CN114782988A - Construction environment-oriented multi-stage safety early warning method - Google Patents
Construction environment-oriented multi-stage safety early warning method Download PDFInfo
- Publication number
- CN114782988A CN114782988A CN202210322348.7A CN202210322348A CN114782988A CN 114782988 A CN114782988 A CN 114782988A CN 202210322348 A CN202210322348 A CN 202210322348A CN 114782988 A CN114782988 A CN 114782988A
- Authority
- CN
- China
- Prior art keywords
- data
- early warning
- behavior
- safety
- data set
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000010276 construction Methods 0.000 title claims abstract description 70
- 238000000034 method Methods 0.000 title claims abstract description 53
- 238000001514 detection method Methods 0.000 claims abstract description 36
- 230000008569 process Effects 0.000 claims abstract description 15
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 6
- 230000006399 behavior Effects 0.000 claims description 70
- 206010000117 Abnormal behaviour Diseases 0.000 claims description 43
- 238000003708 edge detection Methods 0.000 claims description 23
- 238000009826 distribution Methods 0.000 claims description 19
- 230000033001 locomotion Effects 0.000 claims description 14
- 230000002159 abnormal effect Effects 0.000 claims description 12
- 238000004364 calculation method Methods 0.000 claims description 11
- 238000013528 artificial neural network Methods 0.000 claims description 9
- 238000002372 labelling Methods 0.000 claims description 9
- 230000001788 irregular Effects 0.000 claims description 8
- 238000012545 processing Methods 0.000 claims description 7
- 230000004913 activation Effects 0.000 claims description 6
- 230000005856 abnormality Effects 0.000 claims description 5
- 238000004140 cleaning Methods 0.000 claims description 4
- 238000012549 training Methods 0.000 claims description 4
- 230000009194 climbing Effects 0.000 claims description 3
- 238000013135 deep learning Methods 0.000 claims description 3
- 238000001914 filtration Methods 0.000 claims description 3
- 231100001261 hazardous Toxicity 0.000 claims description 3
- 238000010801 machine learning Methods 0.000 claims description 3
- 230000009329 sexual behaviour Effects 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 claims description 2
- 230000007246 mechanism Effects 0.000 abstract description 8
- 238000004458 analytical method Methods 0.000 abstract description 5
- 230000000007 visual effect Effects 0.000 abstract description 3
- 238000005065 mining Methods 0.000 abstract description 2
- 230000000694 effects Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 230000007547 defect Effects 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 210000000988 bone and bone Anatomy 0.000 description 1
- 238000004141 dimensional analysis Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 230000001105 regulatory effect Effects 0.000 description 1
- 230000000391 smoking effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/088—Non-supervised learning, e.g. competitive learning
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Alarm Systems (AREA)
Abstract
The invention discloses a construction environment-oriented multi-stage safety early warning method, and belongs to the field of safety early warning. A multilevel safety early warning method facing to a construction environment adopts a hierarchical early warning mechanism, and improves the systematicness, accuracy and reliability of the multilevel safety early warning of construction big data; the acquired data are labeled and subjected to danger division, and the early warning cost is effectively saved on the premise of ensuring the accuracy and the timeliness by deeply mining the incidence relation among the multi-dimensional multi-element data; in the detection process, a visual depth analysis scheme suitable for project construction scenes is provided by combining the prior advanced algorithm, the invention effectively prevents the occurrence of dead angles in safety detection and endows the system with the capability of dynamically adjusting the early warning level.
Description
Technical Field
The invention belongs to the field of safety early warning, and particularly relates to a construction environment-oriented multistage safety early warning method.
Background
With the continuous advance of the industrial informatization technology, how to early warn, prevent and control in advance aiming at the safety in the project construction process in multiple directions, multiple angles and multiple levels becomes one of the problems to be solved urgently in the safety production. The data of the early warning mechanism mainly come from large data of a construction site, focuses on safety control of the construction site of a construction site, and surrounds key elements such as personnel, machinery, materials, environment and the like, however, the existing related early warning model has the problems of single data, insufficient data linkage, small data processing scale, poor technical expansibility, unclear grading control limit, high cost and the like, so that the system is poor in migratability and expansibility, low in accuracy and reliability and poor in timeliness and linkage, and therefore, from a specific construction scene, the design of an early warning method with high universality, reliability, systematicness and expansibility is particularly important.
Different from a general early warning system, the reasons for the fact that the project construction early warning mechanism is not obvious in effect and poor in usability are mainly divided into the following three aspects: firstly, construction scenes are complex and changeable, the source is single only by means of data collected by a sensor, a leak exists in the detection aspect due to the limitation of the construction scenes, the acquisition rate of effective information is low, the data redundancy is high, and the danger level of the early warning information cannot be accurately divided, extracted and judged; secondly, the relationships between people and the environment, between objects (materials, equipment and the like) and between people and objects (materials, equipment and the like) are varied, and the construction scenes and states are different, so that the existing early warning system is difficult to flexibly adjust aiming at the complex environment, and the difficulty of the early warning system is increased by the access of multiple elements and multiple dimensions; finally, in a construction scene, abnormal behaviors such as people alarming, smoking, chatting and the like are difficult to detect in a conventional mode, and the conventional early warning system lacks the capabilities of global monitoring and large-scale image capturing, processing and analyzing, so that the safety early warning effect of construction big data is seriously influenced.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a construction environment-oriented multi-stage safety early warning method.
In order to achieve the purpose, the invention adopts the following technical scheme to realize the purpose:
a construction environment-oriented multi-stage safety early warning method comprises the following steps;
s1, collecting multi-dimensional construction big data C, wherein the multi-dimensional construction big data C comprises image video and text data C of ambient temperature and humiditytextThe image video is marked by historical data C to be markedoldAnd new data C acquired in real timenewForming;
historical data C to be labeledoldLabeling in different dimensions to obtain a label data set CL(ii) a Tag data set CLData set C containing dangerous areaDaAnd a dangerous behavior data set CDb；
S2, based on the label data set CLPredicting the wearing condition of the safety equipment by utilizing a neural network to obtain a wearing prediction result;
based on the hazardous area data set CDaDemarcating a dangerous area DareaCalculating to obtain the edge detection confidence coefficient Cborder(ii) a Predicting the personnel position information by using a neural network to obtain a personnel position detection confidence coefficient Clocation；
New data C based on real-time acquisitionnewGenerating a historical motion image MHI based on a dangerous behavior data set CDbAcquiring time-space information of normal human body movement, and constructing and updating an abnormal behavior detection model by using a neural network based on the time-space information and a historical moving image MHI to predict the behavior of a human body;
text data C according to ambient temperature and humiditytextCalculating the mean value and the standard deviation of the data;
s3, converting the wearing prediction result into wearing and non-wearing conditions according to a preset threshold, combining the wearing conditions of each safety equipment, and dividing the wearing conditions into 8 levels of dangerous behaviors based on the combined result;
according to edge detection confidence CborderConfidence degree C of detection of position of personlocationThe combined result of the step (2) is divided into the grades of 4 objects and the construction environment;
according to the personnel behavior prediction result and each abnormal behavior-to-danger behavior data set CDbDetermining the abnormal behavior category;
s4, calculating a basic security score S of personnelPEnvironmental safety score SEAnd a global security score SGDynamically adjusting the activation parameters of each safety score to realize dynamic adjustment of the early warning grade to obtain the early warning score
And simultaneously implementing early warning.
Further, step S1 specifically includes:
cleaning the collected data by adopting a machine learning algorithm, deleting redundant data, and processing labeled historical data ColdLabeling to obtain a safety helmet label LhatSeat belt label LbeltMask label LmaskAnd construction vehicle label LcarConversion of the data set C into a label data set CL；
Respectively labeling data acquired by terminal equipment in a dangerous area in the data set C and data with dangerous behaviors to obtain a dangerous area data set CDaAnd a dangerous behavior data set CDb。
Further, step S2 is specifically:
s201, tag data set C is subjected to deep learning network through YOLOLTraining and modeling are carried out to obtain a safety wearing detection model, the wearing conditions of the safety helmet, the safety belt and the mask under the same scene are predicted simultaneously, and a safety helmet wearing prediction score P is obtainedhelmetSeatbelt wearing prediction score PbeltPredicted gauze mask wearing score Pmask；
S202, dangerous area data set CDaCarrying out graying, filtering and edge detection on the data, then adopting Hough line transformation on a straight line boundary, adopting a sliding window on a curve boundary, and utilizing an irregular contour to externally connect a minimum rotating rectangle to replace the irregular contour in the irregular region so as to determine a dangerous region DareaComputing edgesEdge detection confidence Cborder(ii) a Calculating the confidence degree C of the detection of the position of the personnel according to the relationship between the adjacent frames and the position of the person obtained by the safety wearing detection modellocation；
S203, deducting the dangerous behavior data set C from the data set CDbObtaining a normal behavior data set
Will normal behavior data set
Normalizing the ith original posture Pi as spatial information; by aligning normal behavior data sets
Calculating to obtain the motion velocity v of the ith skeleton joint pointiUsing a Rayleigh distribution to distribute v of discrete distributionsiModeling is probability distribution, namely the normal behavior probability distribution of the human body, and the normal behavior probability distribution of the human body is used as time information; from new data C collected in real timenewGenerating a historical moving image MHI;
the method comprises the steps of constructing regularity of a human posture trajectory learned by a spatio-temporal Transformer model by utilizing time information and space information, introducing a historical motion image MHI and a non-negative sparse automatic coding machine, enabling the Transformer model to learn in a self-supervision incremental mode, and predicting by taking the trained Transformer model as an abnormal behavior detection model to obtain an abnormal behavior score F;
s204, text data C based on ambient temperature and humiditytextThe statistical process of SPC is used to judge the standard deviation, determine the data of the uncontrolled state in the batch, and calculate the data abnormality index E.
Further, the confidence of edge detection C in S202borderAnd object position confidence ClocationThe calculation process of (2) is as follows:
wherein k is1Is a filter impact factor; kernel _ size is the size of the gaussian kernel used; k is a radical of2An edge detection impact factor; t is a threshold value for edge detection; h is a total ofSUBThe minimum contour height of the object obtained by the difference between adjacent frames; cur (curve)hA predicted height obtained for the safety wear detection model; beta is a high scale factor.
Further, the person moving speed v in S203iComprises the following steps:
wherein (x)i，j，yi，j) And (x)i+1，j，yi+1，j) Respectively representing the position of the jth joint point at the ith frame and the (i + 1) th frame. 6. The construction environment-oriented multistage safety early warning method as claimed in claim 3, wherein the data anomaly index E in S204 is calculated as follows:
Nnormalnumber of normal data, NabnormalNumber of data representing uncontrolled state.
Further, step S3 specifically includes:
s301, combining wearing prediction results according to the importance degree of different wearing devices in construction, and dividing the wearing prediction results into 8 dangerous behaviors with different levels, wherein the dangerous behaviors are as follows:
wherein, whThe safety helmet is worn on behalf of the wearer,
representing no safety helmet is worn; w is ahRepresenting the wearing of a safety belt,
representing no safety belt is worn; w is amThe mask is worn on the representative of the patient,
representing no wearing mask, wh、wb、wm∈{0，1}；
S302, coordinate of position of personnel and DareaJudging the boundary coordinates to obtain an initial border crossing behavior; according to edge detection confidence CborderIs used to divide the boundary decision factor SbDetecting confidence C according to the position of the personlocationIs used to divide the position determination factor SlBased on a boundary decision factor SbAnd a position determination factor SlThe combination of (1) divides the initial off-boundary behavior into 4 classes of off-boundary confidence levels BiThe method comprises the following steps:
boundary decision factor SbA position determination factor SlAnd cross-border sexual behavior BiThe calculation is as follows:
s303, calculating three types of abnormal behaviors of alarming, falling and climbing in dangerous behavior data set CDbDetermining the abnormal behavior class A based on the abnormal behavior probability and the abnormal behavior score Fj，AjComprises the following steps:
wherein, cfightData set C representing abnormal alarming behavior in dangerous behaviorDbThe number of (2); c. CfallData set C representing abnormal falling behavior at riskDbThe number of (2); c. CsneakData set C representing abnormal behavior of conversation in dangerous behaviorDbThe number of (2); f represents an abnormal behavior score;
s304, setting an alarm threshold value TalarmThe data abnormality index is compared with an alarm threshold value TalarmComparing, and giving early warning to the abnormal state, wherein the definition of the abnormal state is as follows:
wherein S iskRepresenting the k-th state.
Further, a person basic security score S in step S4PEnvironmental safety score SEGlobal security score SGAnd early warning score
The calculation method comprises the following steps:
SP＝D1*Phelmet+D2*Pbelt+D3*Pmask
SE＝Cborder*Clocation
SG＝F
wherein D is1、D2、D3Is a risk factor; f is the abnormal behavior score; a is1、a2、a3To activate the parameters, a1，a2，a3∈{0，1}；w1、w2、w3Is a detection factor; s. theiIs the activation parameter condition under the ith early warning level.
Further, a risk factor D1、D2、D3Comprises the following steps:
wherein N isiRepresents LiBehavior level in tag dataset CLThe statistical amount of (1).
Compared with the prior art, the invention has the following beneficial effects:
the invention provides a construction environment-oriented multi-stage safety early warning method, which adopts a grading early warning mechanism around the difference between construction projects and general projects, and improves the systematicness, accuracy and reliability of construction big data multi-stage safety early warning; the acquired data are labeled and subjected to danger division, and the early warning cost is effectively saved on the premise of ensuring the accuracy and the timeliness by deeply mining the incidence relation among the multi-dimensional multi-element data; in the detection process, a visual depth analysis scheme suitable for project construction scenes is provided by combining the prior advanced algorithm, the invention effectively prevents the occurrence of dead angles in safety detection and endows the system with the capability of dynamically adjusting the early warning level.
Furthermore, in step S1, processing operations such as tagging and risk division are performed on the collected real-time data of the construction site, so that the raw data can provide richer semantic information in the multi-stage early warning, the subsequent operation and processing are facilitated, and the sensing and collection of the data are completed.
Further, step S2 focuses on the semantics of the raw data, and processes and analyzes different tag data sets in a manner of combining an advanced algorithm with a conventional computer vision technology, so as to extract key features of the data and discriminate basic information, thereby greatly reducing the early warning cost of the construction big data.
Further, step S3 focuses on three aspects of standardization of wearing of constructors, incidence relation of objects and environments and personnel behavior distribution, gives consideration to real-time states of construction site materials, constructs early warning behaviors, provides rule guarantee for multi-stage early warning, and further improves early warning capacity while achieving systematicness.
Further, in the step S4, the system early warning score is calculated according to the safety of the calculation personnel, the environmental safety and the global safety score, the early warning grade is dynamically adjusted, the system early warning score can effectively adapt to various occasions, the three-level early warning of dead angles in the early warning is prevented, various abnormal behaviors are effectively distinguished by combining the skeleton information of the human body according to the normal distribution of the normal behaviors, the early warning of the abnormal behaviors is realized by judging the rationality of the human body behaviors in the whole, and the dead angles in the early warning are effectively prevented.
In conclusion, the early warning method disclosed by the invention fully realizes a multilevel safety early warning mechanism suitable for different construction scenes, more systematically relates to a three-level early warning module, improves the utilization rate of construction big data by carrying out feature extraction and training on different tagged data, improves the adaptability of a model by an incremental learning mode, reduces the early warning cost by visual realization, and effectively prevents the occurrence of the condition of dead angles in early warning by the multilevel early warning mechanism.
Drawings
FIG. 1 is a flow chart of a multi-level three-dimensional project construction safety early warning method applied to industrial big data;
FIG. 2 is a diagram of the end-edge-cloud architecture schema of the present invention;
fig. 3 is a diagram of the effect of the implementation.
Detailed Description
In order to make the technical solutions of the present invention better understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that the terms "first," "second," and the like in the description and claims of the present invention and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the invention described herein are capable of operation in other sequences than those illustrated or described herein. Moreover, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
The invention is described in further detail below with reference to the accompanying drawings:
the invention provides a construction environment-oriented multistage safety early warning method, which is used for carrying out correlation and hierarchical analysis on native data, designing a multistage early warning method based on industrial big data, designing a multistage early warning module aiming at the native data acquired by edge intelligent equipment, and firstly predicting normative behaviors worn by personnel to realize primary early warning; aiming at a complex construction site, the aim of secondary early warning is fulfilled by analyzing the incidence relation between the native data and the construction environment; and analyzing abnormal behaviors of personnel in a three-dimensional manner aiming at the global state, thereby realizing three-level early warning facing to the construction big data.
Referring to fig. 1, fig. 1 is a flowchart of a construction environment-oriented multistage safety early warning method, which includes the following specific steps:
s1, collecting multi-dimensional construction big data C, wherein the multi-dimensional construction big data C comprises image video and text data C of ambient temperature and humiditytextThe image video is marked by historical data C to be markedoldAnd new data C acquired in real timenewForming;
historical data C to be labeledoldLabeling with different dimensions to obtain a label data set CL(ii) a Tag data set CLIncluding a data set C of the hazardous areaDaAnd a dangerous behavior data set CDb；
Step S1 specifically includes:
s101, cleaning the collected data by adopting a machine learning algorithm, deleting redundant data, and processing the historical data C to be labeledoldLabeling to obtain a safety helmet label LhatSeat belt label LbeltMask label LmaskAnd construction vehicle label LcarThe data set C is converted into a label data set CL；
Respectively labeling the data collected by the terminal equipment of the dangerous area in the data set C and the data with dangerous behaviors to obtain a dangerous area data set CDaAnd a dangerous behavior data set CDb。
S2, based on the label data set CLPredicting the wearing condition of the safety equipment by utilizing a neural network to obtain a wearing prediction result;
based on the danger area data set CDaDemarcating a dangerous area DareaAnd using the relevant parametersObtaining the confidence coefficient C of edge detectionborder(ii) a Predicting the personnel position information by utilizing a neural network to obtain a personnel position detection confidence coefficient Clocation；
New data C based on real-time acquisitionnewGenerating a historical motion image MHI based on a dangerous behavior data set CDbAcquiring time-space information of normal human motion, and constructing and updating an abnormal behavior detection model by using a neural network based on the time-space information and a historical moving image MHI to predict human behaviors;
text data C according to ambient temperature and humiditytextCalculating the mean value and standard deviation of the data;
step S2 specifically includes:
s201, tag data set C is subjected to deep learning network through YOLOLTraining and modeling are carried out to obtain a safety wearing detection model, the wearing conditions of the safety helmet, the safety belt and the mask under the same scene are predicted simultaneously, and a safety helmet wearing prediction score P is obtainedhelmetPredicted seatbelt wearing score PbeltPredicted gauze mask wearing score Pmask。
S202, setting a dangerous area data set CDaCarrying out graying, filtering and edge detection on the data, then adopting Hough line transformation on a straight line boundary, adopting a sliding window on a curve boundary, and utilizing an irregular contour to externally connect a minimum rotating rectangle to replace the irregular contour in the irregular region so as to determine a dangerous region DareaAnd calculating the confidence coefficient C of the edge detection by using the parameters with strong correlationborderIn addition, the person position detection confidence coefficient C is calculated through the adjacent frame relation and the person position obtained by the safety wearing detection modellocation。
Further, the confidence of edge detection CborderObject position confidence ClocationThe calculation process is as follows:
wherein k is1Is a filter impact factor; kernel _ size is the size of the gaussian kernel used; k is a radical of2An edge detection impact factor; t is a threshold value for edge detection; h is a total ofSUBThe minimum contour height of the object obtained by the difference between adjacent frames; cur (curve)hA predicted height obtained for the safety wear detection model; beta is a high scale factor.
S203, deducting the dangerous behavior data set C from the data set CDbObtaining a normal behavior data set
Will normal behavior data set
Normalizing the ith original posture Pi as spatial information; by aligning normal behavior data sets
Calculating to obtain the motion speed v of the ith skeleton joint pointiV of discrete distribution using Rayleigh distributioniModeling is probability distribution, namely the probability distribution of the normal behaviors of the human body, and the probability distribution of the normal behaviors of the human body is used as time information; from new data C collected in real timenewGenerating a history moving image MHI;
the method comprises the steps of constructing regularity of a human posture trajectory learned by a spatio-temporal Transformer model by utilizing time information and space information, introducing a historical motion image MHI and a non-negative sparse automatic coding machine, enabling the Transformer model to learn in a self-supervision incremental mode, and predicting by taking the trained Transformer model as an abnormal behavior detection model to obtain an abnormal behavior score F.
Further, the speed v of movement of the personiThe calculation process of (c) is as follows:
wherein (x)i，j，yi，j) And (x)i+1，j，yi+1，j) Respectively representing the position of the jth joint point in the ith frame and the (i + 1) th frame;
s204, text data C based on ambient temperature and humiditytextThe statistical process of SPC is used to determine the criteria for the deviation, determine the data for the uncontrolled state in the batch, and calculate the data abnormality index E.
Further, the data anomaly index E is calculated as follows:
wherein, NnormalNumber of normal data, NabnormalNumber of data representing uncontrolled state.
S3, converting the wearing prediction result into wearing and non-wearing conditions according to a preset threshold, combining the wearing conditions of each safety equipment, and dividing the wearing conditions into 8 levels of dangerous behaviors based on the combined result;
according to the edge detection confidence coefficient CborderConfidence degree C of person position detectionlocationThe combined result of the three-dimensional space division is divided into the grades of 4 objects and the construction environment;
according to the personnel behavior prediction result and each abnormal behavior-to-danger behavior data set CDbDetermining the abnormal behavior category;
step S3 specifically includes:
s301, combining wearing prediction results according to the importance degree of different wearing devices in construction, and dividing the wearing prediction results into 8 dangerous behaviors with different levels, wherein the dangerous behaviors are as follows:
wherein, whRepresenting the wearing of the safety helmet,
representing the non-wearing of the safety helmet; w is ahWhich represents the wearing of a safety belt,
representing no harness is worn; w is amThe mask is worn on the representative of the patient,
representing no wearing mask, wh、wb、wm∈{0，1}；
S302, coordinate of position of person and DareaJudging the boundary coordinates to obtain an initial boundary crossing behavior; according to the edge detection confidence coefficient CborderIs used to divide the boundary decision factor SbDetecting confidence C based on the position of the personlocationIs used to divide the position determination factor SlBased on a boundary decision factor SbAnd a position determination factor SlThe combination of (1) divides the initial borderline behavior into 4 classes of borderline confidence levels BiThe method comprises the following steps:
boundary decision factor SbA position determination factor SlAnd borderline sexual behavior BiThe calculation is as follows:
s303, calculating three types of abnormal behaviors of alarming, falling and climbing in dangerous behavior data set CDbDetermining the abnormal behavior class A based on the abnormal behavior probability and the abnormal behavior score Fj，AjComprises the following steps:
wherein, cfightData set C representing abnormal alarming behavior in dangerous behaviorDbThe number of (1); c. CfallData set C representing abnormal falling behavior at riskDbThe number of (1); c. CsneakData set C representing abnormal behavior of conversation in dangerous behaviorDbThe number of (1); f represents an abnormal behavior score;
s304, setting an alarm threshold value TalarmComparing the data anomaly index with an alarm threshold TalarmComparing, and giving early warning to the abnormal state, wherein the definition of the abnormal state is as follows:
wherein S iskRepresenting the k-th state.
S4, calculating a basic security score S of personnelPEnvironmental safety score SEAnd a global security score SGDynamically adjusting the activation parameters of each safety score to realize dynamic adjustment of the early warning grade to obtain the early warning score
And meanwhile, early warning is implemented.
Step S4 specifically includes:
s401, basic personnel safety score SPEnvironmental safety score SEGlobal security score SGAnd early warning score
The calculation method comprises the following steps:
SP＝D1*Phelmet+D2*Pbelt+D3*Pmask
SE＝Cborder*Clocation
SG＝F
wherein D is1、D2、D3Is a risk factor; f is the abnormal behavior score; a is a1、a2、a3To activate a parameter, a1，a2，a3∈{0，1}；w1、w2、w3Is a detection factor; siIs the activation parameter condition under the ith early warning level.
Further, D1、D2、D3The calculation process is as follows:
where Ni represents the statistical number of behavior levels of L _ i in the tag data set C _ L.
In conclusion, the construction environment-oriented multi-stage safety early warning method provided by the invention can continuously utilize newly acquired data to improve the model capability, and can realize multi-scene and three-dimensional efficient dynamic early warning by means of a three-stage early warning method.
Referring to fig. 2, fig. 2 is an end-edge-cloud architecture mode diagram, a data sensing and acquisition-gathering and integration-analysis and reasoning-early warning system is used, an end-edge-cloud mode for industrial big data in a complex construction scene is used as an architecture, regional autonomy and core early warning analysis and regulation are used as design ideas, a novel three-dimensional safety early warning mechanism is mainly formed by three-level early warning, and the method is specifically developed to carry out real-time monitoring on early warning native data, such as normalization of personnel operation and wearing, so as to realize first-level early warning; then, aiming at the complex construction site environment, analyzing the correlation between the primary data related to safety and the construction environment, such as boundary crossing detection of personnel and vehicles, combined detection of terminal temperature and humidity intelligent sensor data and the like, and realizing secondary early warning; and finally, starting from the whole situation, performing three-dimensional analysis on abnormal behaviors of personnel, establishing an abnormal behavior rule base in a construction scene, and defining the abnormal behaviors by adopting an incremental learning method according to the probability distribution of the normal behaviors counted by the bone actions, thereby realizing a three-level early warning mechanism of project construction safety for industrial big data. Meanwhile, in order to overcome the defects of high complexity, wide range, wide distribution and multiple types of construction scenes, the small-range autonomy formed by intelligent terminal equipment is researched and designed, the whole three-dimensional safety early warning 'end-edge-cloud' architecture model regulated and controlled by a cloud end is developed, the safety purpose of 'people-oriented' is realized constantly, the sensing and the acquisition of construction data are completed by the terminal intelligent equipment (end), data preprocessing and preliminary early warning such as data cleaning and marking of a main label are completed by an edge server (edge), and the establishment and the dynamic early warning of a multistage early warning model rule base are realized by the cloud end (cloud).
Examples
The method has the effect in construction sites and railway construction scenes, as shown in fig. 3, the first column is an original picture acquired from the intelligent terminal equipment, the second column is a data representation obtained through a detection model, and the third column is a qualitative description obtained after passing through a rule base.
The above contents are only for illustrating the technical idea of the present invention, and the protection scope of the present invention should not be limited thereby, and any modification made on the basis of the technical idea proposed by the present invention falls within the protection scope of the claims of the present invention.
Claims (9)
1. A construction environment-oriented multi-stage safety early warning method is characterized by comprising the following steps;
s1, collecting multi-dimensional construction big data C, wherein the multi-dimensional construction big data C comprises image videos and text data C of ambient temperature and humiditytextThe image video is marked by historical data C to be markedoldAnd new data C acquired in real timenewForming;
historical data C to be annotatedoldLabeling with different dimensions to obtain a label data set CL(ii) a Tag data set CLData set C containing dangerous areaDaAnd a dangerous behavior data set CDb；
S2, based on the label data set CLPredicting the wearing condition of the safety equipment by utilizing a neural network to obtain a wearing prediction result;
based on the hazardous area data set CDaDemarcating a dangerous area DareaCalculating to obtain the confidence coefficient C of edge detectionborder(ii) a Predicting the personnel position information by utilizing a neural network to obtain a personnel position detection confidence coefficient Clocation；
New data C based on real-time acquisitionnewGenerating a historical motion image MHI based on a dangerous behavior data set CDbAcquiring time-space information of normal human motion, and constructing and updating an abnormal behavior detection model by using a neural network based on the time-space information and a historical moving image MHI to predict human behaviors;
text data C according to ambient temperature and humiditytextCalculating the mean value and standard deviation of the data;
s3, converting the wearing prediction result into two conditions of wearing and not wearing according to a preset threshold, combining the wearing conditions of the safety equipment, and dividing the combined result into 8 levels of dangerous behaviors;
according to edge detection confidence CborderConfidence degree C of detection of position of personlocationThe combined result of the three-dimensional space division is divided into the grades of 4 objects and the construction environment;
according to the personnel behavior prediction result and each abnormal behavior-to-danger behavior data set CDbDetermining the abnormal behavior category;
s4, calculating a basic security score S of personnelPEnvironmental safety score SEAnd a global security score SGDynamically adjusting the activation parameters of each safety score to realize dynamic adjustment of the early warning grade and obtain the early warning score
And simultaneously implementing early warning.
2. The construction environment-oriented multistage safety early warning method according to claim 1, wherein the step S1 specifically comprises:
cleaning the collected data by adopting a machine learning algorithm, deleting redundant data, and processing labeled historical data ColdLabeling to obtain a safety helmet label LhatSeat belt label LbeltMask label LmaskAnd construction vehicle label LcarConversion of the data set C into a label data set CL；
Respectively labeling the data collected by the terminal equipment of the dangerous area in the data set C and the data with dangerous behaviors to obtain a dangerous area data set CDaAnd a dangerous behavior data set CDb。
3. The construction environment-oriented multistage safety early warning method according to claim 1, wherein step S2 specifically comprises:
s201, utilizing a YOLO deep learning network to perform label data set CLTraining and modeling are carried out to obtain a safety wearing detection model, the wearing conditions of the safety helmet, the safety belt and the mask in the same scene are predicted simultaneously, and the wearing prediction of the safety helmet is obtainedFraction PhelmetSeatbelt wearing prediction score PbeltPredicted gauze mask wearing score Pmask；
S202, dangerous area data set CDaCarrying out graying, filtering and edge detection on data in the process, then adopting Hough straight line transformation on straight line boundaries, adopting a sliding window on curve boundaries, and utilizing an irregular contour to circumscribe a minimum rotation rectangle to replace the irregular contour so as to determine a dangerous region DareaComputing an edge detection confidence Cborder(ii) a Calculating the confidence degree C of the detection of the position of the personnel according to the relationship between the adjacent frames and the position of the person obtained by the safety wearing detection modellocation；
S203, deducting the dangerous behavior data set C from the data set CDbObtaining a normal behavior data set
Will normal behavior data set
Normalizing the ith original posture Pi as spatial information; by aligning normal behavior data sets
Calculating to obtain the motion speed v of the ith skeleton joint pointiUsing a Rayleigh distribution to distribute v of discrete distributionsiModeling is probability distribution, namely the normal behavior probability distribution of the human body, and the normal behavior probability distribution of the human body is used as time information; from new data C acquired in real timenewGenerating a historical moving image MHI;
the method comprises the steps of constructing regularity of a human posture trajectory learned by a spatio-temporal Transformer model by utilizing time information and space information, introducing a historical motion image MHI and a non-negative sparse automatic coding machine, enabling the Transformer model to learn in a self-supervision incremental mode, and predicting by taking the trained Transformer model as an abnormal behavior detection model to obtain an abnormal behavior score F;
s204, text data C based on ambient temperature and humiditytextThe statistical process of SPC is used to determine the criteria for the deviation, determine the data for the uncontrolled state in the batch, and calculate the data abnormality index E.
4. The construction environment-oriented multistage safety early warning method as claimed in claim 3, wherein in S202, edge detection confidence CborderAnd object position confidence ClocationThe calculation process of (2) is as follows:
wherein k is1Is a filter impact factor; kernel _ size is the size of the gaussian kernel used; k is a radical of2An edge detection impact factor; t is a threshold value for edge detection; h isSUBThe minimum contour height of the object obtained by the difference between adjacent frames; cur (curve)hA predicted height obtained for the safety wear detection model; beta is a high scale factor.
5. The construction environment-oriented multistage safety precaution method according to claim 3, characterized in that the personnel movement speed v in S203 isiComprises the following steps:
wherein (x)i，j，yi，j) And (x)i+1，j，yi+1，j) Respectively represent the position of the jth joint point at the ith frame and the (i + 1) th frame.
6. The construction environment-oriented multistage safety early warning method as claimed in claim 3, wherein the data anomaly index E in S204 is calculated as follows:
Nnormalnumber of normal data, NabnormalNumber of data representing uncontrolled state.
7. The construction environment-oriented multistage safety early warning method according to claim 3, wherein the step S3 specifically comprises:
s301, combining wearing prediction results according to the importance degree of different wearing devices in construction, and dividing the wearing prediction results into 8 dangerous behaviors with different levels, wherein the dangerous behaviors are as follows:
wherein whThe safety helmet is worn on behalf of the wearer,
representing the non-wearing of the safety helmet; w is ahWhich represents the wearing of a safety belt,
representing no harness is worn; w is amThe mask is worn on the representative of the patient,
representing no wearing mask, wh、wb、wm∈{0，1}；
S302, coordinate of position of person and DareaJudging the boundary coordinates to obtain an initial border crossing behavior; according to edge detection confidence CborderTo divide the boundary decision factor SbDetecting confidence C according to the position of the personlocationIs used to divide the position determination factor SlBased on a boundary decision factor SbAnd a position determination factor SlThe combination of (1) divides the initial borderline behavior into 4 classes of borderline confidence levels BiThe method comprises the following steps:
boundary decision factor SbA position determination factor SlAnd cross-border sexual behavior BiThe calculation is as follows:
s303, calculating three types of abnormal behaviors of alarming, falling and climbing in dangerous behavior data set CDbBased on the abnormal behavior probability and the abnormal behavior score F, determining the abnormal behavior class Aj，AjComprises the following steps:
wherein, cfightData set C representing abnormal alarming behavior in dangerous behaviorDbThe number of (2); c. CfallData set C representing abnormal falling behavior at riskDbThe number of (1); c. CsneakData set C representing abnormal behavior of conversation in dangerous behaviorDbThe number of (1); f represents an abnormal behavior score;
s304, setting an alarm thresholdValue TalarmThe data abnormality index is compared with an alarm threshold value TalarmAnd comparing, and early warning the abnormal state, wherein the definition of the abnormal state is as follows:
wherein S iskRepresenting the kth state.
8. The construction environment-oriented multistage safety early warning method as claimed in claim 1, wherein in step S4, the personnel basic safety score S isPEnvironmental safety score SEGlobal security score SGAnd early warning score
The calculating method comprises the following steps:
SP＝D1*Phelmet+D2*Pbelt+D3*Pmask
SE＝Cborder*Clocation
SG＝F
wherein D is1、D2、D3Is a risk factor; f is the abnormal behavior score; a is1、a2、a3To activate the parameters, a1，a2，a3∈{0，1}；w1、w2、w3Is a detection factor; siIs the activation parameter condition under the ith early warning level.
9. The construction environment-oriented multi-stage safety precaution method according to claim 8, characterized by a risk factor D1、D2、D3Comprises the following steps:
wherein N isiRepresents LiBehavior level in tag dataset CLThe statistical amount of (1).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210322348.7A CN114782988B (en) | 2022-03-29 | 2022-03-29 | Multistage safety early warning method oriented to construction environment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210322348.7A CN114782988B (en) | 2022-03-29 | 2022-03-29 | Multistage safety early warning method oriented to construction environment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114782988A true CN114782988A (en) | 2022-07-22 |
| CN114782988B CN114782988B (en) | 2024-04-02 |
Family
ID=82427554
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210322348.7A Active CN114782988B (en) | 2022-03-29 | 2022-03-29 | Multistage safety early warning method oriented to construction environment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114782988B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116503797A (en) * | 2023-04-10 | 2023-07-28 | 盐城工学院 | Medical waste treatment personnel protection tool wearing detection method based on target detection |
| CN117041502A (en) * | 2023-10-10 | 2023-11-10 | 湖南睿图智能科技有限公司 | Dangerous scene analysis and monitoring system and method based on machine vision |
| CN117168633A (en) * | 2023-10-20 | 2023-12-05 | 南通豪强电器设备有限公司 | High-low voltage complete equipment protection method and system based on temperature monitoring |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110119656A (en) * | 2018-02-07 | 2019-08-13 | 中国石油化工股份有限公司 | Intelligent monitor system and the scene monitoring method violating the regulations of operation field personnel violating the regulations |
| CN110121053A (en) * | 2018-02-07 | 2019-08-13 | 中国石油化工股份有限公司 | A kind of video monitoring method of situ of drilling well risk stratification early warning |
| US20200117903A1 (en) * | 2018-10-10 | 2020-04-16 | Autodesk, Inc. | Architecture, engineering and construction (aec) construction safety risk analysis system and method for interactive visualization and capture |
| CN112785798A (en) * | 2020-12-04 | 2021-05-11 | 国网江苏省电力工程咨询有限公司 | Behavior analysis method for construction project constructors of electric power substation engineering |
| CN113052125A (en) * | 2021-04-09 | 2021-06-29 | 内蒙古科电数据服务有限公司 | Construction site violation image recognition and alarm method |
| KR20220007460A (en) * | 2020-07-10 | 2022-01-18 | 주식회사 포스코 | Method for determining whether to wear personal protective equipment and server for performing the same |
-
2022
- 2022-03-29 CN CN202210322348.7A patent/CN114782988B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110119656A (en) * | 2018-02-07 | 2019-08-13 | 中国石油化工股份有限公司 | Intelligent monitor system and the scene monitoring method violating the regulations of operation field personnel violating the regulations |
| CN110121053A (en) * | 2018-02-07 | 2019-08-13 | 中国石油化工股份有限公司 | A kind of video monitoring method of situ of drilling well risk stratification early warning |
| US20200117903A1 (en) * | 2018-10-10 | 2020-04-16 | Autodesk, Inc. | Architecture, engineering and construction (aec) construction safety risk analysis system and method for interactive visualization and capture |
| KR20220007460A (en) * | 2020-07-10 | 2022-01-18 | 주식회사 포스코 | Method for determining whether to wear personal protective equipment and server for performing the same |
| CN112785798A (en) * | 2020-12-04 | 2021-05-11 | 国网江苏省电力工程咨询有限公司 | Behavior analysis method for construction project constructors of electric power substation engineering |
| CN113052125A (en) * | 2021-04-09 | 2021-06-29 | 内蒙古科电数据服务有限公司 | Construction site violation image recognition and alarm method |
Non-Patent Citations (2)
| Title |
|---|
| 王伟;吕山可;张雨果;赵楚楠;何华刚;: "基于BIM与机器视觉技术结合的建筑施工危险区域入侵预警研究", 安全与环境工程, no. 02, 30 March 2020 (2020-03-30) * |
| 郭亨聪;黄玲;何斌;: "山区高速公路安全生产视频监控系统关键技术研究", 湖南交通科技, no. 03, 25 September 2018 (2018-09-25) * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116503797A (en) * | 2023-04-10 | 2023-07-28 | 盐城工学院 | Medical waste treatment personnel protection tool wearing detection method based on target detection |
| CN116503797B (en) * | 2023-04-10 | 2023-10-27 | 盐城工学院 | Medical waste treatment personnel protection tool wearing detection method based on target detection |
| CN117041502A (en) * | 2023-10-10 | 2023-11-10 | 湖南睿图智能科技有限公司 | Dangerous scene analysis and monitoring system and method based on machine vision |
| CN117041502B (en) * | 2023-10-10 | 2023-12-08 | 湖南睿图智能科技有限公司 | Dangerous scene analysis and monitoring system and method based on machine vision |
| CN117168633A (en) * | 2023-10-20 | 2023-12-05 | 南通豪强电器设备有限公司 | High-low voltage complete equipment protection method and system based on temperature monitoring |
| CN117168633B (en) * | 2023-10-20 | 2024-02-02 | 南通豪强电器设备有限公司 | High-low voltage complete equipment protection method and system based on temperature monitoring |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114782988B (en) | 2024-04-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114782988B (en) | Multistage safety early warning method oriented to construction environment | |
| CN110738127B (en) | Helmet identification method based on unsupervised deep learning neural network algorithm | |
| CN111898514B (en) | Multi-target visual supervision method based on target detection and action recognition | |
| CN108564069B (en) | Video detection method for industrial safety helmet | |
| CN111986228B (en) | Pedestrian tracking method, device and medium based on LSTM model escalator scene | |
| CN110414400B (en) | Automatic detection method and system for wearing of safety helmet on construction site | |
| Junaedi et al. | Driver drowsiness detection based on face feature and PERCLOS | |
| CN105070053A (en) | Intelligent traffic monitoring camera for identifying vehicle illegal movement modes | |
| CN107133569A (en) | The many granularity mask methods of monitor video based on extensive Multi-label learning | |
| CN110728252A (en) | Face detection method applied to regional personnel motion trail monitoring | |
| Rentao et al. | Indoor smoking behavior detection based on yolov3-tiny | |
| CN115223249A (en) | Quick analysis and identification method for unsafe behaviors of underground personnel based on machine vision | |
| CN104077591A (en) | Intelligent and automatic computer monitoring system | |
| Wang et al. | Vision-based highway traffic accident detection | |
| CN117423157A (en) | Mine abnormal video action understanding method combining migration learning and regional invasion | |
| CN112183532A (en) | Safety helmet identification method based on weak supervision collaborative learning algorithm and storage medium | |
| CN117492814A (en) | Production operation and maintenance changing method and device | |
| Lee et al. | Modeling of human walking trajectories for surveillance | |
| CN116503809A (en) | Post-processing method for intelligent factory behavior wearing false alarm filtering | |
| CN113379247B (en) | Modeling method and system for enterprise potential safety hazard tracking model | |
| CN115169673A (en) | Intelligent campus epidemic risk monitoring and early warning system and method | |
| CN115147930A (en) | Big data video AI analytic system based on artificial intelligence | |
| CN114792437A (en) | Method and system for analyzing safe driving behavior based on facial features | |
| CN113191273A (en) | Oil field well site video target detection and identification method and system based on neural network | |
| Zhang et al. | Semantic segmentation of point clouds of field obstacle-crossing terrain for multi-legged rescue equipment based on random forest |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |