A method for optimizing the deployment of visual sensor networks in dynamic construction environments based on risk classification maps

Abstract

This invention belongs to the field of information science and technology, specifically a method for optimizing the deployment of visual sensor networks in dynamic construction environments based on risk grading maps. It includes a building information model (BIM) module, a dynamic multi-obstacle visual node optimization deployment data modeling module, and an optimization algorithm and visualization module. The BIM module, the dynamic multi-obstacle visual node optimization deployment data modeling module, and the optimization algorithm and visualization module transmit data via a risk map. The invention also includes the following specific steps: extracting information such as space, structures, and obstacles from the BIM model and generating a risk map. This invention considers the dynamic construction process and the differences in coverage requirements across different areas, grading different areas to allocate resources and prioritize areas with high coverage requirements, thus deriving the optimal visual sensor network deployment scheme throughout the entire process; and achieving visualization of the optimization results.

Description

Technical Field

[0001] This invention relates to the field of information science and technology, specifically to a method for optimizing the deployment of a dynamic construction environment visual sensor network based on a risk classification map. Background Technology

[0002] Existing optimization methods for visual sensor networks at construction sites have improved the optimization methods from multiple perspectives, such as algorithm development, site scene, location, and constraints. However, these methods are mainly based on static construction sites. In reality, construction sites are constantly changing and complex environments, and the importance of different areas is not the same. Once the location of the visual sensor is determined, it is very costly to change the location later.

[0003] BIM, or Building Information Modeling, encompasses geometry, spatial relationships, geographic information systems, and the properties and quantities of various building components (such as detailed supplier information). BIM can be used to represent the entire product lifecycle of a building, including construction and operation. Compared to traditional construction models, BIM possesses non-geometric information not found in other models, including attributes such as spatial relationships, physical characteristics, and functional features, reflecting more accurate construction project information. BIM can store various types of information, such as three-dimensional spatial information and four-dimensional schedule information.

[0004] Parallel Genetic Algorithm (PGMA) refers to a genetic algorithm designed in parallel. It is a multi-population parallel evolutionary genetic algorithm suitable for complex optimization problems. This algorithm can effectively overcome the premature convergence problem of standard genetic algorithms and has strong global search capabilities.

[0005] The current technical solution, as described in reference [1] (Hu Yao, Feng Yong, Li Yingna, et al. Deployment method of directed charger for wireless sensor network based on virtual potential field in multi-obstacle environment [J]. Journal of Sensor Technology, 2020, 33(9): 1333-1340. DOI: 10.3969 / j.issn.1004-1699.2020.09.017.), uses a heterogeneous particle swarm algorithm to solve the deployment optimization problem of heterogeneous sensor network. It mainly aims to adapt to the heterogeneity of the network and improve the network performance, but does not consider the influence of obstacles. However, unlike traditional wireless sensor networks (WSN), the sensing range of VSN nodes is very sensitive to obstacles, and the presence of obstacles greatly reduces the sensing range of nodes. In past studies, due to the limitations of multi-obstacle scene model construction, many scholars have idealized the optimization scene as an obstacle-free model, but this makes it difficult to apply the proposed optimization method to the actual situation.

[0006] Reference [2] (Wang Changzheng, Mao Jianlin, Fu Lixia, et al. Coverage optimization algorithm for directed heterogeneous sensor networks with multiple obstacles [J]. Computer Engineering and Applications,

[0007] 2017, 53(12):92-98.DOI:10.3778 / j.issn.1002-8331.1601-0143.) proposed a PCMODH coverage optimization algorithm to solve the coverage optimization problem of sensor networks in multi-obstacle environments. However, it still uses manually input scene simulation data, which cannot accurately provide real scene data. This not only makes the model unable to fit the real scene, but also consumes a lot of time.

[0008] Existing optimization methods are only applicable to large spaces with discrete priority regions. These methods focus only on local changes in simple obstacles, neglecting the global field of view and project progress. In summary, few studies address the dynamic nature of construction because they focus only on static and single scenes to optimize visual sensor placement. Therefore, this paper proposes an optimization method for visual sensor network deployment in dynamic construction environments based on risk grading maps. Summary of the Invention

[0009] In view of the problems existing in the above and / or the existing method for optimizing the deployment of visual sensor networks in dynamic construction environments based on risk classification maps, this invention is proposed.

[0010] Therefore, the purpose of this invention is to provide a method for optimizing the deployment of a dynamic construction environment visual sensor network based on a risk grading map. By using the risk map as a data medium, the BIM module, mathematical module, and optimization module are connected to visualize discrete locations and optimal results, thereby solving the aforementioned existing problems.

[0011] To address the aforementioned technical problems, according to one aspect of the present invention, the present invention provides the following technical solution:

[0012] A method for optimizing the deployment of a dynamic construction environment visual sensor network based on a risk grading map includes a building information modeling (BIM) module, a dynamic multi-obstacle visual node optimization deployment data modeling module, and an optimization algorithm and visualization module. The BIM module, the dynamic multi-obstacle visual node optimization deployment data modeling module, and the optimization algorithm and visualization module transmit data via a risk map. The method also includes the following specific steps:

[0013] Step 1, Building Information Modeling (BIM) module: Extract information such as space, structures, and obstacles from the BIM model and generate a risk map;

[0014] Step 2, Dynamic Multi-obstacle Visual Node Optimization Deployment Data Modeling Module: Establish an analysis model of the effective perception area of ​​nodes and an optimization model for the deployment of the visual sensor network;

[0015] Step 3, Algorithm and Visualization Module Optimization: Parallel genetic algorithms are used for optimization.

[0016] As a preferred embodiment of the dynamic construction environment visual sensor network deployment optimization method based on risk classification map described in this invention, the specific steps of the building information model module are as follows: the building information model is saved as an IFC format file, and the location information of objects such as space and structures is extracted from the IFC file through the Python external library IfcOpenShell. The space is discretized into a W*L grid set. Given a grid side length of a, the a*a area in the site will be mapped to pixels in the risk map. The smaller the resolution, the higher the mapping accuracy. Subsequently, the structures are represented in the risk map.

[0017] As a preferred embodiment of the dynamic construction environment visual sensor network deployment optimization method based on risk classification map described in this invention, wherein: W and L can be obtained from the following:

[0018]

[0019]

[0020] Linear structures are represented as:

[0021]

[0022] Columnar structures are represented as:

[0023]

[0024]

[0025] Structures are marked in gray, while other open spaces are marked in white. The structures are then assigned an importance level and a risk level of v. k Measure the Euclidean distance from each grid point (x, y) to the structure, and then select the minimum distance d to the nearest structure and the corresponding risk level v. k We obtain the risk level vector [d, k], and then map the obtained risk level vector onto the risk map, given a risk level range vector T = t1, t2, ..., t Q-1 The risk map value RM(x,y) is assigned to each grid point (x,y), and the functions defining the risk level vector [d,k] and risk level range vector T are as follows:

[0026]

[0027] Where 1 < k ≤ Q.

[0028] As a preferred embodiment of the dynamic construction environment visual sensor network deployment optimization method based on risk classification map described in this invention, the algorithm flow for whether grid points can be effectively perceived by nodes in the dynamic multi-obstacle visual node optimization deployment data modeling module is as follows:

[0029] Step 1: First, the following condition must be met: the grid points should fall within the effective sensing radius R of the node, where R = r / a, and r is the actual radiation distance of the visual node. Assume the visual node is located at (x... c y c ), then grid point (x q y q ) needs to satisfy (x) q -x c ) 2 +(y q -y c ) 2 ≤R 2 ;

[0030] Step 2: Calculate the obstacle point (x0, y0) and the region point (x0, y0) respectively. q y q The angle from the visual node is as follows:

[0031]

[0032]

[0033] If angle qc ≠angle oc Then the region points can be perceived, if angle qc =angle oc Then calculate the Euclidean distance d to each node. qc d oc ;

[0034] Step 3: If d qc ≥d oc If d, then the region cannot be perceived. qc <d oc Then the area can be perceived;

[0035] Step 4: Repeat the above steps for the grid points within the node's sensing range to obtain the final effective sensing range of the node;

[0036] Step 5: Repeat the above steps for each node to obtain the perception range of the visual sensor network;

[0037] A formula for the coverage of a multi-stage visual sensor network is established, with the optimization objective being to maximize it.

[0038]

[0039] The constraints are:

[0040]

[0041] in:

[0042]

[0043] Where, p it Indicates whether the i-th working area is covered by the candidate visual sensor group, Q tk This represents the number of regions with risk level k in stage t, where K is the number of risk levels, T is the stage number, and w k Z represents the weight under risk k. j This indicates whether the j-th visual sensor has been installed. G is the total number of candidate visual sensors, BG is the available budget for the monitoring visual sensor system, and CP is the unit price of the monitoring visual sensor system. Different combinations of weight sets can be used for different security needs. k The larger the value, the higher the requirement for visual sensor coverage in that area.

[0044] As a preferred embodiment of the dynamic construction environment visual sensor network deployment optimization method based on risk grading map described in this invention, the specific steps of the optimization algorithm and visualization module are as follows: Since the camera layout problem is an NP-hard problem, classic genetic algorithms often have weak convergence ability and long computation time. A parallel improved genetic algorithm is adopted to obtain better performance. Because the fitness calculation of each individual is independent, the genetic algorithm can distribute the fitness calculation of individuals to multiple CPU cores. First, an initial population of N individuals is generated. Second, if the stopping condition is not met, individuals will crossover and mutate with a certain probability to generate a new population. Next, the new individuals are combined with the original individuals. The fitness is calculated using the multi-stage visual sensor network coverage formula in the dynamic multi-obstacle visual node optimization deployment data modeling module. Based on the fitness, probability P is used... i Selecting N individuals to create the next population, in this modified algorithm, the mutation probability can be set to a higher value to ensure population diversity. Finally, the optimal solution is obtained and presented in the building information model for visualization.

[0045] As a preferred embodiment of the dynamic construction environment visual sensor network deployment optimization method based on risk classification map described in this invention, wherein: the probability P i The formula is as follows:

[0046]

[0047] Among them, P i F represents the choice probability of the i-th individual. i P represents the fitness of the i-th individual. i The value is between 0 and 1.

[0048] Compared with existing technologies:

[0049] This invention takes into account the dynamic construction process and the differences in coverage requirements in different areas. It classifies different areas to allocate resources to prioritize areas with high coverage requirements, thereby deriving the optimal visual sensor network deployment scheme throughout the entire process and realizing the visualization of the optimization results. Attached Figure Description

[0050] Figure 1 This is a schematic diagram of the visual sensor network deployment optimization framework of the present invention;

[0051] Figure 2 This invention discretizes the space into a graph.

[0052] Figure 3 This is a schematic diagram of the risk map generation process of the present invention;

[0053] Figure 4 This is a diagram of the parallel genetic algorithm of the present invention;

[0054] Figure 5 This is a visualization of the results of the present invention. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0056] This invention provides a method for optimizing the deployment of visual sensor networks in dynamic construction environments based on risk grading maps. Please refer to [link / reference]. Figures 1-5 It includes a building information modeling module, a dynamic multi-obstacle visual node optimization deployment data modeling module, and an optimization algorithm and visualization module. The building information modeling module, the dynamic multi-obstacle visual node optimization deployment data modeling module, and the optimization algorithm and visualization module transmit data through a risk map.

[0057] The specific steps also include the following:

[0058] Step 1, Building Information Modeling (BIM) module: Extract information such as space, structures, and obstacles from the BIM model and generate a risk map;

[0059] The specific steps of the Building Information Modeling (BIM) module are as follows: Save the BIM model as an IFC format file. Using the Python add-in IfcOpenShell, extract the location information of spaces, structures, and other objects from the IFC file. Discretize the space into a W*L grid set. Given a grid side length 'a', an area of ​​'a'*a' in the site will be mapped to pixels in the risk map. The smaller the resolution, the higher the mapping accuracy. Then, represent the structures in the risk map (e.g., ...). Figure 2 (as shown);

[0060] W and L can be obtained from the following:

[0061]

[0062]

[0063] Linear structures are represented as:

[0064]

[0065] Columnar structures are represented as:

[0066]

[0067]

[0068] Structures are marked in gray, while other open spaces are marked in white. The structures are then labeled with their importance level (e.g., ...). Figure 3 As shown), assign a risk level v to the structure. k Measure the Euclidean distance from each grid point (x, y) to the structure, and then select the minimum distance d to the nearest structure and the corresponding risk level v. k We obtain the risk level vector [d, k], and then map the obtained risk level vector onto the risk map, given a risk level range vector T = t1, t2, ..., t Q-1 The risk map value RM(x,y) is assigned to each grid point (x,y), and the functions defining the risk level vector [d,k] and risk level range vector T are as follows:

[0069]

[0070] Where 1 < k ≤ Q;

[0071] Step 2, Dynamic Multi-obstacle Visual Node Optimization Deployment Data Modeling Module: Establish an analysis model of the effective perception area of ​​nodes and an optimization model for the deployment of the visual sensor network;

[0072] The algorithm flow for determining whether grid points can be effectively perceived by nodes in the dynamic multi-obstacle visual node optimization deployment data modeling module is as follows:

[0073] Step 1: First, the following condition must be met: the grid points should fall within the effective sensing radius R of the node, where R = r / a, and r is the actual radiation distance of the visual node. Assume the visual node is located at (x... c y c ), then grid point (x q y q ) needs to satisfy (x) q -x c ) 2 +(y q -y c ) 2 ≤R 2 ;

[0074] Step 2: Calculate the obstacle point (x0, y0) and the region point (x0, y0) respectively. q y q The angle from the visual node is as follows:

[0075]

[0076]

[0077] If angle qc ≠angle oc Then the region points can be perceived, if angle qc =angle oc Then calculate the Euclidean distance d to each node. qc d oc ;

[0078] Step 3: If d qc ≥d oc If d, then the region cannot be perceived. qc <d oc Then the area can be perceived;

[0079] Step 4: Repeat the above steps for the grid points within the node's sensing range to obtain the final effective sensing range of the node;

[0080] Step 5: Repeat the above steps for each node to obtain the perception range of the visual sensor network;

[0081] A formula for the coverage of a multi-stage visual sensor network is established, with the optimization objective being to maximize it.

[0082]

[0083] The constraints are:

[0084]

[0085] in:

[0086]

[0087] Where, p it Indicates whether the i-th working area is covered by the candidate visual sensor group, Q tk This represents the number of regions with risk level k in stage t, where K is the number of risk levels, T is the stage number, and w k Z represents the weight under risk k. j This indicates whether the j-th visual sensor has been installed. G is the total number of candidate visual sensors, BG is the available budget for the monitoring visual sensor system, and CP is the unit price of the monitoring visual sensor system. Different combinations of weight sets can be used for different security needs. k The larger the value, the higher the requirement for visual sensor coverage in that area;

[0088] Step 3, Algorithm and Visualization Module Optimization: A parallel genetic algorithm is used for optimization;

[0089] The specific steps for optimizing the algorithm and visualization module are as follows (e.g.) Figure 4 (As shown): Since the camera placement problem is an NP-hard problem, classic genetic algorithms often have weak convergence ability and long computation time. Parallel improvements to the genetic algorithm are adopted to achieve better performance. Because the fitness calculation of each individual is independent, the genetic algorithm can distribute the fitness calculation of individuals to multiple CPU cores. First, an initial population of N individuals is generated. Second, if the stopping condition is not met, individuals will crossover and mutate with a certain probability to generate a new population. Next, the new individuals are combined with the original individuals. The fitness is calculated using the multi-stage visual sensor network coverage formula in the dynamic multi-obstacle visual node optimization deployment data modeling module. Based on the fitness, the probability P is used... i Selecting N individuals to create the next population, in this modified algorithm, allows for a higher mutation probability to ensure population diversity. The final optimal solution is then presented in the Building Information Model (BIM) for visualization (e.g., ...). Figure 5 (As shown).

[0090] probability P i The formula is as follows:

[0091]

[0092] Among them, P i F represents the choice probability of the i-th individual. i P represents the fitness of the i-th individual. i The value is between 0 and 1.

[0093] Among them, such as Figure 3As shown, (a) the boundaries of the structure and the open space are determined; (b) the risk level of the structure is determined; and (c) the risk level of the open space is determined based on the Euclidean distance to the structure.

[0094] Although the present invention has been described above with reference to embodiments, various modifications can be made and components can be replaced with equivalents without departing from the scope of the invention. In particular, as long as there is no structural conflict, the features in the disclosed embodiments can be combined with each other in any manner. The lack of an exhaustive description of these combinations in this specification is merely for the sake of brevity and resource conservation. Therefore, the present invention is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. A method for optimizing the deployment of a dynamic construction environment visual sensor network based on a risk grading map, comprising a building information modeling module, a dynamic multi-obstacle visual node optimization deployment data modeling module, and an optimization algorithm and visualization module, wherein the building information modeling module, the dynamic multi-obstacle visual node optimization deployment data modeling module, and the optimization algorithm and visualization module transmit data through a risk map, characterized in that, The specific steps also include the following: Step 1, Building Information Modeling (BIM) module: Extract information on space, structures, and obstacles from the BIM model and generate a risk map; Step 2, Dynamic Multi-obstacle Visual Node Optimization Deployment Data Modeling Module: Establish an analysis model of the effective perception area of ​​nodes and an optimization model for the deployment of the visual sensor network; Step 3, Algorithm and Visualization Module Optimization: A parallel genetic algorithm is used for optimization; The algorithm flow for determining whether grid points can be effectively perceived by nodes in the dynamic multi-obstacle visual node optimization deployment data modeling module is as follows: Step 1: First, the following condition must be met: the grid points should fall within the effective sensing radius R of the node, where R = r / a, and r is the actual radiation distance of the visual node. Assume the visual node is located at (x... c y c ), then grid point (x q y q ) needs to satisfy (x) q -x c ) 2 +(y q -y c ) 2 ≤R 2 ; Step 2: Calculate the obstacle point (x0, y0) and the region point (x0, y0) respectively. q y q The angle from the visual node is as follows: If angle qc ≠angle oc Then the region points can be perceived, if angle qc =angle oc Then calculate the Euclidean distance d to each node. qc d oc ; Step 3: If d qc ≥d oc If d, then the region cannot be perceived. qc <d oc Then the area can be perceived; Step 4: Repeat the above steps for the grid points within the node's sensing range to obtain the final effective sensing range of the node; Step 5: Repeat the above steps for each node to obtain the perception range of the visual sensor network; A formula for the coverage of a multi-stage visual sensor network is established, with the optimization objective being to maximize it. The constraints are: in: Where, p it Indicates whether the i-th working area is covered by the candidate visual sensor group, Q tk This represents the number of regions with risk level k in stage t, where K is the number of risk levels, T is the stage number, and w k Z represents the weight under risk k. j This indicates whether the j-th visual sensor has been installed. G is the total number of candidate visual sensors, BG is the available budget for the monitoring visual sensor system, and CP is the unit price of the monitoring visual sensor system. Different combinations of weight sets can be used for different security needs. k The larger the value, the higher the requirement for visual sensor coverage in that area; The specific steps of the optimization algorithm and visualization module are as follows: Since the camera layout problem is an NP-hard problem, classic genetic algorithms often have weak convergence ability and long computation time. A parallel improved genetic algorithm is adopted to obtain better performance. Because the fitness calculation of each individual is independent, the genetic algorithm can distribute the fitness calculation of individuals to multiple CPU cores. First, an initial population of N individuals is generated. Second, if the stopping condition is not met, individuals will crossover and mutate with a certain probability to generate a new population. Next, the new individuals are combined with the original individuals. The fitness is calculated using the multi-stage visual sensor network coverage formula in the dynamic multi-obstacle visual node optimization deployment data modeling module. Based on the fitness, probability P is used... i Selecting N individuals to create the next population, in this modified algorithm, the mutation probability can be set to a higher value to ensure population diversity. Finally, the optimal solution is presented in the building information model for visualization. The probability P i The formula is as follows: Among them, P i F represents the choice probability of the i-th individual. i P represents the fitness of the i-th individual. i The value is between 0 and 1.

2. The method for optimizing the deployment of a dynamic construction environment visual sensor network based on a risk classification map according to claim 1, characterized in that, The specific steps of the building information model module are as follows: save the building information model as an IFC format file, extract the spatial and structural object location information from the IFC file using the Python external library IfcOpenShell, discretize the space into a W*L grid set, and given a grid side length a, the a*a area in the site will be mapped to pixels in the risk map. The smaller the resolution, the higher the mapping accuracy. Then, the structures are represented in the risk map.

3. The method for optimizing the deployment of a dynamic construction environment visual sensor network based on a risk classification map according to claim 2, characterized in that, The values ​​of W and L can be obtained from the following: Linear structures are represented as: Columnar structures are represented as: Structures are marked in gray, while other open spaces are marked in white. The structures are then assigned an importance level and a risk level of v. k Measure the Euclidean distance from each grid point (x, y) to the structure, and then select the minimum distance d to the nearest structure and the corresponding risk level v. k We obtain the risk level vector [d, k], and then map the obtained risk level vector onto the risk map, given a risk level range vector T = t1, t2, ..., t Q-1 The risk map value RM(x,y) is assigned to each grid point (x,y), and the functions defining the risk level vector [d,k] and risk level range vector T are as follows: Where 1 < k ≤ Q.

Images