阅读说明：本技术 通过大数据进行智慧工地的安全行为分析方法 (Safety behavior analysis method for intelligent construction site through big data ) 是由 万里 熊榆 洪敏� 白金龙 胡宇 唐良艳 于 2020-11-03 设计创作，主要内容包括：本发明提供了一种通过大数据进行智慧工地的安全行为分析方法,包括：通过建筑工地图像的等比例放大图中获取工地人员图像数据,构建空间直方图数组对纹理刻画完成的工地人员图像转化成灰度图Grey,建立行为限制函数D-i逐渐提取安全行为和危险行为。(The invention provides a method for analyzing safety behaviors of an intelligent construction site through big data, which comprises the following steps: building site personnel image data are obtained through the equal proportion enlarged image of the building site image, and a space histogram array is constructed Converting the image of the construction site personnel with the described texture into a Grey scale graph Grey, and establishing a behavior restriction function D i Safety behaviors and dangerous behaviors are gradually extracted.)

1. A safety behavior analysis method for an intelligent construction site through big data is characterized by comprising the following steps:

s1, acquiring the image data of the construction site personnel through the equal proportion enlarged image of the construction site image, and constructing a space histogram arrayQ(I_j) Extraction of characteristic images for the workers at the construction site, I_jFor the enlarged image of the workers on the construction site, extraction is carried out by the trisection method, the symbols are U and [ 2 ]]All represent a combination of the histograms,

carrying out accumulation summation on the square image array H to form an enhanced image; extracting texture features in the histogram array on the basis of a histogram of gray levels of the enhanced image,

H_feature＝125[f(I_X/X_n)-f(I_Y/Y_n)]·116[f(I_X/X_n)+f(I_Y/Y_n)]counting the gray level of the texture features of the enhanced image, and setting X_nIs 35.879, Y_nTo 113.245, by calculating the channel image function f (I) of the X-axis_X/X_n) Thereby depicting the texture of the X-axis and calculating the channel image function f (I) of the Y-axis_Y/Y_n) Thereby characterizing the Y-axis texture.

2. The method of claim 1 for analyzing safety behaviors of an intelligent construction site through big data, comprising the steps of:

and S2, converting the texture-delineated construction site personnel image into a Grey scale graph Grey, wherein the formula is as follows:

the behavior monitoring model for RGB is defined as:

E_i,j(σ₁and theta) represents given (sigma)₁θ) obtaining values of the behavior quantity i and the gray level j within the window of the safety behavior of the building site personnel, wherein; sigma₁To obtain adjusted distance values for security behavior pixels, F_i,j(σ₂And theta) represents given (sigma)₂θ) number of actions i and gray within the worksite personnel hazardous actions Window of the pairObtaining values of degree j, wherein; sigma₁In order to obtain the adjustment distance value of the dangerous behavior pixel point, in order to improve the operation efficiency, the gray level is subjected to degradation processing; the distance adjustment value is equally divided and jumped according to the gray level definition; theta is selected from 0 degree, 90 degrees, 180 degrees and 270 degrees.

3. The method of claim 1 for analyzing safety behaviors of an intelligent construction site through big data, comprising the steps of:

s3, establishing a behavior limiting function D_iThe safe behavior and the dangerous behavior are gradually extracted,

k_ifor secure action image frame feature sets, p_i+1Detecting index phi for the next image frame feature set of dangerous behaviors through the safety behavior of the personnel in the construction site_iRespectively associated with the X-axis gray scale feature points L_xAnd Y-axis gray feature point M_yAfter multiplication, comparing the two through the construction site personnel characteristic set F to obtain a time vector t of the safety behavior₁And a safety behavior characteristic factor u₁Squaring convolution vector of safety behavior and time vector t of dangerous behavior₂And a dangerous behavior characteristic factor u₂Squaring the convolution vector of security behavior, where O_tDensity means representing the performance of the staff at the worksite.

4. The method of claim 1 for analyzing safety behaviors of an intelligent construction site through big data, comprising the steps of:

s4, clustering feature sets F of building site safety behavior images in the safety behavior identification process, and after preliminary screening, performing model judgment on image data of different building site personnel feature sets C; extracting attribute values of the specific characteristic image, constructing texture information,

helmet attribute valueWherein psi_iFor a construction site personnel helmet wearing behavior attribute value at a certain time,for wearing the safety helmet control factor in the safety behavior,the helmet control factor is not worn in dangerous behaviors,

attribute value of walking speedWherein r is_iFor the attribute value of the walking speed behavior of the building site personnel at a certain moment,a control factor for the normal walking speed in the safe behavior,an abnormal walking speed control factor in dangerous behaviors,

attribute value of reflective vestWherein gamma is_iFor the staff wearing the reflective vest at a certain time,in order to wear reflective vest control factors in safety activities,in order to avoid wearing the reflective vest control factor in dangerous behaviors,

hand-held item attribute valuesWherein eta_iFor the property value of the behavior of the building site personnel holding the article at a certain moment,for the hand-held article control factor in security activities,the control factors for the articles are not held by hands in the dangerous behaviors, the corresponding moving radius is also considered due to the influence of the length and the width of the articles held by the building site personnel, if the building site personnel hold bricks by hands, the moving radii of other building site personnel are relatively large, if the building site personnel hold reinforcing steel bars by shoulders, the moving radii of other building site personnel are relatively small, and the behavior control factors judged by the attribute values are fast in convergence, so that the dangerous behaviors are prevented;

violation retrograde attribute valueWhereinFor a worker violation retrograde behavior attribute value at a time,for non-violating retrograde control factors in the security activities,for the violation of the retrograde control factor in the dangerous behavior,

different construction sites have different site conditions, and different attribute values are adjusted to adapt to different site behavior identification, for example, some sites do not need to wear luminous waistcoats, and some sites do not have safe walking guide lines, so that illegal retrograde motion does not exist; and adjusting the actually generated attribute value according to the actual situation.

5. The method of claim 1 for analyzing safety behaviors of an intelligent construction site through big data, comprising the steps of:

s5, extracting the texture characteristics of the helmet attribute value, the walking speed attribute value, the reflective waistcoat attribute value, the handheld article attribute value and the illegal converse attribute value, screening by corresponding control factors,

in optimally obtaining site personnel safety behavior image distance attributes

Through construction site personnel safety behavior weight vector J and safety behavior vector transposition J^TAnd safety behavior extraction sample u_iObtaining the distance attribute of the safety behavior image through the convolution calculation after multiplication, and performing dangerous behavior weight vector K and dangerous behavior vector transposition K on the construction site personnel^TSample c taken in conjunction with dangerous behavior_iAnd performing the multiplied convolution calculation to obtain the distance attribute of the dangerous behavior image.

6. The method of claim 1 for analyzing safety behaviors of an intelligent construction site through big data, comprising the steps of:

since the corresponding behavior attribute has been determined in S4, the behavior is subjected to attribute determination through step S5, and the distance attribute is used to determine a critical value of the security behavior image, that is, whether the corresponding picture attribute reaches the degree of security behavior determination, and is referred to as a distance attribute;

optimized safety behavior convergence function

And optimizing the dangerous behavior convergence function

7. The method of claim 1 for analyzing safety behaviors of an intelligent construction site through big data, comprising the steps of:

s6, judging according to the safety behavior estimation value

Wherein tau is a result control operator;

and if the safety behavior is higher than the safety behavior estimation value, the safety behavior is listed as the safety behavior, and if the safety behavior is lower than the safety behavior estimation value, the dangerous behavior is listed as the dangerous behavior.

Technical Field

The invention relates to the field of image recognition, in particular to a safety behavior analysis method for an intelligent construction site through big data.

Background

Because the site safety needs to be monitored in real time in the building engineering construction process, but the omission is caused by manual inspection, but the danger degree in the building engineering is increased, although some building engineering construction main bodies are provided with monitors or cameras, the monitors or the cameras can not automatically judge the safety behavior only by real-time checking of monitoring room personnel, and even if a software method for behavior identification is provided, the extracted behavior data is inaccurate and incomplete,

disclosure of Invention

The invention aims to at least solve the technical problems in the prior art, and particularly innovatively provides a novel

In order to achieve the above object, the present invention provides a method for analyzing safety behavior of an intelligent construction site through big data, comprising:

s1, acquiring the image data of the construction site personnel through the equal proportion enlarged image of the construction site image, and constructing a space histogram arrayQ(I_j) Extraction of characteristic images for the workers at the construction site, I_jFor the enlarged image of the workers on the construction site, extraction is carried out by the trisection method, the symbols are U and [ 2 ]]All represent a combination of the histograms,

carrying out accumulation summation on the square image array H to form an enhanced image; extracting texture features in the histogram array on the basis of a histogram of gray levels of the enhanced image,

H_feature＝125[f(I_X/X_n)-f(I_Y/Y_n)]·116[f(I_X/X_n)+f(I_Y/Y_n)]counting the gray level of the texture features of the enhanced image, and setting X_nIs 35.879, Y_nTo 113.245, by calculating the channel image function f (I) of the X-axis_X/X_n) Thereby depicting the texture of the X-axis and calculating the channel image function f (I) of the Y-axis_Y/Y_n) Thereby depicting the Y-axis texture;

and S2, converting the texture-delineated construction site personnel image into a Grey scale graph Grey, wherein the formula is as follows:

the behavior monitoring model for RGB is defined as:

E_i,j(σ₁and theta) represents given (sigma)₁θ) obtaining values of the behavior quantity i and the gray level j within the window of the safety behavior of the building site personnel, wherein; sigma₁To obtain adjusted distance values for security behavior pixels, F_i,j(σ₂And theta) represents given (sigma)₂Theta) obtaining values of the behavior quantity i and the gray level j in the construction site personnel dangerous behavior window, wherein; sigma₁In order to obtain the adjustment distance value of the dangerous behavior pixel point, in order to improve the operation efficiency, the gray level is subjected to degradation processing; the distance adjustment value is equally divided and jumped according to the gray level definition; selecting theta at 0 degree, 90 degrees, 180 degrees and 270 degrees;

s3, establishing a behavior limiting function D_iThe safe behavior and the dangerous behavior are gradually extracted,

k_ifor secure action image frame feature sets, p_i+1Detecting index phi for the next image frame feature set of dangerous behaviors through the safety behavior of the personnel in the construction site_iRespectively associated with the X-axis gray scale feature points L_xAnd Y-axis gray feature point M_yAfter multiplication, comparing the two through the construction site personnel characteristic set F to obtain a time vector t of the safety behavior₁And a safety behavior characteristic factor u₁Squaring convolution vector of safety behavior and time vector t of dangerous behavior₂And a dangerous behavior characteristic factor u₂Squaring the convolution vector of security behavior, where O_tA density average representing the behavior of the worksite personnel,

s4, clustering feature sets F of building site safety behavior images in the safety behavior identification process, and after preliminary screening, performing model judgment on image data of different building site personnel feature sets C; extracting attribute values of the specific characteristic image, constructing texture information,

helmet attribute valueWherein psi_iFor a construction site personnel helmet wearing behavior attribute value at a certain time,for wearing the safety helmet control factor in the safety behavior,the helmet control factor is not worn in dangerous behaviors,

attribute value of walking speedWherein r is_iFor the attribute value of the walking speed behavior of the building site personnel at a certain moment,a control factor for the normal walking speed in the safe behavior,an abnormal walking speed control factor in dangerous behaviors,

attribute value of reflective vestWherein gamma is_iFor the staff wearing the reflective vest at a certain time,in order to wear reflective vest control factors in safety activities,not wearing reflective horse for dangerous behaviorThe factor for controlling the growth of the nail,

hand-held item attribute valuesWherein eta_iFor the property value of the behavior of the building site personnel holding the article at a certain moment,for the hand-held article control factor in security activities,the control factors for the articles are not held by hands in the dangerous behaviors, the corresponding moving radius is also considered due to the influence of the length and the width of the articles held by the building site personnel, if the building site personnel hold bricks by hands, the moving radii of other building site personnel are relatively large, if the building site personnel hold reinforcing steel bars by shoulders, the moving radii of other building site personnel are relatively small, and the behavior control factors judged by the attribute values are fast in convergence, so that the dangerous behaviors are prevented;

violation retrograde attribute valueWhereinFor a worker violation retrograde behavior attribute value at a time,for non-violating retrograde control factors in the security activities,for the violation of the retrograde control factor in the dangerous behavior,

different construction sites have different site conditions, and different attribute values are adjusted to adapt to different site behavior identification, for example, some sites do not need to wear luminous waistcoats, and some sites do not have safe walking guide lines, so that illegal retrograde motion does not exist; adjusting the actually generated attribute value according to the actual situation;

s5, extracting the texture characteristics of the helmet attribute value, the walking speed attribute value, the reflective waistcoat attribute value, the handheld article attribute value and the illegal converse attribute value, screening by corresponding control factors,

in optimally obtaining site personnel safety behavior image distance attributes

Through construction site personnel safety behavior weight vector J and safety behavior vector transposition J^TAnd safety behavior extraction sample u_iObtaining the distance attribute of the safety behavior image through the convolution calculation after multiplication, and performing dangerous behavior weight vector K and dangerous behavior vector transposition K on the construction site personnel^TSample c taken in conjunction with dangerous behavior_iThe multiplied convolution calculation obtains the distance attribute of the dangerous behavior image,

since the corresponding behavior attribute has been determined in S4, the behavior is subjected to attribute determination through step S5, and the distance attribute is used to determine a critical value of the security behavior image, that is, whether the corresponding picture attribute reaches the degree of security behavior determination, and is referred to as a distance attribute;

optimized safety behavior convergence function

And optimizing the dangerous behavior convergence function

S6, judging according to the safety behavior estimation value

Wherein tau is a result control operator;

and if the safety behavior is higher than the safety behavior estimation value, the safety behavior is listed as the safety behavior, and if the safety behavior is lower than the safety behavior estimation value, the dangerous behavior is listed as the dangerous behavior.

In summary, due to the adoption of the technical scheme, the invention has the beneficial effects that:

the method comprises the steps of obtaining building site image data through an image acquisition module, carrying out screening operation on building site personnel images, obtaining area frame images of building site personnel features, carrying out equal-proportion image amplification, extracting and identifying through a human body feature histogram, effectively measuring human body information, carrying out image clustering, judging behavior types of the building site, quickly judging behaviors, and refining dangerous behaviors.

Additional aspects and advantages of the invention will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the invention.

Drawings

The above and/or additional aspects and advantages of the present invention will become apparent and readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings of which:

fig. 1 is a general schematic of the present invention.

Detailed Description

Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to the same or similar elements or elements having the same or similar function throughout. The embodiments described below with reference to the accompanying drawings are illustrative only for the purpose of explaining the present invention, and are not to be construed as limiting the present invention.

As shown in fig. 1, the invention discloses a method for analyzing safety behaviors of an intelligent construction site through big data, which comprises the following steps:

building site image acquisition is carried out through an image acquisition module, and image screening preparation is synchronously carried out on block chain link points; acquiring human body dynamic behavior data by using the construction site image, and capturing the image with the human body dynamic behavior data; in the grabbing process, a human body characteristic diagram is usedImage weight vector C_iExtracting personnel from the construction site image, and estimating the number of workers on the construction site through similarity discrimination, wherein M is the conditional probability of the basic feature a of the workers on the construction site, F is the feature set of the workers on the construction site, b is the behavior feature of the workers on the construction site, and eta is a behavior limiting factor, wherein F belongs to { hellet, bodyspeed, vest, arm, anti-diagnosis }, and collecting the characteristics of behaviors of helmet wearing, walking speed, reflective waistcoat, handheld articles, illegal converse and the like;

forming expected estimation relations

Wherein u is_iRepresenting a worksite personnel performance result; i represents the number of actions taking place, including safety and hazardous actions; v. of_iRepresenting the probability of occurrence of dangerous behaviors of the staff at the construction site; l (u)_i) Is a worksite personnel number weight function, which is related to the probability of dangerous behavior occurrence; q (v)_i) Is a gain function of the dangerous behavior of the staff in the construction site, is an overall evaluation of the deviation of the dangerous behavior from a reference point, w_iPreliminarily screening conditional probability values of the construction site personnel region frames in the construction site images, wherein beta is a construction site personnel behavior correction parameter;

the worksite personnel number weighting function represents a probabilistic measure of the number of worksite personnel acquired from the construction site image, influences the design worksite personnel safety behavior decisions by anticipating changes in the estimated relationship,

where S is the safety behavior expectation, Δ P_i ^δExtracting deviation value of safety behavior image from construction site image, delta is safety behavior deviation correction value, and superscript W_iFitting parameters for safety behavior, superscript Z_iFitting parameters for the dangerous behavior, epsilon is a dangerous behavior deviation correction value,

the occurrence probability of the image safety behaviors of the construction site is influenced by judgment and evaluation of working environment by workers and habits of safety behaviors of other workers, so that the behavior targets are corrected through the dangerous behavior deviation correction values, and more accurate weight functions can be fitted.

Dividing the obtained area frame image into continuous windows through the construction site image, carrying out amplification operation on each window containing the construction site personnel image, and calculating a construction site image window pair (N)_k,N_k+1) Average value of each channel in color space; form U e (N)_k,N_k+1)；

The equal-scale magnification weight function is

Wherein, label (N)_k) Obtaining a function, label (N), for a current frame of a construction site image_k+1) Obtaining a function for the next frame of the building site image, m is greater than 1, mu is an image frame constraint factor, omega₁Adjusting parameters, omega, for a current frame of a construction site image₂The parameters are adjusted for the next frame of the building site image,

wherein q is₁For safety action preference coefficient, q₂A risk behavior preference coefficient;

comparing the accumulated number of the construction site personnel obtained from the current construction site image window pair, amplifying the construction site image window pair when the change is more than a threshold value T, and moving to the next construction site image window pair to continuously search the construction site personnel if the change is not more than the threshold value T;

s1, acquiring the image data of the construction site personnel through the equal proportion enlarged image of the construction site image, and constructing a space histogram arrayQ(I_j) Is composed ofExtraction of characteristic images by the workers at the construction site, I_jFor the enlarged image of the workers on the construction site, extraction is carried out by the trisection method, the symbols are U and [ 2 ]]All represent a combination of the histograms,

carrying out accumulation summation on the square image array H to form an enhanced image; extracting texture features in the histogram array on the basis of a histogram of gray levels of the enhanced image,

H_feature＝125[f(I_X/X_n)-f(I_Y/Y_n)]·116[f(I_X/X_n)+f(I_Y/Y_n)]

counting the gray level of the texture features of the enhanced image, and setting X_nIs 35.879, Y_nTo 113.245, by calculating the channel image function f (I) of the X-axis_X/X_n) Thereby depicting the texture of the X-axis and calculating the channel image function f (I) of the Y-axis_Y/Y_n) Thereby depicting the Y-axis texture;

and S2, converting the texture-delineated construction site personnel image into a Grey scale graph Grey, wherein the formula is as follows:

the behavior monitoring model for RGB is defined as:

E_i,j(σ₁and theta) represents given (sigma)₁θ) obtaining values of the behavior quantity i and the gray level j within the window of the safety behavior of the building site personnel, wherein; sigma₁To obtain adjusted distance values for security behavior pixels, F_i,j(σ₂And theta) represents given (sigma)₂Theta) obtaining values of the behavior quantity i and the gray level j in the construction site personnel dangerous behavior window, wherein; sigma₁In order to obtain the adjustment distance value of the dangerous behavior pixel point, in order to improve the operation efficiency, the gray level is subjected to degradation processing; the distance adjustment value performs equal jump according to the gray level definitionRotating; selecting theta at 0 degree, 90 degrees, 180 degrees and 270 degrees;

s3, establishing a behavior limiting function D_iThe safe behavior and the dangerous behavior are gradually extracted,

k_ifor secure action image frame feature sets, p_i+1Detecting index phi for the next image frame feature set of dangerous behaviors through the safety behavior of the personnel in the construction site_iRespectively associated with the X-axis gray scale feature points L_xAnd Y-axis gray feature point M_yAfter multiplication, comparing the two through the construction site personnel characteristic set F to obtain a time vector t of the safety behavior₁And a safety behavior characteristic factor u₁Squaring convolution vector of safety behavior and time vector t of dangerous behavior₂And a dangerous behavior characteristic factor u₂Squaring the convolution vector of security behavior, where O_tA density average representing the behavior of the worksite personnel,

s4, clustering feature sets F of building site safety behavior images in the safety behavior identification process, and after preliminary screening, performing model judgment on image data of different building site personnel feature sets C; extracting attribute values of the specific characteristic image, constructing texture information,

helmet attribute valueWherein psi_iFor a construction site personnel helmet wearing behavior attribute value at a certain time,for wearing the safety helmet control factor in the safety behavior,the helmet control factor is not worn in dangerous behaviors,

attribute value of walking speedWherein r is_iFor the attribute value of the walking speed behavior of the building site personnel at a certain moment,a control factor for the normal walking speed in the safe behavior,an abnormal walking speed control factor in dangerous behaviors,

attribute value of reflective vestWherein gamma is_iFor the staff wearing the reflective vest at a certain time,in order to wear reflective vest control factors in safety activities,in order to avoid wearing the reflective vest control factor in dangerous behaviors,

hand-held item attribute valuesWherein eta_iFor the property value of the behavior of the building site personnel holding the article at a certain moment,for the hand-held article control factor in security activities,for the control factor of the article not held by the hand in the dangerous behaviors, the corresponding moving radius is also considered due to the influence of the length and the width of the article held by the building site personnel, if the building site personnel hold bricks by hands, the moving radius of other building site personnel is relatively larger, if the building site personnel hold the reinforcing steel bars by shoulders, the moving radius of other building site personnel is relatively smaller, and the judgment is carried out according to the attribute valueThe convergence of the disconnected behavior control factor is fast, so that dangerous behaviors are prevented;

violation retrograde attribute valueWhereinFor a worker violation retrograde behavior attribute value at a time,for non-violating retrograde control factors in the security activities,for the violation of the retrograde control factor in the dangerous behavior,

different construction sites have different site conditions, and different attribute values are adjusted to adapt to different site behavior identification, for example, some sites do not need to wear luminous waistcoats, and some sites do not have safe walking guide lines, so that illegal retrograde motion does not exist; adjusting the actually generated attribute value according to the actual situation;

s5, extracting the texture characteristics of the helmet attribute value, the walking speed attribute value, the reflective waistcoat attribute value, the handheld article attribute value and the illegal converse attribute value, screening by corresponding control factors,

in optimally obtaining site personnel safety behavior image distance attributes

Through construction site personnel safety behavior weight vector J and safety behavior vector transposition J^TAnd safety behavior extraction sample u_iObtaining the distance attribute of the safety behavior image through the convolution calculation after multiplication, and performing dangerous behavior weight vector K and dangerous behavior vector transposition K on the construction site personnel^TSample c taken in conjunction with dangerous behavior_iCalculating the distance of the acquired dangerous behavior image by the multiplied convolutionThe attribute of the tag is set to be off,

since the corresponding behavior attribute has been determined in S4, the behavior is subjected to attribute determination through step S5, and the distance attribute is used to determine a critical value of the security behavior image, that is, whether the corresponding picture attribute reaches the degree of security behavior determination, and is referred to as a distance attribute;

optimized safety behavior convergence function

And optimizing the dangerous behavior convergence function

S6, judging according to the safety behavior estimation value

Wherein tau is a result control operator;

if the safety behavior estimated value is higher than the safety behavior estimated value, the safety behavior is classified as safety behavior, if the safety behavior estimated value is lower than the safety behavior estimated value, the safety behavior is classified as dangerous behavior, and meanwhile, the safety behavior estimated value is synchronized at the block link point in real time.

While embodiments of the invention have been shown and described, it will be understood by those of ordinary skill in the art that: various changes, modifications, substitutions and alterations can be made to the embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.