A construction site monitoring system based on artificial intelligence

Abstract

The present application discloses an artificial intelligence-based construction site monitoring system, which relates to the technical field of safety identification systems and solves the technical problem that the working status of staff members and some other dangerous situations in the construction area are not fully considered for identification, resulting in low efficiency of safety monitoring of the construction area; a number of project labels are obtained by inputting the preprocessed monitoring image into a label recognition model; an image anomaly coefficient corresponding to the monitoring image is generated according to the project label; a work alarm signal is generated according to the image anomaly coefficient; an environmental anomaly coefficient is generated according to the preprocessed monitoring data; an environmental anomaly alarm signal is generated according to the environmental anomaly coefficient; monitoring videos and monitoring data are packaged and uploaded according to the image anomaly coefficient and the environmental anomaly coefficient, thereby increasing the working status of personnel monitoring the construction site, dynamically adjusting the content size and upload time of the data packet, and improving the efficiency of safety monitoring in the construction area and the feedback efficiency in emergency situations.

Description

Technical Field

[0001] The present application belongs to the technical field of security identification systems, and relates to security identification system technology, specifically a construction site monitoring system based on artificial intelligence. Background Art

[0002] The AI-based security identification system is different from the traditional security system. It not only relies on video surveillance, but also takes deep neural network computer vision AI technology as its core. It uses machine vision to replace human visual supervision, truly liberating manpower and providing 24-hour seamless supervision without blind spots. It greatly saves human resources while making security monitoring methods more efficient and diversified.

[0003] The prior art (the invention patent application with publication number CN116193083A) discloses a safety identification system based on artificial intelligence, which relates to the technical field of safety identification systems. The safety identification system based on artificial intelligence includes a construction site camera node, an image recognition control terminal, an intelligent monitoring and identification module, an information entry management module and a construction site management module. The construction site camera node is connected to the image recognition control terminal to monitor the construction site operation scene in the entire area through multi-node construction site cameras, and collects monitoring information on the wearing of safety helmets and releases alarm information through the image recognition control terminal. By using the construction site camera node, image recognition control terminal, intelligent monitoring and identification module, information entry management module and construction site management module to cooperate with each other, the wearing of safety helmets on the construction site and the handling of emergency events are realized, and at the same time, artificial intelligence is used to liberate the labor of construction monitoring, effectively ensuring safety and construction efficiency.

[0004] The above system uses image recognition technology to determine whether the staff are wearing safety helmets, and promptly reminds relevant personnel to wear safety helmets to ensure the safety of personnel in the construction area. The above system is difficult to identify the working status of the staff and some other dangerous situations in the construction area, resulting in low efficiency in safety monitoring of the construction area. Therefore, a construction site monitoring system based on artificial intelligence is needed. Summary of the invention

[0005] The present application aims to solve at least one of the technical problems existing in the prior art; to this end, the present application proposes an artificial intelligence-based construction site monitoring system to solve the technical problem that the prior art does not fully consider the working status of the staff and some other dangerous situations in the construction area, resulting in low efficiency in safety monitoring of the construction area.

[0006] To achieve the above-mentioned purpose, the first aspect of the present application provides a construction site monitoring system based on artificial intelligence, including: a function expansion module, a data acquisition module, a data analysis module, a data upload module, an early warning module and a data storage module;

[0007] The function expansion module is provided with a number of expansion interfaces of different models for connecting to the expanded data acquisition equipment;

[0008] The data acquisition module: obtains the construction monitoring video and real-time monitoring image of the construction area through the video monitoring equipment connected to it; obtains the monitoring data through the data acquisition equipment connected to the function expansion module connected to it; and pre-processes the monitoring image and monitoring data;

[0009] The data analysis module: obtains the pre-processed monitoring image; inputs the monitoring image into the label recognition model to obtain a number of project labels; generates the image abnormality coefficient corresponding to the monitoring image according to the project label; generates a work alarm signal according to the image abnormality coefficient; the label recognition model is obtained by training the artificial intelligence model; and,

[0010] Acquire the pre-processed monitoring data, generate an environmental anomaly coefficient according to the monitoring data; generate an environmental anomaly alarm signal according to the environmental anomaly coefficient;

[0011] The data uploading module is used to obtain an image anomaly coefficient and an environmental anomaly coefficient, and to package and upload the monitoring video and monitoring data according to the image anomaly coefficient and the environmental anomaly coefficient;

[0012] The early warning module sends out corresponding alarm prompts according to the alarm signal and the early warning signal, and contacts the management personnel according to the alarm signal.

[0013] The present application obtains monitoring images and monitoring data through corresponding data acquisition equipment, and preprocesses them; inputs the preprocessed monitoring images into a label recognition model to obtain a number of project labels; generates an image anomaly coefficient corresponding to the monitoring image according to the project label; generates a work alarm signal according to the image anomaly coefficient; and generates an environmental anomaly coefficient according to the preprocessed monitoring data; generates an environmental anomaly alarm signal according to the environmental anomaly coefficient; and packages the monitoring video and monitoring data for upload according to the image anomaly coefficient and the environmental anomaly coefficient, thereby increasing the working status of personnel monitoring the construction site, dynamically adjusting the content size and upload time of the data packet, and improving the efficiency of safety monitoring in the construction area and the feedback efficiency in emergency situations.

[0014] Furthermore, the preprocessing includes: monitoring image preprocessing and monitoring data preprocessing.

[0015] Furthermore, the label recognition model is obtained by training an artificial intelligence model, including:

[0016] A number of historical monitoring images and a number of monitoring items, as well as their corresponding project labels, are obtained through the data storage module; the monitoring items include monitoring identification items; the project labels corresponding to the monitoring identification items are person monitoring or scene monitoring; the monitoring identification items also include ID identification items and a number of dangerous monitoring items; the project labels corresponding to the ID identification items are the corresponding personnel IDs; the ID identification items include a number of behavior identification items and attribute identification items; the project labels corresponding to the behavior identification items, attribute identification items and dangerous monitoring items include dangerous or normal;

[0017] Several historical monitoring images are divided according to monitoring projects, and each monitoring project is divided into corresponding training data, verification data and test data according to the project label

[0018] Preprocess the training data, verification data and test data to obtain the training set, verification set and test set;

[0019] The artificial intelligence model ResNet is selected as the backbone network, and multi-task branches are constructed, including monitoring and recognition branches, ID recognition branches, attribute recognition branches, behavior recognition branches, and danger monitoring branches, so as to construct the basic network model; the monitoring and recognition branch is connected with the ID recognition branch and the danger monitoring branch, and the ID recognition branch is connected with several attribute recognition branches and behavior recognition branches;

[0020] Use the corresponding training set to train each branch of the basic model, adjust the parameters on the validation set, and test the model on the test data. Finally, we get a label recognition model that takes monitoring images as input and outputs several project labels.

[0021] Furthermore, generating the image anomaly coefficient corresponding to the monitoring image according to a plurality of project labels includes:

[0022] Get the project labels corresponding to each monitoring project;

[0023] Determine whether the project label corresponding to the monitoring and identification project is person monitoring;

[0024] Yes, obtain the personnel ID of each item label corresponding to the ID identification item; and the item label corresponding to each behavior identification item and attribute identification item corresponding to each personnel ID; extract the item danger level YDu of several behavior identification items and attribute identification items whose item label is dangerous;

[0025] The abnormal coefficient of each personnel ID is calculated by the formula RYXi=∑YDu; the abnormal coefficient of personnel is calculated by the formula RYX=∑RYXi;

[0026] No, obtain several hazardous monitoring items with item labels as hazardous and their corresponding item hazard levels YDj;

[0027] The scenario anomaly coefficient CYX is calculated by the formula, CYX=∑YDj; where YDj represents the project danger level of the jth project label, j∈[1,P], and P represents the total number of danger monitoring project labels;

[0028] The image abnormality coefficient includes a personnel abnormality coefficient or a scene abnormality coefficient.

[0029] Further, generating the working alarm signal according to the image abnormality coefficient includes:

[0030] Obtain image abnormality coefficient;

[0031] When the image abnormality coefficient is a personnel abnormality coefficient: determine whether the image abnormality coefficient is greater than the personnel alarm threshold; if yes, generate a personnel abnormality alarm signal; if no, generate a personnel abnormality warning signal;

[0032] When the image anomaly coefficient is a scene anomaly coefficient: determine whether the image anomaly coefficient is greater than the scene alarm threshold; if so, generate a scene anomaly alarm signal; if not, generate a scene anomaly warning signal.

[0033] Further, generating the environmental anomaly coefficient according to the monitoring data includes:

[0034] Obtain the corresponding data change amount Bk of several monitoring data in unit time and the important proportion ZBk of several monitoring data;

[0035] Calculate the abnormal coefficient YXk corresponding to several monitoring data through the formula;

[0036] YXk=β×arctan(|Bk / DBk|);

[0037] The environmental anomaly coefficient is calculated using the formula HYX=∑(ZBk×YXk); where β is the adjustment factor and DBk is the corresponding monitoring unit data.

[0038] Further, generating the environmental abnormality alarm signal according to the environmental abnormality coefficient includes:

[0039] Obtain environmental anomaly coefficient;

[0040] Determine whether the environmental anomaly coefficient is greater than the environmental alarm threshold:

[0041] Yes, generate an environmental abnormality alarm signal;

[0042] No, obtain the abnormal coefficients corresponding to the several monitoring data; when the abnormal coefficients corresponding to the several monitoring data are greater than their corresponding abnormal thresholds; generate the corresponding monitoring data abnormal alarm signals.

[0043] Furthermore, the monitoring video and monitoring data are packaged and uploaded according to the image anomaly coefficient and the environmental anomaly coefficient, including:

[0044] Obtain the image anomaly coefficient TYX and the environment anomaly coefficient HYX; obtain the standard interval durations of normal data packet upload under monitoring image, scene monitoring and monitoring data, which are BDSr, BDSc and BDSs respectively;

[0045] When the monitoring image is for human monitoring, the interval duration is calculated by the formula DSr=BDSr×exp(-(b1×TYX)), and the corresponding supplementary duration is equal to the interval duration;

[0046] When the monitoring image is scene monitoring, the interval duration is calculated by the formula DSc=BDSc×exp(-(b2×TYX)), and the corresponding supplementary duration is equal to the interval duration;

[0047] The interval duration under the monitoring data is calculated by the formula DSs=BDSs×exp(-(b3×HYX)), and the corresponding supplementary duration is equal to the interval duration, where b1, b2 and b3 are proportional coefficients.

[0048] Packing the monitoring video within the interval time before the real-time monitoring image corresponding to the image abnormality coefficient TYX into an abnormal reference data packet; after the supplementary time, packing the monitoring video within the interval supplementary time after the real-time monitoring image corresponding to the image abnormality coefficient TYX into an abnormal supplementary data packet; generating the transmission priority coefficient of the corresponding abnormal reference data packet or abnormal supplementary data packet according to the image abnormality coefficient TYX;

[0049] The monitoring data within the interval time before the real-time monitoring data corresponding to the environmental anomaly coefficient HYX are packaged into an abnormal reference data packet; after the supplementary time period has passed, the monitoring data within the interval supplementary time after the real-time monitoring data corresponding to the environmental anomaly coefficient HYX are packaged into an abnormal supplementary data packet; a transmission priority coefficient of a corresponding abnormal reference data packet or an abnormal supplementary data packet is generated according to the environmental anomaly coefficient HYX; and the corresponding abnormal reference data packet or abnormal supplementary data packet is uploaded according to the transmission priority coefficient.

[0050] This application obtains the standard interval duration of normal uploaded data packets under monitoring images, scene labels, and monitoring data, as well as image anomaly coefficients and environmental anomaly coefficients; calculates the interval duration and supplementary duration based on the anomaly coefficients and their corresponding standard interval durations, and adjusts the size of the corresponding uploaded data packets through the dynamic changes of the anomaly coefficients, thereby improving the feedback efficiency of emergency processing under construction conditions.

[0051] Furthermore, the transmission priority coefficient of the corresponding abnormal reference data packet and the transmission priority coefficient of the abnormal supplementary data packet are generated according to the image abnormality coefficient TYX, including:

[0052] Obtain image abnormality coefficient TYX;

[0053] Determine whether the image abnormality coefficient is the personnel abnormality coefficient:

[0054] Yes, the transmission priority coefficient of the abnormal reference data packet is calculated by the formula CYX=g1×(1 / (1+exp(-TYX))), and the transmission priority coefficient of the abnormal supplementary data packet is calculated by the formula BYX=g2×(1 / (1+exp(-TYX)));

[0055] No, the transmission priority coefficient of the abnormal reference data packet is calculated by the formula CYX=g3×(1 / (1+exp(-TYX))), and the transmission priority coefficient of the abnormal supplementary data packet is calculated by the formula BYX=g4×(1 / (1+exp(-TYX))), where g1, g2, g3 and g4 are proportional coefficients, and g1>g2, g3>g4.

[0056] Further, the transmission priority coefficient of the corresponding abnormal reference data packet and the transmission priority coefficient of the abnormal supplementary data packet are generated according to the environmental abnormality coefficient HYX, including:

[0057] Get the environmental anomaly coefficient HYX;

[0058] The transmission priority coefficient of the abnormal reference data packet is calculated by the formula CYX=g5×(1 / (1+exp(-HYX))), and the transmission priority coefficient of the abnormal supplementary data packet is calculated by the formula BYX=g6×(1 / (1+exp(-HYX))), where g5 and g6 are proportional coefficients, and g5>g6.

[0059] Further, uploading the corresponding abnormal reference data packet or abnormal supplementary data packet according to the transmission priority coefficient includes:

[0060] When an abnormal reference data packet or an abnormal supplementary data packet is generated, it is detected whether there are other data packets to be uploaded at this time;

[0061] If yes, then obtain the transmission priority coefficients corresponding to other data packets; when the transmission priority coefficients corresponding to other data packets are greater than the transmission priority coefficients of the abnormal reference data packets or abnormal supplementary data packets that need to be transmitted this time, wait for their transmission to be completed before transmitting the abnormal reference data packet or abnormal supplementary data packet; otherwise, obtain their corresponding transmission progress; when the transmission progress is less than the progress threshold, terminate the transmission of other abnormal reference data packets or abnormal supplementary data packets; immediately upload the abnormal reference data packet or abnormal supplementary data packet; otherwise, wait for the transmission of other abnormal reference data packets or abnormal supplementary data packets to be completed, and then immediately upload the abnormal reference data packet or abnormal supplementary data packet;

[0062] No, directly upload the abnormal reference data package or abnormal supplementary data package.

[0063] The present application calculates the transmission priority coefficient of each abnormal reference data packet or abnormal supplementary data packet, compares the priority coefficients of other abnormal reference data packets or abnormal supplementary data packets with the current abnormal reference data packet or abnormal supplementary data packet, and adaptively adjusts the data packet upload method according to the priority coefficient and transmission progress, thereby improving the feedback efficiency of emergency processing under construction conditions.

[0064] Compared with the prior art, the beneficial effects of this application are:

[0065] 1. The present application obtains monitoring images and monitoring data through corresponding data acquisition equipment, and preprocesses them; inputs the preprocessed monitoring images into a label recognition model to obtain a number of project labels; generates an image anomaly coefficient corresponding to the monitoring image according to the project label; generates a work alarm signal according to the image anomaly coefficient; and generates an environmental anomaly coefficient according to the preprocessed monitoring data; generates an environmental anomaly alarm signal according to the environmental anomaly coefficient; and packages the monitoring video and monitoring data according to the image anomaly coefficient and the environmental anomaly coefficient for uploading, thereby increasing the working status of personnel monitoring the construction site, dynamically adjusting the content size and upload time of the data packet, and improving the efficiency of safety monitoring in the construction area and the feedback efficiency in emergency situations.

[0066] 2. This application obtains the standard interval duration of normal upload data packets under monitoring images, scene labels, and monitoring data, as well as image anomaly coefficients and environmental anomaly coefficients; calculates the interval duration and supplementary duration based on the anomaly coefficient and its corresponding standard interval duration, and adjusts the size of the corresponding uploaded data packet through the dynamic change of the anomaly coefficient, thereby improving the upload speed of the data packet.

[0067] 3. The present application calculates the transmission priority coefficient of each abnormal reference data packet or abnormal supplementary data packet, compares the priority coefficients of other abnormal reference data packets or abnormal supplementary data packets with the current abnormal reference data packet or abnormal supplementary data packet, and adaptively adjusts the data packet upload method according to the priority coefficient and transmission progress. By comparing the transmission priority coefficient and transmission progress of the data packet, the efficiency of data packet transmission in dangerous emergency situations is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0069] Figure 1 This is a schematic diagram of the principle of a construction site monitoring system based on artificial intelligence in this application;

[0070] Figure 2 A method flow chart of a construction site monitoring system based on artificial intelligence in this application;

[0071] Figure 3 This is a flow chart of generating an alarm signal according to an abnormal coefficient of the present application. DETAILED DESCRIPTION

[0072] The technical solution of the present application will be described clearly and completely below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present application.

[0073] See also Figure 1-Figure 2 , the first aspect of the present application provides an artificial intelligence-based construction site monitoring system, including: a function expansion module, a data acquisition module, a data analysis module, a data upload module, an early warning module, a data storage module, a self-organizing network module, an autonomous power supply module and a remote management module;

[0074] Functional expansion module: There are several expansion interfaces of different models for connecting the expanded data acquisition devices; the data acquisition devices include temperature acquisition devices, rainfall acquisition devices and wind speed acquisition devices, etc. Specifically, when the data acquisition device is connected to the expansion interface, select the corresponding number of expansion interfaces in the drop-down box of the data acquisition device interface in the visual interface, and click the device add button, the system automatically identifies the data acquisition device type and allocates data storage space;

[0075] Data acquisition module: obtains construction monitoring videos and real-time monitoring images of the construction area through the video monitoring equipment connected to it; the video monitoring equipment can be ultra-clear cameras, etc.; obtains monitoring data through the data acquisition equipment connected to the function expansion module connected to it; when the function expansion module is connected to a single data acquisition device, its corresponding monitoring data is the data collected by the corresponding data acquisition device; if the data acquisition device is a temperature acquisition device, its corresponding monitoring data is temperature data; when the function expansion module is connected to multiple data acquisition devices, the monitoring data includes the corresponding data collected by each data acquisition device; if the data acquisition device is a temperature acquisition device and an air pressure acquisition device, the monitoring data includes temperature data and air pressure data; and pre-processes the monitoring image and monitoring data;

[0076] Data analysis module: obtain preprocessed monitoring images; input the monitoring images into the label recognition model to obtain several project labels; generate image anomaly coefficients corresponding to the monitoring images according to the project labels; the image anomaly coefficient refers to the degree to which the project in the monitoring image affects the construction, and the image anomaly coefficient is a personnel anomaly coefficient or a scene anomaly coefficient; generate work alarm signals according to the image anomaly coefficient; the work alarm signals include personnel anomaly alarm signals, personnel anomaly warning signals, scene anomaly alarm signals and scene anomaly warning signals; the label recognition model is obtained through training of the artificial intelligence model; and,

[0077] Obtaining pre-processed monitoring data, including temperature data, rainfall data, and wind speed data; generating an environmental anomaly coefficient based on the monitoring data; the environmental anomaly coefficient refers to the degree to which the monitoring data affects the construction; generating an environmental anomaly alarm signal based on the environmental anomaly coefficient;

[0078] Data upload module: obtains image anomaly coefficients and environmental anomaly coefficients, and packages and uploads monitoring videos and monitoring data according to the image anomaly coefficients and environmental anomaly coefficients;

[0079] Early warning module: issues corresponding alarm prompts according to alarm signals and early warning signals, and contacts management personnel according to alarm signals;

[0080] Self-organizing network module: using domestic advanced 4G / 5G technology to create a low-latency and high-efficiency network transmission architecture to achieve real-time on-site image transmission;

[0081] The autonomous power supply module uses a low-light solar power generation system, which can provide power support for the system outdoors even in rainy weather;

[0082] Remote management module, users can log in to the system remotely through the Internet to remotely monitor, collect data and set up sensor devices.

[0083] This embodiment obtains monitoring images and monitoring data through corresponding data acquisition equipment, and preprocesses them; inputs the preprocessed monitoring images into the label recognition model to obtain a number of project labels; generates an image anomaly coefficient corresponding to the monitoring image according to the project label; generates a work alarm signal according to the image anomaly coefficient; and generates an environmental anomaly coefficient according to the preprocessed monitoring data; generates an environmental anomaly alarm signal according to the environmental anomaly coefficient; and packages the monitoring video and monitoring data according to the image anomaly coefficient and the environmental anomaly coefficient for uploading, thereby increasing the working status of personnel monitoring the construction site, dynamically adjusting the content size and upload time of the data packet, and improving the efficiency of safety monitoring in the construction area and the feedback efficiency in emergency situations.

[0084] Preprocessing includes: monitoring image preprocessing and monitoring data preprocessing;

[0085] Monitoring image preprocessing: performing denoising operation on the monitoring image to obtain a denoised image; performing image enhancement operation on the denoised image to achieve monitoring image preprocessing;

[0086] Monitoring data preprocessing includes missing value processing and normalization processing: directly eliminating missing values ​​in monitoring data to obtain normal data; using normalization operation on normal data to implement monitoring data preprocessing.

[0087] In this embodiment, the maximum and minimum normalization method is selected to preprocess the monitoring data: Y=(X-Xmin) / (Xmax-Xmin), where X represents the data to be normalized, Xmax and Xmin represent the maximum and minimum values ​​in the original data respectively, and Y represents the normalized data; for example, the temperature to be normalized is 32 degrees Celsius, the highest temperature in the original data is 50 degrees Celsius, and the lowest temperature is 20 degrees Celsius, then after the maximum and minimum normalization method, Y=(32-20) / (50-20)=0.4.

[0088] The label recognition model is trained through an artificial intelligence model, including:

[0089] A number of historical monitoring images and a number of monitoring projects, as well as their corresponding project labels, are obtained through the data storage module; the monitoring projects include monitoring and identification projects; the project labels corresponding to the monitoring and identification projects are person monitoring or scene monitoring, and can also be "1" or "0"; "1" is person monitoring, indicating that there are people in the monitoring image; "0" is scene monitoring, indicating that there are no people in the picture; the monitoring and identification projects also include ID identification projects and a number of dangerous monitoring projects; the project labels corresponding to the ID identification projects are the corresponding personnel IDs; the ID identification projects include a number of behavior identification projects and attribute identification projects; the project labels corresponding to the behavior identification projects, attribute identification projects and dangerous monitoring projects include dangerous or normal; they can also be "1" and "0"; "1" is dangerous, indicating that the corresponding behavior identification project or attribute identification project is unqualified; "0" is normal, indicating that the corresponding behavior identification project or attribute identification project is qualified; in this embodiment, the behavior identification projects include smoking identification projects and fatigue identification projects; the attribute identification projects include helmet wearing identification projects and reflective clothing identification projects, etc.; the dangerous monitoring projects include item placement monitoring projects and operation process monitoring projects, etc.;

[0090] Several historical monitoring images are divided according to monitoring projects, and each monitoring project is divided into corresponding training data, verification data and test data according to the project label; the training data, verification data and test data are divided in a ratio of 7:2:1;

[0091] Preprocess the training data, verification data and test data to obtain the training set, verification set and test set;

[0092] The artificial intelligence model ResNet is selected as the backbone network, and multi-task branches are constructed, including monitoring and recognition branches, ID recognition branches, attribute recognition branches, behavior recognition branches, and danger monitoring branches, so as to construct the basic network model; the monitoring and recognition branch is connected with the ID recognition branch and the danger monitoring branch, and the ID recognition branch is connected with several attribute recognition branches and behavior recognition branches;

[0093] Use the corresponding training set to train each branch of the basic model, adjust the parameters on the validation set, and test the model on the test data to finally obtain a label recognition model that inputs monitoring images and outputs several project labels; the goal of the training phase is to enable the model to capture the characteristic relationship between monitoring images and project labels, so that the project labels can be accurately output in subsequent stages; the verification phase is to evaluate and optimize the performance of the model, and adjust the parameters or structure of the model according to the verification results, such as modifying the learning rate, adding regularization terms, adjusting the number of network layers, etc.; the label recognition model can efficiently process the input monitoring images and give several corresponding project labels.

[0094] Generate image anomaly coefficients corresponding to monitoring images based on several project labels, including:

[0095] Get the project labels corresponding to each monitoring project;

[0096] Determine whether the project label corresponding to the monitoring and identification project is person monitoring;

[0097] Yes, obtain the personnel ID of each project label corresponding to the ID identification project; and the project label corresponding to each behavior identification project and attribute identification project corresponding to each personnel ID; extract the project danger level YDu of several behavior identification projects and attribute identification projects with dangerous project labels; the project danger level is the degree of influence of the project label on the construction;

[0098] The abnormal coefficient of each personnel ID is calculated by the formula RYXi=∑YDu; the abnormal coefficient of personnel is calculated by the formula RYX=∑RYXi;

[0099] Where YDu represents the project danger level of the u-th project label, u∈[1,n], n represents the total number of attribute project labels and behavior project labels; i is the number of personnel ID, i∈[0,m], m is the total number of personnel; the personnel abnormality coefficient increases with the increase of the number of people and the increase of project labels.

[0100] In this embodiment, a group of project labels are obtained, including ID1 {helmet identification item: dangerous, reflective clothing identification item: normal, smoking identification item: dangerous, fatigue identification item: normal}, ID2 {helmet identification item: normal, reflective clothing identification item: normal, smoking identification item: dangerous, fatigue identification item: dangerous}, and the project danger levels corresponding to the behavior identification items and attribute identification items are set according to experience. In this embodiment, the danger level corresponding to the helmet identification item is set to 10; the danger level corresponding to the reflective clothing identification item is 8; the danger level corresponding to the smoking identification item is 4; the danger level corresponding to the fatigue identification item is 6; therefore, the abnormal coefficient RYX1 of ID1=10+0+4+0=14; the abnormal coefficient RYX2 of ID2=0+0+4+6=10; the personnel abnormal coefficient RYX=RYX1+RYX2=14+10=24; the image abnormal coefficient TYX=RYX=24;

[0101] No, obtain several hazardous monitoring items with item labels as hazardous and their corresponding item hazard levels YDj;

[0102] The scenario anomaly coefficient CYX is calculated by the formula, CYX=∑YDj; where YDj represents the project danger level of the jth project label, j∈[1,P], and P represents the total number of dangerous monitoring project labels; the scenario anomaly coefficient increases with the increase in the number of dangerous project labels, and is related to the project danger level corresponding to the dangerous monitoring project with dangerous project labels;

[0103] The image abnormality coefficient includes a personnel abnormality coefficient or a scene abnormality coefficient.

[0104] See also Figure 3 , generating a working alarm signal according to the image abnormality coefficient, including: obtaining the image abnormality coefficient;

[0105] When the image abnormality coefficient is a personnel abnormality coefficient: determine whether the image abnormality coefficient is greater than the personnel alarm threshold; if yes, generate a personnel abnormality alarm signal; if no, generate a personnel abnormality warning signal; the personnel alarm threshold is set based on experience;

[0106] When the image anomaly coefficient is a scene anomaly coefficient: determine whether the image anomaly coefficient is greater than the scene alarm threshold, which is set based on experience; if yes, generate a scene anomaly alarm signal; if no, generate a scene anomaly warning signal.

[0107] Generate environmental anomaly coefficients based on monitoring data, including:

[0108] Obtain the corresponding data change amount Bk of several monitoring data in unit time and the important proportion ZBk of several monitoring data; the important proportion refers to the importance of several monitoring data to construction safety, which is set according to experience; several monitoring data include temperature data, rainfall data and wind speed data;

[0109] The abnormal coefficient corresponding to several monitoring data is calculated by the formula YXk=β×arctan(|Bk / DBk|);

[0110] The environmental anomaly coefficient is calculated by the formula HYX=∑(ZBk×YXk); wherein β is the adjustment factor, which is set according to experience; DBk is the corresponding monitoring unit data, which is set according to experience; wherein k represents the number of each monitoring data, k∈[0,Q], Q is the total number of monitoring data; the anomaly coefficient corresponding to several monitoring data increases with the increase of the change amount of the corresponding monitoring data; the environmental anomaly coefficient increases with the increase of the anomaly coefficient corresponding to several detection data.

[0111] In this embodiment, a set of monitoring data is obtained, including a temperature change of 5°C, a rainfall change of 3mm, and a wind speed change of 2m / s; β is set according to experience, and in this embodiment, β is set to 0.6, the abnormal coefficient of temperature data YK1=0.6×arctan(|5°C / 1°C|)=0.6×1.37=0.822, and the abnormal coefficient of rainfall data YK2=0.6×arctan(|3mm / 1mm|)=0.6×1.25= 0.75, the anomaly coefficient of wind speed data YK3=0.6×arctan(|(2m / s) / (1m / s)|)=0.6×1.107=0.664; the important proportions of several monitoring data in this embodiment are set to ZB1=0.5, ZB2=0.2, ZB3=0.3 according to experience; the environmental anomaly coefficient HYX is calculated by the formula = 0.5×0.822+0.2×0.75+0.3×0.664=0.760.

[0112] Generating an environmental abnormality alarm signal according to the environmental abnormality coefficient includes: obtaining the environmental abnormality coefficient;

[0113] Determine whether the environmental anomaly coefficient is greater than the environmental alarm threshold, which is set based on experience; if yes, generate an environmental anomaly alarm signal;

[0114] No, obtain the abnormal coefficients corresponding to the several monitoring data; when the abnormal coefficients corresponding to the several monitoring data are greater than the corresponding abnormal thresholds, the abnormal thresholds corresponding to the several monitoring data are set according to experience; and generate abnormal alarm signals for the corresponding monitoring data.

[0115] Pack and upload monitoring videos and monitoring data according to image anomaly coefficients and environmental anomaly coefficients, including:

[0116] Obtain the image anomaly coefficient TYX and the environment anomaly coefficient HYX; obtain the standard interval durations of normal data packet upload under monitoring image, scene monitoring and monitoring data, which are BDSr, BDSc and BDSs respectively;

[0117] When the monitoring image is for human monitoring, the interval duration is calculated by the formula DSr=BDSr×exp(-(b1×TYX)), and the corresponding supplementary duration is equal to the interval duration;

[0118] When the monitoring image is scene monitoring, the interval duration is calculated by the formula DSc=BDSc×exp(-(b2×TYX)), and the corresponding supplementary duration is equal to the interval duration;

[0119] The interval duration under the monitoring data is calculated by the formula DSs=BDSs×exp(-(b3×HYX)), and the corresponding supplementary duration is equal to the interval duration, where b1, b2 and b3 are proportional coefficients and are set according to experience; with the increase of the image anomaly coefficient or the environmental anomaly coefficient, the probability of danger at the construction site will increase. In order to upload the data packet as soon as possible, the interval duration and supplementary duration of the data packet are shortened, so that the size of the data packet is reduced, thereby increasing the upload speed of the data packet; and the impact on construction safety under different anomaly coefficients is different, so b1, b2 and b3 are different.

[0120] Pack the monitoring video within the interval time before the real-time monitoring image corresponding to the image abnormality coefficient TYX into an abnormal reference data packet; after the supplementary time, pack the monitoring video within the interval supplementary time after the real-time monitoring image corresponding to the image abnormality coefficient TYX into an abnormal supplementary data packet; generate the transmission priority coefficient of the corresponding abnormal reference data packet or abnormal supplementary data packet according to the image abnormality coefficient TYX;

[0121] The monitoring data within the interval time before the real-time monitoring data corresponding to the environmental anomaly coefficient HYX are packaged into an abnormal reference data packet; after the supplementary time has passed, the monitoring data within the interval supplementary time after the real-time monitoring data corresponding to the environmental anomaly coefficient HYX are packaged into an abnormal supplementary data packet; the transmission priority coefficient of the corresponding abnormal reference data packet or abnormal supplementary data packet is generated according to the environmental anomaly coefficient HYX; the corresponding abnormal reference data packet or abnormal supplementary data packet is uploaded according to the transmission priority coefficient.

[0122] This embodiment obtains the standard interval duration of normal uploaded data packets under the monitoring image, scene labels and monitoring data, as well as the image anomaly coefficient and the environmental anomaly coefficient; calculates the interval duration and the supplementary duration according to the anomaly coefficient and its corresponding standard interval duration, and adjusts the size of the corresponding uploaded data packet through the dynamic change of the anomaly coefficient, thereby improving the feedback efficiency of emergency processing under construction conditions.

[0123] The transmission priority coefficient of the corresponding abnormal reference data packet and the transmission priority coefficient of the abnormal supplementary data packet are generated according to the image abnormality coefficient TYX, including:

[0124] Obtain image abnormality coefficient TYX;

[0125] Determine whether the image abnormality coefficient is the personnel abnormality coefficient:

[0126] Yes, the transmission priority coefficient of the abnormal reference data packet is calculated by the formula CYX=g1×(1 / (1+exp(-TYX))), and the transmission priority coefficient of the abnormal supplementary data packet is calculated by the formula BYX=g2×(1 / (1+exp(-TYX)));

[0127] No, the transmission priority coefficient of the abnormal reference data packet is calculated by the formula CYX=g3×(1 / (1+exp(-TYX))), and the transmission priority coefficient of the abnormal supplementary data packet is calculated by the formula BYX=g4×(1 / (1+exp(-TYX))), where g1, g2, g3 and g4 are proportional coefficients, and g1>g2, g3>g4; and g1, g2, g3 and g4 are set according to experience; as the image anomaly coefficient increases, the probability of safety hazards to construction will increase. In order to upload the data packet in time, the transmission priority coefficient of the data packet needs to be increased; and the importance of the data packet when an abnormality occurs is higher than the importance of the subsequent supplementary data packet, so g1>g2, g3>g4.

[0128] The transmission priority coefficient of the corresponding abnormal reference data packet and the transmission priority coefficient of the abnormal supplementary data packet are generated according to the environmental abnormality coefficient HYX, including:

[0129] Get the environmental anomaly coefficient HYX;

[0130] The transmission priority coefficient of the abnormal reference data packet is calculated by the formula CYX=g5×(1 / (1+exp(-HYX))), and the transmission priority coefficient of the abnormal supplementary data packet is calculated by the formula BYX=g6×(1 / (1+exp(-HYX))), where g5 and g6 are proportional coefficients, and g5>g6; as the environmental anomaly coefficient increases, the probability of safety hazards to construction will also increase. In order to upload the data packet in time, it is necessary to increase the transmission priority coefficient of the data packet; and when an anomaly occurs, the importance of the data packet is higher than that of the subsequent supplementary data packet, so g5>g6, and the importance of the data packet transmission priority coefficient generated under different anomaly coefficients is different, so g1, g3 and g5 are all different, and g2, g4 and g6 are also all different.

[0131] In this embodiment, the same strategy is adopted for calculating the transmission priority coefficients of the abnormal reference data packets and the abnormal supplementary data packets, that is, the content time range of each abnormal reference data packet is consistent with that of its corresponding abnormal supplementary data packet.

[0132] Uploading the corresponding abnormal reference data packet or abnormal supplementary data packet according to the transmission priority coefficient includes:

[0133] When an abnormal reference data packet or an abnormal supplementary data packet is generated, it is detected whether there are other data packets to be uploaded at this time;

[0134] If yes, obtain the transmission priority coefficients corresponding to other data packets; when the transmission priority coefficients corresponding to other data packets are greater than the transmission priority coefficients of the abnormal reference data packets or abnormal supplementary data packets that need to be transmitted this time, wait for their transmission to be completed before transmitting the abnormal reference data packet or abnormal supplementary data packet; otherwise, obtain the corresponding transmission progress; when the transmission progress is less than the progress threshold, terminate the transmission of other abnormal reference data packets or abnormal supplementary data packets; immediately upload the abnormal reference data packet or abnormal supplementary data packet; otherwise, wait for the transmission of other abnormal reference data packets or abnormal supplementary data packets to be completed, and then immediately upload the abnormal reference data packet or abnormal supplementary data packet.

[0135] No, directly upload the abnormal reference data package or abnormal supplementary data package.

[0136] In this embodiment, the transmission priority coefficient of the abnormal reference data packet obtained this time is 0.786, and the transmission priority coefficient of other abnormal reference data packets obtained is 0.688. When the abnormal reference data packets are generated this time, the transmission progress of other abnormal reference data packets is 75%. The progress threshold is set according to experience. In this embodiment, the progress threshold is set to 70%. Although the transmission priority coefficient of the abnormal reference data packet this time is greater than the transmission priority coefficients of other abnormal reference data packets, since the transmission progress of other abnormal reference data packets is greater than the progress threshold and the transmission is about to be completed, it is necessary to wait for the transmission of other abnormal reference data packets to be completed before uploading this abnormal reference data packet.

[0137] This embodiment calculates the transmission priority coefficient of each abnormal reference data packet or abnormal supplementary data packet, compares the priority coefficients of other abnormal reference data packets or abnormal supplementary data packets with the current abnormal reference data packet or abnormal supplementary data packet, and adaptively adjusts the data packet upload method according to the priority coefficient and transmission progress, thereby improving the feedback efficiency of emergency processing under construction conditions.

[0138] Part of the data in the above formula is calculated by removing the dimension and taking its numerical value. The formula is a formula closest to the actual situation obtained by software simulation of a large amount of collected data; the preset parameters and preset thresholds in the formula are set by technical personnel in this field according to actual conditions or obtained through simulation of a large amount of data.

[0139] The working principle of the present application is as follows: monitoring videos, monitoring images and monitoring data are acquired through corresponding data acquisition equipment, and the monitoring images and monitoring data are preprocessed; the preprocessed monitoring images are input into a label recognition model to obtain a number of project labels; an image anomaly coefficient corresponding to the monitoring image is generated according to the project label; a working alarm signal is generated according to the image anomaly coefficient; and, an environmental anomaly coefficient is generated according to the preprocessed monitoring data; an environmental anomaly alarm signal is generated according to the environmental anomaly coefficient; monitoring videos and monitoring data are packaged and uploaded according to the image anomaly coefficient and the environmental anomaly coefficient; corresponding alarm prompts are issued according to the alarm signal and the early warning signal, and the management personnel are contacted according to the alarm signal, the working status of the personnel on the construction site is increased, the content size and upload time of the data packet are dynamically adjusted, the efficiency of safety monitoring in the construction area and the feedback efficiency in emergency situations are improved, and the problem of low efficiency of safety monitoring in the construction area due to insufficient consideration of the working status of the staff and the identification of some other dangerous situations in the construction area is avoided.

[0140] The above embodiments are only used to illustrate the technical method of the present application and are not intended to limit it. Although the present application has been described in detail with reference to the preferred embodiments, a person of ordinary skill in the art should understand that the technical method of the present application may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present application.

Claims

1. A construction site monitoring system based on artificial intelligence, characterized in that: include: Function expansion module, data collection module, data analysis module, data upload module, early warning module and data storage module; The function expansion module is provided with a number of expansion interfaces of different models for connecting to the expanded data acquisition equipment; The data acquisition module: obtains the construction monitoring video and real-time monitoring image of the construction area through the video monitoring equipment connected to it; obtains the monitoring data through the data acquisition equipment connected to the function expansion module connected to it; and pre-processes the monitoring image and monitoring data; The data analysis module: obtains the pre-processed monitoring image; inputs the monitoring image into the label recognition model to obtain a number of project labels; generates the image abnormality coefficient corresponding to the monitoring image according to the project label; generates a work alarm signal according to the image abnormality coefficient; the label recognition model is obtained by training the artificial intelligence model; and, Acquire the pre-processed monitoring data, generate an environmental anomaly coefficient according to the monitoring data; generate an environmental anomaly alarm signal according to the environmental anomaly coefficient; The data uploading module is used to obtain an image anomaly coefficient and an environmental anomaly coefficient, and to package and upload the monitoring video and monitoring data according to the image anomaly coefficient and the environmental anomaly coefficient; The early warning module: issues corresponding alarm prompts according to the alarm signal and the early warning signal, and contacts the management personnel according to the alarm signal; Generating the image anomaly coefficient corresponding to the monitoring image according to a number of project labels includes: Get the project labels corresponding to each monitoring project; Determine whether the project label corresponding to the monitoring and identification project is person monitoring; Yes, obtain the personnel ID of each item label corresponding to the ID identification item; and the item label corresponding to each behavior identification item and attribute identification item corresponding to each personnel ID; extract the item danger level YDu of several behavior identification items and attribute identification items whose item label is dangerous; The abnormal coefficient of each personnel ID is calculated by the formula RYXi=∑YDu; where u is the number of the project label, u∈[1,n], n represents the total number of attribute project labels and behavior project labels; i is the number of the personnel ID; the personnel abnormal coefficient is calculated by the formula RYX=∑RYXi; i∈[0,m], m is the total number of personnel; No, obtain several hazardous monitoring items with item labels as hazardous and their corresponding item hazard levels YDj; The scenario anomaly coefficient CYX is calculated by the formula, CYX=∑YDj; where YDj represents the project danger level of the jth project label, j∈[1,P], and P represents the total number of danger monitoring project labels; The image abnormality coefficient includes a personnel abnormality coefficient or a scene abnormality coefficient; The environmental anomaly coefficient is generated according to the monitoring data, including: Obtain the corresponding data change amount Bk of several monitoring data in unit time and the important proportion ZBk of several monitoring data; Calculate the abnormal coefficient YXk corresponding to several monitoring data through the formula; YXk=β×arctan(|Bk / DBk|); The environmental anomaly coefficient is calculated by the formula HYX=∑(ZBk×YXk); where β is the adjustment factor, DBk is the corresponding monitoring unit data; k represents the number of each monitoring data, k∈[0,Q], and Q is the total number of monitoring data.

2. The construction site monitoring system based on artificial intelligence according to claim 1 is characterized in that: The preprocessing includes: monitoring image preprocessing and monitoring data preprocessing.

3. The construction site monitoring system based on artificial intelligence according to claim 1 is characterized in that: The label recognition model is obtained through artificial intelligence model training, including: A number of historical monitoring images and a number of monitoring items, as well as their corresponding project labels, are obtained through the data storage module; the monitoring items include monitoring identification items; the project labels corresponding to the monitoring identification items are person monitoring or scene monitoring; the monitoring identification items also include ID identification items and a number of dangerous monitoring items; the project labels corresponding to the ID identification items are the corresponding personnel IDs; the ID identification items also include a number of behavior identification items and attribute identification items; the project labels corresponding to the behavior identification items, attribute identification items and dangerous monitoring items include dangerous or normal; Several historical monitoring images are divided according to monitoring projects, and each monitoring project is divided into corresponding training data, verification data and test data according to the project label; the training data, verification data and test data are divided in a ratio of 7:2:1; Preprocess the training data, verification data and test data to obtain the training set, verification set and test set; The artificial intelligence model ResNet is selected as the backbone network, and multi-task branches are constructed, including monitoring and recognition branches, ID recognition branches, attribute recognition branches, behavior recognition branches, and danger monitoring branches, so as to construct the basic network model; the monitoring and recognition branch is connected with the ID recognition branch and the danger monitoring branch, and the ID recognition branch is connected with several attribute recognition branches and behavior recognition branches; Use the corresponding training set to train each branch of the basic model, adjust the parameters on the validation set, and test the model on the test data. Finally, we get a label recognition model that takes monitoring images as input and outputs several project labels.

4. The construction site monitoring system based on artificial intelligence according to claim 1 is characterized in that: Generating the working alarm signal according to the image abnormality coefficient includes: Obtain image abnormality coefficient; When the image abnormality coefficient is a personnel abnormality coefficient: determine whether the image abnormality coefficient is greater than the personnel alarm threshold; if yes, generate a personnel abnormality alarm signal; if no, generate a personnel abnormality warning signal; When the image anomaly coefficient is a scene anomaly coefficient: determine whether the image anomaly coefficient is greater than the scene alarm threshold; if so, generate a scene anomaly alarm signal; if not, generate a scene anomaly warning signal.

5. The construction site monitoring system based on artificial intelligence according to claim 1 is characterized in that: Generating the environmental abnormality alarm signal according to the environmental abnormality coefficient includes: Obtain environmental anomaly coefficient; Determine whether the environmental anomaly coefficient is greater than the environmental alarm threshold: Yes, generate an environmental abnormality alarm signal; No, obtain the abnormal coefficients corresponding to the several monitoring data; when the abnormal coefficients corresponding to the several monitoring data are greater than their corresponding abnormal thresholds; generate the corresponding monitoring data abnormal alarm signals.

6. The construction site monitoring system based on artificial intelligence according to claim 1 is characterized in that: Pack and upload monitoring videos and monitoring data according to image anomaly coefficients and environmental anomaly coefficients, including: Obtain the image anomaly coefficient TYX and the environment anomaly coefficient HYX; obtain the standard interval durations of normal data packet upload under monitoring image, scene monitoring and monitoring data, which are BDSr, BDSc and BDSs respectively; When the monitoring image is for human monitoring, the interval duration is calculated by the formula DSr=BDSr×exp(-(b1×TYX)), and the corresponding supplementary duration is equal to the interval duration; When the monitoring image is scene monitoring, the interval duration is calculated by the formula DSc=BDSc×exp(-(b2×TYX)), and the corresponding supplementary duration is equal to the interval duration; The interval duration under the monitoring data is calculated by the formula DSs=BDSs×exp(-(b3×HYX)), and the corresponding supplementary duration is equal to the interval duration, where b1, b2 and b3 are proportional coefficients; Packing the monitoring video within the interval time before the real-time monitoring image corresponding to the image abnormality coefficient TYX into an abnormal reference data packet; after the supplementary time, packing the monitoring video within the interval supplementary time after the real-time monitoring image corresponding to the image abnormality coefficient TYX into an abnormal supplementary data packet; generating the transmission priority coefficient of the corresponding abnormal reference data packet or abnormal supplementary data packet according to the image abnormality coefficient TYX; The monitoring data within the interval time before the real-time monitoring data corresponding to the environmental anomaly coefficient HYX are packaged into an abnormal reference data packet; after the supplementary time period has passed, the monitoring data within the interval supplementary time after the real-time monitoring data corresponding to the environmental anomaly coefficient HYX are packaged into an abnormal supplementary data packet; a transmission priority coefficient of a corresponding abnormal reference data packet or an abnormal supplementary data packet is generated according to the environmental anomaly coefficient HYX; and the corresponding abnormal reference data packet or abnormal supplementary data packet is uploaded according to the transmission priority coefficient.

7. The construction site monitoring system based on artificial intelligence according to claim 6 is characterized in that: The transmission priority coefficient of the corresponding abnormal reference data packet and the transmission priority coefficient of the abnormal supplementary data packet are generated according to the image abnormality coefficient TYX, including: Obtain image abnormality coefficient TYX; Determine whether the image abnormality coefficient is the personnel abnormality coefficient: Yes, the transmission priority coefficient of the abnormal reference data packet is calculated by the formula CYX=g1×(1 / (1+exp(-TYX))), and the transmission priority coefficient of the abnormal supplementary data packet is calculated by the formula BYX=g2×(1 / (1+exp(-TYX))); No, the transmission priority coefficient of the abnormal reference data packet is calculated by the formula CYX=g3×(1 / (1+exp(-TYX))), and the transmission priority coefficient of the abnormal supplementary data packet is calculated by the formula BYX=g4×(1 / (1+exp(-TYX))), where g1, g2, g3 and g4 are proportional coefficients, and g1>g2, g3>g4.

8. The construction site monitoring system based on artificial intelligence according to claim 6 is characterized in that: The uploading of the corresponding abnormal reference data packet or abnormal supplementary data packet according to the transmission priority coefficient includes: When an abnormal reference data packet or an abnormal supplementary data packet is generated, it is detected whether there are other data packets to be uploaded at this time; If yes, then obtain the transmission priority coefficients corresponding to other data packets; when the transmission priority coefficients corresponding to other data packets are greater than the transmission priority coefficients of the abnormal reference data packets or abnormal supplementary data packets that need to be transmitted this time, wait for their transmission to be completed before transmitting the abnormal reference data packet or abnormal supplementary data packet; otherwise, obtain their corresponding transmission progress; when the transmission progress is less than the progress threshold, terminate the transmission of other abnormal reference data packets or abnormal supplementary data packets; immediately upload the abnormal reference data packet or abnormal supplementary data packet; otherwise, wait for the transmission of other abnormal reference data packets or abnormal supplementary data packets to be completed, and then immediately upload the abnormal reference data packet or abnormal supplementary data packet; No, directly upload the abnormal reference data package or abnormal supplementary data package.

Images