Large-scale historic building gridding early warning method for multiple data sources

Abstract

The invention discloses a large-scale historic building gridding early warning method for multiple data sources, and the method comprises the steps: collecting and sorting fire risk indexes, and building a fire risk index system; solving a fire risk index weight and a final risk model, namely comparing two indexes of a certain level to obtain a judgment matrix of the level, performing consistency check, and obtaining the final risk model containing the indexes according to a weight vector Y of the judgment matrix; designing a grid division algorithm based on density inverse proportion weight to complete digitization of grids: converting discrete information data of the historic building into continuous change, and compiling the index weight calculated by an analytic hierarchy process into the attributes of the final divided grids to complete digitization of the grids; and through visualization of the risk grid, early warning is carried out on the fire risk of the ancient building. According to the invention, a gridding and visualized fire early warning method can be realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of fire early warning, and in particular to a fire risk gridding early warning method for large ancient buildings based on multiple data sources. Background Art

[0002] Ancient buildings contain cultural relics and cultural heritage of great value. Once they are burned in a fire, it will be a great loss to human cultural heritage. Ancient buildings carry the changes of history and have high artistic and cultural value. The older the age, the more valuable the historical and cultural information they contain. The fire warning technology of ancient buildings has always been a complex issue involving multiple influencing factors. Large ancient buildings are generally distributed in groups, and their area generally reaches 300,000 square meters. Once a fire hazard occurs, the impact is severe. Especially considering that most ancient buildings use wooden structures with poor fire resistance, coupled with years of natural erosion and a large number of artificial development and utilization, they face a very high probability of fire. The historical relics in ancient buildings are non-renewable cultural heritage. The rich historical and cultural value they contain is an important data for studying ancient history, science, politics and economy. Once a fire hazard occurs, the result will cause irreparable huge losses. Therefore, appropriate fire warning technology, that is, fire risk visualization technology, is crucial for the protection of cultural relics and ancient buildings.

[0003] The current mainstream application technology is the fire warning technology that combines smoke sensors and image processing algorithms. Smoke information data is collected by smoke sensors, and then the corresponding fire probability values ​​are trained through image processing algorithms such as binarization and grayscale processing. For example, Wang Tao and Wan Yuemin. Application of independent photoelectric intelligent fire detection technology in ancient buildings in Tibetan areas [J]. Science and Technology Innovation, 2024, (3): 162-6. An intelligent detection algorithm based on photoelectric smoke sensing is designed. According to the characteristics of the floor area and environment of ancient buildings in Tibetan areas, insufficient fire detection and early warning needs, a smoke sensing algorithm based on particle size feature extraction is designed through the scattering characteristics of black and white smoke particles, thereby improving the sensitivity of fire detection. However, there is a problem of threshold alarm delay, and the timeliness of the algorithm is poor. There are still many problems in applying smoke sensors and image processing algorithms to ancient buildings;

[0004] 1. Slow response speed. Since the main body is a variety of data sources of large ancient buildings, if image processing is performed on multiple areas, the amount of calculation is very large and the processing speed will be very limited. In addition, most ancient buildings are brick and wood structures, with low fire resistance and lack of flame retardant materials, which will cause the fire to spread faster and make it more difficult to extinguish. In addition, some large ancient buildings are popular tourist attractions with dense crowds. Evacuation work also takes a certain amount of time.

[0005] 2. Sensor threshold delayed alarm. The sensor alarm principle currently used is mostly that when the fire occurs to a certain extent, it will reach the sensor threshold and trigger the alarm, which has a certain time and space delay.

[0006] 3. The visibility of risk level is poor. First, there are many interference factors in the environment of ancient buildings, which will affect the detection accuracy of the sensor; second, the factors considered are relatively single, only the smoke factor is considered, and the objective effect cannot be achieved; finally, image processing technology is more suitable for single-frame image analysis, and for complex images of large ancient buildings, it is impossible to dynamically give an overall visual distribution effect.

[0007] Due to the potential problems mentioned above, the currently widely used image processing technology is not ideal when applied to large ancient buildings and has poor practicality. Summary of the invention

[0008] The purpose of the present invention is to provide a grid-based early warning method for large-scale ancient buildings based on multiple data sources, so as to realize a grid-based and visualized fire early warning method.

[0009] The technical solution to achieve the purpose of the present invention is:

[0010] A grid-based early warning method for large ancient buildings based on multiple data sources, including:

[0011] Step 1: Collect and organize fire risk indicators and establish a fire risk indicator system through the analytic hierarchy process;

[0012] Step 2: Solve the fire risk index weights and the final risk model through the hierarchical analysis method, including comparing two indicators at a certain level, obtaining the judgment matrix of the level, and performing consistency test, and obtaining the final risk model containing various indicators according to the weight vector Y of the judgment matrix;

[0013] Step 3. Design a grid division algorithm based on density inverse proportional weights to complete the grid digitization: transform the information data of the ancient buildings from discrete to continuous changes, and compile the indicator weights calculated by the hierarchical analysis method into the attributes of the final grid division to complete the grid digitization; through the visualization of the risk grid, warn of the fire risk of the ancient buildings.

[0014] Compared with the prior art, the present invention has the following significant advantages:

[0015] (1) This method takes into account multiple data sources and reflects them into grid risk through the hierarchical analysis method, which has a higher accuracy rate than single smoke sensor for fire detection. (2) (Gridding) Through the analysis of ancient building areas, a grid division algorithm based on density inverse proportional weight is proposed to solve the problem of poor visibility of risk level. (3) According to the periodicity of sensor timer, sensor information data is dynamically uploaded to reflect the fire risk distribution in real time and solve the problem of delayed alarm of sensor threshold. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is the flow chart of the meshing algorithm based on density inverse proportional weight;

[0017] Figure 2 A vector file image of a selected ancient building;

[0018] Figure 3 Find the partition diagram for the grid grouping (for example 3x3);

[0019] Figure 4 Vector diagram of meshing after density compensation optimization;

[0020] Figure 5 Divide the vector map into equally spaced grids;

[0021] Figure 6 It is the final risk grid vector diagram combined with AHP weight data;

[0022] Figure 7 Risk grid vector diagram for equally spaced grids; DETAILED DESCRIPTION

[0023] Combination Figure 1 The present invention provides a grid-based early warning method for large-scale ancient buildings based on multiple data sources, comprising the following steps:

[0024] Step 1: Collect and organize fire risk indicators and establish a fire risk indicator system:

[0025] The AHP structure consists of three layers: target layer (A), criterion layer (B), and indicator layer (C). The target layer is the fire risk model of large ancient buildings. The selection of the criterion layer and the indicator layer should ensure that the indicator system is fair, objective, and operational, and should adhere to the principles of hierarchical integrity, qualitative representativeness, feasibility, and objectivity.

[0026] It is assumed that the final fire risk index system is as shown in Table 1.

[0027] Table 1 Hierarchical model

[0028]

[0029] Step 2: Solve the fire risk index weights through the analytic hierarchy process (AHP):

[0030] Step 2.1 Construct the judgment matrix:

[0031] After comparing two indicators at a certain level, the judgment matrix of this level is obtained: the two indicators are compared and judged by the 1-9 scale method, and the relative scale is used to evaluate the level according to the importance, b ij The comparison results of the importance of indicator i and indicator j are shown in Table 2. Taking the calculation of criterion layer B as an example, the judgment matrix obtained through expert scoring is shown in Table 3.

[0032]

[0033] Step 2.2: Consistency check:

[0034] Taking the judgment matrix of the criterion layer about the target layer A in Table 3 as an example, calculate the hierarchical single ranking and perform consistency check. The steps are as follows:

[0035] (1) The judgment matrix X is expressed in matrix form:

[0036]

[0037] (2) The geometric mean method is used to calculate the weight vector Y of the judgment matrix X, and formula (2) is:

[0038]

[0039] Where: x ij is the element in the i-th row and j-th column of the matrix X, and n is the number of indicators in the criterion layer or indicator layer.

[0040] (3) Perform consistency check on the judgment matrix and find the maximum eigenvalue λ of the judgment matrix max , calculate its consistency index CI and consistency ratio CR,

[0041]

[0042] Where: Y k is the kth element in the weight vector matrix obtained by formula (2).

[0043]

[0044] At this time, the random consistency index RI is obtained by looking up the hierarchical analysis table when the n value is obtained, and the consistency ratio CR is:

[0045]

[0046] If CR<0.1, the scheme meets the requirements of the consistency test; otherwise, it does not meet the consistency test and the judgment matrix needs to be reconstructed.

[0047] Step 2.3: Solve the final risk model:

[0048] The weight calculation of indicator layer C with respect to criterion layer B can be taken as an example of the above criterion layer calculation, and Matlab and other related software can be used to assist in the calculation. If the final result CR values ​​all meet the consistency test, the final risk model is expressed as:

[0049] S=Val*y b1 +Pro*y b2 +Fi*y b3 +Res*y b4 +Per*y b5 +Eva*y b6 (8)

[0050] Among them, S represents the final risk factor, Val represents the weights of various indicators under the value index criterion layer, Pro represents the weights of various indicators under the fire possibility criterion layer, Fi represents the weights of various indicators under the fire spread speed criterion layer, Res represents the weights of various indicators under the rescue speed index criterion layer, Per represents the weights of various indicators under the fire perception ability index criterion layer, and Eva represents the weights of various indicators under the evacuation speed index criterion layer.

[0051] Step 3: Design a grid partitioning algorithm based on density inverse proportional weights:

[0052] Several common gridding methods include equidistant gridding, nearest neighbor gridding, inverse distance method (IDW), Kriging and radial basis function. In actual work, appropriate methods are often selected for data gridding according to the research purpose and data characteristics. In view of the discrete type of fire risk gridding characteristic data and the complexity of buildings, a grid division algorithm based on density inverse proportional weight is designed to transform the information data of ancient buildings from discrete to continuous changes, and the indicator weights calculated by AHP analysis are compiled into the attributes of the final grid division to complete the grid data.

[0053] (1) Design density weight function:

[0054] Considering the huge area of ​​ancient buildings, if a single equally spaced grid is used, it will lead to a lack of mapping of characteristic data such as building density and pedestrian flow, ignoring the non-uniformity of characteristic attributes in the region, resulting in excessive or low risks in some areas. Therefore, the equally spaced grid division and the IDW algorithm are combined, with equally spaced grid division as the main method and the IDW method as the auxiliary method. In the IDW method part, the building density is used as the weight function for unknown prediction, taking full account of the differences in the building density of ancient buildings and the differences in pedestrian flow. The larger the weight assigned, the greater the building density and the smaller the assigned grid. It can be seen from the IDW method that the weight function is the building density. In the two-dimensional plane supported by the coordinate longitude and latitude, the number of buildings in the grid area BuildNum p The method of determining is as follows:

[0055]

[0056] In the formula, contain means that the grid contains the building, exclude means that the grid does not contain the building, N is the total number of buildings in the entire area, and N p is the total number of buildings in the pth grid group.

[0057]

[0058] Where q is the power exponent of the density index, usually 1 or 2, α p is the weight assigned to the pth grid group, W0 is the initial grid division data, W p Predicted values ​​of the gridded data for the pth grid group.

[0059] (2) Group search grid

[0060] Combined with the implementation example Figure 3 As shown in the figure, after multiple tests of the global equidistant grid partitioning algorithm, it was found that when the amount of input data is large, the algorithm takes a long time to operate, and the results of multiple operations have certain errors, and its search efficiency is difficult to meet user needs. Assuming that the size of the data level is T1, the algorithm uses the traditional double traversal method to perform search processing, traversing the building vector file based on the grid to determine the spatial relationship between the grid and the building, and then assigning eigenvalues, so its algorithm complexity is O(T1 2 / 2), the operation time is relatively long. In order to improve the running speed and search speed of the algorithm, a grid grouping search algorithm is proposed based on the idea of ​​global equidistant grid division algorithm and human-computer interaction performance. First, the gridding idea of ​​the equidistant grid division algorithm is used to divide some characteristic grid areas, which are called grid groups. In the application of ancient buildings, the characteristic information can be determined according to the building density distribution, and the overall area is divided into grid areas based on different building densities. For example, the densely populated area is divided into a grid group, and the area without buildings is divided into a grid group, etc.; then, the spatial relationship of the building vector file is judged for the grid group in turn, including whether the grid area contains, borders or is separated from a small building; finally, the data of the buildings separated from the grid group are shielded. When performing double grid division in the subsequent time, it is only necessary to traverse the buildings contained or bordered by the grid group, and select one of the grid groups to give it the initial grid division data W0. If the number of grids in a grid group is T2, the algorithm complexity in the grid group is O(T2 2 / 2), obviously T2< <T1,O(T2 2 / 2)< <O(T1 2 / 2).

[0061] (3) Density compensation optimization

[0062] The effect of combining implementation examples Figure 4 As shown in the figure, different weight coefficients are set for different grid groups. If the building distribution in some areas of ancient buildings is unbalanced, it will lead to the inclination of attribute data such as population density and number of cultural relics. In order to solve this imbalance problem, density compensation optimization measures are added. Specifically, the divided grid data W is adjusted according to the density differences of different types of grid groups. p , increase the number of uneven sides, so as to achieve the purpose of balancing density distribution. The absolute value of the length and width difference of the grid coordinates d is selected as the density compensation factor of the algorithm, that is,

[0063] d=|xLenth-yLenth| (13)

[0064] addnum=d*min(xnum,ynum) / min(xLenth,yLenth) (14)

[0065] min(xnum,ynum)=min(xnum,ynum)+addnum (15)

[0066] Where xLenth and yLenth are the coordinate length and width of the grid group, respectively; xnum and ynum are the grid division data of the grid group in the x and y directions; d is the density compensation factor, and addnum is the density compensation amount.

[0067] (4) Fusion indicator weight

[0068] After step 2, the weights of the indicators in Table 1 have been determined. Figure 1 In the process of traversing the building vector file, code is written to pass the detection data of each indicator of indicator layer C through the algorithm interface of step 3, calculate the final risk factor size according to the final risk model formula (8), and overwrite it to the attribute key "attribute value" of the grid to support the subsequent risk visualization step.

[0069] In summary, the fire risk grid of ancient buildings can be obtained through the above-mentioned hierarchical analysis method and the grid division algorithm based on inverse density proportional weight. Then, the final visual risk grid can be obtained in conjunction with the GDAL library, GEOS library, MHMapGIS software, etc., which can provide early warning for the fire risk of ancient buildings.

[0070] Example:

[0071] A grid-based early warning method for large-scale ancient buildings based on multiple data sources in this embodiment includes the following steps:

[0072] Step 1: Collect and organize fire risk indicators and establish a fire risk indicator system:

[0073] The AHP structure consists of three layers: target layer (A), criterion layer (B), and indicator layer (C). Considering that most ancient buildings in my country cover a large area, are old, and have a strong historical atmosphere, six indicators are selected on the criterion layer based on the actual situation: value index, fire probability, fire spread speed, rescue speed index, fire perception ability, and evacuation speed index. The value index mainly refers to the value reflection of ancient buildings, including the number of cultural relics and the value of cultural relics. Most ancient buildings in my country are brick and wood structures, with low fire resistance and large fire load. Therefore, the possibility of fire needs to include a series of indicators that may cause fires, such as building materials, fire resistance, building structure, and electrical distribution. Another factor that causes fires in ancient buildings is the fast spread of fires and the poor sensitivity of fire detectors. On the one hand, there is a lack of corresponding flame retardant treatment, and there are many flammable materials such as curtain fabrics and wood in the building, which will cause the fire to spread rapidly; on the other hand, the number of fire detectors is limited, the distribution position is unreasonable, and traditional fire detectors are not enough to cope with complex environments. Therefore, a series of indicators such as building flame retardant level, number of cotton and linen products, temperature and humidity need to be included in the criterion layer of fire spread speed and perception ability index. In addition, rescue and evacuation work in the event of a fire cannot be ignored. Some large ancient building complexes are also important tourist attractions with large flow of people, making rescue and evacuation work difficult. Therefore, the rescue and evacuation speed index criterion layer needs to include relevant indicators such as the number of firefighters, building area, and fire accessibility.

[0074] In order to further refine the risk indicators, the Delphi method was used to organize the fire risk indicator system with reference to relevant books and literature: Li Zhimin. Research on Disaster Prevention and Protection of Ancient Buildings in Jin Temple [D]. Taiyuan University of Technology, 2019. The process is to sort out, summarize and count the opinions of experts, and then anonymously feedback the indicators to the experts for a second round of consultation. The indicator system is revised and improved according to the final opinions. The final fire risk indicator system is shown in Table 4.

[0075] Table 4 Hierarchical model

[0076]

[0077] Step 2: Solve the fire risk index weights through the analytic hierarchy process (AHP):

[0078] Step 2.1 Construct the judgment matrix:

[0079] After comparing two indicators at a certain level, the judgment matrix of this level is obtained: Since there are only two indicators in the criterion layer of Table 4, namely, the value index and the fire perception capability index, the weights can be directly judged to be 0.5 respectively, and the other four criterion layers use the 1-9 scale method to compare and judge the two indicators. Since there are many indicators in the selected indicator layer, only the criterion layer calculation is taken as an example, and its judgment matrix is ​​obtained through expert scoring as shown in Table 5.

[0080] Table 5 Judgment matrix of the criterion layer about A

[0081]

[0082]

[0083] Step 2.2: Consistency check:

[0084] Taking the judgment matrix of the criterion layer about the target layer A in Table 5 as an example, calculate the hierarchical single ranking and perform consistency check. The steps are as follows:

[0085] (1) The judgment matrix X is expressed in matrix form:

[0086]

[0087] (2) The geometric mean method is used to calculate the weight vector Y of the judgment matrix X, and formula (17) is:

[0088]

[0089] Where: x ij is the element in the i-th row and j-th column of the matrix X, and n is the number of indicators in the criterion layer or indicator layer.

[0090] (3) Perform consistency check on the judgment matrix and find the maximum eigenvalue λ of the judgment matrix max , calculate its consistency index CI and consistency ratio CR,

[0091]

[0092] Where: Y k is the kth element in the weight vector matrix obtained by formula (17).

[0093]

[0094] At this time, when n=6, the random consistency index RI=1.24 is obtained by looking up the table, and the consistency ratio CR is:

[0095]

[0096] Therefore, this solution meets the requirements of consistency verification.

[0097] Step 2.3: Solve the final risk model:

[0098] The weight calculation of indicator layer C with respect to criterion layer B can be taken as an example of the above criterion layer calculation, and the calculation is assisted by Matlab and other related software. The CR values ​​meet the consistency test, and the weights of each indicator are shown in Table 6. If the criterion layer weights are expressed, the final risk model is:

[0099] S risk =Val*0.10+Pro*0.32+Fi*0.21+Res*0.14+Per*0.09+Eva*0.14 (23)

[0100] Among them, S risk It represents the final risk factor, Val represents the weights of various indicators under the value index criterion layer, Pro represents the weights of various indicators under the fire possibility criterion layer, Fi represents the weights of various indicators under the fire spread speed criterion layer, Res represents the weights of various indicators under the rescue speed index criterion layer, Per represents the weights of various indicators under the fire perception ability index criterion layer, and Eva represents the weights of various indicators under the evacuation speed index criterion layer.

[0101] Table 6 Weights of various indicators

[0102]

[0103] Step 3: Design a grid partitioning algorithm based on density inverse proportional weights:

[0104] (1) Design density weight function:

[0105] Considering the huge area of ​​ancient buildings, if a single equally spaced grid is used, it will lead to a lack of mapping of characteristic data such as building density and pedestrian flow, ignoring the non-uniformity of characteristic attributes in the region, resulting in excessive or low risks in some areas. Therefore, combining equally spaced grid division and IDW algorithm, with equally spaced grid division as the main method and IDW method as the auxiliary method, in the IDW method part, building density is used as the weight function for unknown prediction, taking full account of the differences in building density of ancient buildings and the differences in pedestrian flow. The larger the weight assigned, the greater the building density and the smaller the assigned grid. It can be seen from the IDW method that the weight function is building density. In the two-dimensional plane supported by coordinate longitude and latitude, the method for determining the number of buildings in the grid area is as follows:

[0106]

[0107] In the formula, contain means that the grid contains the building, exclude means that the grid does not contain the building, N is the total number of buildings in the entire area, and N p is the total number of buildings in the pth grid group.

[0108]

[0109] Where q is the power exponent of the density index, usually 1 or 2, α p The weight assigned to the pth grid group, W0 is the initial grid division data, W p Predicted values ​​of the gridded data for the pth grid group.

[0110] (2) Group search grid

[0111] In order to improve the running speed and search speed of the algorithm, a grid grouping search algorithm is proposed based on the idea of ​​global equidistant grid algorithm and human-computer interaction performance. First, the gridding idea of ​​the equidistant grid algorithm is used to divide some characteristic grid areas, which are called grid groups. In the application of ancient buildings, the characteristic information can be determined according to the building density distribution, and the overall area is divided into grid areas based on different building densities. For example, the densely built area is divided into a grid group, and the area without buildings is divided into a grid group. Then, the spatial relationship of the building vector file is judged according to the grid group, including whether the grid area contains, borders or is separated from a small building; finally, the data of the building separated from the grid group is shielded. When performing double grid division in the subsequent time, only the buildings contained or bordered by the grid group need to be traversed, and one of the grid groups is selected to give it the initial grid division data W0.

[0112] (3) Density compensation optimization

[0113] The effect of combining implementation examples Figure 4As shown in the figure, different weight coefficients are set for different grid groups. If the building distribution in some areas of ancient buildings is unbalanced, it will lead to the inclination of attribute data such as population density and number of cultural relics. In order to solve this imbalance problem, density compensation optimization measures are added. Specifically, the divided grid data W is adjusted according to the density differences of different types of grid groups. p , increase the number of uneven sides, so as to achieve the purpose of balancing density distribution. The absolute value of the length and width difference of the grid coordinates d is selected as the density compensation factor of the algorithm, that is,

[0114] d=|xLenth-yLenth| (28)

[0115] addnum=d*min(xnum,ynum) / min(xLenth,yLenth) (29)

[0116] min(xnum,ynum)=min(xnum,ynum)+addnum (30)

[0117] Where xLenth and yLenth are the coordinate length and width of the grid group, respectively; xnum and ynum are the grid division data of the grid group in the x and y directions; d is the density compensation factor, and addnum is the density compensation amount.

[0118] (4) Fusion indicator weight

[0119] After step 2, the weights of the indicators in Table 6 are obtained. Figure 1 In the process of traversing the building vector file, code is written to pass the detection data of each indicator of indicator layer C through the algorithm interface of step 3, calculate the final risk factor size according to the final risk model formula (23), and overwrite it to the attribute key "attribute value" of the grid to support the subsequent risk visualization step.

[0120] Step 4: Configure the development environment:

[0121] This algorithm is deployed in MHMapGIS software suitable for spatial information analysis and processing. This algorithm is developed using the C++ programming language, integrating third-party libraries such as GDAL, GEOS, and PROJ, and generating a dynamic link library .dll file through the language compiler of Visual Studio 2022 version. The algorithm is integrated using the dynamic link library method of overloading the required interface. When compiling the GDAL library, the GEOS library and the PROJ library need to be integrated.

[0122] In terms of software parameters, the driver program selects "ESPI Shapefile", and the output file format of the driver program is shown in Table 7. The spatial reference coordinate system setting selects "WGS84". Considering the human-computer interaction, an interface window is added for dynamic data input. This setting can be completed by adding the required parameters in the declaration function of extern "C" __declspec (dllexport).

[0123] Table 7 Output file format

[0124]

[0125] Step 5: Divide the risk grid according to the algorithm in step 3:

[0126] The GDAL library provides many operation functions for vector files, among which the core target function is the function of traversing vector file features: GetLayerDefn() obtains the basic information of the layer; the GetFieldCount() function obtains the number of attributes in the layer attribute table; the GetFieldDefn() function obtains the attribute information of the specified attribute, including the attribute name, attribute length and other information; GetNextFeature() searches for the next feature attribute of the layer; GetFieldAsInteger() obtains the characteristic value of the attribute, etc.

[0127] The building vector file is traversed twice. In the first traversal, the initial grid group is divided according to the initial division data provided by humans, and the GDAL library function is called to add grid division data, building name, AHP weight and other attribute information in the layer of the building vector file; in the second traversal, each grid group is optimized for density compensation, and the optimized grid division information data is overwritten into the attribute. Figure 1 The specific measures are as follows:

[0128] (1) The ancient building vector file is passed into the algorithm interface and divided into grid groups according to the initial X and Y direction grids;

[0129] (2) traverse and generate the grid file, traverse the building vector file for each grid group, calculate the characteristic value of the number of ancient buildings contained in the group, and then use the inverse proportional density weight algorithm to determine the grid division data of the group;

[0130] (3) The grid grouping in the generated grid file is not a standard equal-width grid, and the density compensation factor is used to compensate and optimize the non-uniform edge grid division data;

[0131] (4) Divide the grid again according to the optimized grid division data and assign the weight obtained by the AHP method to the grid, and generate the risk grid through SetField();

[0132] Step 6: Visualize the result of step 5 and update it dynamically:

[0133] The color rendering function of MHMapGIS is called to render the entire area into a fire risk level distribution map according to the final risk size. The fire risk range is shown in Table 8. In addition, referring to the circular queue idea, various sensors installed in the ancient building run at a certain timer cycle to collect feature information data, which are uploaded to the host computer to dynamically update the overall gridded risk status of each area. The final effect diagram of the equal-line spacing grid division and the use of this algorithm is shown in the implementation example Figure 6 , Figure 7 shown.

[0134] Table 8 Risk interval table

[0135]

[0136] The ancient buildings are selected from a typical large ancient building, and its vector diagram is as follows Figure 2 As shown in the figure, the relevant characteristic data such as the number of cultural relics and the flow of people are obtained through online public information and project research. Figure 6 , Figure 7 shown.

[0137] To facilitate data analysis, Figure 3 The grid partitions in the grid are marked. In order to compare the effect of the methods, the control group used the equal spacing grid partition algorithm, and the final effect is as follows Figure 7 Analysis and comparison Figure 6 and Figure 7 When the building distribution is highly uneven, the effect is more obvious when the grid division method based on density inverse weight is used. The risk distribution of the upper part of the building complex is more obvious and detailed than the equal-spaced grid division method, which fully considers the non-uniformity of the region. In addition, there are also certain efficiency improvements in early warning measures, computing performance, and search speed. For example, Figure 6 The middle area 2 has not yet received a red-level risk warning due to factors such as open space and less traffic. Figure 7 Due to the uniformity of equal spacing and failure to consider actual factors such as building geographical conditions, the area is given a very high risk level. As a result, more rescue and security personnel will be dispatched and more attention time will be paid to early warning measures, reducing manpower efficiency.

[0138] In addition, from Figure 6The result shows that the places with high fire risk are mostly distributed in areas 5, 7, 8, and 9, because on the one hand, the building density in this area is high, the streets are narrow, and the flow of people is relatively concentrated; on the other hand, the number of cultural relics stored in this area is large and valuable. It is recommended to increase the number of fire detectors and arrange and arrange staff reasonably or take diversion measures, and focus on rescue and evacuation work. In areas 1 and 3, there are mostly open areas with relatively low risks. Area 2 is located in the gate fortress and other places. Considering the actual situation, the population density is large and the risk is high. However, there are still shortcomings such as relatively rough indicator data and sparse grids, which leads to insufficiently detailed risk distribution. In the future, the number of grids can be increased and the indicators can be refined to show the distribution of fire risks more clearly and more carefully. Due to the real-time nature of software and risk factors, early warning measures should focus more on areas with obvious red features and regular appearance. In principle, measures should adhere to prevention first. The ancient buildings should not be damaged during the renovation process, nor should the original form, pattern and style of the ancient buildings be affected. For key protection areas and high-risk areas, when wired connection between the detector and the controller is difficult, wireless communication can be used to transmit information, such as LiFi technology (visible light wireless communication), WiFi technology (wireless communication technology), ZigBee technology (a low-speed, short-distance transmission wireless network protocol), etc.

Claims

1. A grid-based early warning method for large ancient buildings based on multiple data sources, characterized in that: include: Step 1: Collect and organize fire risk indicators and establish a fire risk indicator system through the analytic hierarchy process; Step 2: Solve the fire risk index weights and the final risk model through the hierarchical analysis method, including comparing two indicators at a certain level, obtaining the judgment matrix of the level, and performing consistency test, and obtaining the final risk model containing various indicators according to the weight vector Y of the judgment matrix; Step 3. Design a grid division algorithm based on density inverse proportional weights to complete the digitization of the grid: transform the information data of the ancient buildings from discrete to continuous changes, and compile the indicator weights calculated by the hierarchical analysis method into the attributes of the final grid division to generate a digitized risk grid; through the visualization of the risk grid, warn of the fire risk of the ancient buildings.

2. The large-scale ancient building gridding early warning method for multiple data sources according to claim 1 is characterized in that: The target layer is the ancient building fire risk model; the criterion layer selects six indicators: value index, fire possibility, fire spread speed, rescue speed index, fire perception ability and evacuation speed index; the indicator layer is the various indicators under the criterion layer.

3. The large-scale ancient building gridding early warning method for multiple data sources according to claim 2 is characterized in that: The final risk model is expressed as: S=Val*y b1 +Pro*y b2 +Fi*y b3 +Res*y b4 +Per*y b5 +Eva*y b6 Among them, S represents the final risk factor, Val represents the weight of each indicator under the value index criterion layer, Pro represents the weight of each indicator under the fire possibility criterion layer, Fi represents the weight of each indicator under the fire spread speed criterion layer, Res represents the weight of each indicator under the rescue speed index criterion layer, Per represents the weight of each indicator under the fire perception ability index criterion layer, Eva represents the weight of each indicator under the evacuation speed index criterion layer, y b1 ~y b6 are the elements of the weight vector Y.

4. The large-scale ancient building gridding early warning method for multiple data sources according to claim 1 is characterized in that: Weight vector Y: Where: x ij is the element in the i-th row and j-th column of the judgment matrix, and n is the number of indicators in the criterion layer or indicator layer.

5. The large-scale ancient building gridding early warning method for multiple data sources according to claim 1 is characterized in that: Design a grid partitioning algorithm based on density inverse proportional weight to complete the grid digitization, including: (1) Design density weight function: Use building density as the weight function to predict unknowns. The greater the building density, the greater the weight assigned and the smaller the assigned grid. (2) Grouping grid search: The whole area is divided into grid areas based on different building densities. The spatial relationship of the building vector file is judged for each grid group, including whether the grid area contains, borders, or is separated from a small building. Finally, the data of the buildings separated from the grid group are shielded. When performing double grid division in the future, only the buildings contained or bordered by the grid group need to be traversed, and one of the grid groups is selected to be assigned the initial grid division data W0. (3) Density compensation optimization: Adjust the divided grid data W according to the density differences of different types of grid groups. p , increase the number of uneven sides to achieve the purpose of balancing the density distribution; (4) Fusion indicator weight: The final risk factor size is obtained through the algorithm code interface according to the fire risk model in step 2, and overwritten to the "attribute value" attribute of the grid for subsequent risk grid visualization.

6. The large-scale ancient building gridding early warning method for multiple data sources according to claim 5 is characterized in that: The number of buildings in the grid areaBuildNum p The method of determining is as follows: In the formula, contain means that the grid contains the building, exclude means that the grid does not contain the building, N is the total number of buildings in the entire area, and N p is the total number of buildings in the pth grid group; Where q is the power exponent of the density index, α p is the weight assigned to the pth grid group, W0 is the initial grid division data, W p Predicted values ​​of the gridded data for the pth grid group.

7. The large-scale ancient building gridding early warning method for multiple data sources according to claim 5 is characterized in that: Select the absolute value d of the grid coordinate length and width difference as the density compensation factor for density compensation optimization: d=|xLenth-yLenth| addnum=d*min(xnum,ynum) / min(xLenth,yLenth) min(xnum,ynum)=min(xnum,ynum)+addnum Where xLenth and yLenth are the coordinate length and width of the grid group, respectively; xnum and ynum are the grid division data of the grid group in the x and y directions; d is the density compensation factor, and addnum is the density compensation amount.

8. The large-scale ancient building gridding early warning method for multiple data sources according to claim 1 is characterized in that: Also includes: Step 4: Configure the development environment; Step 5, divide the risk grid according to step 3: perform two traversals on the building vector file. In the first traversal, perform initial division of the grid group according to the initial division data provided manually, and call the GDAL library function to add grid division data, building name, and AHP weight attribute information to the layer of the building vector file; in the second traversal, perform density compensation optimization on each grid group, and overwrite the optimized grid division information data into the attribute; Step 6: Visualize the result of step 5 and update it dynamically: call the color rendering function of MHMapGIS to render the entire area into a fire risk level distribution map based on the final risk size.

9. The large-scale ancient building grid-based early warning method for multiple data sources according to claim 8 is characterized in that: The specific process of step 5 is as follows: (1) Import the ancient building vector file into the interface and divide it into grid groups according to the initial X and Y direction grids; (2) traverse and generate the grid file, traverse the building vector file for each grid group, calculate the characteristic value of the number of ancient buildings contained in the group, and then use the inverse proportional density weight algorithm to determine the grid division data of the group; (3) The grid grouping in the generated grid file is not a standard equal-width grid, and the density compensation factor is used to compensate and optimize the non-uniform edge grid division data; (4) Divide the grid again according to the optimized grid division data and assign the weights obtained by the AHP method to the grid, and generate the risk grid through SetField().

Images