A comprehensive management and control method and system for road construction based on the Internet of Things

Abstract

This invention discloses a comprehensive road construction management and control method and system based on the Internet of Things (IoT), relating to the field of IoT technology. It constructs an occlusion feature vector set Vft, and further fuses it through a model to form a comprehensive occlusion visual field index Ovf, outputting an occlusion risk score set OvfList with station location distribution characteristics. Compared to traditional construction safety management methods that rely solely on static deployment and personnel experience, this invention automatically filters high-occlusion and high-conflict points by setting a risk threshold Tovf, outputting a high-risk point index set Rls and generating a targeted list of enclosure structure optimization suggestions Opl. Finally, after optimization, the generated improvement score Dsc is re-evaluated to quantify the actual improvement in optimization effectiveness. This effectively solves problems such as misjudgment of direction caused by obstructed vision due to enclosures, high risk of collisions between pedestrians or between pedestrians and non-motorized vehicles, and poor predictability of path intersection areas.

Description

Technical Field

[0001] This invention relates to the field of Internet of Things (IoT) technology, specifically to a comprehensive management and control method and system for road construction based on IoT. Background Technology

[0002] With the accelerating pace of urbanization, the construction of smart cities has become an important component of the current urban governance system. Within the macro-system of smart cities, intelligent management of urban infrastructure is one of its core supports, and "road construction and maintenance," as a crucial link in ensuring urban traffic operation and spatial renewal, is gradually being incorporated into a more refined and dynamic scheduling system. Especially in large and medium-sized cities, road construction operations involving main roads, transportation hubs, or densely populated areas are no longer simply civil engineering activities, but rather an intelligent management and control process that requires coordinated response with the city's operational status. Therefore, "comprehensive management and control of road construction" is gradually evolving into a multi-module integrated system combining perception, analysis, and response, and the Internet of Things (IoT) technology, acting as a bridge between the perception and interaction layers, is key to realizing this system.

[0003] In actual urban road construction, while temporary physical barriers ensure the enclosure of construction areas and operational safety, most deployment plans do not adequately consider their impact on the continuity of sight between people and between people and vehicles. Particularly at intersections, narrow sidewalks, and temporary detours, barriers create numerous blind spots, preventing pedestrians from different directions from anticipating oncoming pedestrians and greatly increasing the risk of collisions, scrapes, and corner impacts. More seriously, these visual obstructions can also lead to sudden encounters between pedestrians and non-motorized vehicles, bicycles, scooters, and other small, fast-moving vehicles at turns. Unable to predict the direction and distance of the other vehicle, pedestrians often experience dangerous instantaneous collisions or falls while trying to avoid them. These issues have gradually become a major cause of pedestrian safety incidents, yet they are often considered non-priority items in most construction planning schemes, lacking specific monitoring or predictive mechanisms. This makes it difficult for management to identify "high-risk obstruction points" in a timely manner, let alone dynamically optimize the system. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a comprehensive road construction management and control method and system based on the Internet of Things, which solves the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a comprehensive road construction management and control method based on the Internet of Things, comprising the following steps:

[0006] S1. Based on the Internet of Things, collect spatial structure and pedestrian behavior data of road construction areas, extract the structural parameter subset StructSet and the behavioral parameter subset BehavSet, and use them to form the original visual domain dataset Vfd of the construction area.

[0007] S2. After extracting features based on the original visual field dataset Vfd, construct an occlusion feature vector set Vft by calculating the feature vector of each station point, including occlusion rate Sor, path conflict factor Cof, and visual blur Blu.

[0008] S3. The feature vector Vft is input into the occlusion risk scoring model for training, and a comprehensive occlusion field index Ovf is constructed, which is combined into an occlusion risk score set OvfList for each station.

[0009] S4. Using the preset occlusion risk threshold Tovf, the occlusion risk score set OvfList is traversed and compared to determine the location of all stations marked as high-risk points by the determination results, forming a high-risk point index set Rls, and generating a fence structure optimization suggestion list Opl based on the high-risk point index set Rls.

[0010] S5. Implement the optimization suggestion list Opl for the enclosure structure and collect a new set of occlusion risk scores OvfList (New). Then compare it with the occlusion risk score set OvfList of the site before implementation and obtain the optimization improvement score Dsc. Based on the optimization improvement score Dsc, determine the effectiveness status of the measures after implementation.

[0011] Preferably, S1 includes S11;

[0012] S11. Collect spatial projection data of the road construction area fence and the visible range information of the station locations through IoT devices, and extract the spatial structure parameters of each station location in the construction area, including the actual occlusion area Aov, the theoretical visible range of the station location Avw, the pedestrian traffic direction angle Dag, and the main viewing direction of the station location Vag, to form a subset of structural parameters of all station locations StructSet={Aov(k),Avw(k),Dag(k),Vag(k)|k∈N}, where N represents the total number of station locations, and Aov(k), Avw(k), Dag(k), and Vag(k) represent the actual occlusion area Aov, the theoretical visible range Avw, the pedestrian traffic direction angle Dag, and the main viewing direction Vag of the kth station location, respectively;

[0013] Among them, IoT devices include panoramic wide-angle cameras and 3D laser scanners;

[0014] The actual occlusion area Aov is obtained by acquiring the projected images of the construction fence at different angles in real time through a panoramic wide-angle camera and a 3D laser scanner installed above the construction area. The boundary of the occlusion area is extracted by an image processing algorithm, which includes a graphic contour filling algorithm and a point cloud bounding volume division method.

[0015] The theoretical visible range Avw of the station location is simulated by using a 3D laser scanner installed and a high-precision map device deployed at the boundary of the construction area to perform a laser scan in a 180° direction in front of the station location without considering the obstruction of the fence. Combined with the existing road structure map, the theoretical open field of view area is calculated by fan-shaped expansion.

[0016] The main view direction Vag of the station point is obtained by measuring the maximum continuous unobstructed angle from the left and right boundaries with the station point as the center and the angle of the projection ray preset.

[0017] The pedestrian direction angle Dag is obtained by capturing the movement trajectory of the crowd through a panoramic wide-angle camera and combining it with the Ak path fitting algorithm to calculate the angle between the direction of the crowd movement trajectory and the direction of the station position.

[0018] Preferably, S1 includes S12;

[0019] S12. By deploying sensing devices in the road construction area, obtain the traffic behavior feature data of the crowd in the construction site, extract the behavioral interaction parameters required to build the occlusion conflict analysis model, specifically including the path intersection angle Xag and the pedestrian density Dpt per unit time, and form a behavioral parameter subset BehavSet={Dpt, Xag(m, w)|(m, w)∈M*M}, m≠w, and then integrate it with the structural parameter subset StructSet to obtain the original visual domain dataset Vfd of the construction area;

[0020] The path intersection angle Xag is obtained by continuously acquiring video images through a panoramic wide-angle camera; the edge module detects pedestrians in the images and performs multi-target tracking; the trajectory point sequence of each pedestrian is recorded to form a path set {P1, P2, ..., Pm|m∈M}, where M represents the total number of paths and Pm represents the m-th path; then, a linear fit is performed on each path in the path set to obtain the unit direction vector v, and the path intersection angle Xag is obtained by calculating the unit direction vector v of any two paths.

[0021] The unit direction vector v is obtained using the following formula:

[0022] ;

[0023] In the formula, v(m) represents the unit direction vector of the m-th path, and x(m, end) and y(m, end) and x(m, start) and y(m, start) represent the coordinates of the end point and the start point of the m-th path, respectively.

[0024] The path intersection angle Xag is obtained using the following formula:

[0025] ;

[0026] In the formula, Xag(m, w) represents the path intersection angle between the m-th path and the w-th path, and arccos represents the inverse cosine function, which specifically converts the cosine value into an angle value;

[0027] The pedestrian density per unit time, Dpt, is obtained by measuring the ratio of the number of pedestrians in the road construction area to the time interval within a fixed statistical time window.

[0028] Preferably, S2 includes S21;

[0029] S21. Based on the original visual field dataset Vfd, feature extraction is performed. After extracting the structural parameter subset StructSet of each station, the actual occlusion area Aov and the theoretical unobstructed visible range Avw are extracted. The occlusion rate Sor of each station is calculated to quantify the degree to which each station is occluded by the enclosure structure.

[0030] The occlusion rate Sor is obtained using the following formula:

[0031] ;

[0032] In the formula, Sor(k) represents the occlusion rate of the k-th station, and Aov(k) and Avw(k) represent the actual occlusion area Aov and the theoretical unobstructed visible range Avw of the k-th station, respectively.

[0033] Preferably, S2 includes S22;

[0034] S22. Based on the original visual field dataset Vfd, feature extraction is performed. By extracting the behavioral parameter subset BehavSet and structural parameter subset StructSet of each station, key dynamic data reflecting the relationship between the crowd's behavioral path and direction are obtained. Then, the pedestrian density per unit time Dpt, path intersection angle Xag, passage direction angle Dag, and station main view direction Vag are extracted to construct the path conflict factor Cof and visual blur Blu. Then, they are integrated with the occlusion rate Sor to obtain the occlusion feature vector set Vft.

[0035] The path conflict factor Cof is obtained using the following formula:

[0036] ;

[0037] In the formula, Cof(m, w, k) represents the path conflict factor between path m and path w at the k-th station, skn represents the sine function, Dpt(k) represents the pedestrian density per unit time at the tangent point of the k-th station, Dag(k) represents the direction angle of passage at the k-th station, and Vag(k) represents the main view direction of the k-th station.

[0038] Blu, the visual blur level, is obtained using the following formula:

[0039] ;

[0040] In the formula, Blu(k) represents the visual blur at the k-th station.

[0041] Preferably, S3 includes S31;

[0042] S31. Normalize the feature vector Vft to eliminate the difference in units between different data, and obtain the standard path conflict factor Cofn, standard visual blur Blun, and standard occlusion rate Sorn. Then input them into the occlusion risk scoring model and perform fusion training with each station k as a unit to construct a comprehensive index that reflects the degree of influence of multi-dimensional occlusion risk. This index is marked as the comprehensive occlusion visual field index Ovf(k) of the kth station. After integrating all stations, obtain the occlusion risk scoring set OvfList={Ovf(k)|k∈N}, where N represents the total number of stations.

[0043] The comprehensive occlusion field index Ovf(k) of the kth station is obtained through the following model fusion training formula:

[0044] ;

[0045] In the formula, N(k) represents the number of path pairs existing at the k-th station, which is obtained by the formula N(k) = |M(k)|*(||M(k)-1) / 2, M represents the total number of paths, Sorn(k) represents the standard occlusion rate of the k-th station, Cofn(k) represents the standard path conflict factor of the k-th station, and Blun(k) represents the standard visual blur of the k-th station.

[0046] Preferably, S4 includes S41;

[0047] S41. Using the total number of stations N as the traversal boundary, each comprehensive occlusion field index Ovf(k) in the occlusion risk score set OvfList is judged. The judgment is made by comparing it with the preset occlusion risk threshold Tovf, and marking high-risk occlusion stations according to the comparison result. After the traversal is completed, all high-risk occlusion stations are integrated to obtain the high-risk point index set Rls.

[0048] High-risk site locations with obstruction are marked using the following comparison method:

[0049] When the comprehensive occlusion field index Ovf(k) > the occlusion risk threshold Tovf, the comparison result is the risk result, and the kth station is marked as a high-risk occlusion station.

[0050] When the comprehensive occlusion field index Ovf(k) ≤ occlusion risk threshold Tovf, the comparison result is normal, the kth station is not marked as a high-risk occlusion station, and the kth station is removed.

[0051] Preferably, S4 includes S42;

[0052] S42. Based on the high-risk point index set Rls and combined with the original visual domain dataset Vfd of the high-risk station sites, generate hoarding structure optimization suggestions for each high-risk station site. The hoarding structure optimization suggestions include adjusting the hoarding layout shape of the high-risk station sites, adjusting the perspective window and semi-transparent visual domain area of ​​the high-risk station sites, and adjusting the number of eye-catching signs guiding pedestrian flow through the visual priority channel, and generate a hoarding structure optimization suggestion list Opl.

[0053] Preferably, S5 includes S51;

[0054] S51. Implement and apply the optimization suggestion list Opl for the enclosure structure. The implementation and application include sending a notification to relevant operators for actual deployment, collecting the original field of view dataset Vfd for each high-risk occlusion site after actual deployment, calculating and obtaining a new occlusion risk score set OvfList (New) after actual deployment, comparing it with the occlusion risk score set OvfList before actual deployment, summarizing the difference, obtaining the optimization and improvement score Dsc, and then judging the effectiveness status of the measures after actual deployment based on the optimization and improvement score Dsc.

[0055] The measures implemented after actual deployment are judged in the following ways:

[0056] When the optimization and improvement score Dsc > 0, it indicates that the implementation and application of the fence structure optimization suggestion list Opl is effective;

[0057] When the optimization and improvement score Dsc≤0, it means that the implementation of the fence structure optimization suggestion list Opl is invalid, and optimization measures such as fence shape adjustment, perspective window setting, addition of prompt signs, and reconstruction of path visual guidance need to be carried out again.

[0058] A comprehensive road construction management and control system based on the Internet of Things includes a road construction network data acquisition module, a construction feature extraction module, an occlusion risk analysis module, an occlusion judgment module, and an iterative optimization module;

[0059] The road construction network data acquisition module collects spatial structure and pedestrian behavior data of the road construction area based on the Internet of Things, forming the original visual domain dataset Vfd of the construction area.

[0060] After the construction feature extraction module extracts features based on the original visual domain dataset Vfd, it constructs an occlusion feature vector set Vft, which includes occlusion rate Sor, path conflict factor Cof, and visual blur Blu.

[0061] The feature vector Vft of the occlusion risk analysis module is input into the occlusion risk scoring model for training, and a comprehensive occlusion field index Ovf is constructed, which is combined into an occlusion risk score set OvfList for each station.

[0062] The occlusion judgment module uses the preset occlusion risk threshold Tovf to traverse and compare the occlusion risk score set OvfList, obtains all the site locations marked as high-risk points by the judgment results, forms a high-risk point index set Rls, and generates a fence structure optimization suggestion list Opl based on the high-risk point index set Rls.

[0063] The iterative optimization module applies the optimization suggestion list Opl for the fence structure, and then compares the occlusion risk score set OvfList of the site before and after the application to obtain the optimization improvement score Dsc.

[0064] This invention provides a comprehensive road construction management and control method and system based on the Internet of Things, which has the following beneficial effects:

[0065] (1) Construct a set of occlusion feature vectors Vft, and further construct a comprehensive occlusion field index Ovf through model fusion, outputting an occlusion risk score set OvfList with station distribution characteristics. Compared with the traditional construction safety management method that is guided only by static layout and personnel experience, this method automatically filters high occlusion and high conflict points by setting a risk threshold Tovf, outputs a high-risk point index set Rls, and generates a targeted fence structure optimization suggestion list Opl. Finally, the improvement score Dsc generated after optimization is re-evaluated to quantify the actual improvement of the optimization effect. The overall method effectively solves the problems mentioned in the background technology, such as misjudgment of direction caused by the obstruction of vision caused by fences, high risk of collision between pedestrians or between pedestrians and non-motorized vehicles, and poor predictability of path intersection areas.

[0066] (2) By using panoramic wide-angle cameras and 3D laser scanners in IoT devices to achieve real-time multi-angle acquisition, not only can spatial geometric parameters such as the actual occlusion area Aov and the theoretical visible range Avw of the station location be obtained, but also the station location's main viewpoint direction Vag and the pedestrian direction angle Daag reflected by the dynamic trajectory of the crowd can be accurately extracted. At the same time, by using behavioral interaction parameters such as the path intersection angle Xag and the pedestrian flow density Dpt per unit time calculated based on the crowd trajectory, a set of behavioral features reflecting the potential for traffic conflict can be further constructed. After the above multi-dimensional parameters form a structured dataset Vfd, they provide a highly realistic, quantifiable, and traceable decision-making basis for subsequent occlusion risk modeling. It is particularly noteworthy that parameters such as the path intersection angle Xag are derived from the dynamic trajectory angle calculation and have long been neglected in existing traffic safety technologies. Because they rely on continuous visual perception and direction vector modeling, they are difficult to obtain in traditional solutions. This significantly improves the ability to identify the causes of local occlusion misjudgment and the path of crowd conflict, and has strong interpretability and high adaptability.

[0067] (3) By normalizing the multidimensional parameters in the occlusion feature vector set Vft, a comprehensive occlusion field index Ovf(k) is constructed, realizing the fusion expression of different types of occlusion risk factors under a unified scoring scale, effectively solving the problem of the inability to quantify and unify structural occlusion factors and dynamic traffic interference. Based on the occlusion risk scoring set OvfList, with the total number of station points N as the index boundary, high-risk occlusion station points are efficiently screened based on the set occlusion risk threshold Tovf, forming a high-risk point index set Rls. Then, combined with the specific structural layout of each risk point in the original field dataset Vfd, a fence structure optimization suggestion list Opl is output, and the occlusion risk scoring set OvfList(New) after re-collection is compared with the original scoring set OvfList to form an optimization improvement score Dsc, which is used to measure the actual effect after deployment. This method not only achieves accurate identification of highly occluded areas, but also forms a risk-policy-feedback closed loop path with the comprehensive occlusion field index Ovf(k) as the core. It has dynamic response capability and verifiable effect, which is significantly better than traditional manual experience judgment and single path optimization strategy. Attached Figure Description

[0068] Figure 1 This is a schematic diagram illustrating the steps of a comprehensive road construction management and control method based on the Internet of Things according to the present invention;

[0069] Figure 2 This is a schematic diagram of a road construction integrated management and control system based on the Internet of Things according to the present invention;

[0070] Figure 3 This is a schematic diagram showing the distribution trend of the comprehensive occlusion field index Ovf(k) at various stations. Detailed Implementation

[0071] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0072] Example 1: This invention provides a comprehensive road construction management and control method based on the Internet of Things. Please refer to [link / reference]. Figure 1 This includes the following steps:

[0073] S1. Based on the Internet of Things, collect spatial structure and pedestrian behavior data of road construction areas, extract the structural parameter subset StructSet and the behavioral parameter subset BehavSet, and use them to form the original visual domain dataset Vfd of the construction area.

[0074] S2. After extracting features based on the original visual field dataset Vfd, construct an occlusion feature vector set Vft by calculating the feature vector of each station point, including occlusion rate Sor, path conflict factor Cof, and visual blur Blu.

[0075] S3. The feature vector Vft is input into the occlusion risk scoring model for training, and a comprehensive occlusion field index Ovf is constructed, which is combined into an occlusion risk score set OvfList for each station.

[0076] S4. Using the preset occlusion risk threshold Tovf, the occlusion risk score set OvfList is traversed and compared to determine the location of all stations marked as high-risk points by the determination results, forming a high-risk point index set Rls, and generating a fence structure optimization suggestion list Opl based on the high-risk point index set Rls.

[0077] S5. Implement the optimization suggestion list Opl for the enclosure structure and collect a new set of occlusion risk scores OvfList (New). Then compare it with the occlusion risk score set OvfList of the site before implementation and obtain the optimization improvement score Dsc. Based on the optimization improvement score Dsc, determine the effectiveness status of the measures after implementation.

[0078] In this embodiment, the original visual domain dataset Vfd of the construction area is collected by an IoT terminal. Combining structural parameter subsets and behavioral parameter subsets, the occlusion rate Sor, path conflict factor Cof, and visual blur Blu are extracted to construct an occlusion feature vector set Vft. This is further fused through a model to form a comprehensive occlusion visual domain index Ovf, outputting an occlusion risk score set OvfList with station location distribution characteristics. Compared to traditional construction safety management methods that rely solely on static deployment and personnel experience, this invention automatically filters high-occlusion, high-conflict points by setting a risk threshold Tovf, outputting a high-risk point index set Rls, and generating a targeted fence structure optimization suggestion list Opl. Finally, after optimization, the generated improvement score Dsc is re-evaluated to quantify the actual improvement in optimization effectiveness. The overall approach effectively solves the problems mentioned in the background technology, such as misjudgment of direction caused by obstructed vision due to construction barriers, high risk of collisions between pedestrians or between pedestrians and non-motorized vehicles, and poor predictability of path intersection areas. It has significant advantages such as quantifiable risk modeling, visualized control strategies, and iterative optimization feedback. It is especially suitable for intelligent, refined, and dynamic management scenarios of visual safety for pedestrian passage in complex construction environments.

[0079] Example 2: Specifically: S1 includes S11;

[0080] S11. Collect spatial projection data of the road construction area fence and the visible range information of the station locations through IoT devices, and extract the spatial structure parameters of each station location in the construction area, including the actual occlusion area Aov, the theoretical visible range of the station location Avw, the pedestrian traffic direction angle Dag, and the main viewing direction of the station location Vag, to form a subset of structural parameters of all station locations StructSet={Aov(k),Avw(k),Dag(k),Vag(k)|k∈N}, where N represents the total number of station locations, and Aov(k), Avw(k), Dag(k), and Vag(k) represent the actual occlusion area Aov, the theoretical visible range Avw, the pedestrian traffic direction angle Dag, and the main viewing direction Vag of the kth station location, respectively;

[0081] Among them, IoT devices include panoramic wide-angle cameras and 3D laser scanners;

[0082] The actual occlusion area Aov is obtained by acquiring the projected images of the construction fence at different angles in real time through a panoramic wide-angle camera and a 3D laser scanner installed above the construction area. The boundary of the occlusion area is extracted by an image processing algorithm, which includes a graphic contour filling algorithm and a point cloud bounding volume division method.

[0083] The theoretical visible range Avw of the station location is simulated by using a 3D laser scanner installed and a high-precision map device deployed at the boundary of the construction area to perform a laser scan in a 180° direction in front of the station location without considering the obstruction of the fence. Combined with the existing road structure map, the theoretical open field of view area is calculated by fan-shaped expansion.

[0084] The main view direction Vag of the station point is obtained by measuring the maximum continuous unobstructed angle from the left and right boundaries with the station point as the center and the angle of the projection ray preset.

[0085] The pedestrian direction angle Dag is obtained by capturing the movement trajectory of the crowd through a panoramic wide-angle camera and combining it with the Ak path fitting algorithm, which includes using SORT trajectory tracking to calculate the angle between the direction of the crowd movement trajectory and the direction of the station position.

[0086] S1 includes S12;

[0087] S12. By deploying sensing devices in the road construction area, obtain the traffic behavior feature data of the crowd in the construction site, extract the behavioral interaction parameters required to build the occlusion conflict analysis model, specifically including the path intersection angle Xag and the pedestrian density Dpt per unit time, and form a behavioral parameter subset BehavSet={Dpt, Xag(m, w)|(m, w)∈M*M}, m≠w, and then integrate it with the structural parameter subset StructSet to obtain the original visual domain dataset Vfd of the construction area;

[0088] The path intersection angle Xag is obtained by continuously acquiring video images through a panoramic wide-angle camera; the edge module detects pedestrians in the images and performs multi-target tracking; the trajectory point sequence of each pedestrian is recorded to form a path set {P1, P2, ..., Pm|m∈M}, where M represents the total number of paths and Pm represents the m-th path; then, a linear fit is performed on each path in the path set to obtain the unit direction vector v, and the path intersection angle Xag is obtained by calculating the unit direction vector v of any two paths.

[0089] The unit direction vector v is obtained using the following formula:

[0090] ;

[0091] In the formula, v(m) represents the unit direction vector of the m-th path, which is obtained by dividing the direction vector formed by the starting point and ending point of the path by its magnitude. The magnitude is the Euclidean norm of the direction vector, which represents its geometric length in the Cartesian coordinate system. x(m, end) and y(m, end) and x(m, start) and y(m, start) represent the coordinates of the end point and the starting point of the m-th path, respectively. The coordinates of the end point are specifically represented as (x(m, end), y(m, end)), where x and y represent the x-axis coordinates and y-axis coordinates of the end point, respectively. The coordinates of the starting point are specifically represented as (x(m, start), y(m, start)), where x and y represent the x-axis coordinates and y-axis coordinates of the starting point, respectively.

[0092] The path intersection angle Xag is obtained using the following formula:

[0093] ;

[0094] In the formula, Xag(m, w) represents the path intersection angle between the m-th path and the w-th path, and arccos represents the inverse cosine function, specifically converting the cosine value into an angle value. The smaller the path intersection angle Xag(m, w) between the m-th path and the w-th path, the closer the two directions are, and the paths may travel in parallel. When it is close to 90°, it is a high-risk intersection angle, indicating that there may be a collision point between the paths.

[0095] The pedestrian density per unit time (Dpt) is obtained by measuring the ratio of the number of pedestrians in the road construction area to the time interval within a fixed statistical time window. It is used to reflect the instantaneous pedestrian flow intensity and traffic density in the road construction area.

[0096] In this embodiment, by constructing an original visual domain dataset Vfd, including a subset of structural parameters (StructSet) and a subset of behavioral parameters (BehavSet), a two-dimensional fusion modeling foundation for occlusion structures and crowd behavior patterns within the construction area is achieved. Compared to previous methods of traffic safety assessment relying on planar drawings or static path presets, the method of this invention uses panoramic wide-angle cameras and 3D laser scanners in IoT devices to achieve multi-angle real-time acquisition. This not only obtains spatial geometric parameters such as the actual occlusion area (Aov) and the theoretical visible range (Avw) of the station location, but also accurately extracts the station location's main viewpoint direction (Vag) and the pedestrian traffic direction angle (Dag) reflected by the dynamic trajectory of the crowd. Simultaneously, by using behavioral interaction parameters such as the path intersection angle (Xag) and the pedestrian density (Dpt) calculated based on the crowd trajectory, a behavioral feature set reflecting the potential for traffic conflicts is further constructed. After forming the structured dataset Vfd, these multi-dimensional parameters provide a highly realistic, quantifiable, and traceable decision-making basis for subsequent occlusion risk modeling. It is particularly noteworthy that parameters such as the path intersection angle Xag are derived from dynamic trajectory angle calculations and have long been neglected in existing traffic safety technologies. Because they rely on continuous visual perception and direction vector modeling, they are difficult to obtain in traditional solutions. By integrating these parameters into the dynamic modeling process of construction scenarios for the first time, the ability to identify the causes of partial occlusion misjudgment and the path of crowd conflict has been significantly improved, demonstrating strong interpretability and high adaptability.

[0097] Example 3: Specifically: S2 includes S21;

[0098] S21. Based on the original visual field dataset Vfd, feature extraction is performed. After extracting the structural parameter subset StructSet of each station, the actual occlusion area Aov and the theoretical unobstructed visible range Avw are extracted. The occlusion rate Sor of each station is calculated to quantify the degree to which each station is occluded by the enclosure structure.

[0099] The occlusion rate Sor is obtained using the following formula:

[0100] ;

[0101] In the formula, Sor(k) represents the occlusion rate of the k-th station, and Aov(k) and Avw(k) represent the actual occlusion area Aov and the theoretical unobstructed visible range Avw of the k-th station, respectively.

[0102] S2 includes S22;

[0103] S22. Based on the original visual domain dataset Vfd, feature extraction is performed. By extracting the behavioral parameter subset BehavSet and structural parameter subset StructSet of each station, key dynamic data reflecting the relationship between the pedestrian behavior path and direction are obtained. Then, the pedestrian density per unit time Dpt, path intersection angle Xag, passage direction angle Dag, and station main view direction Vag are extracted to construct the path conflict factor Cof and visual blur Blu, which are used as the measurement parameters of behavioral intersection intensity and visual perception error, respectively. Through angle calculation and density correction, the conflict risk caused by pedestrian view occlusion and path overlap in real construction scenarios is reflected. Then, it is integrated with the occlusion rate Sor to form a multi-dimensional risk factor, comprehensively modeling the occlusion passage safety status of the station, and obtaining the occlusion feature vector set Vft.

[0104] The path conflict factor Cof is obtained using the following formula:

[0105] ;

[0106] In the formula, Cof(m, w, k) represents the path conflict factor between path m and path w at the k-th station, skn represents the sine function, Dpt(k) represents the pedestrian density per unit time at the tangent point of the k-th station, Dag(k) represents the direction angle of travel at the k-th station, and Vag(k) represents the main viewpoint direction of the k-th station. Adding 1 to the denominator is to avoid the denominator being 0 and to suppress high deviation directions. The actual physical meaning of this formula is: it actually measures the intersection angle formed between two paths in the spatial structure × the current pedestrian density × the degree of matching between behavior and viewpoint. When the path intersection angle is large, the pedestrian density is high, and the consistency between viewpoint and direction is high, the conflict factor value is the largest, indicating that there is a significant potential for path intersection collision.

[0107] Blu, the visual blur level, is obtained using the following formula:

[0108] ;

[0109] In the formula, Blu(k) represents the visual ambiguity of the kth station point; the actual physical meaning of this formula is: it measures the degree of deviation between the structural visual direction and the actual pedestrian flow direction of a station point. The larger this value is, the more inconsistent the visual "direction" of this position is with the actual pedestrian flow direction, which is likely to lead to pedestrians misjudging the path, not being able to see oncoming people, and risks such as stopping or conflict.

[0110] In this embodiment, the occlusion rate Sor is constructed by the ratio of the actual occlusion area Aov to the theoretical unobstructed visible range Avw in the structural parameter subset StructSet, achieving for the first time a refined quantification of the degree of occlusion of a station under a specific viewing direction. A path conflict factor Cof and a line-of-sight ambiguity Blu are constructed to form an occlusion feature vector set Vft containing three-dimensional risk factors. This not only achieves unified modeling of static occlusion degree and dynamic behavioral conflict, but also innovatively incorporates pedestrian directional error into the occlusion modeling logic, possessing a sensitive ability to capture path intersection risks, directional confusion, and the aggregation trend of occlusion in high-density areas. Compared to the coarse-grained processing method of traditional solutions that can only perceive "whether there is occlusion," the three-factor modeling strategy provides a highly structured, multi-variable expression dimension for occlusion risk, making it particularly suitable for accurate and quantifiable pre-assessment of risks in highly dynamic intersection scenarios.

[0111] Example 4: Please refer to Figure 1 and Figure 3 Specifically: S3 includes S31;

[0112] S31. Normalize the feature vector Vft to eliminate the difference in units between different data, and obtain the standard path conflict factor Cofn, standard visual blur Blun, and standard occlusion rate Sorn. Then input them into the occlusion risk scoring model and perform fusion training with each station k as a unit to construct a comprehensive index that reflects the degree of influence of multi-dimensional occlusion risk. This index is marked as the comprehensive occlusion visual field index Ovf(k) of the kth station. After integrating all stations, obtain the occlusion risk scoring set OvfList={Ovf(k)|k∈N}, where N represents the total number of stations.

[0113] The comprehensive occlusion field index Ovf(k) of the kth station is obtained through the following model fusion training formula:

[0114] ;

[0115] In the formula, N(k) represents the number of path pairs existing at the k-th station, which is obtained by the formula N(k) = |M(k)|*(||M(k)-1) / 2, M represents the total number of paths, Sorn(k) represents the standard occlusion rate of the k-th station, Cofn(k) represents the standard path conflict factor of the k-th station, and Blun(k) represents the standard visual blur of the k-th station.

[0116] Example of calculating the comprehensive occlusion field of view index Ovf(k) for station k=002:

[0117] The corresponding path set is M(k) = {P1, P2, P3}, meaning there are 3 paths.

[0118] The possible path pairs are: (P1,P2), (P1,P3), (P2,P3) and (P1,P2), which is a total of 3 pairs;

[0119] Therefore, the number of path pairs existing at the k-th station is N(k) = 3;

[0120] The standard visual blur is obtained as Blun(k) = 0.46; the standard occlusion rate is Sorn(k) = 0.58; and the standard path conflict factor is Cofn(k) = {Cofn(P1,P2,k) = 0.71, Cofn(P1,P3,k) = 0.64, Cofn(P2,P3,k) = 0.82}.

[0121] Substituting the comprehensive occlusion field index Ovf(k) of the k-th station location, the following model is used for fusion and training:

[0122] Substitute the summation term:

[0123] ∑(0.58+0.71+0.46)+(0.58+0.64+0.46)+(0.58+0.82+0.46)=1.75+168+1.86=5.29;

[0124] Divide by the number of path pairs: = 5.29 / 3 ≈ 1.763;

[0125] S4 includes S41;

[0126] S41. Using the total number of stations N as the traversal boundary, each comprehensive occlusion field index Ovf(k) in the occlusion risk score set OvfList is judged. The judgment is made by comparing it with the preset occlusion risk threshold Tovf, and marking high-risk occlusion stations according to the comparison result. After the traversal is completed, all high-risk occlusion stations are integrated to obtain the high-risk point index set Rls.

[0127] High-risk site locations with obstruction are marked using the following comparison method:

[0128] When the comprehensive occlusion field index Ovf(k) > the occlusion risk threshold Tovf, the comparison result is the risk result, and the k-th station is marked as a high-risk occlusion station.

[0129] When the comprehensive occlusion field index Ovf(k) ≤ occlusion risk threshold Tovf, the comparison result is a normal result. The kth station is not marked as a high-risk occlusion station and is removed.

[0130] S4 includes S42;

[0131] S42. Based on the high-risk point index set Rls and combined with the original visual domain dataset Vfd of the high-risk station sites, generate hoarding structure optimization suggestions for each high-risk station site. The hoarding structure optimization suggestions include adjusting the hoarding layout shape of the high-risk station sites, adjusting the perspective window and semi-transparent visual domain area of ​​the high-risk station sites, and adjusting the number of eye-catching signs guiding pedestrian flow through the visual priority channel, and generate a hoarding structure optimization suggestion list Opl.

[0132] S5 includes S51;

[0133] S51. Implement and apply the optimization suggestion list Opl for the enclosure structure. The implementation and application include sending a notification to relevant operators for actual deployment, collecting the original field of view dataset Vfd for each high-risk occlusion site after actual deployment, calculating and obtaining a new occlusion risk score set OvfList (New) after actual deployment, comparing it with the occlusion risk score set OvfList before actual deployment, summarizing the difference, obtaining the optimization and improvement score Dsc, and then judging the effectiveness status of the measures after actual deployment based on the optimization and improvement score Dsc.

[0134] The measures implemented after actual deployment are judged in the following ways:

[0135] When the optimization and improvement score Dsc > 0, it indicates that the implementation and application of the fence structure optimization suggestion list Opl is effective;

[0136] When the optimization and improvement score Dsc≤0, it means that the implementation of the fence structure optimization suggestion list Opl is invalid, and optimization measures such as fence shape adjustment, perspective window setting, addition of prompt signs, and reconstruction of path visual guidance need to be carried out again.

[0137] In this embodiment, by normalizing the multidimensional parameters in the occlusion feature vector set Vft, a comprehensive occlusion view index Ovf(k) is constructed. This achieves a fusion expression of different types of occlusion risk factors under a unified scoring scale, effectively solving the problem of the inability to quantify and unify structural occlusion factors and dynamic traffic interference. Based on the occlusion risk scoring set OvfList, and using the total number of stations N as the index boundary, high-risk occlusion stations are efficiently screened based on the set occlusion risk threshold Tovf, forming a high-risk point index set Rls. Furthermore, combining the specific structural layout of each risk point in the original view dataset Vfd, a list of proposed fence structure optimization suggestions Opl is output to guide differentiated governance measures, including fence layout adjustments, perspective window modifications, and the addition of guidance prompts. The re-collected occlusion risk scoring set OvfList(New) is compared with the original scoring set OvfList to form an optimization and improvement score Dsc, used to measure the actual effectiveness after deployment. This method not only achieves accurate identification of highly occluded areas, but also forms a risk-strategy-feedback closed-loop path with the comprehensive occlusion field index Ovf(k) as the core. It has dynamic response capabilities and verifiable effects, which are significantly better than traditional manual experience judgment and single path optimization strategies. It is especially suitable for urban construction areas with complex line-of-sight intersections and serious pedestrian and vehicle traffic, and realizes continuous self-assessment and strategy regeneration capabilities for construction traffic safety.

[0138] Example 5: A comprehensive road construction management and control system based on the Internet of Things (IoT). Please refer to... Figure 2 Specifically, it includes a road construction network data collection module, a construction feature extraction module, an occlusion risk analysis module, an occlusion judgment module, and an iterative optimization module;

[0139] The road construction network data acquisition module collects spatial structure and pedestrian behavior data of the road construction area based on the Internet of Things, forming the original visual domain dataset Vfd of the construction area.

[0140] After the construction feature extraction module extracts features based on the original visual domain dataset Vfd, it constructs an occlusion feature vector set Vft, which includes occlusion rate Sor, path conflict factor Cof, and visual blur Blu.

[0141] The feature vector Vft of the occlusion risk analysis module is input into the occlusion risk scoring model for training, and a comprehensive occlusion field index Ovf is constructed, which is combined into an occlusion risk score set OvfList for each station.

[0142] The occlusion judgment module uses the preset occlusion risk threshold Tovf to traverse and compare the occlusion risk score set OvfList, obtains all the site locations marked as high-risk points by the judgment results, forms a high-risk point index set Rls, and generates a fence structure optimization suggestion list Opl based on the high-risk point index set Rls.

[0143] The iterative optimization module applies the optimization suggestion list Opl for the fence structure, and then compares the occlusion risk score set OvfList of the site before and after the application to obtain the optimization improvement score Dsc.

[0144] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A comprehensive management and control method for road construction based on the Internet of Things, characterized in that: Includes the following steps: S1. Based on the Internet of Things, collect spatial structure and pedestrian behavior data of the road construction area, extract the structural parameter subset StructSet and the behavioral parameter subset BehavSet to form the original visual domain dataset Vfd of the construction area. S2. After extracting features based on the original visual field dataset Vfd, construct an occlusion feature vector set Vft by calculating the feature vector of each station point, including occlusion rate Sor, path conflict factor Cof, and visual blur Blu. S3. The occlusion feature vector set Vft is input into the occlusion risk scoring model for training, and a comprehensive occlusion visual field index Ovf is constructed and combined into an occlusion risk score set OvfList for each station. S4. Using the preset occlusion risk threshold Tovf, the occlusion risk score set OvfList is traversed and compared to determine the location of all stations marked as high-risk points by the determination results, forming a high-risk point index set Rls, and generating a fence structure optimization suggestion list Opl based on the high-risk point index set Rls. S5. Implement the optimization suggestion list Opl for the enclosure structure and collect a new set of occlusion risk scores OvfList (New). Then compare it with the occlusion risk score set OvfList of the site before implementation and obtain the optimization improvement score Dsc. Based on the optimization improvement score Dsc, determine the effectiveness status of the measures after implementation.

2. The comprehensive road construction management and control method based on the Internet of Things according to claim 1, characterized in that: S1 includes S11; S11. Collect spatial projection data of the road construction area fence and the visible range information of the station locations through IoT devices, and extract the spatial structure parameters of each station location in the construction area, including the actual occlusion area Aov, the theoretical visible range of the station location Avw, the pedestrian traffic direction angle Dag, and the main viewing direction of the station location Vag, to form a subset of structural parameters of all station locations StructSet={Aov(k),Avw(k),Dag(k),Vag(k)|k∈N}, where N represents the total number of station locations, and Aov(k), Avw(k), Dag(k), and Vag(k) represent the actual occlusion area Aov, the theoretical visible range Avw, the pedestrian traffic direction angle Dag, and the main viewing direction Vag of the kth station location, respectively; Among them, IoT devices include panoramic wide-angle cameras and 3D laser scanners; The actual occlusion area Aov is obtained by acquiring the projected images of the construction fence at different angles in real time through a panoramic wide-angle camera and a 3D laser scanner installed above the construction area. The boundary of the occlusion area is extracted by an image processing algorithm, which includes a graphic contour filling algorithm and a point cloud bounding volume division method. The theoretical visible range Avw of the station location is simulated by using a 3D laser scanner installed and a high-precision map device deployed at the boundary of the construction area to perform a laser scan in a 180° direction in front of the station location without considering the obstruction of the fence. Combined with the existing road structure map, the theoretical open field of view area is calculated by fan-shaped expansion. The main view direction Vag of the station point is obtained by measuring the maximum continuous unobstructed angle from the left and right boundaries with the station point as the center and the angle of the projection ray preset. The pedestrian direction angle Dag is obtained by capturing the movement trajectory of the crowd through a panoramic wide-angle camera and combining it with the Ak path fitting algorithm to calculate the angle between the direction of the crowd movement trajectory and the direction of the station position.

3. The comprehensive road construction management and control method based on the Internet of Things according to claim 2, characterized in that: S1 includes S12; S12. By deploying sensing devices in the road construction area, obtain the traffic behavior feature data of the crowd in the construction site, extract the behavioral interaction parameters required to build the occlusion conflict analysis model, specifically including the path intersection angle Xag and the pedestrian density Dpt per unit time, and form a behavioral parameter subset BehavSet={Dpt, Xag(m, w)|(m, w)∈M*M}, m≠w, and then integrate it with the structural parameter subset StructSet to obtain the original visual domain dataset Vfd of the construction area; The path intersection angle Xag is continuously captured by a panoramic wide-angle camera; The edge module detects pedestrians and performs multi-target tracking in the image; it records the trajectory point sequence of each pedestrian to form a path set {P1, P2, ..., Pm|m∈M}, where M represents the total number of paths and Pm represents the m-th path; then it performs linear fitting on each path in the path set to obtain the unit direction vector v, and calculates the path intersection angle Xag between any two paths by calculating the unit direction vector v. The unit direction vector v is obtained using the following formula: ； In the formula, v(m) represents the unit direction vector of the m-th path, and x(m, end) and y(m, end) and x(m, start) and y(m, start) represent the coordinates of the end point and the start point of the m-th path, respectively. The path intersection angle Xag is obtained using the following formula: ； In the formula, Xag(m, w) represents the path intersection angle between the m-th path and the w-th path, and arccos represents the inverse cosine function, which specifically converts the cosine value into an angle value; The pedestrian density per unit time, Dpt, is obtained by measuring the ratio of the number of pedestrians in the road construction area to the time interval within a fixed statistical time window.

4. The comprehensive road construction management and control method based on the Internet of Things according to claim 3, characterized in that: S2 includes S21; S21. Based on the original visual field dataset Vfd, feature extraction is performed. After extracting the structural parameter subset StructSet of each station, the actual occlusion area Aov and the theoretical unobstructed visible range Avw are extracted. The occlusion rate Sor of each station is calculated to quantify the degree to which each station is occluded by the enclosure structure. The occlusion rate Sor is obtained using the following formula: ； In the formula, Sor(k) represents the occlusion rate of the k-th station, and Aov(k) and Avw(k) represent the actual occlusion area Aov and the theoretical unobstructed visible range Avw of the k-th station, respectively.

5. The comprehensive road construction management and control method based on the Internet of Things according to claim 4, characterized in that: S2 includes S22; S22. Based on the original visual field dataset Vfd, feature extraction is performed. By extracting the behavioral parameter subset BehavSet and structural parameter subset StructSet of each station, key dynamic data reflecting the relationship between the crowd's behavioral path and direction are obtained. Then, the pedestrian density per unit time Dpt, path intersection angle Xag, passage direction angle Dag, and station main view direction Vag are extracted to construct the path conflict factor Cof and visual blur Blu. Then, they are integrated with the occlusion rate Sor to obtain the occlusion feature vector set Vft. The path conflict factor Cof is obtained using the following formula: ； In the formula, Cof(m, w, k) represents the path conflict factor between path m and path w at the k-th station, sin represents the sine function, Dpt(k) represents the pedestrian density per unit time at the tangent point of the k-th station, Dag(k) represents the direction angle of passage at the k-th station, and Vag(k) represents the main view direction of the k-th station. Blu, the visual blur level, is obtained using the following formula: ； In the formula, Blu(k) represents the visual blur at the k-th station.

6. The comprehensive road construction management and control method based on the Internet of Things according to claim 5, characterized in that: S3 includes S31; S31. Normalize the occlusion feature vector set Vft to eliminate the dimensional differences between different data, and obtain the standard path conflict factor Cofn, standard visual blur Blun, and standard occlusion rate Sorn. Then input them into the occlusion risk scoring model and perform fusion training with each station k as a unit to construct a comprehensive index that reflects the degree of influence of multidimensional occlusion risk. This index is marked as the comprehensive occlusion visual field index Ovf(k) of the kth station. After integrating all stations, obtain the occlusion risk scoring set OvfList={Ovf(k)|k∈N}, where N represents the total number of stations. The comprehensive occlusion field index Ovf(k) of the kth station is obtained through the following model fusion training formula: ； In the formula, N(k) represents the number of path pairs at the k-th station, which is obtained by the formula N(k) = |M(k)|*(||M(k)-1) / 2. M represents the total number of paths. Sorn(k) represents the standard occlusion rate of the k-th station. Blun(k) represents the standard visual blur of the k-th station. Cofn(m,w,k) represents the standard path conflict factor between path m and path w at the k-th station.

7. The comprehensive road construction management and control method based on the Internet of Things according to claim 6, characterized in that: S4 includes S41; S41. Using the total number of stations N as the traversal boundary, compare each comprehensive occlusion field index Ovf(k) in the occlusion risk score set OvfList with the preset occlusion risk threshold Tovf, and mark high-risk occlusion stations according to the comparison results. After the traversal is completed, integrate all high-risk occlusion stations and obtain the high-risk point index set Rls. High-risk site locations with obstruction are marked using the following comparison method: When the comprehensive occlusion field index Ovf(k) > the occlusion risk threshold Tovf, the comparison result is the risk result, and the kth station is marked as a high-risk occlusion station. When the comprehensive occlusion field index Ovf(k) ≤ occlusion risk threshold Tovf, the comparison result is a normal result. The kth station is not marked as a high-risk occlusion station and is removed.

8. The comprehensive road construction management and control method based on the Internet of Things according to claim 7, characterized in that: S4 includes S42; S42. Based on the high-risk point index set Rls and combined with the original visual domain dataset Vfd of the high-risk station sites, generate hoarding structure optimization suggestions for each high-risk station site. The hoarding structure optimization suggestions include adjusting the hoarding layout shape of the high-risk station sites, adjusting the perspective window and semi-transparent visual domain area of ​​the high-risk station sites, and adjusting the number of eye-catching signs guiding pedestrian flow through the visual priority channel, and generate a hoarding structure optimization suggestion list Opl.

9. The comprehensive management and control method for road construction based on the Internet of Things according to claim 1, characterized in that: S5 includes S51; S51. Implement and apply the optimization suggestion list Opl for the enclosure structure. The implementation and application include sending a notification to relevant operators for actual deployment, collecting the original field of view dataset Vfd for each high-risk occlusion site after actual deployment, calculating and obtaining a new occlusion risk score set OvfList (New) after actual deployment, comparing it with the occlusion risk score set OvfList before actual deployment, summarizing the difference, obtaining the optimization and improvement score Dsc, and then judging the effectiveness status of the measures after actual deployment based on the optimization and improvement score Dsc. The measures implemented after actual deployment are judged in the following ways: When the optimization and improvement score Dsc > 0, it indicates that the implementation and application of the fence structure optimization suggestion list Opl is effective; When the optimization and improvement score Dsc≤0, it means that the implementation of the fence structure optimization suggestion list Opl is invalid, and optimization measures such as fence shape adjustment, perspective window setting, addition of prompt signs, and reconstruction of path visual guidance need to be carried out again.

10. A road construction integrated management and control system based on the Internet of Things (IoT), applied to the road construction integrated management and control method based on the IoT as described in any one of claims 1 to 9, characterized in that: It includes a road construction network data collection module, a construction feature extraction module, an occlusion risk analysis module, an occlusion judgment module, and an iterative optimization module; The road construction network data acquisition module collects spatial structure and pedestrian behavior data of the road construction area based on the Internet of Things, forming the original visual domain dataset Vfd of the construction area. After the construction feature extraction module extracts features based on the original visual domain dataset Vfd, it constructs an occlusion feature vector set Vft, which includes occlusion rate Sor, path conflict factor Cof, and visual blur Blu. The occlusion feature vector set Vft of the occlusion risk analysis module is input into the occlusion risk scoring model for training, and a comprehensive occlusion field index Ovf is constructed, which is combined into an occlusion risk score set OvfList for each station. The occlusion judgment module uses the preset occlusion risk threshold Tovf to traverse and compare the occlusion risk score set OvfList, obtains all the site locations marked as high-risk points by the judgment results, forms a high-risk point index set Rls, and generates a fence structure optimization suggestion list Opl based on the high-risk point index set Rls. The iterative optimization module applies the optimization suggestion list Opl for the fence structure, and then compares the occlusion risk score set OvfList of the site before and after the application to obtain the optimization improvement score Dsc.

Images