CN112017384A - Automatic alarm method and system for real-time area monitoring - Google Patents
Automatic alarm method and system for real-time area monitoring Download PDFInfo
- Publication number
- CN112017384A CN112017384A CN202010778621.8A CN202010778621A CN112017384A CN 112017384 A CN112017384 A CN 112017384A CN 202010778621 A CN202010778621 A CN 202010778621A CN 112017384 A CN112017384 A CN 112017384A
- Authority
- CN
- China
- Prior art keywords
- neural network
- classification
- image
- real
- anchor point
- 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
Images
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B13/00—Burglar, theft or intruder alarms
- G08B13/18—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength
- G08B13/189—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems
- G08B13/194—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems using image scanning and comparing systems
- G08B13/196—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems using image scanning and comparing systems using television cameras
- G08B13/19602—Image analysis to detect motion of the intruder, e.g. by frame subtraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/103—Static body considered as a whole, e.g. static pedestrian or occupant recognition
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B13/00—Burglar, theft or intruder alarms
- G08B13/18—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength
- G08B13/189—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems
- G08B13/194—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems using image scanning and comparing systems
- G08B13/196—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems using image scanning and comparing systems using television cameras
- G08B13/19602—Image analysis to detect motion of the intruder, e.g. by frame subtraction
- G08B13/19613—Recognition of a predetermined image pattern or behaviour pattern indicating theft or intrusion
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B13/00—Burglar, theft or intruder alarms
- G08B13/18—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength
- G08B13/189—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems
- G08B13/194—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems using image scanning and comparing systems
- G08B13/196—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems using image scanning and comparing systems using television cameras
- G08B13/19665—Details related to the storage of video surveillance data
- G08B13/19669—Event triggers storage or change of storage policy
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B13/00—Burglar, theft or intruder alarms
- G08B13/18—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength
- G08B13/189—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems
- G08B13/194—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems using image scanning and comparing systems
- G08B13/196—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems using image scanning and comparing systems using television cameras
- G08B13/19678—User interface
- G08B13/19682—Graphic User Interface [GUI] presenting system data to the user, e.g. information on a screen helping a user interacting with an alarm system
Abstract
The disclosure provides a regional real-time monitoring automatic alarm method and a system, which are used for acquiring video information in a region; intercepting a frame, transmitting data into a trained neural network model, and acquiring a selected anchor point frame and a classification result predicted value; setting a classification threshold, if the classification feasibility is smaller than the threshold, determining that no object exists, and re-acquiring the detection image in the region, if the classification possibility is larger than the threshold, determining that the object exists; the picture is intercepted and stored, and the alarm message is sent.
Description
Technical Field
The disclosure belongs to the technical field of automatic monitoring, and relates to an automatic alarm method and system for real-time monitoring of an area.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
The living safety is always an important consideration for residents as a basic life guarantee. A large number of residential areas with courtyards are distributed in suburbs of cities and vast rural areas in China, and the number of monitoring facilities is slowly increased in the parts of the areas due to high prices of monitoring software and equipment and the indifference of people to house monitoring in the early period. The method has the advantages that the recognition efficiency is improved, certain accuracy is guaranteed, meanwhile, the price of monitoring software is reduced, and the popularization rate of community monitoring facilities can be effectively increased.
The traditional monitoring mostly adopts a mode of combining a sensor and a camera, although the mode can play a certain precaution role, the monitoring video is not touched, a large amount of useful information in the video is not utilized, and an improvement margin exists.
Disclosure of Invention
The invention provides an automatic alarm method and system for area real-time monitoring, which can quickly find intruders, give an alarm and ensure safety.
According to some embodiments, the following technical scheme is adopted in the disclosure:
an automatic alarm method for real-time monitoring of an area comprises the following steps:
acquiring an image in a monitoring area;
inputting the trained neural network model, and acquiring the position information and classification result of the target in the image.
Setting a classification threshold, if the classification feasibility is smaller than the threshold, determining that no object exists, and re-acquiring the detection image in the region, if the classification possibility is larger than the threshold, determining that the object exists;
and intercepting and storing the picture, and sending an alarm message.
As an alternative embodiment, the training process of the neural network model includes:
selecting a designated human body database, and cleaning data;
scaling an image to a specified size;
inputting image data into a neural network core layer 1; wherein the neural network core layer includes:
the convolution layer obtains the bottom layer characteristics of the image;
increasing the degree of neural network nonlinearity by using a linear rectification function;
a pooling function is used for down-sampling, so that the image dimensionality is reduced, and the calculated amount is reduced;
and copying the pooled result to one copy for standby.
Again using convolutional layers, linear rectification functions, and pooling functions, where the pooled results that have been replicated as described above are combined.
Data is input to the neural network core layer 2, and the type of the internal network layer is the same as that of the neural network core layer 1.
Data is input to the neural network core layer 3, and the type of the internal network layer is the same as that of the neural network core layer 1.
Data is input to the neural network core layer 4, and the type of the internal network layer is the same as that of the neural network core layer 1.
Further feature extraction using a convolution kernel and a linear rectification function;
using a specific convolution kernel to perform feature dimension reduction;
copying the data after dimensionality reduction for use;
setting anchor point frames with specified size and number;
carrying out foreground and background classification on the data subjected to dimensionality reduction by using a normalization index function;
pooling image data in the frame selection area by using a pooling function in combination with the copied dimension reduction data, the anchor point frame and the classification result;
and inputting the pooling result and the classification result into a full-connection layer to perform anchor point frame screening and target classification.
And performing appointed loss function calculation on the anchor point frame screening result and the classification result, the fact anchor point frame and the classification label, performing back propagation, calculating partial derivatives of all parameters, and correcting all parameters by using an appointed optimization function.
By way of further limitation, the basic information includes image length, width, latitude, and type of object involved.
As a further limitation, the specified human database is the VOC2007+2012 database.
By way of further limitation, the data cleaning method is to delete data such as type names, position information, actual anchor point information and the like except for human bodies in the label file of each picture.
By way of further limitation, the image scaling specifies a size of 600 × 600.
By way of further limitation, the linear rectification function is
Wherein x is an input parameter, and f (x) is an output parameter.
By way of further limitation, the pooling function is
Wherein, the image size is W × H, W is the image width, H is the image height, the convolution kernel size is F × F, and S is the step size.
By way of further limitation, the particular convolution kernel is a 3 × 3 convolution kernel.
As a further limitation, the anchor blocks of the specified size use 3 groups of anchor blocks, the area of the anchor blocks is sequentially reduced, and the ratio of each group of 5 anchor blocks is sequentially 1:2,1:2,1:1,2:1, and 2: 1. The specific dimensions are (536, 384), (384, 536), (454 ), (384, 192), (192, 384), (272 ), (192, 96), (96, 192), (136 ).
By way of further limitation, the normalized exponential function is
P (y ═ j) is the probability that the sample vector belongs to the jth class, x is the sample vector, K is the total number of samples, and W is the weight matrix.
By way of further limitation, the specified loss function is a cross-entropy loss function, and in particular, is a cross-entropy loss function
Wherein p (x) is the probability of the true distribution, q (x) is the probability estimate calculated by the model through the data, and n is the number of types.
By way of further limitation, the specified optimization function is a stochastic gradient descent algorithm, in particular
Wherein, theta is a parameter to be optimized, eta is a learning rate, and i is a parameter index to be learned.
By way of further limitation, the training process of the neural network model is continuously carried out until the optimization result is unchanged or the change amplitude is small.
An area real-time monitoring automatic alarm system, comprising:
the monitoring module comprises at least one camera and is used for detecting images in an area to be monitored;
the network model module is used for processing the image based on the trained neural network model to obtain the selected anchor point frame and the classification result predicted value;
the judging module is used for setting a classification threshold, if the classification feasibility is smaller than the threshold, the object is considered not to exist, the detected image in the region is obtained again, and if the classification possibility is larger than the threshold, the object is considered to exist;
and the alarm module is used for sending an alarm message when judging that the object exists.
As a further limitation, the system further comprises a display module for displaying the image and the alarm information;
the storage module is used for storing the pictures judged to have the object;
and the communication module is used for being connected with the user intelligent terminal and sending alarm information.
A computer readable storage medium having stored therein a plurality of instructions adapted to be loaded by a processor of a terminal device and to execute said a method of area real-time monitoring and automatic alerting.
A terminal device comprising a processor and a computer readable storage medium, the processor being configured to implement instructions; the computer readable storage medium is used for storing a plurality of instructions, and the instructions are suitable for being loaded by a processor and executing the automatic alarm method for real-time monitoring of the area.
Compared with the prior art, the beneficial effect of this disclosure is:
the invention provides five specific network anchor points with the proportion of 1:2,1:2,1:1,2:1 and 2:1 by designing a neural network structure, and the specific network anchor points are used for fitting the shape of a human body and improving the accuracy rate of human body identification.
The real-time monitoring system provided by the disclosure continuously stores images when people are detected to invade, gives an alarm, and simultaneously sends an alarm message to the user intelligent terminal in real time, so that the convenience is increased while the real-time performance is met.
Drawings
The accompanying drawings, which are included to provide a further understanding of the disclosure, illustrate embodiments of the disclosure and together with the description serve to explain the disclosure and are not to limit the disclosure.
FIG. 1 is a schematic flow chart of the present embodiment;
FIG. 2 is a neural network training flow diagram;
FIG. 3 is an anchor point shape diagram;
FIG. 4 is a flow chart of a monitoring system main routine;
fig. 5 and 6 are an original image and an identification effect image of scene one, respectively;
fig. 7 and 8 are an original image and a recognition effect image of a second scene, respectively;
fig. 9 and 10 are an original drawing and a recognition effect drawing of scene three, respectively;
fig. 11 and 12 are an original image and a recognition effect image of scene four, respectively;
FIG. 13 is a visualization interface effect diagram;
FIG. 14 is a flow chart of the operation of the visualization interface.
The specific implementation mode is as follows:
the present disclosure is further described with reference to the following drawings and examples.
It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the disclosure. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments according to the present disclosure. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.
An automatic alarm method for real-time area monitoring includes
Acquiring an image in a monitoring area;
inputting the trained neural network model, and acquiring the position information and classification result of the target in the image.
Setting a classification threshold, if the classification feasibility is smaller than the threshold, determining that no object exists, and re-acquiring the detection image in the region, if the classification possibility is larger than the threshold, determining that the object exists;
and intercepting and storing the picture, and sending an alarm message.
As an alternative embodiment, the training process of the neural network model includes:
selecting a designated human body database, and cleaning data;
scaling an image to a specified size;
inputting image data into a neural network core layer 1; wherein the neural network core layer includes:
the convolution layer obtains the bottom layer characteristics of the image;
increasing the degree of neural network nonlinearity by using a linear rectification function;
a pooling function is used for down-sampling, so that the image dimensionality is reduced, and the calculated amount is reduced;
and copying the pooled result to one copy for standby.
Again using convolutional layers, linear rectification functions, and pooling functions, where the pooled results that have been replicated as described above are combined.
Data is input to the neural network core layer 2, and the type of the internal network layer is the same as that of the neural network core layer 1.
Data is input to the neural network core layer 3, and the type of the internal network layer is the same as that of the neural network core layer 1.
Data is input to the neural network core layer 4, and the type of the internal network layer is the same as that of the neural network core layer 1.
Further feature extraction using a convolution kernel and a linear rectification function;
using a specific convolution kernel to perform feature dimension reduction;
copying the data after dimensionality reduction for use;
setting anchor point frames with specified size and number;
carrying out foreground and background classification on the data subjected to dimensionality reduction by using a normalization index function;
pooling image data in the frame selection area by using a pooling function in combination with the copied dimension reduction data, the anchor point frame and the classification result;
and inputting the pooling result and the classification result into a full-connection layer to perform anchor point frame screening and target classification.
And performing appointed loss function calculation on the anchor point frame screening result and the classification result, the fact anchor point frame and the classification label, performing back propagation, calculating partial derivatives of all parameters, and correcting all parameters by using an appointed optimization function.
By way of further limitation, the basic information includes image length, width, latitude, and type of object involved.
As a further limitation, the specified human database is the VOC2007+2012 database.
By way of further limitation, the data cleaning method is to delete data such as type names, position information, actual anchor point information and the like except for human bodies in the label file of each picture.
By way of further limitation, the image scaling specifies a size of 600 × 600.
By way of further limitation, the linear rectification function is
Wherein x is an input parameter, and f (x) is an output parameter.
By way of further limitation, the pooling function is
Wherein, the image size is W × H, W is the image width, H is the image height, the convolution kernel size is F × F, and S is the step size.
By way of further limitation, the particular convolution kernel is a 3 × 3 convolution kernel.
As a further limitation, the anchor blocks of the specified size use 3 groups of anchor blocks, the area of the anchor blocks is sequentially reduced, and the ratio of each group of 5 anchor blocks is sequentially 1:2,1:2,1:1,2:1, and 2: 1. The specific dimensions are (536, 384), (384, 536), (454 ), (384, 192), (192, 384), (272 ), (192, 96), (96, 192), (136 ).
By way of further limitation, the normalized exponential function is
P (y ═ j) is the probability that the sample vector belongs to the jth class, x is the sample vector, K is the total number of samples, and W is the weight matrix.
By way of further limitation, the specified loss function is a cross-entropy loss function, and in particular, is a cross-entropy loss function
Wherein p (x) is the probability of the true distribution, q (x) is the probability estimate calculated by the model through the data, and n is the number of types.
By way of further limitation, the specified optimization function is a stochastic gradient descent algorithm, in particular
Wherein, theta is a parameter to be optimized, eta is a learning rate, and i is a parameter index to be learned.
By way of further limitation, the training process of the neural network model is continuously carried out until the optimization result is unchanged or the change amplitude is small.
Taking a courtyard as an example for illustration, of course, the monitoring area of the present disclosure may be applied to a factory building, a school district, a community, a room, etc., rather than a courtyard.
(1) As shown in fig. 1, selecting default parameters, selecting a training period of 7, setting a training check point to 10021, starting a GPU acceleration sub-module, setting a model loading path, setting a default state of a camera to "on", and entering a main program of the monitoring system shown in fig. 5.
(2) In the detection process, data are input into a neural network model, wherein the training of the model is realized through the following steps:
1) selecting a designated human body database, and cleaning data;
2) scaling an image to a specified size;
3) inputting image data into a neural network core layer 1; wherein the neural network core layer includes:
firstly, acquiring bottom layer characteristics of an image by a convolution layer;
increasing the nonlinear degree of the neural network by using a linear rectification function;
using a pooling function to perform downsampling, reducing image dimensionality and reducing calculated amount;
fourthly, duplicating the pooling result to be reserved for standby.
Re-using the convolution layer, the linear rectification function and the pooling function, and combining the result of the replicated pooling as shown in the above-mentioned (r).
4) Data is input to the neural network core layer 2, and the type of the internal network layer is the same as that of the neural network core layer 1.
5) Data is input to the neural network core layer 3, and the type of the internal network layer is the same as that of the neural network core layer 1.
6) Data is input to the neural network core layer 4, and the type of the internal network layer is the same as that of the neural network core layer 1.
7) Further feature extraction using a convolution kernel and a linear rectification function;
8) using a specific convolution kernel to perform feature dimension reduction;
9) copying the data after dimensionality reduction for use;
10) setting anchor point frames with specified size and number;
11) carrying out foreground and background classification on the data subjected to dimensionality reduction by using a normalization index function;
12) pooling image data in the frame selection area by using a pooling function in combination with the copied dimension reduction data in the step 9), the anchor point frame in the step 10) and the classification result in the step 11);
13) and inputting the pooling result and the classification result into a full-connection layer to perform anchor point frame screening and target classification.
14) And performing appointed loss function calculation on the anchor point frame screening result and the classification result, the fact anchor point frame and the classification label, performing back propagation, calculating partial derivatives of all parameters, and correcting all parameters by using an appointed optimization function. And returning to the step 3) to perform a new round of training.
(3) The camera is turned on and, of course, is arranged in the yard in advance.
(4) Reading a frame of data.
(5) And transmitting the picture into the trained neural network model, acquiring the selected anchor point frame and the classification result predicted value, and displaying the result on a screen.
Fig. 5-6, 7-8, 9-10, 11-12 are diagrams of the original image and recognition effect of scene one, two, three, and four, respectively.
(6) Setting the classification threshold value to be 60%, if the classification feasibility is smaller than the threshold value, namely, no object is considered to exist, then, the step (4) is carried out, if the classification possibility is larger than the threshold value, namely, the object is considered to exist, then, the step (7) is carried out.
(7) And (5) intercepting the picture, storing the picture in a specified folder, returning the main thread to the step (4), starting the sub-thread, and entering the step (8).
(8) And re-starting a sub thread, and calling the deep-wechat Wechat client to send an alarm message to the user.
Of course, in other embodiments, other threads or modules may be used to send information to the user intelligent terminal.
The embodiment provides a display module and an interaction module, and as shown in fig. 13, the visualized operation interface style is composed of a title bar, a prompt bar and a button area.
Fig. 14 shows an operation flowchart, which includes the following steps:
1) title bar display "courtyard monitoring automatic WeChat alarm System";
2) the prompt bar displays "prompt: when the 'start' button is pressed to start the program to exit, a video window is selected firstly, and then a 'q' key is pressed to exit the video interface;
3) the button area consists of three buttons, namely 'start', 'view screenshot' and 'exit'. Pressing a start button, enabling the system to enter a main program of the monitoring system, and starting the monitoring program; pressing a screen capture viewing button, opening a file browser by the system, entering a specified folder, and viewing the acquired screen capture; and exiting the program.
In the embodiment, by designing a neural network model, five specific anchor points with the proportions of 1:2,1:2,1:1,2:1 and 2:1 are provided, and the anchor points are used for fitting the shape of a human body and improving the human body identification accuracy. The original human body identification accuracy is improved by 7%. A whole set of courtyard real-time monitoring system is designed, images are continuously stored when people are detected to invade, and alarm messages are sent to mobile phone WeChat of users in real time, so that convenience is increased while real-time performance is met. Experiments show that the system can find the target within 120 milliseconds when an intruder enters the visual field, and can send out the alarm message, thereby meeting the real-time design requirement of the system.
As will be appreciated by one skilled in the art, embodiments of the present disclosure may be provided as a method, system, or computer program product. Accordingly, the present disclosure may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present disclosure may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and so forth) having computer-usable program code embodied therein.
The present disclosure is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the disclosure. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The above description is only a preferred embodiment of the present disclosure and is not intended to limit the present disclosure, and various modifications and changes may be made to the present disclosure by those skilled in the art. Any modification, equivalent replacement, improvement and the like made within the spirit and principle of the present disclosure should be included in the protection scope of the present disclosure.
Although the present disclosure has been described with reference to specific embodiments, it should be understood that the scope of the present disclosure is not limited thereto, and those skilled in the art will appreciate that various modifications and changes can be made without departing from the spirit and scope of the present disclosure.
Claims (10)
1. An automatic alarm method for real-time monitoring of an area is characterized in that: the method comprises the following steps:
acquiring a detection image in the region;
transmitting the picture into a trained neural network model, and acquiring an anchor point frame and a classification result predicted value;
setting a classification threshold, if the classification feasibility is smaller than the threshold, determining that no object exists, and re-acquiring the detection image in the region, if the classification possibility is larger than the threshold, determining that the object exists;
and intercepting and storing the picture, and sending an alarm message.
2. The area real-time monitoring automatic alarm method as claimed in claim 1, characterized in that: the training process of the neural network model comprises the following steps:
selecting a database and carrying out data cleaning;
scaling an image to a specified size;
inputting image data into a neural network core layer 1; wherein the neural network core layer includes:
(1) the convolution layer obtains the bottom layer characteristics of the image;
(2) increasing the degree of neural network nonlinearity by using a linear rectification function;
(3) down-sampling using a pooling function;
(4) copying the pooling result to one copy for later use;
(5) again using the convolutional layer, linear rectification function, and pooling function, where the pooled results that have been replicated as described above are combined;
inputting data into a neural network core layer 2, wherein the type of an internal network layer is the same as that of the neural network core layer 1;
inputting data into a neural network core layer 3, wherein the type of an internal network layer is the same as that of the neural network core layer 1;
inputting data into a neural network core layer 4, wherein the type of an internal network layer is the same as that of the neural network core layer 1;
further feature extraction using a convolution kernel and a linear rectification function;
using a 3 multiplied by 3 convolution kernel to carry out feature dimensionality reduction;
copying the data after dimensionality reduction for use;
setting anchor point frames with specified size and number;
carrying out foreground and background classification on the data subjected to dimensionality reduction by using a normalization index function;
pooling image data in the frame selection area by using a pooling function in combination with the copied dimension reduction data, the anchor point frame and the classification result;
inputting the pooling result and the classification result into a full-connection layer to carry out anchor point frame screening and target classification;
and performing appointed loss function calculation on the anchor point frame screening result and the classification result, the fact anchor point frame and the classification label, performing back propagation, calculating partial derivatives of all parameters, and correcting all parameters by using an appointed optimization function.
3. The area real-time monitoring automatic alarm method as claimed in claim 2, characterized in that: the basic information includes image length, width, latitude and the type of the included object.
4. The area real-time monitoring automatic alarm method as claimed in claim 2, characterized in that: the specific process of modifying each parameter includes refining each parameter value using a stochastic gradient descent algorithm.
5. The area real-time monitoring automatic alarm method as claimed in claim 2, characterized in that: and continuously training the neural network model until the optimization result is unchanged or the change amplitude is small.
6. The area real-time monitoring automatic alarm method as claimed in claim 2, characterized in that: the anchor point is designed into five proportions of 1:2,1:2,1:1,2:1 and 2:1 respectively.
7. An automatic alarm system for area real-time monitoring is characterized in that: the method comprises the following steps:
the monitoring module comprises at least one camera and is used for detecting images in an area to be monitored;
the network model module is used for processing the image based on the trained neural network model to obtain the selected anchor point frame and the classification result predicted value;
the judging module is used for setting a classification threshold, if the classification feasibility is smaller than the threshold, the object is considered not to exist, the detected image in the region is obtained again, and if the classification possibility is larger than the threshold, the object is considered to exist;
and the alarm module is used for sending an alarm message when judging that the object exists.
8. The system according to claim 7, wherein the system comprises: the display module is used for displaying the image and the alarm information;
the storage module is used for storing the pictures judged to have the object;
and the communication module is used for being connected with the user intelligent terminal and sending alarm information.
9. A computer-readable storage medium characterized by: a plurality of instructions stored therein, the instructions being adapted to be loaded by a processor of a terminal device and to perform a method of real-time monitoring and automatic alerting of an area as claimed in any one of claims 1 to 6.
10. A terminal device is characterized in that: the system comprises a processor and a computer readable storage medium, wherein the processor is used for realizing instructions; a computer readable storage medium for storing a plurality of instructions adapted to be loaded by a processor and to perform a method of real-time monitoring and automatic alerting of an area as claimed in any one of claims 1-6.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010778621.8A CN112017384A (en) | 2020-08-05 | 2020-08-05 | Automatic alarm method and system for real-time area monitoring |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010778621.8A CN112017384A (en) | 2020-08-05 | 2020-08-05 | Automatic alarm method and system for real-time area monitoring |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112017384A true CN112017384A (en) | 2020-12-01 |
Family
ID=73500064
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010778621.8A Pending CN112017384A (en) | 2020-08-05 | 2020-08-05 | Automatic alarm method and system for real-time area monitoring |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112017384A (en) |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103824054A (en) * | 2014-02-17 | 2014-05-28 | 北京旷视科技有限公司 | Cascaded depth neural network-based face attribute recognition method |
CN108647559A (en) * | 2018-03-21 | 2018-10-12 | 四川弘和通讯有限公司 | A kind of danger recognition methods based on deep learning |
CN108960067A (en) * | 2018-06-05 | 2018-12-07 | 北京华纵科技有限公司 | Real-time train driver motion recognition system and method based on deep learning |
CN109376693A (en) * | 2018-11-22 | 2019-02-22 | 四川长虹电器股份有限公司 | Method for detecting human face and system |
CN109583315A (en) * | 2018-11-02 | 2019-04-05 | 北京工商大学 | A kind of quick human posture recognition method of multichannel towards intelligent video monitoring |
CN109800802A (en) * | 2019-01-10 | 2019-05-24 | 深圳绿米联创科技有限公司 | Visual sensor and object detecting method and device applied to visual sensor |
CN110633641A (en) * | 2019-08-15 | 2019-12-31 | 河北工业大学 | Intelligent security pedestrian detection method, system and device and storage medium |
WO2020061489A1 (en) * | 2018-09-20 | 2020-03-26 | Nvidia Corporation | Training neural networks for vehicle re-identification |
-
2020
- 2020-08-05 CN CN202010778621.8A patent/CN112017384A/en active Pending
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103824054A (en) * | 2014-02-17 | 2014-05-28 | 北京旷视科技有限公司 | Cascaded depth neural network-based face attribute recognition method |
CN108647559A (en) * | 2018-03-21 | 2018-10-12 | 四川弘和通讯有限公司 | A kind of danger recognition methods based on deep learning |
CN108960067A (en) * | 2018-06-05 | 2018-12-07 | 北京华纵科技有限公司 | Real-time train driver motion recognition system and method based on deep learning |
WO2020061489A1 (en) * | 2018-09-20 | 2020-03-26 | Nvidia Corporation | Training neural networks for vehicle re-identification |
CN109583315A (en) * | 2018-11-02 | 2019-04-05 | 北京工商大学 | A kind of quick human posture recognition method of multichannel towards intelligent video monitoring |
CN109376693A (en) * | 2018-11-22 | 2019-02-22 | 四川长虹电器股份有限公司 | Method for detecting human face and system |
CN109800802A (en) * | 2019-01-10 | 2019-05-24 | 深圳绿米联创科技有限公司 | Visual sensor and object detecting method and device applied to visual sensor |
CN110633641A (en) * | 2019-08-15 | 2019-12-31 | 河北工业大学 | Intelligent security pedestrian detection method, system and device and storage medium |
Non-Patent Citations (7)
Title |
---|
吴建鑫: "《模式识别》", 31 March 2020, pages: 285 - 287 * |
李在铭: "《数字图像处理压缩与识别技术》", 30 November 2000, pages: 257 - 258 * |
焦李成: "《计算智能导论》", 30 September 2019, pages: 279 - 280 * |
王浩: "基于视觉的行人检测技术研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
王浩: "基于视觉的行人检测技术研究", 《中国优秀硕士学位论文全文数据库信息科技辑》, 15 December 2018 (2018-12-15) * |
门鑫: "基于深度学习的视频内容识别和搜索算法研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
门鑫: "基于深度学习的视频内容识别和搜索算法研究", 《中国优秀硕士学位论文全文数据库信息科技辑》, 15 August 2019 (2019-08-15) * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106527892B (en) | Screen capturing method and system of electronic equipment | |
CN109905782B (en) | Control method and device | |
CA3061745A1 (en) | System and method for generative model for stochastic point processes | |
CN109800483A (en) | A kind of prediction technique, device, electronic equipment and computer readable storage medium | |
CN108416314B (en) | Picture important face detection method | |
CN113807276B (en) | Smoking behavior identification method based on optimized YOLOv4 model | |
CN112312087B (en) | Method and system for quickly positioning event occurrence time in long-term monitoring video | |
CN109816745A (en) | Human body thermodynamic chart methods of exhibiting and Related product | |
CN111046821A (en) | Video behavior identification method and system and electronic equipment | |
CN110427824B (en) | Automatic security testing method and system for artificial intelligent virtual scene | |
Cheng et al. | Smoke detection and trend prediction method based on Deeplabv3+ and generative adversarial network | |
WO2020103293A1 (en) | Method, device, and electronic device for presenting individual search information | |
CN110740298A (en) | Distributed classroom discipline behavior detection system, method and medium | |
CN115002414A (en) | Monitoring method, monitoring device, server and computer readable storage medium | |
CN111860457A (en) | Fighting behavior recognition early warning method and recognition early warning system thereof | |
CN111951390A (en) | Warning situation display method, system, device and storage medium | |
CN111046213A (en) | Knowledge base construction method based on image recognition | |
CN114187541A (en) | Intelligent video analysis method and storage device for user-defined service scene | |
CN113553952A (en) | Abnormal behavior recognition method and device, equipment, storage medium and program product | |
WO2024045948A1 (en) | Interface display method and apparatus, terminal and storage medium | |
CN112819509A (en) | Method, system, electronic device and storage medium for automatically screening advertisement pictures | |
CN113052127A (en) | Behavior detection method, behavior detection system, computer equipment and machine readable medium | |
CN116189063B (en) | Key frame optimization method and device for intelligent video monitoring | |
CN112017384A (en) | Automatic alarm method and system for real-time area monitoring | |
CN112487961A (en) | Traffic accident detection method, storage medium and equipment |
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 |
Application publication date: 20201201 |
|
RJ01 | Rejection of invention patent application after publication |