 | 2017-11-10 | 东华理工大学 | A kind of image classification method based on convolutional neural networks |
-
2020
- 2020-02-28 CN CN202010133407.7A patent/CN111400572A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20140300758A1 (en) * | 2013-04-04 | 2014-10-09 | Bao Tran | Video processing systems and methods |
CN105022835A (en) * | 2015-08-14 | 2015-11-04 | 武汉大学 | Public safety recognition method and system for crowd sensing big data |
CN107104978A (en) * | 2017-05-24 | 2017-08-29 | 赖洪昌 | A kind of network risks method for early warning based on deep learning |
CN107341518A (en) * | 2017-07-07 | 2017-11-10 | 东华理工大学 | A kind of image classification method based on convolutional neural networks |
Non-Patent Citations (1)
Title |
---|
石少东: "基于计算机视觉的安全检测技术研究与应用", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112418556A (en) * | 2020-12-10 | 2021-02-26 | 华东交通大学 | Gridding service system based on Internet sharing platform |
CN112597828A (en) * | 2020-12-11 | 2021-04-02 | 京东数字科技控股股份有限公司 | Training method and device of webpage recognition model and webpage recognition method |
CN112560029A (en) * | 2020-12-25 | 2021-03-26 | 中国南方电网有限责任公司超高压输电公司 | Website content monitoring and automatic response protection method based on intelligent analysis technology |
CN112906712A (en) * | 2021-03-02 | 2021-06-04 | 湖南金烽信息科技有限公司 | Neural network image preprocessing method based on light intensity analysis |
CN112906712B (en) * | 2021-03-02 | 2023-05-02 | 湖南金烽信息科技有限公司 | Neural network image preprocessing method based on light intensity analysis |
CN114148103A (en) * | 2021-11-22 | 2022-03-08 | 北京柏瑞安电子技术有限公司 | Solder paste printing mark identification method and system based on neural network learning |
CN114148103B (en) * | 2021-11-22 | 2023-09-26 | 北京柏瑞安电子技术有限公司 | Solder paste printing mark recognition method and system based on neural network learning |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111368896B (en) | Hyperspectral remote sensing image classification method based on dense residual three-dimensional convolutional neural network | |
CN109584248B (en) | Infrared target instance segmentation method based on feature fusion and dense connection network | |
Xie et al. | Multilevel cloud detection in remote sensing images based on deep learning | |
CN109977918B (en) | Target detection positioning optimization method based on unsupervised domain adaptation | |
CN111400572A (en) | Content safety monitoring system and method for realizing image feature recognition based on convolutional neural network | |
TW202207077A (en) | Text area positioning method and device | |
CN110532417A (en) | Image search method, device and terminal device based on depth Hash | |
CN114969405A (en) | Cross-modal image-text mutual inspection method | |
CN108491856B (en) | Image scene classification method based on multi-scale feature convolutional neural network | |
CN112801015A (en) | Multi-mode face recognition method based on attention mechanism | |
CN113191204B (en) | Multi-scale blocking pedestrian detection method and system | |
CN113888461A (en) | Method, system and equipment for detecting defects of hardware parts based on deep learning | |
CN113159067A (en) | Fine-grained image identification method and device based on multi-grained local feature soft association aggregation | |
CN115439694A (en) | High-precision point cloud completion method and device based on deep learning | |
CN111368775A (en) | Complex scene dense target detection method based on local context sensing | |
CN115937552A (en) | Image matching method based on fusion of manual features and depth features | |
JP2015036939A (en) | Feature extraction program and information processing apparatus | |
CN114626476A (en) | Bird fine-grained image recognition method and device based on Transformer and component feature fusion | |
Yaohua et al. | A SAR oil spill image recognition method based on densenet convolutional neural network | |
CN112668662B (en) | Outdoor mountain forest environment target detection method based on improved YOLOv3 network | |
CN105844299B (en) | A kind of image classification method based on bag of words | |
CN109583584B (en) | Method and system for enabling CNN with full connection layer to accept indefinite shape input | |
CN116824330A (en) | Small sample cross-domain target detection method based on deep learning | |
Li et al. | A new algorithm of vehicle license plate location based on convolutional neural network | |
CN115063831A (en) | High-performance pedestrian retrieval and re-identification method and device |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200710 |