CN115170059A - Intelligent safety monitoring system for outdoor construction site and working method - Google Patents
Intelligent safety monitoring system for outdoor construction site and working method Download PDFInfo
- Publication number
- CN115170059A CN115170059A CN202210671538.XA CN202210671538A CN115170059A CN 115170059 A CN115170059 A CN 115170059A CN 202210671538 A CN202210671538 A CN 202210671538A CN 115170059 A CN115170059 A CN 115170059A
- Authority
- CN
- China
- Prior art keywords
- face
- processing module
- alarm
- module
- central processing
- 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 46
- 238000010276 construction Methods 0.000 title claims abstract description 29
- 238000000034 method Methods 0.000 title claims description 30
- 238000012545 processing Methods 0.000 claims abstract description 72
- 230000005540 biological transmission Effects 0.000 claims abstract description 25
- 238000012549 training Methods 0.000 claims abstract description 9
- 230000001815 facial effect Effects 0.000 claims abstract description 4
- 238000001514 detection method Methods 0.000 claims description 35
- 239000011159 matrix material Substances 0.000 claims description 33
- 238000005070 sampling Methods 0.000 claims description 21
- 239000000779 smoke Substances 0.000 claims description 17
- 239000013598 vector Substances 0.000 claims description 12
- 238000004364 calculation method Methods 0.000 claims description 9
- 230000008569 process Effects 0.000 claims description 9
- 230000009467 reduction Effects 0.000 claims description 6
- 230000001629 suppression Effects 0.000 claims description 6
- 230000001960 triggered effect Effects 0.000 claims description 6
- 238000000513 principal component analysis Methods 0.000 claims description 5
- 230000008859 change Effects 0.000 claims description 4
- 238000009825 accumulation Methods 0.000 claims description 3
- 238000000354 decomposition reaction Methods 0.000 claims description 3
- 238000010586 diagram Methods 0.000 claims description 3
- 239000000284 extract Substances 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 3
- 239000013589 supplement Substances 0.000 claims description 3
- 230000002776 aggregation Effects 0.000 claims description 2
- 238000004220 aggregation Methods 0.000 claims description 2
- 238000005516 engineering process Methods 0.000 description 5
- 238000004458 analytical method Methods 0.000 description 2
- 238000009435 building construction Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 230000007257 malfunction Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000007935 neutral effect Effects 0.000 description 1
- 230000017105 transposition Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/10—Office automation; Time management
- G06Q10/103—Workflow collaboration or project management
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
- G06Q50/265—Personal security, identity or safety
-
- 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
- G06V10/765—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects using rules for classification or partitioning the feature space
-
- 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
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/161—Detection; Localisation; Normalisation
-
- 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
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/168—Feature extraction; Face representation
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B19/00—Alarms responsive to two or more different undesired or abnormal conditions, e.g. burglary and fire, abnormal temperature and abnormal rate of flow
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B25/00—Alarm systems in which the location of the alarm condition is signalled to a central station, e.g. fire or police telegraphic systems
- G08B25/01—Alarm systems in which the location of the alarm condition is signalled to a central station, e.g. fire or police telegraphic systems characterised by the transmission medium
- G08B25/08—Alarm systems in which the location of the alarm condition is signalled to a central station, e.g. fire or police telegraphic systems characterised by the transmission medium using communication transmission lines
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B3/00—Audible signalling systems; Audible personal calling systems
- G08B3/10—Audible signalling systems; Audible personal calling systems using electric transmission; using electromagnetic transmission
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N7/00—Television systems
- H04N7/18—Closed-circuit television [CCTV] systems, i.e. systems in which the video signal is not broadcast
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- Multimedia (AREA)
- General Health & Medical Sciences (AREA)
- Tourism & Hospitality (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Pure & Applied Mathematics (AREA)
- Human Computer Interaction (AREA)
- General Business, Economics & Management (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Databases & Information Systems (AREA)
- Marketing (AREA)
- Computational Mathematics (AREA)
- Software Systems (AREA)
- Mathematical Optimization (AREA)
- Mathematical Analysis (AREA)
- Computing Systems (AREA)
- Entrepreneurship & Innovation (AREA)
- Economics (AREA)
- Evolutionary Computation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Medical Informatics (AREA)
- Artificial Intelligence (AREA)
- Computer Security & Cryptography (AREA)
- Development Economics (AREA)
- Educational Administration (AREA)
- Primary Health Care (AREA)
- Signal Processing (AREA)
- Electromagnetism (AREA)
- Algebra (AREA)
- General Engineering & Computer Science (AREA)
Abstract
The system comprises a face acquisition module, a cloud processing module, a video acquisition module, a central processing module, a wireless transmission module and an alarm module. Firstly, a system administrator compares and logs in a safety monitoring system through facial recognition with an authorized face database established at a cloud end, and starts the monitoring system; secondly, the system center processing module calls a monitoring camera to acquire real-time construction site information, and performs matching monitoring with a safety helmet and a firework model established by training; and finally, if the situation that the personnel do not wear the safety helmet or a fire disaster happens is monitored, the alarm module is activated to carry out voice prompt or trigger an alarm system, and meanwhile, short message prompt is carried out on the security management personnel. The invention judges whether the user has the authority to enter the safety monitoring system through face recognition, designs a pre-alarm system aiming at safety helmet wearing and fire according to potential safety hazards possibly existing in the field aiming at the characteristics of outdoor construction scenes, can effectively prevent the occurrence of dangers in outdoor construction sites, and ensures the safety of personnel and property.
Description
Technical Field
The invention relates to an intelligent safety detection system for an outdoor construction site and a working method, and belongs to the technical field of monitoring systems.
Background
With the development of society and the progress of information science and technology, the building industry has become an important component of national economy in China. In recent years, under the combined efforts of the nation, government and industry subjects, the accident rate of building construction safety production is reduced year by year, and the safety measures are gradually improved, but the potential safety hazard still exists in the construction site due to an open construction scene, a complicated space structure and a large number of personnel. Because the current safety systems are mostly installed on a construction site for real-time monitoring, risk alarming and prompting are lacked, and a dispatcher cannot avoid risks timely and accurately.
The patent with patent number CN107798805A and invented name "a safety monitoring system based on information technology" discloses a fire monitoring module which uses smoke concentration sensors, temperature sensors and smoke color sensors to monitor whether fire occurs in real time, and the related hardware sensors are too many and are not suitable for being installed in outdoor sites, especially outdoor construction sites, which may possibly result in the failure of the temperature sensors and the smoke concentration sensors, and the proposed smoke color sensors are very susceptible to the influence of light to cause the failure. In patent No. CN210955410U entitled "automatic fire alarm system", the proposed solution also uses related sensors to perform real-time fire monitoring, but temperature sensors and smoke sensors are prone to malfunction in outdoor sites, and even cannot be applied to building construction sites.
For the construction aspect, personal safety is also particularly important, and whether the safety helmet of the constructor is worn correctly as required is monitored in real time. The patent number CN210955410U, entitled "an intelligent construction site safety helmet detection system and method" provides a system which cannot realize the instant reminding of people who do not correctly wear the safety helmet. Aiming at the problems, the invention provides an intelligent safety monitoring system for an outdoor construction site and a working method, which are used for collecting real-time information of a construction site, carrying out joint monitoring on the problems of whether a person wears a safety helmet and whether smoke and fire occur or not and transmitting the occurring risk information to a mobile phone end in time to realize an alarm function; meanwhile, managers can enter the monitoring system only through face verification, and the safety of the system is guaranteed. Not only can be applied to indoor construction site, also can be applied to outdoor construction site, and place flexibility and adaptability are better, and the problem that relevant sensor easily receives environmental impact in the traditional scheme has effectually been avoided moreover
Disclosure of Invention
According to the defects and shortcomings of the prior art and solutions, the invention provides an intelligent safety monitoring system for an outdoor construction site and a working method.
The technical scheme of the invention is as follows:
the system comprises a face acquisition module, a cloud processing module, a video acquisition module, a central processing module, a wireless transmission module and an alarm module: the face acquisition module is a camera equipped for the central processing module and can acquire face information; the central processing module is connected with the wireless transmission module and is communicated with the cloud processing module through the wireless transmission module, the cloud processing module comprises an established face database, the central processing module performs data transmission through the wireless transmission module and enters the face database according to the IP address of the face database of the cloud processing module to perform data processing; the construction method comprises the following steps that a video acquisition module with a wireless transmission module is configured on a construction site, acquired real-time information is transmitted to a central processing module in a wireless mode, the central processing module judges whether personnel wear safety helmets and whether fire risks exist in the construction site or not according to the acquired real-time information, if the fire risks exist, an alarm is triggered, alarm information is transmitted to an alarm module through an mqtt protocol in the wireless transmission module, and a system is triggered to alarm, wherein the working method of the system comprises the following steps:
1) The face acquisition module acquires face information of a user and extracts facial features of the user;
2) The face information of the user is uploaded to a cloud processing module, the face information is compared with a face database established at the cloud, the identity information of the user is determined, the authority of the user is judged, and if the user is successfully matched with the database, the user is allowed to enter a safety helmet and a firework safety monitoring interface of a central processing module of a safety monitoring system for subsequent monitoring; if the matching fails, the user is not allowed to enter the safety monitoring system and feeds the information back to the face acquisition module to wait for the next recognition;
the face recognition part adopts a face recognition algorithm based on opencv, and the steps are as follows:
(1) The central processing module performs frame reading on the face data set, performs heterogeneous stretching processing on an original image by adopting a histogram equalization method to expand the pixel value interval, homogenizes the pixel quantity of each gray scale range through an accumulation function, enhances the image contrast and improves the face recognition accuracy;
(2) Adopting a projection algorithm based on Linear Discriminant Analysis (LDA) to reduce dimension of the face picture, and assuming that the existing sample data matrix X = [ X ] 1 ,x 2 ,...,x m ]Wherein the sample x i For any n-dimensional vector, divide X into k classes of samples, i.e. { C 1 ,C 2 ,…,C k H, i belongs to 1,2,. K; definition of N j 、X j And mu j Are respectively j type samples C j The number, set and mean value of j ∈ 1, 2., k, LDA dimension reduction operation comprises the following four steps,
(1) calculating an intra-class divergence matrix S b For indicating the degree of scatter of data points within each class,
wherein (.) T Representing a transposition;
(2) meterInterspinal divergence matrix S w For indicating the degree of aggregation of data points within each class,
(3) computing matrices
Performing characteristic decomposition on the characteristic vector, and calculating characteristic vectors corresponding to the largest d characteristic values to form a matrix W;
(4) calculating the data point Y = W after projection according to the sample matrix X T X, finishing the dimensionality reduction of the data;
(3) The method for searching the main characteristic points of the image by adopting a Principal Component Analysis (PCA) algorithm comprises the following five steps,
(1) data preprocessing: preparing a face data set, forming an N multiplied by M dimensional matrix after linear identification LDA algorithm, representing a face image as a pixel matrix during processing, and representing the pixel value of one face image in each row of the matrix;
(2) calculating the average value of the face pixels: calculating the average value of each column to finally obtain an M-dimensional row vector, and generating an average pixel value of the face image, namely an average face;
(3) sample centralization: subtracting the average face from each row of the original data matrix, and transposing, wherein each transposed column contains pixel information specific to each face;
(4) calculating the characteristic quantity of the face pixels: calculating an eigenvalue and an eigenvector of the covariance matrix, and forming an eigenface by the eigenvector matrix;
(5) the face recognition is carried out through a camera shooting function integrated by the central processing module so as to confirm the identity of the user;
3) After the user identity is confirmed, the user enters a safety helmet and a smoke and fire monitoring interface of the safety monitoring system, the central processing module calls a video acquisition module in a construction site, and picture video information acquired by a monitoring camera in real time is received through the wireless transmission module;
4) The potential safety hazard judgment comprises safety helmet detection and smoke and fire detection, a detection target is identified through the behavior and the posture of an object in a picture by using a corresponding identification algorithm through a central processing module according to picture video information collected by a video collecting module, and whether the potential safety hazard exists in the picture video information is judged through the safety helmet and smoke and fire characteristics led into a model;
the helmet and firework detection part adopts a training detection algorithm based on a Yolov5 network, firstly, a Yolov5 network (Yolov 5s, yolov5m, yolov5l, yolov5x network) is used for training a target detection model, and target detection of the helmet and firework is completed through a generated target detection model, wherein the Yolov5 network comprises an input end, a backhaul, a neutral and an output end:
(1) the input end comprises Mosaic data enhancement, self-adaptive anchor frame calculation and self-adaptive picture scaling, wherein the Mosaic data enhancement is mainly used for solving the problem of small target detection and enriching a data set;
(2) the Backbone part adopts a Focus structure and a cross-phase local network (CSPnet) structure, an original 608 multiplied by 3 image is input into the Focus structure, a slicing operation is carried out to change the original 608 multiplied by 3 image into a feature map of 304 multiplied by 12, and the original 608 multiplied by 3 image is finally changed into a feature map of 304 multiplied by 32 through a convolution operation of 32 convolution kernels;
(3) the method comprises the following steps that a Neck part adopts a Feature Pyramid Network (FPN) + bottom-up pyramid network (PAN) structure to carry out up-sampling and down-sampling on a picture, the FPN carries out up-sampling on the picture from top to bottom, the FPN structure fuses high-level features of the picture through the up-sampling and low-level features to obtain a predicted feature map, the PAN structure adds a pyramid network from bottom to top behind the FPN structure to supplement the FPN structure and transmit positioning features of the low level upwards, the PAN structure firstly copies the bottom level of the FPN structure to serve as the bottom level of a new feature pyramid and carries out down-sampling on the bottom level of the new feature pyramid, then the second layer of the reciprocal of the original feature pyramid carries out 3 x 3 convolution calculation with the step size of 2, and finally, the PAN structure is transversely connected with the bottom level of the new feature pyramid after the down-sampling operation, the 3 x 3 convolution calculation is carried out after the addition of the two layers to fuse the features of the pyramid, and the up-down sampling refers to the scaling of the picture;
(4) the method comprises the steps that a Bounding box loss function and an nms non-maximum suppression algorithm are adopted at an output end, wherein the Bounding box loss function is the sum of coordinate loss, target confidence loss and classification loss, the nms non-maximum suppression algorithm is used for screening target frames, the specific process is that a plurality of candidate frames are obtained by depending on a target detection classifier (a classifier which learns classification rules by using given classes and known training data and is used for classifying or predicting unknown data), probability values corresponding to all the candidate frames are ranked according to the probability values of the category to which the candidate frames belong, a candidate frame A with the highest probability value is selected, other candidate frames are traversed, if the overlapping area of the candidate frame B and the candidate frame A with the highest current probability value is larger than or equal to a set threshold value, the candidate frame B is deleted, if the overlapping area is smaller than the threshold value, processing is not carried out, then a highest probability value is continuously selected from the unprocessed candidate frames, the process is repeated, and only one candidate frame which is not screened after all the candidate frames are traversed, and the probability value of the candidate frame corresponding to the confidence value of the candidate frame is the target frame;
5) Warning prompt, if the identification result has potential safety hazard, the central processing module triggers warning to make the warning module send warning short message to give a warning or play voice prompt alarm; if no potential safety hazard exists, the central processing module continues to identify the video information of the subsequent pictures;
the alarm part sets a recognition alarm threshold value, and the setting process is as follows: the central processing module monitors the received video information, respectively obtains confidence degrees corresponding to the safety helmet and the firework monitoring target, if the confidence degree is larger than or equal to an alarm threshold value, the safety hazard is identified, voice prompt is played or an alarm short message is sent to a mobile phone end of an administrator for alarming, if the confidence degree is smaller than the alarm threshold value, no safety hazard exists, and the next video information is monitored, wherein the confidence degree represents the confidence degree of the detection target existing in the framed square and the confidence degree of the whole object including all characteristics of the framed square.
The invention has the advantages that: the performance of outdoor building site intelligent security monitoring system under complicated place is effectual, mainly to the building site. In terms of implementation environment, the system is hardly influenced by external factors except that the influence of illumination is large. The system meets the requirements of real-time monitoring of fire in a construction site and also meets the requirements of real-time monitoring and reminding whether constructors correctly wear safety helmets or not, and the safety of a construction site is fully ensured; and the designed target monitoring system applies the technology without the sensor, and the computer vision technology is involved, so that the flexibility of the specific implementation application is higher. The system applies a face recognition technology on authority management, thereby ensuring the privacy and safety of system information and preventing persons without authority from invading the system.
Drawings
FIG. 1 is a block diagram showing the structure of the monitoring system of the present invention.
FIG. 2 is a flow chart of a method of use of the monitoring system of the present invention.
Detailed Description
The invention is further described below, but not limited to, with reference to the following figures and examples.
Example (b):
the technical scheme of the invention is as follows:
the system comprises a face acquisition module, a cloud processing module, a video acquisition module, a central processing module, a wireless transmission module and an alarm module: the face acquisition module is a camera equipped for the central processing module and can acquire face information; the central processing module is connected with the wireless transmission module and is communicated with the cloud processing module through the wireless transmission module, the cloud processing module comprises an established face database, the central processing module performs data transmission through the wireless transmission module and enters the face database according to the IP address of the face database of the cloud processing module to perform data processing; a video acquisition module with a wireless transmission module is configured on a construction site, acquired real-time information is transmitted to a central processing module in a wireless mode, the central processing module judges whether a person wears a safety helmet or not and whether fire risks exist in the construction site or not according to the acquired real-time information, if the fire risks exist, an alarm is triggered, alarm information is transmitted to an alarm module through an mqtt protocol in the wireless transmission module, and a system is triggered to alarm; the working method of the system comprises the following steps:
1) The face acquisition module acquires face information of a user and extracts facial features of the user;
2) Uploading the face information of the user to a cloud processing module, comparing the face information with a face database established at the cloud, determining the identity information of the user, judging the authority of the user, and allowing the user to enter a safety helmet and a smoke and fire safety monitoring interface of a central processing module of a safety monitoring system for subsequent monitoring if the user is successfully matched with the database; if the matching fails, the user is not allowed to enter the safety monitoring system and feeds the information back to the face acquisition module to wait for the next recognition;
the face recognition part adopts a face recognition algorithm based on opencv, and the steps are as follows:
(1) The central processing module performs frame reading on the face data set, performs heterogeneous stretching processing on an original image by adopting a histogram equalization method to expand the pixel value interval, homogenizes the pixel quantity of each gray scale range through an accumulation function, enhances the image contrast and improves the face recognition accuracy;
(2) Adopting a projection algorithm based on Linear Discriminant Analysis (LDA) to reduce dimension of the face picture, and assuming that the existing sample data matrix X = [ X ] 1 ,x 2 ,...,x m ]Wherein the sample x i For any n-dimensional vector, divide X into k classes of samples, i.e. { C 1 ,C 2 ,…,C k J, i ∈ 1, 2., k; definition of N j 、X j And mu j Are respectively j-th type samples C j The number, the set and the mean value of the LDA, j belongs to 1, 2.. K, the LDA dimension reduction operation comprises the following four steps,
(1) calculating an intra-class divergence matrix S b For indicating the degree of scatter of data points within each class,
wherein (.) T Representing a transpose;
(2) calculating an inter-class divergence matrix S w For representing the course of clustering data points within various classesThe degree of the magnetic field is measured,
(3) computing matrices
Performing characteristic decomposition on the characteristic vector, and calculating characteristic vectors corresponding to the largest d characteristic values to form a matrix W;
(4) calculating the data point Y = W after projection according to the sample matrix X T X, completing the dimensionality reduction of the data;
(3) The Principal Component Analysis (PCA) algorithm is adopted to find the main characteristic points of the image, and the method comprises the following five steps,
(1) data preprocessing: preparing a face data set, forming an N multiplied by M dimensional matrix after linear identification (LDA) algorithm, representing a face image as a pixel matrix when processing, and representing the pixel value of one face image by each line of the matrix;
(2) calculating the average value of the face pixels: calculating the average value of each column to finally obtain an M-dimensional row vector, and generating an average pixel value of the face image, namely an average face;
(3) sample centralization: subtracting the average face from each row of the original data matrix, and then transposing, wherein each transposed column contains pixel information specific to each face;
(4) calculating the characteristic quantity of the face pixels: calculating an eigenvalue and an eigenvector of the covariance matrix, and forming an eigenface by the eigenvector matrix;
(5) the face recognition is carried out through a camera shooting function integrated by the central processing module so as to confirm the identity of the user;
3) After the user identity is confirmed, the user enters a safety helmet and a smoke and fire monitoring interface of the safety monitoring system, the central processing module calls a video acquisition module in a construction site, and picture video information acquired by a monitoring camera in real time is received through the wireless transmission module;
4) The potential safety hazard judgment comprises safety helmet detection and smoke and fire detection, a detection target is identified through the behavior and the posture of an object in a picture by using a corresponding identification algorithm through a central processing module according to picture video information collected by a video collecting module, and whether the potential safety hazard exists in the picture video information is judged through the safety helmet and smoke and fire characteristics led into a model;
the safety cap and firework detection part adopts a Yolov5 network-based training detection algorithm, firstly, a Yolov5 network (Yolov 5s, yolov5m, yolov5l, yolov5x network) is used for training a target detection model, and the target detection of the safety cap and firework is completed through the generated target detection model, wherein the Yolov5 network comprises four parts of an input end, a backhaul, a Neck and an output end:
(1) the input end comprises Mosaic data enhancement, self-adaptive anchor frame calculation and self-adaptive picture scaling, wherein the Mosaic data enhancement is mainly used for solving the problem of small target detection and enriching a data set;
(2) the Backbone part adopts a Focus structure and a cross-phase local network (CSPnet) structure, an original 608 multiplied by 3 image is input into the Focus structure, a slicing operation is carried out to change the original 608 multiplied by 3 image into a feature map of 304 multiplied by 12, and the original 608 multiplied by 3 image is finally changed into a feature map of 304 multiplied by 32 through a convolution operation of 32 convolution kernels;
(3) the method comprises the following steps that a Neck part adopts a characteristic pyramid network (FPN) + bottom-up pyramid network (PAN) structure to carry out up-sampling and down-sampling on a picture, the FPN carries out up-sampling on the picture from top to bottom, the FPN structure fuses high-level characteristics of the picture with low-level characteristics to obtain a predicted characteristic diagram, the PAN structure adds a pyramid network from bottom to top behind the FPN structure to supplement the FPN structure and transmit positioning characteristics of the low level upwards, the PAN structure firstly copies the bottom level of the FPN structure as the bottom level of a new characteristic pyramid and carries out down-sampling operation on the bottom level of the new characteristic pyramid, then the second layer from the bottom to the top of the original characteristic pyramid carries out 3 x 3 convolution calculation with the step of 2, and finally, the PAN structure is transversely connected with the bottom level of the new characteristic pyramid after the down-sampling operation, the 3 x 3 convolution calculation is carried out to fuse the characteristics of the PAN structure after the sum of the two, and the up-down-sampling refers to the scaling of the picture;
(4) the method comprises the steps that a Bounding box loss function and an nms non-maximum suppression algorithm are adopted at an output end, wherein the Bounding box loss function is the sum of coordinate loss, target confidence coefficient loss and classification loss, the nms non-maximum suppression algorithm is used for screening target boxes, the specific process is that a plurality of candidate boxes are obtained by depending on a target detection classifier (a classifier which learns classification rules by using given classes and known training data and is used for classifying or predicting unknown data), probability values corresponding to all the candidate boxes are ranked according to the class to which the candidate boxes belong, a candidate box A with the highest probability value is selected, the rest candidate boxes are traversed, if the overlapping area of the candidate box B and the current highest candidate box A is more than or equal to a set threshold value, the candidate box B is deleted, if the overlapping area is less than the threshold value, processing is not carried out, then a highest probability value is continuously selected from the unprocessed candidate boxes, the process is repeated, and only an unscreentered candidate box is left after traversing all the candidate boxes, and the probability value of the candidate box is the confidence coefficient value of the target box corresponding to which is the probability value of the target box;
5) Alarming prompt, if the identification result has potential safety hazard, the central processing module triggers an alarm to enable the alarm module to send an alarm short message for alarming or play a voice prompt alarm; if no potential safety hazard exists, the central processing module continues to identify the video information of the subsequent pictures;
the alarm part sets a recognition alarm threshold value, and the setting process is as follows: the central processing module monitors the received video information, respectively obtains confidence degrees corresponding to the safety helmet and the firework monitoring target, if the confidence degree is larger than or equal to an alarm threshold value, the safety hazard is identified, voice prompt is played or an alarm short message is sent to a mobile phone end of an administrator to alarm, if the confidence degree is smaller than the alarm threshold value, no safety hazard exists, and next video information is monitored, wherein the confidence degree indicates the confidence degree of the detection target in the framed square and the confidence degree that the framed square includes all the characteristics of the whole object.
Claims (1)
1. The system comprises a face acquisition module, a cloud processing module, a video acquisition module, a central processing module, a wireless transmission module and an alarm module: the face acquisition module is a camera equipped for the central processing module and can acquire face information; the central processing module is connected with the wireless transmission module and is communicated with the cloud processing module through the wireless transmission module, the cloud processing module comprises an established face database, the central processing module performs data transmission through the wireless transmission module and enters the face database according to the IP address of the face database of the cloud processing module to perform data processing; a video acquisition module with a wireless transmission module is configured on a construction site, acquired real-time information is transmitted to a central processing module in a wireless mode, the central processing module judges whether a person wears a safety helmet or not and whether fire risks exist in the construction site or not according to the acquired real-time information, if the fire risks exist, an alarm is triggered, alarm information is transmitted to an alarm module through an mqtt protocol in the wireless transmission module, and a system is triggered to alarm; the working method of the system comprises the following steps:
1) The face acquisition module acquires face information of a user and extracts facial features of the user;
2) Uploading the face information of the user to a cloud processing module, comparing the face information with a face database established at the cloud, determining the identity information of the user, judging the authority of the user, and allowing the user to enter a safety helmet and a smoke and fire safety monitoring interface of a central processing module of a safety monitoring system for subsequent monitoring if the user is successfully matched with the database; if the matching fails, the user is not allowed to enter the safety monitoring system and feeds the information back to the face acquisition module to wait for the next recognition;
the face recognition part adopts a face recognition algorithm based on opencv, and the steps are as follows:
(1) The central processing module performs frame reading on the face data set, performs heterogeneous stretching processing on an original image by adopting a histogram equalization method to expand the pixel value interval, homogenizes the pixel quantity of each gray scale range through an accumulation function, enhances the image contrast and improves the face recognition accuracy;
(2) Adopting a linear identification projection algorithm to reduce the dimension of the face picture, and assuming that the existing sample data matrix X = [ X ] 1 ,x 2 ,...,x m ]Wherein the sample x i For any n-dimensional vector, divide X into k classes of samples, i.e. { C 1 ,C 2 ,…,C k J, i ∈ 1, 2., k; definition of N j 、X j And mu j Are respectively j type samples C j The number, set and mean value of j ∈ 1, 2.., k, the dimensionality reduction operation comprises the following four steps,
(1) calculating an intra-class divergence matrix S b For indicating the degree of scatter of data points within each class,
wherein (.) T Representing a transpose;
(2) calculating an inter-class divergence matrix S w For indicating the degree of aggregation of data points within each class,
(3) computing matrices
Performing characteristic decomposition on the characteristic vector, and calculating characteristic vectors corresponding to the largest d characteristic values to form a matrix W;
(4) calculating the data point Y = W after projection according to the sample matrix X T X, finishing the dimensionality reduction of the data;
(3) Searching the main characteristic points of the image by adopting a principal component analysis algorithm comprises the following five steps,
(1) data preprocessing: preparing a face data set, forming an N multiplied by M dimensional matrix after a linear identification algorithm, representing a face image as a pixel matrix during processing, and representing the pixel value of one face image by each line of the matrix;
(2) calculating the average value of the face pixels: calculating the average value of each column to finally obtain an M-dimensional row vector, and generating an average pixel value of the human face image, namely an average face;
(3) sample centralization: subtracting the average face from each row of the original data matrix, and transposing, wherein each transposed column contains pixel information specific to each face;
(4) calculating the characteristic quantity of the face pixels: calculating an eigenvalue and an eigenvector of the covariance matrix, and forming an eigenface by the eigenvector matrix;
(5) the face recognition is carried out through a camera shooting function integrated by the central processing module so as to confirm the identity of the user;
3) After the user identity is confirmed, the user enters a safety helmet and a smoke and fire monitoring interface of the safety monitoring system, the central processing module calls a video acquisition module in a construction site, and picture video information acquired by a monitoring camera in real time is received through the wireless transmission module;
4) The potential safety hazard judgment comprises safety helmet detection and smoke and fire detection, a detection target is identified through the behavior and the posture of an object in a picture by using a corresponding identification algorithm through a central processing module according to picture video information collected by a video collecting module, and whether the potential safety hazard exists in the picture video information is judged through the safety helmet and smoke and fire characteristics led into a model;
the safety helmet and firework detection part adopts a Yolov5 network-based training detection algorithm, firstly, a Yolov5 network is used for training a target detection model, and target detection on the safety helmet and firework is completed through the generated target detection model, wherein the Yolov5 network consists of four parts, namely an input end, a backhaul, a Neck and an output end:
(1) the input end comprises Mosaic data enhancement, self-adaptive anchor frame calculation and self-adaptive picture scaling, wherein the Mosaic data enhancement is mainly used for solving the problem of small target detection and enriching a data set;
(2) the Backbone part adopts a Focus structure and a cross-stage local network structure, an original 608 × 608 × 3 image is input into the Focus structure, a slicing operation is performed to change the original 608 × 608 × 3 image into a 304 × 304 × 12 feature map, and the original 608 × 608 × 3 image is subjected to a convolution operation of 32 convolution kernels to finally change the original 608 × 304 × 32 feature map into a 304 × 304 × 32 feature map;
(3) the method comprises the following steps that a Neck part adopts a characteristic pyramid network, namely an FPN + bottom-up pyramid network, namely a PAN structure to carry out up-sampling and down-sampling on a picture, the FPN carries out up-sampling on the picture from top to bottom, the FPN structure carries out up-sampling on the picture, the high-level characteristics of the picture are fused through the up-sampling and the low-level characteristics to obtain a predicted characteristic diagram, the PAN structure adds a pyramid network from bottom to top behind the FPN structure to supplement the FPN structure and transmit the positioning characteristics of the low level upwards, the PAN structure firstly copies the bottom level of the FPN structure to serve as the bottom level of a new characteristic pyramid and carries out down-sampling operation on the bottom level of the new characteristic pyramid, then the second-last level of the original characteristic pyramid carries out 3 x 3 convolution calculation, the step is 2, and finally, the PAN structure is transversely connected with the bottom level of the new characteristic pyramid after the down-sampling operation, and then carries out 3 x 3 convolution calculation to fuse the characteristics after the addition, and the up-down sampling refers to the scaling of the picture;
(4) the method comprises the steps that a Bounding box loss function and an nms non-maximum value suppression algorithm are adopted at an output end, wherein the Bounding box loss function is the sum of coordinate loss, target confidence loss and classification loss, the nms non-maximum value suppression algorithm is used for completing screening of target frames, the specific process includes that a plurality of candidate frames are obtained by means of a target detection classifier, probability values corresponding to all the candidate frames are ranked according to probability values of the attribution categories of the candidate frames, a candidate frame A with the highest probability value is selected, other candidate frames are traversed, if the overlapping area of the candidate frame B and the candidate frame A with the highest current probability value is larger than or equal to a set threshold value, the candidate frame B is deleted, if the overlapping area is smaller than the threshold value, no processing is carried out, then a candidate frame with the highest probability value is selected from the unprocessed candidate frames, the process is repeated, only one candidate frame which is not screened is left after all the candidate frames are traversed, and the probability value of the candidate frame is the confidence coefficient of a target corresponding to the candidate frame;
5) Alarming prompt, if the identification result has potential safety hazard, the central processing module triggers an alarm to enable the alarm module to send an alarm short message for alarming or play a voice prompt alarm; if no potential safety hazard exists, the central processing module continues to identify the video information of the subsequent pictures;
the alarm part sets a recognition alarm threshold value, and the setting process is as follows: the central processing module monitors the received video information, respectively obtains confidence degrees corresponding to the safety helmet and the firework monitoring target, if the confidence degree is larger than or equal to an alarm threshold value, the safety hazard is identified, voice prompt is played or an alarm short message is sent to a mobile phone end of an administrator to alarm, if the confidence degree is smaller than the alarm threshold value, no safety hazard exists, and next video information is monitored, wherein the confidence degree indicates the confidence degree of the detection target in the framed square and the confidence degree that the framed square includes all the characteristics of the whole object.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210671538.XA CN115170059A (en) | 2022-06-14 | 2022-06-14 | Intelligent safety monitoring system for outdoor construction site and working method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210671538.XA CN115170059A (en) | 2022-06-14 | 2022-06-14 | Intelligent safety monitoring system for outdoor construction site and working method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115170059A true CN115170059A (en) | 2022-10-11 |
Family
ID=83485679
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210671538.XA Pending CN115170059A (en) | 2022-06-14 | 2022-06-14 | Intelligent safety monitoring system for outdoor construction site and working method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115170059A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116449752A (en) * | 2023-04-11 | 2023-07-18 | 国电建投内蒙古能源有限公司 | Intelligent monitoring alarm system of power plant |
| CN117523773A (en) * | 2023-11-23 | 2024-02-06 | 江苏南北木屋文化科技有限公司 | Intelligent log cabin anomaly detection method and system |
-
2022
- 2022-06-14 CN CN202210671538.XA patent/CN115170059A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116449752A (en) * | 2023-04-11 | 2023-07-18 | 国电建投内蒙古能源有限公司 | Intelligent monitoring alarm system of power plant |
| CN117523773A (en) * | 2023-11-23 | 2024-02-06 | 江苏南北木屋文化科技有限公司 | Intelligent log cabin anomaly detection method and system |
| CN117523773B (en) * | 2023-11-23 | 2024-05-10 | 江苏南北木屋文化科技有限公司 | Intelligent log cabin anomaly detection method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhang et al. | Edge video analytics for public safety: A review | |
| CN109819208B (en) | Intensive population security monitoring management method based on artificial intelligence dynamic monitoring | |
| CN109670446B (en) | Abnormal behavior detection method based on linear dynamic system and deep network | |
| US11735018B2 (en) | Security system with face recognition | |
| CN108009473B (en) | Video structuralization processing method, system and storage device based on target behavior attribute | |
| CN109271554B (en) | Intelligent video identification system and application thereof | |
| CN108062349B (en) | Video monitoring method and system based on video structured data and deep learning | |
| CN108052859B (en) | Abnormal behavior detection method, system and device based on clustering optical flow characteristics | |
| CN110298278B (en) | Underground parking garage pedestrian and vehicle monitoring method based on artificial intelligence | |
| US10366595B2 (en) | Surveillance method and system based on human behavior recognition | |
| KR101215948B1 (en) | Image information masking method of monitoring system based on face recognition and body information | |
| CN108647582A (en) | Goal behavior identification and prediction technique under a kind of complex dynamic environment | |
| CN110705500B (en) | Attention detection method and system for personnel working image based on deep learning | |
| CN111814638B (en) | Security scene flame detection method based on deep learning | |
| CN111127830A (en) | Alarm method, alarm system and readable storage medium based on monitoring equipment | |
| KR102149832B1 (en) | Automated Violence Detecting System based on Deep Learning | |
| Yimyam et al. | Face detection criminals through CCTV cameras | |
| CN115170059A (en) | Intelligent safety monitoring system for outdoor construction site and working method | |
| CN109190475A (en) | A kind of recognition of face network and pedestrian identify network cooperating training method again | |
| CN107330414A (en) | Act of violence monitoring method | |
| CN116994390A (en) | Security monitoring system and method based on Internet of things | |
| CN112183472A (en) | Method for detecting whether test field personnel wear work clothes or not based on improved RetinaNet | |
| CN115620471A (en) | Image identification security system based on big data screening | |
| CN111160149A (en) | Vehicle-mounted face recognition system and method based on motion scene and deep learning | |
| US10939120B1 (en) | Video upload in limited bandwidth |
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 |