A Computer Vision-Based Risk Assessment Method for Worker Intrusion into Hazardous Construction Areas

Abstract

This invention relates to a computer vision-based method for risk assessment of worker intrusion into hazardous construction areas, comprising: acquiring video information of the worker's construction area; performing worker target detection and hazardous area labeling on the video information, extracting the coordinates of key points of the human skeleton and the coordinates of the hazardous area boundary line; extracting the distance of the human body from the hazardous area boundary line, the direction of human walking movement, the step frequency of human walking movement, and the gait periodicity stability of human walking movement based on the coordinates of the key points of the human skeleton and the coordinates of the hazardous area boundary line, thereby obtaining risk assessment data for each indicator; calculating the weight of each indicator based on the volatility of the risk assessment data for each indicator; and obtaining the risk assessment result of worker intrusion into hazardous construction areas based on the risk assessment data of each indicator and its corresponding weight. Compared with the prior art, this invention has the advantages of being able to obtain refined risk assessment results for worker intrusion into hazardous construction areas, taking into account individual differences among workers.

Description

Technical Field

[0001] This invention relates to the field of risk assessment technology for worker intrusion into hazardous construction areas, and in particular to a computer vision-based method for risk assessment of worker intrusion into hazardous construction areas. Background Technology

[0002] Worker intrusion into hazardous construction areas has always been a key focus of construction safety management and a major cause of various human-caused safety accidents. Construction sites are characterized by complex environments, diverse area types, large numbers of workers and machinery, and constantly changing movement patterns. Therefore, identifying and assessing the risks of worker intrusion into hazardous construction areas is crucial for improving the efficiency of construction site safety management and reducing the occurrence of accidents. Computer vision-based identification methods, due to their low cost and non-intrusive nature, are increasingly attracting the attention of industry scholars and construction safety managers.

[0003] In recent years, with the continuous improvement of construction site monitoring systems and the development of computer vision technology, computer vision-based methods for worker behavior recognition and feature extraction have emerged and are gradually being applied in practice. Based on the layout of hazardous areas on construction sites and combined with trajectory intersection theory, computer vision technology can be used to identify and conduct preliminary risk assessments of workers' intrusion into hazardous areas. However, due to the complexity of human movement, simply identifying and analyzing the overall behavior of workers is insufficient to improve the precision of construction safety management.

[0004] Previous studies have primarily focused on qualitative risk assessments of worker intrusion into hazardous construction areas, based on safety regulations or rules. Quantitative risk assessments, however, have been less frequently discussed due to a lack of effective evaluation indicators. Identification methods that treat workers as a collective target fail to consider individual differences in worker movement, hindering the efficiency of risk assessments for such intrusions. Furthermore, while human walking movements exhibit periodic variations, no solutions have been proposed for utilizing this movement characteristic in risk assessments of unsafe behaviors such as worker intrusion into hazardous construction areas. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the existing technology and provide a computer vision-based risk assessment method for worker intrusion into construction hazard areas that takes into account the periodic changes in human walking movements (actions), thereby achieving a more accurate risk assessment of worker intrusion into construction hazard areas that takes into account individual differences.

[0006] The objective of this invention can be achieved through the following technical solutions:

[0007] A computer vision-based risk assessment method for worker intrusion into hazardous construction areas includes the following steps:

[0008] Obtain video information of the workers' construction area;

[0009] Worker target detection and hazardous area labeling are performed on the video information, and the coordinates of key points of the human skeleton and the coordinates of the boundary line of the hazardous area are extracted.

[0010] Based on the coordinates of key points on the human skeleton and the coordinates of the boundary line of the danger zone, the distance between the human body and the boundary line of the danger zone, the direction of human walking movement, the step frequency of human walking movement, and the stability of the gait cycle of human walking movement are extracted to obtain risk assessment data for each indicator.

[0011] The weight of each indicator is calculated based on the volatility of the risk assessment data for each indicator.

[0012] Based on the risk assessment data and corresponding weights of each indicator, the risk assessment results of workers intruding into the dangerous construction area are obtained.

[0013] Furthermore, the coordinates of key points of the human skeleton include the coordinates of key points of the neck, shoulders, hips, and feet.

[0014] The calculation process of the distance between the human body and the boundary line of the danger zone includes: taking the coordinates of the key point of the human neck as the human body reference point and the coordinates of the boundary line of the danger zone as the area reference line, calculating the shortest distance between the human body reference point and the area reference line on the horizontal axis.

[0015] The coordinates of key points in the hip include the coordinates of key points at the center of the body and on both sides of the hip joint.

[0016] The calculation process for the direction of human walking includes: obtaining the direction of human walking based on the line connecting the coordinates of key points on the left and right sides of the human body at the hip joint;

[0017] The calculation process of the human walking frequency includes: calculating the human walking frequency based on the periodic change frequency of the distance between the coordinates of the hip key point and the coordinates of the foot key point.

[0018] The calculation process for the gait cycle stability of human walking includes: taking the coordinates of the key point at the center of the body at the hip joint as the center of gravity of the human body, selecting the vertical distance relationship between the center of gravity of the human body and the coordinates of the key points of the feet, and calculating the gait cycle stability of human walking.

[0019] Furthermore, the video information is also marked with buffer boundary lines, and the method further includes cropping the video information to obtain risk assessment data for each indicator based on the cropped video information;

[0020] The interception process includes:

[0021] Based on the worker target detection results, the starting point of time is taken as the coordinates of any point in the key points of the worker's human skeleton entering the buffer zone coordinate range, the intermediate point is taken as the time t when the worker enters the buffer zone and turns 45°, and the ending point is taken as the time t after passing the intermediate point and experiencing the same time t again. The captured video information is obtained.

[0022] Furthermore, the distribution of the buffer zone boundary lines is consistent with the boundary of the danger zone, and the area corresponding to the buffer zone boundary lines is the area extending outward from the boundary of the danger zone. The width of the buffer zone boundary lines is obtained by adding the movement space of a human walking and the reaction space of a safety response.

[0023] Furthermore, the entropy weight method is used to calculate the weight of each indicator based on the volatility of the risk assessment data of each indicator.

[0024] Furthermore, the entropy weight method includes the following steps:

[0025] S401: Normalize the risk assessment data for each indicator;

[0026] S402: Calculate the information entropy of each indicator after data normalization;

[0027] S403: Calculate the weight of each indicator based on the information entropy calculation results of each indicator.

[0028] Furthermore, the calculation expression for the data normalization is as follows:

[0029]

[0030] In the formula, min and max are the minimum and maximum values ​​in the sample, v ij α represents the j-th value in the i-th sample after matrix normalization, and α is the normalization coefficient;

[0031] The expression for calculating the information entropy is:

[0032]

[0033] In the formula, E j is the information entropy value of the j-th indicator, k is the information entropy coefficient, m is the total number of sample values, and n is the total number of samples;

[0034] The formulas for calculating the weights of each indicator are as follows:

[0035]

[0036] In the formula, W j Let be the weight of the j-th indicator.

[0037] Furthermore, the method also includes obtaining risk assessment data and corresponding weights for each individual in the target group, thereby obtaining the overall indicator weight of the target group based on the weight data of each individual, and then conducting an intrusion behavior risk assessment for each individual in the target group based on the overall indicator weight.

[0038] Furthermore, the entropy weight method is used to calculate the overall index weight of the target group based on the weight data of each individual.

[0039] Furthermore, the calculation expression for the intrusion behavior risk assessment of each individual in the target group is as follows:

[0040]

[0041] In the formula, S i To give the final score for the risk assessment of the intrusion behavior of the i-th person, w j p represents the target group weight result for the j-th indicator. ij Let m be the score of the j-th risk assessment indicator for the i-th person, and m be the total number of indicators.

[0042] Compared with the prior art, the present invention has the following advantages:

[0043] (1) Considering that the behavior of workers invading dangerous construction areas will change periodically, this invention sets up four risk assessment indicators for intrusion behavior that take into account the gait movement characteristics of human walking. The risk level of intrusion behavior is assessed from the perspectives of distance, direction, step frequency and gait period stability, respectively. Based on this, the risk avoidance ability of the intruder is calculated, thereby quantifying the risk of intrusion behavior.

[0044] This application takes into account the individual differences in the worker movement and can obtain more accurate and reliable risk assessment results for intrusion into dangerous construction areas based on the characteristics of individual workers.

[0045] (2) This application determines the corresponding indicator weights based on the volatility of each risk assessment indicator. The smaller the volatility, the higher the stability of the behavior, and the lower the risk of the intrusion behavior, thus accurately determining the corresponding risk assessment results.

[0046] (3) This application performs secondary weighting calculations of risk assessment indicators for the target group. If the secondary weight of a certain risk assessment indicator is high, it indicates that the data difference of the primary weight result of the indicator in the group is large, that is, the overall performance of the indicator is unbalanced and can reflect the risk situation of the group. This application further uses the secondary weighting of the risk assessment indicators of the group to assess the risk of individual intrusion behavior. Based on the general principle (assumption) that the majority is better than the minority, this method can reduce the influence of environmental or group similarity by considering the characteristics of the group and improve the accuracy of individual risk assessment. Attached Figure Description

[0047] Figure 1 This is a schematic diagram illustrating the calculation of various risk assessment indicators provided in this embodiment of the invention;

[0048] Figure 2 This is a flowchart illustrating a computer vision-based risk assessment method for worker intrusion into hazardous construction areas, as provided in an embodiment of the present invention.

[0049] Figure 3 This is a schematic diagram of a worker's work area, a danger zone, and a buffer zone provided in an embodiment of the present invention. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0051] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0052] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0053] Example 1

[0054] like Figure 2 As shown, this embodiment provides a computer vision-based method for risk assessment of worker intrusion into hazardous construction areas, including the following steps:

[0055] S1: Acquire video information of workers moving in construction areas near hazardous areas;

[0056] S2: Perform worker target detection and hazardous area labeling on video information, and extract the coordinates of key points of the human skeleton and the coordinates of the boundary line of the hazardous area;

[0057] The coordinates of key points in the human skeleton specifically include the coordinates of key points in the neck, shoulders, hips, and feet.

[0058] S3: Based on the coordinates of key points of the human skeleton and the boundary line of the danger zone, extract the distance of the human body from the boundary line of the danger zone, the direction of human walking movement, the step frequency of human walking movement, and the stability of the gait cycle of human walking movement, thereby obtaining risk assessment data for each indicator;

[0059] The calculation process of the distance between the human body and the boundary line of the danger zone includes: taking the coordinates of the key points of the human neck as the human body reference point and the coordinates of the boundary line of the danger zone as the area reference line, calculating the shortest distance between the human body reference point and the area reference line on the horizontal axis.

[0060] Key point coordinates of the hip include the coordinates of key points at the center of the body and on both sides of the hip joint.

[0061] The calculation process for the direction of human walking motion includes: obtaining the direction of human walking motion by connecting the coordinates of key points on the left and right sides of the body at the hip joint;

[0062] The process of calculating the cadence of human walking includes: calculating the cadence of human walking based on the periodic change frequency of the distance between the coordinates of the hip key point and the coordinates of the foot key point.

[0063] The calculation process of gait cycle stability of human walking includes: taking the coordinates of the key point at the center of the body at the hip joint as the center of gravity of the human body, selecting the vertical distance relationship between the center of gravity of the human body and the coordinates of the key points of the feet, and calculating the gait cycle stability of human walking.

[0064] S4: Calculate the weight of each indicator based on the volatility of the risk assessment data for each indicator;

[0065] The preferred method is to use the entropy weight method to calculate the weight of each indicator based on the volatility of the risk assessment data of each indicator.

[0066] The entropy weight method includes the following steps:

[0067] S401: Normalize the risk assessment data for each indicator;

[0068] S402: Calculate the information entropy of each indicator after data normalization;

[0069] S403: Calculate the weight of each indicator based on the information entropy calculation results of each indicator.

[0070] S5: Based on the risk assessment data and corresponding weights of each indicator, obtain the risk assessment results of workers intruding into the construction hazard area.

[0071] As a preferred implementation, considering the computational requirements of the entropy weight method, the amount of data input for each assessment object should be the same. Therefore, buffer boundary lines are also marked in the video information. The method also includes cropping the video information to obtain risk assessment data for each indicator based on the cropped video information.

[0072] The interception process includes:

[0073] Based on the worker target detection results, the starting point of time is taken as the coordinates of any point in the key points of the worker's human skeleton entering the buffer zone coordinate range, the intermediate point is taken as the time t when the worker enters the buffer zone and turns 45°, and the ending point is taken as the time t after passing the intermediate point and experiencing the same time t again. The captured video information is obtained.

[0074] The distribution of the buffer zone boundary line is consistent with the boundary of the danger zone. The area corresponding to the buffer zone boundary line is the area extending outward from the boundary of the danger zone. The width of the buffer zone boundary line is obtained by adding the movement space of a human walking and the reaction space of a safety response.

[0075] Preferably, to improve the reliability of the above weights, the method further includes:

[0076] S6: Obtain risk assessment data and corresponding weights for each individual in the target group, and then obtain the overall indicator weight of the target group based on the weight data of each individual, and conduct intrusion risk assessment for each individual in the target group based on the overall indicator weight.

[0077] The overall index weight of the target group is calculated using the entropy weight method based on the weight data of each individual.

[0078] The formula for calculating the intrusion risk assessment of each individual in the target group is as follows:

[0079]

[0080] In the formula, S i To give the final score for the risk assessment of the intrusion behavior of the i-th person, w j p represents the weight result of the j-th indicator in the target group. ij Let be the score of the j-th risk assessment indicator for the i-th person. This score can be obtained by normalizing the risk assessment data or by human scoring. m is the total number of indicators.

[0081] The following is a detailed description of each of the above steps:

[0082] S1: By deploying high-definition cameras near hazardous areas of the construction site, data on workers' walking postures are collected. The skeleton information of workers is extracted using a computer vision-based target detection method, and a database of workers' potential intrusion into hazardous construction areas is established.

[0083] S2: The collected video stream data is transmitted to a computer device equipped with a target detection algorithm, and worker targets and hazardous areas are detected and marked using the Mask-R-CNN detection algorithm.

[0084] Based on the method for setting up safety management buffer zones, buffer zone markings are performed in video images in conjunction with the annotations of hazardous areas. The annotation information primarily consists of 2D coordinate information from the image, such as... Figure 3 As shown.

[0085] The distribution of the buffer zone is consistent with the boundary of the danger zone, based on the area extending outward from the boundary of the danger zone, with the specific width subject to actual conditions.

[0086] The width of the buffer zone is composed of the sum of the movement space (step length) of a human walking and the reaction space of a safety response.

[0087] The average human stride is approximately 0.45 to 0.6 meters. The reaction space is calculated based on the average walking speed v (1.2 to 1.5 m / s) and reaction time t (0.2 to 0.3 s), and is approximately 0.24 to 0.45 meters.

[0088] In practical applications, the above parameters of the buffer zone can be adjusted according to factors such as the characteristics of the worker group and site conditions. In the following implementation steps, the buffer zone width is set to 1m.

[0089] By using the Openpose3D skeleton recognition algorithm, worker skeleton information is extracted from the video stream and recorded together with the coordinate information of dangerous areas and buffer zones to establish a motion database of intrusion behavior.

[0090] The intrusion behavior motion database stores the 2D coordinates of key points of the human skeleton, boundary lines of danger zones, and boundary lines of buffer zones. In this invention, the key points of the human skeleton consist of key points of the neck, shoulders, hips, and feet that involve the aforementioned four types of risk assessment indicators, including distance, direction, cadence, and gait cycle stability.

[0091] S3: After entering the buffer zone, the target personnel will make a turning (approximately 90°) movement based on their safety awareness to avoid entering the danger zone, thus entering the intrusion behavior risk assessment and observation process of "walking in - turning - parallel movement".

[0092] Considering the computational requirements of the entropy weight method, the amount of data input for each assessment object should be the same. Therefore, the midpoint for extracting risk assessment index data is the time point when the worker enters the buffer and turns 45°, with an extraction time of t; the cutoff point is the time point after passing the midpoint and experiencing the same time t again. Thus, the total duration of the final intrusion behavior video stream is 2t.

[0093] By extracting coordinate information frame by frame from the above video stream at a rate of 60 frames per second, continuous motion information of the evaluated object within the evaluation time period can be obtained.

[0094] Because the video streams from different target personnel have different durations, the amount of coordinate information data varies. To meet the calculation requirements of the entropy weight method, the video stream is divided into 10 equal parts, which yields equal amounts of motion data for each target personnel when performing the same action (walking in—turning—parallel movement), and a risk assessment database is established based on this data.

[0095] By considering the gait characteristics of human walking, the risk assessment index extraction rules for intrusion behavior are used to extract four indicators, including the distance between the human body and the edge of the danger zone (hereinafter referred to as distance), the direction of human walking movement (hereinafter referred to as direction), the gait frequency of human walking movement (hereinafter referred to as gait frequency), and the gait cycle stability of human walking movement (hereinafter referred to as gait cycle stability). Figure 1 As shown.

[0096] Extraction and calculation of distance metrics:

[0097] Using the key points of the human neck as reference points and the marked lines at the edge of the danger zone as regional reference lines, the shortest distance (X-axis direction) between the two in a 2D image is used as the distance indicator for intrusion behavior. Dynamic changes in this distance indicator are used as input for intrusion risk assessment.

[0098] When workers are active around hazardous areas, changes in the aforementioned distance indicators can represent the changing trend of intrusion risk. Based on the entropy weight method, the resulting information entropy reflects the worker's behavioral stability regarding distance indicators and the degree of intrusion risk.

[0099] Extraction and calculation of directional indicators:

[0100] Compared to the uncertainty of upper limb movement, this invention uses the direction of lower limb movement as the basis for calculating the actual direction of human movement. This invention selects the line connecting three key points of the hip joint as a reference for calculating the direction of human movement.

[0101] In the data extracted from the human skeleton, the coordinate information of key points on both sides of the body at the hip joint can be used to calculate the current direction of human movement.

[0102] In intrusion risk assessment, when a worker's movement direction is facing the danger zone, the risk of intrusion is low because the worker has a good ability to observe and identify dangers / risks, and vice versa.

[0103] Using the dynamic changes of the aforementioned directional indicators as input for intrusion risk assessment, the risk level of intrusion behavior in terms of directional indicators can be obtained by employing the entropy weight method.

[0104] Step frequency index extraction and calculation:

[0105] Due to individual differences, different workers have different gait frequencies when walking. Furthermore, for walking, a low-speed movement, gait frequency differences affect the risk assessment of intrusion behavior. Considering the gait characteristics of human walking, this study uses the periodic changes in lower limb movements to calculate the variation in workers' gait frequency and incorporates this into the risk assessment of intrusion behavior.

[0106] The gait characteristics of human walking are mainly described by three key points on each side of the lower limbs. Considering the characteristics of human walking, the relative distances between the knee key point and the hip and foot key points remain constant, while the distance between the hip and foot key points exhibits periodic changes. By observing the continuous changes in this walking cycle, the gait frequency, or step frequency, can be calculated.

[0107] Using the dynamic change data of the above step frequency index as input for the risk assessment of intrusion behavior, the risk level of intrusion behavior in terms of step frequency index can be obtained by using the entropy weight method.

[0108] Extraction and calculation of gait period stability indices:

[0109] Similar to the extraction and calculation principles of gait frequency, human gait exhibits periodic changes during walking. Even when gait frequency changes are relatively small, gait can still vary due to external physical environment and internal factors such as fatigue and health. A stable gait facilitates risk avoidance behaviors. Therefore, considering the three indicators mentioned above, selecting gait periodic stability as an indicator can improve the identification of gait information related to intrusion behavior and the assessment of intrusion risk.

[0110] According to the theory of human kinesiology, the stability of gait during walking is generally characterized by the distance relationship between points such as the body's center of gravity and the contact points of the feet. The body's center of gravity is close to the center point of the hip among key skeletal points; therefore, this invention uses the center point of the hip to represent the body's center of gravity.

[0111] Furthermore, the vertical distance relationship between the center of gravity of the human body and the key points of both feet, that is, the coordinate distance between their ground projection points, is selected to calculate the stability of the gait cycle.

[0112] The calculation data for the above four indicators all come from the human 2D skeleton information in a unified coordinate system, which has a high degree of uniformity and convenience.

[0113] S4: By calculating the degree of information dispersion of the above four indicators using the entropy weight method, the changing state of worker intrusion behavior can be measured, and the risk assessment results of intrusion behavior of different personnel can be obtained, which are reflected in the form of four types of assessment indicators, namely four weighted results w.

[0114] The main steps of the entropy weight method include:

[0115] The first step is data normalization, which aims to convert different types of data into data with a unified unit of measurement and establish a decision matrix V. The specific formula is as follows:

[0116]

[0117] In the above formula, min and max values ​​are the minimum and maximum values ​​in the sample, v ij This represents the j-th value in the i-th sample after matrix normalization.

[0118] The second step is to calculate the information entropy. To assign weight to each evaluation indicator, substitute it into the following formula:

[0119]

[0120] E in the above formula j Let be the information entropy value of the j-th indicator, and let it satisfy (0≤E). j ≤1).

[0121] The third step is to calculate the weights, using the following formula:

[0122]

[0123] The data format of the evaluation indicators and weights is shown in Table 1 below:

[0124] Table 1

[0125]

[0126] S5: The following is an analysis and comprehensive calculation of the risk assessment results.

[0127] The weight calculation result of each risk assessment indicator represents the data volatility of the person's movement characteristics reflected by that indicator. The smaller the volatility, the higher the stability of the behavior, and the lower the risk of the intrusion behavior.

[0128] For example, in distance index analysis, the weight result is w1. If this weight is generally smaller than other indicators, it indicates that the distance index of the person in the intrusion behavior has less fluctuation and higher stability, and vice versa.

[0129] S6: Obtain the weighted result data of all target personnel, calculate the average value of the aforementioned four types of risk assessment indicators for all target personnel, and use it as a reference value for intrusion behavior risk assessment.

[0130] By directly comparing data, one can obtain the level of risk of an individual's intrusion behavior based on a certain indicator.

[0131] Furthermore, by using the continuous weight results of the aforementioned four risk assessment indicators for all personnel as input values ​​for the secondary calculation of the entropy weight method, a secondary assessment of intrusion behavior risk can be conducted on groups composed of different personnel.

[0132] The data format of the secondary assessment indicators and weights for intrusion risk is shown in Table 2.

[0133]

[0134] The results of the secondary risk assessment of intrusion behavior include the weight values ​​of the aforementioned four risk assessment indicators. If the secondary weight of a certain risk assessment indicator is high, it indicates that the data differences in the primary weight results of that group on that indicator are large, that is, it indicates that the overall performance of that indicator is uneven.

[0135] Based on the above results, the importance of the four types of risk assessment indicators in intrusion risk assessment can be ranked. If a certain risk assessment indicator has a large secondary weight, it indicates that the indicator has a significant impact on intrusion risk assessment.

[0136] For a defined target group and individual, the results of the first and second entropy weights can be used as the basis for quantitative assessment of their intrusion risk.

[0137] At the individual level, based on the aforementioned four types of risk assessment indicators within a specific group, a comprehensive risk assessment score for an individual's intrusion behavior can be calculated. Among them, S i w is the final score for the intrusion risk assessment of the i-th person. j For the target group weight result of the j-th risk assessment indicator, p ij Let be the score of the j-th risk assessment indicator for the i-th person.

[0138] Individual-level weights reflect the differences in risk indicators across individuals, while group-level weights assess a group of people exhibiting similar behaviors within a work team or construction site. This overall assessment of the group reveals its characteristics and can serve as a prerequisite for individual risk assessment. Based on this group characteristic, further weighting of individuals leads to more accurate risk assessments. Judging solely by individual weights can sometimes be inaccurate across different scenarios. If a characteristic is common to the group, the risk may be low. In short, with only individual data, we can initially determine the significance of an individual on certain risk indicators based on their weights. However, in practice, this is susceptible to external factors. Significant differences within a group may indicate low individual risk for that indicator. In such cases, the distribution characteristics of group indicators can be used to adjust individual risk weights, reducing the impact of environmental factors. When individual weight calculations show a high-risk indicator, but it's not a common characteristic of the group, that indicator more accurately reflects the true risk situation.

[0139] Advantages: Based on the general principle (assumption) that the majority is superior to the minority, this method can improve the accuracy of individual risk assessment by considering group characteristics, reducing the influence of environmental or group similarity.

[0140] The above quantitative assessment values ​​can be used as the results of risk assessment for intrusion behavior of individuals and groups.

[0141] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. A computer vision-based method for risk assessment of worker intrusion into hazardous construction areas, characterized in that, Includes the following steps: Obtain video information of the workers' construction area; Worker target detection and hazardous area labeling are performed on the video information, and the coordinates of key points of the human skeleton and the coordinates of the boundary line of the hazardous area are extracted. Based on the coordinates of key points on the human skeleton and the coordinates of the boundary line of the danger zone, the distance between the human body and the boundary line of the danger zone, the direction of human walking, the cadence of human walking, and the stability of the gait cycle of human walking are extracted, thereby obtaining risk assessment data on the distance between the human body and the boundary line of the danger zone, the direction of human walking, the cadence of human walking, and the stability of the gait cycle of human walking. The weights of the distance from the human body to the boundary of the danger zone, the direction of human walking, the cadence of human walking, and the stability of the gait cycle of human walking are calculated based on the volatility of risk assessment data. Based on the risk assessment data and corresponding weights of the distance between the human body and the boundary line of the dangerous area, the direction of human walking movement, the step frequency of human walking movement, and the stability of the gait cycle of human walking movement, the risk assessment results of workers intruding into the construction dangerous area are obtained. The entropy weight method is used to calculate the weight of each indicator based on the volatility of the risk assessment data of each indicator; The entropy weight method includes the following steps: S401: Normalize the risk assessment data for each indicator; S402: Calculate the information entropy of each indicator after data normalization; S403: Calculate the weight of each indicator based on the information entropy calculation results of each indicator.

2. The method for risk assessment of worker intrusion into hazardous construction areas based on computer vision according to claim 1, characterized in that, The coordinates of key points on the human skeleton include the coordinates of key points on the neck, shoulders, hips, and feet. The calculation process of the distance between the human body and the boundary line of the danger zone includes: taking the coordinates of the key point of the human neck as the human body reference point and the coordinates of the boundary line of the danger zone as the area reference line, calculating the shortest distance between the human body reference point and the area reference line on the horizontal axis. The coordinates of key points in the hip include the coordinates of key points at the center of the body and on both sides of the hip joint. The calculation process for the direction of human walking includes: obtaining the direction of human walking based on the line connecting the coordinates of key points on the left and right sides of the human body at the hip joint; The calculation process of the human walking frequency includes: calculating the human walking frequency based on the periodic change frequency of the distance between the coordinates of the hip key point and the coordinates of the foot key point. The calculation process for the gait cycle stability of human walking includes: taking the coordinates of the key point at the center of the body at the hip joint as the center of gravity of the human body, selecting the vertical distance relationship between the center of gravity of the human body and the coordinates of the key points of the feet, and calculating the gait cycle stability of human walking.

3. The method for risk assessment of worker intrusion into hazardous construction areas based on computer vision according to claim 1, characterized in that, The video information also includes a buffer boundary line. The method further includes cropping the video information to obtain risk assessment data for each indicator based on the cropped video information. The interception process includes: Based on the worker target detection results, the starting point of time is defined as the point in time from when the coordinates of any point in the key points of the worker's human skeleton enter the buffer zone, to when the worker turns 45°. t Using the midpoint as an example, the same amount of time is required after passing through the midpoint. t The cutoff point is the time point used to obtain the captured video information.

4. The method for risk assessment of worker intrusion into hazardous construction areas based on computer vision according to claim 3, characterized in that, The distribution of the buffer zone boundary lines is consistent with the boundary of the danger zone. The area corresponding to the buffer zone boundary lines is the area extending outward from the boundary of the danger zone. The width of the buffer zone boundary lines is obtained by adding the movement space of a human walking and the reaction space of a safety response.

5. The method for risk assessment of worker intrusion into hazardous construction areas based on computer vision according to claim 1, characterized in that, The calculation expression for the data normalization is: In the formula, min and max are the minimum and maximum values ​​in the sample, Represents the normalized matrix of the first... i The first sample j A number, These are the normalization coefficients; The expression for calculating the information entropy is: In the formula, It is the first t The information entropy value of each indicator The information entropy coefficient, The total number of values ​​in the sample. The total number of samples; The formula for calculating the weight of each indicator is as follows: In the formula, For the first t The weight of each indicator.

6. The method for risk assessment of worker intrusion into hazardous construction areas based on computer vision according to claim 1, characterized in that, The method also includes obtaining risk assessment data and corresponding indicator weights for each individual in the target group, thereby obtaining the overall indicator weight of the target group based on the weight data of each individual, and then conducting intrusion risk assessment on each individual in the target group based on the overall indicator weight.

7. The method for risk assessment of worker intrusion into hazardous construction areas based on computer vision according to claim 6, characterized in that, The overall index weight of the target group is calculated using the entropy weight method based on the weight data of each individual.

8. The method for risk assessment of worker intrusion into hazardous construction areas based on computer vision according to claim 6, characterized in that, The calculation formula for the intrusion risk assessment of each individual in the target group is as follows: In the formula, For the first i The final score for the risk assessment of an individual's intrusion behavior. w j For the first j The target group weight results for each indicator p ij For the first i The first person j The scores of each risk assessment indicator This represents the total number of indicators.

Images