Public video monitoring point selection method and system based on reasoning large model

Abstract

The application discloses a public video monitoring point selection method and system based on a reasoning large model, which comprises the following steps: arranging and importing the basic data of the prevention and control area; performing spatial analysis on the prevention and control area and collecting candidate points; identifying the levels and dynamically adjusting the weights of the candidate points, and calculating the comprehensive importance score S; sequencing based on the comprehensive importance score S and obtaining a planning scheme; and outputting the planning scheme, a visual map and a data analysis report. The abstract police prevention and control experience is converted into a rigorous mathematical model by the system, new data can be regularly imported, and the risk level is recalculated, so that the self-evolution and optimization of the prevention and control network are realized. The candidate points are evaluated and scored in a short time by the system, the planning efficiency is improved, and a quantitative basis is provided for investment decision, construction priority and later performance evaluation. The name and equipment configuration of the candidate points have reliable basis, and the standardization and normalization of the video monitoring construction in the regional range are realized.

Description

Technical Field

[0001] This invention relates to the field of smart city and public safety technology, and in particular to a method and system for selecting locations for public video surveillance based on a large inference model. Background Technology

[0002] According to Article 6 of the "Regulations on the Administration of Public Security Video Image Information Systems", local people's governments at or above the county level shall strengthen the overall planning for the construction of public security video systems, make full use of existing resources, and avoid duplication of construction.

[0003] Currently, the planning of public video surveillance networks relies heavily on human experience, resulting in issues such as strong subjectivity, a lack of unified standards, unscientific resource allocation, and difficulty in quantifying and evaluating prevention and control effectiveness. While some auxiliary planning tools exist, most only perform simple recording and management, failing to provide intelligent generation, quantitative evaluation, and optimized decision-making from the perspective of a macro-level prevention and control system. Traditional methods lack deep semantic understanding and associative reasoning capabilities for multi-source data (such as GIS, road networks, police reports, and POIs), cannot comprehend abstract business rules like "circles, blocks, grids, lines, and points," and cannot simulate the comprehensive decision-making thinking of experts in complex spatial environments. Therefore, there is an urgent need for a scientific, intelligent, and quantifiable planning method to improve the prevention and control effectiveness of the entire video surveillance network.

[0004] For the reasons mentioned above, it is necessary to improve the existing technology. Summary of the Invention

[0005] I. Technical problems to be solved This invention addresses the aforementioned deficiencies in existing technologies by proposing a public video surveillance site selection method and system based on a large inference model, thereby resolving the problems mentioned in the background section.

[0006] II. Technical Solution To address the aforementioned technical problems, this invention provides a method for selecting locations for public video surveillance based on a large inference model, which further includes the following steps: Step 1: Organize and access the basic data of the prevention and control area. The basic data includes the established video surveillance point table, the "circle, block, grid, line, point" rules, the equipment selection and configuration rule library for the construction of the circle prevention and control system, the point type-dimensional weight mapping table, the comprehensive importance score calculation formula, the strategic value calculation formula, the risk level calculation formula, and the functional criticality calculation formula. Step 2: Based on the basic data, perform spatial analysis on the prevention and control area, automatically generate candidate points and summarize them into a first candidate point set, and deduplicate the first candidate point set according to the established video surveillance point table to obtain a second candidate point set; Step 3: Based on the basic data and the second candidate point set, the candidate points in the second candidate point set are sequentially subjected to identification level, dynamic weight adjustment, strategic value calculation, risk level calculation, functional criticality calculation and comprehensive importance calculation to obtain the comprehensive importance score S; Step 4: Based on the comprehensive importance score S of each candidate point, sort each candidate point, and reduce the comprehensive importance score S of each subsequent batch by a fixed amount according to the multiple batches of construction at the same point to form a construction priority list. Then, based on the basic data, assign a standard name, latitude and longitude, and recommended equipment type and quantity to each candidate point. Step 5: Based on the basic data and the construction priority list, output a planning scheme, a visualization map, and a data analysis report that includes site priority sorting, detailed equipment configuration suggestions, and standard naming.

[0007] In the above technical solution, the basic data also includes constraints, GIS map data, road network data, key target POI data, historical police incident data, population heat map data, permanent and temporary resident population data, intelligent task weight table, key target value preset table, and environmental vulnerability assessment rule set.

[0008] In the above technical solution, the constraints include project budget and number of equipment.

[0009] In the above technical solution, step 2 includes the following steps: Step 2.1: Based on the basic data, perform spatial analysis to select all suitable candidate locations for setting up public video surveillance in the prevention and control area, and summarize them into the first candidate location set; Step 2.2: Based on the established video surveillance point table, compare each candidate point with it, delete the overlapping candidate points, and summarize the non-overlapping candidate points into the second candidate point set.

[0010] In the above technical solution, step 3 includes the following steps: Step 3.1: Based on the rules of "circles, blocks, grids, lines, and points", determine the level of each candidate point in the second candidate point set and mark it with a level label to realize the identification level; Step 3.2: Based on the hierarchical labels, query and assign the corresponding weight coefficients from the point type-dimension weight mapping table to realize the dynamic weight adjustment; Step 3.3: Calculate the strategic value of the candidate points based on the strategic value calculation formula to obtain the strategic value Sv of the candidate points; Step 3.4: Calculate the risk level of the candidate locations based on the risk level calculation formula to obtain the risk level Rr; Step 3.5: Calculate the functional criticality of the candidate points based on the aforementioned functional criticality calculation formula to obtain the functional criticality Fc; Step 3.6: Based on the comprehensive importance score calculation formula and the weight coefficients corresponding to the candidate points, calculate the comprehensive importance score of the candidate points to obtain the comprehensive importance score S.

[0011] In the above technical solution, step 4 includes the following steps: Step 4.1: Based on the comprehensive importance score S of each candidate point, perform an initial sorting from high to low according to the comprehensive importance score S; Step 4.2: Based on the list obtained from the initial sorting, for multiple batches of construction at the same location in the list, the comprehensive importance score S of each subsequent batch is reduced by a fixed amount to form a construction priority list, where the fixed amount is 10%. Step 4.3: Based on the basic data, assign a standard name, latitude and longitude, and recommended equipment type and quantity to each candidate point in the construction priority list.

[0012] On the other hand, the present invention provides a public video surveillance site selection system based on a large inference model, comprising: A data access module, used to import the basic data; The reasoning model includes a rule base and knowledge base module, a point generation unit, a spatial analysis unit, a scoring calculation unit, and a scheme generation unit. The rule base and knowledge base module receives and stores the basic data. The point generation unit generates candidate points based on the "circle, block, grid, line, point" rules and removes duplicates from the candidate points based on the established video surveillance point table. The spatial analysis unit performs spatial analysis based on the basic data. The scoring calculation unit calculates the strategic value Sv, risk level Rr, functional criticality Fc, and comprehensive importance score S of the candidate points based on the point type-dimensional weight mapping table, the comprehensive importance score calculation formula, the strategic value calculation formula, the risk level calculation formula, and the functional criticality calculation formula. The scheme generation unit sorts all the candidate points according to the comprehensive importance score S and generates a planning scheme based on the basic data. The output and visualization module outputs a planning scheme, a visualization map, and a data analysis report based on the basic data and the scheme generation unit.

[0013] III. Beneficial Effects Compared with the prior art, the present invention has the following beneficial effects: 1. By leveraging the reasoning capabilities of large-scale reasoning models, abstract police prevention and control experiences can be transformed into rigorous mathematical models, upgrading the planning process from "experience-driven" to "data and model-driven".

[0014] 2. The names and equipment configurations of the generated candidate locations are all based on verifiable data, achieving standardization and normalization of video surveillance construction within the region.

[0015] 3. The public video surveillance site selection system can evaluate a large number of candidate sites in a short time, which greatly improves planning efficiency.

[0016] 4. Each candidate site has a precise score indicating its importance, providing a quantitative basis for investment decisions, construction priorities, and post-construction performance evaluation.

[0017] 5. The public video surveillance site selection system can periodically import new alarm data and recalculate the risk level, thereby realizing the self-evolution and optimization of the prevention and control network. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the point selection method of the present invention. Detailed Implementation

[0019] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and should not be construed as limiting the scope of the invention.

[0020] This invention provides a public video surveillance site selection system based on a large inference model, including a data access module for importing the basic data; The reasoning model includes a rule base and knowledge base module, a point generation unit, a spatial analysis unit, a scoring calculation unit, and a scheme generation unit. The rule base and knowledge base module receives and stores the basic data. The point generation unit generates candidate points based on the "circle, block, grid, line, point" rules and removes duplicates from the candidate points based on the established video surveillance point table. The spatial analysis unit performs spatial analysis based on the basic data. The scoring calculation unit calculates the strategic value Sv, risk level Rr, functional criticality Fc, and comprehensive importance score S of the candidate points based on the point type-dimensional weight mapping table, the comprehensive importance score calculation formula, the strategic value calculation formula, the risk level calculation formula, and the functional criticality calculation formula. The scheme generation unit sorts all the candidate points according to the comprehensive importance score S and generates a planning scheme based on the basic data. The output and visualization module outputs a planning scheme, a visualization map, and a data analysis report based on the basic data and the scheme generation unit.

[0021] See Figure 1 The present invention also provides a method for selecting locations for public video surveillance based on a large inference model, which further includes the following steps: Step 1: Organize the basic data of the prevention and control area. Import the basic data into the rule base and knowledge base modules of the reasoning model through the data access module. The basic data includes the established video surveillance point table, the "circle, block, grid, line, point" rules (see Table 2 in the specific implementation), the equipment selection and configuration rule base for the construction of the circle prevention and control system (see Table 3 in the specific implementation), the point type-dimensional weight mapping table (see Table 4 in the specific implementation), the comprehensive importance score calculation formula, the strategic value calculation formula, the risk level calculation formula, and the functional criticality calculation formula. The basic data also includes constraints, GIS map data, road network data, key target POI data, historical police data, population heat map data, permanent and temporary resident population data, intelligent task weight table, key target value preset table, and environmental vulnerability assessment rule set. The constraints include project budget and equipment quantity.

[0022] Step 2: Based on the basic data, perform spatial analysis on the prevention and control area, automatically generate candidate points and summarize them into a first candidate point set, and deduplicate the first candidate point set according to the established video surveillance point table to obtain a second candidate point set; Step 2 includes the following steps: Step 2.1: The inference model performs spatial analysis based on the imported basic data through the spatial analysis unit, and then uses the point generation unit to filter out all the candidate points in the prevention and control area that are suitable for setting up public video surveillance, and summarizes them into the first candidate point set; Step 2.2: Based on the established video surveillance point table, the point generation unit compares each candidate point with the established video surveillance point table, deletes overlapping candidate points, and summarizes non-overlapping candidate points into the second candidate point set.

[0023] Step 3: Based on the basic data and the second candidate point set, the candidate points in the second candidate point set are sequentially subjected to identification level, dynamic weight adjustment, strategic value calculation, risk level calculation, functional criticality calculation and comprehensive importance calculation to obtain the comprehensive importance score S; Step 3 includes the following steps: Step 3.1: Based on the rules of "circles, blocks, grids, lines, and points," for each candidate point in the second candidate point set, the inference big model analyzes the multi-source information of the point to determine its level within the "circles, blocks, grids, lines, and points" hierarchy, and marks the level label to achieve the identification level. If a point simultaneously satisfies multiple level features, the big model generates a recommendation level based on business priority and marks boundary situations requiring manual review. As mentioned in the background technology, traditional pure mathematical models cannot understand abstract business rules such as "circles, blocks, grids, lines, and points." In this step, the inference big model plays a core role. It transforms the multi-source data (GIS coordinates, road network nodes, etc.) of candidate points into prompts that the model can understand. By analyzing the semantic association between this information and the rules of "circles, blocks, grids, lines, and points" (Table 2), it outputs level labels. This process involves complex logic such as spatial relationship reasoning and rule matching, and ensures interpretability in the form of a thought chain. Its output is the basis for all subsequent calculations.

[0024] Step 3.2: Based on the hierarchical labels, query and assign the corresponding weight coefficients from the point type-dimension weight mapping table (Table 4) to realize the dynamic weight adjustment. The weight coefficients include strategic weight (α), risk weight (β) and functional weight (γ). Step 3.3: Calculate the strategic value of the candidate locations based on the aforementioned strategic value calculation formula, thereby obtaining the strategic value Sv of the candidate locations; The formula for calculating strategic value is as follows: Sv=(W_bc*f_bc+W_c*f_c)*100; In the above formula, W_bc and W_c are weight coefficients. Experimental verification shows that when the weight coefficients W_bc are between 0.6 and 0.8 and W_c are between 0.2 and 0.4 (and satisfy W_bc + W_c = 1), the model can effectively distinguish the differences in strategic value of locations. Among them, when W_bc = 0.7 and W_c = 0.3, the model has the highest goodness of fit with historical data, which is the optimal value in this embodiment. f_bc represents the network betweenness centrality. This value is obtained by calculating the road network of the block to which node v belongs and its neighboring blocks using graph theory algorithms. It measures how many pairs of shortest paths the node is on and is an objective mathematical indicator. Its calculation formula is as follows: f_bc(v) = BC(v) / Max_BC; In the above formula, v is a given node, f_bc(v) is the network betweenness centrality of the selected node v, BC is betweenness centrality, a core concept in graph theory used to quantify a node's control over communication between other nodes in a network or its hub status. BC(v) is the centrality score of the selected node v, which is the hub status score of node v in its block network. The betweenness centrality calculation formula is used to calculate the sum of the proportions of all node pairs in the network where node v appears on their shortest paths. The calculation formula is as follows: BC(v)=Σ(σ(s,t|v) / σ(s,t)), (s≠v≠t∈V); In the above formula, s and t are any two distinct nodes in the network (source node and target node), which are also any two distinct intersections in the city block. σ(s,t) is the total number of shortest paths between node s and node t, that is, how many shortest feasible routes (such as the fastest or shortest routes) are there from intersection s to intersection t. σ(s,t|v) is the number of paths from node s to node t that must pass through the intermediate node v, and the number of paths from node s to node t that pass through node v. Σ (summation) sums over all possible node pairs (s,t) in the network, which is to calculate the sum of the importance of intersection v to all pairwise traffic between intersections in the block. In the formula for calculating the network betweenness centrality, Max_BC is the maximum betweenness centrality BC value among all nodes v. The resulting f_bc(v) is a value between [0,1], which can be directly substituted into the formula for the strategic value Sv. f_c represents the blind zone coverage efficiency, which is calculated by dividing the newly covered blind zone area by the total blind zone area. It can be simplified to consider areas without monitoring within 100 meters as blind zones, and the calculation formula is 1 / (1+number of monitoring units within 100 meters). Step 3.4: The reasoning model calculates the risk level of the candidate points based on the risk level calculation formula, thereby obtaining the risk level Rr; The formula for calculating the risk level is as follows: Rr = (W_kde*f_kde + W_t*f_t + W_e*f_e)*100; In the above formula, W_kde, W_t, and W_e are weighting coefficients, satisfying W_kde + W_t + W_e = 1. The analysis results show that the frequency of historical incidents is the most significant indicator for predicting future risks. Therefore, W_kde is given the highest weight, with a suggested value range of 0.45~0.60. The attractiveness of key targets is the second highest, with W_t suggested to be in the range of 0.25~0.35. Environmental vulnerability is an important correction term, with W_e suggested to be in the range of 0.15~0.25. In this embodiment, W_kde = 0.5, W_t = 0.3, and W_e = 0.2 are the best-fit values ​​verified by empirical studies. f_kde is the historical alert kernel density estimate, calculated as follows: f_kde=KDE_Value(v) / Max_KDE_Value; KDE_Value(v) is the number of incidents in the grid where node v is located in the past 2 years, and Max_KDE_Value is the highest number of incidents in the past 2 years among all blocks. The calculation result is normalized to the interval [0,1]. f_t represents the value density of key targets, calculated as f_t = TVD_Value(v) / Max_TVD_Value. The result of f_t needs to be normalized to the interval [0,1]. TVD_Value(v) = Σ[Value(j) / (1+Distance(v,j))] is the sum of the values ​​of all key targets j within a 2-kilometer radius, where j is a key target. The value coefficient Value(j) is a predefined value coefficient for each type of key target. Distance(v,j) is the distance between v and j (in kilometers) and does not exceed 2 kilometers. Max_TVD_Value is the maximum TVD_Value(v) value among all points v. f_e is the environmental vulnerability score, calculated using the following formula: f_e = (Score_management(v) + Score_Issues(v) + ... + Score_n(v)) / Max_Score; where Score_* (* represents the identifier of various evaluation levels) represents the score of each evaluation dimension, which is obtained after judging multi-source data based on a predefined rule base. n is the total number of evaluation rules used (i.e., the maximum possible score), and Max_Score is the maximum value used to normalize f_e to the [0,1] interval. Score_management(v) = rule: if (it is an open old community or an area without property management) then score = 1. Score_Issues(v) = rule: within 50 meters of the monitoring blind spot proposed by the grassroots, get 1 point, otherwise 0 points. Other items can be added flexibly. Step 3.5: The inference big model calculates the functional criticality of the candidate points based on the functional criticality calculation formula, thereby obtaining the functional criticality Fc; The formula for calculating functional criticality is as follows: Fc = (W_f * f_d + W_i * f_i) * 100; Where W_f and W_i are weighting coefficients, satisfying Wf + W_i = 1. The determination of the weighting coefficients (W_f, W_i) comprehensively adopted the expert survey method (Delphi method) and actual business flow analysis. Experts from departments such as police and urban management unanimously agreed that, in the context of smart cities, the contribution of monitoring points to structured intelligence data (such as face and license plate capture) is usually more effective than simply covering densely populated areas. Therefore, W_i is given a higher weight, with a suggested value range of 0.55~0.70; the suggested range for W_f is 0.30~0.45. The W_f=0.4 and W_i=0.6 used in this embodiment are optimized values ​​determined after multiple rounds of expert demonstration. f_d is a population density index, calculated using the following formula: f_d(v)=Density(C) / Max_Density_In_Region; Where Density(C) = Pop(C) / Area(C), c is the community described by node v, Pop(C) is the number of permanent residents in the community, Area(C) is the area of ​​the community, Density(C) is the population density of the community, and Max_Density_In_Region is the maximum value of all population densities. f_i represents the intelligence contribution rate, calculated using the following formula: f_i = Sum_Weight(v) / Sum_Weight(All_Tasks), where Sum_Weight is the sum of the task weights of the equipment built at this point, and Sum_Weight(All_Tasks) is the sum of all weights, as shown in Table 1 of the specific implementation. Step 3.6: Based on the comprehensive importance score calculation formula and the weight coefficients corresponding to the candidate points, calculate the comprehensive importance score of the candidate points to obtain the comprehensive importance score S; The formula for calculating the overall importance score is as follows: S = α*Sv + β*Rr + γ*Fc, where α, β, and γ are weight coefficients, and α + β + γ = 1. The values ​​of the weight coefficients (α, β, γ) are dynamically adjusted according to the "circle, block, grid, line, point" level to which the point belongs. The large model generates weight adjustment suggestions and accompanying reasoning explanations by analyzing data such as population flow patterns and police trends. Step 4: Based on the comprehensive importance score S of each candidate point, sort each candidate point, and reduce the comprehensive importance score S of each subsequent batch by a fixed amount according to the multiple batches of construction at the same point to form a construction priority list. Then, based on the basic data, assign a standard name, latitude and longitude, and recommended equipment type and quantity to each candidate point. Step 4 is further divided into the following sub-steps: Step 4.1: Based on the comprehensive importance score S of each candidate point, perform an initial sorting from high to low according to the comprehensive importance score S; Step 4.2: Based on the list obtained from the initial sorting, for multiple batches of construction at the same location in the list, the comprehensive importance score S of each subsequent batch is reduced by a fixed amount to form a construction priority list, where the fixed amount is 10%. Step 4.3: Based on the aforementioned basic data, assign a standard name, latitude and longitude, and recommended equipment type and quantity to each candidate location in the construction priority list; Step 5: Based on the basic data and the construction priority list, output a planning scheme, a visualization map, and a data analysis report that includes site priority sorting, detailed equipment configuration suggestions, and standard naming.

[0025] The overall effect is as follows: 1. Transform abstract police prevention and control experience into rigorous mathematical models, upgrading the planning process from "experience-driven" to "data and model-driven".

[0026] 2. The names and equipment configurations of the generated candidate sites are all based on verifiable data, achieving standardization and normalization of construction throughout the region.

[0027] 3. The public video surveillance site selection system can evaluate a large number of candidate sites in a short time, which greatly improves planning efficiency.

[0028] 4. Each candidate site has a precise score indicating its importance, providing a quantitative basis for investment decisions, construction priorities, and post-construction performance evaluation.

[0029] The public video surveillance site selection system can periodically import new alarm data and recalculate the risk level, thereby enabling the self-evolution and optimization of the prevention and control network.

[0030] ; Table 1 (Smart Task Weight Table) ; ; Table 2 (Rules for "Circles, Blocks, Grids, Lines, and Dots") ; ; Table 3 (Equipment Selection and Configuration Rule Library for Layered Prevention and Control System Construction) ; Table 4 (Point Type - Dimension Weight Mapping Table) The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for selecting locations in public video surveillance based on a large inference model, characterized in that, It also includes the following steps: Step 1: Organize and access the basic data of the prevention and control area. The basic data includes the established video surveillance point table, the "circle, block, grid, line, point" rules, the equipment selection and configuration rule library for the construction of the circle prevention and control system, the point type-dimensional weight mapping table, the comprehensive importance score calculation formula, the strategic value calculation formula, the risk level calculation formula, and the functional criticality calculation formula. Step 2: Based on the basic data, perform spatial analysis on the prevention and control area, automatically generate candidate points and summarize them into a first candidate point set, and deduplicate the first candidate point set according to the established video surveillance point table to obtain a second candidate point set; Step 3: Based on the basic data and the second candidate point set, the candidate points in the second candidate point set are sequentially subjected to identification level, dynamic weight adjustment, strategic value calculation, risk level calculation, functional criticality calculation and comprehensive importance calculation to obtain the comprehensive importance score S; Step 4: Based on the comprehensive importance score S of each candidate point, sort each candidate point, and reduce the comprehensive importance score S of each subsequent batch by a fixed amount according to the multiple batches of construction at the same point to form a construction priority list. Then, based on the basic data, assign a standard name, latitude and longitude, and recommended equipment type and quantity to each candidate point. Step 5: Based on the basic data and the construction priority list, output a planning scheme, a visualization map, and a data analysis report that includes site priority sorting, detailed equipment configuration suggestions, and standard naming.

2. The method for selecting locations for public video surveillance based on a large inference model as described in claim 1, characterized in that: The basic data also includes constraints, GIS map data, road network data, key target POI data, historical police incident data, population heat map data, permanent and temporary resident population data, intelligent task weight table, key target value preset table, and environmental vulnerability assessment rule set.

3. The method for selecting locations for public video surveillance based on a large inference model as described in claim 2, characterized in that: The constraints include the project budget and the number of equipment.

4. The method for selecting locations for public video surveillance based on a large inference model as described in claim 1, characterized in that: Step 2 includes the following steps: Step 2.1: Based on the basic data, perform spatial analysis to select all suitable candidate locations for setting up public video surveillance in the prevention and control area, and summarize them into the first candidate location set; Step 2.2: Based on the established video surveillance point table, compare each candidate point with it, delete the overlapping candidate points, and summarize the non-overlapping candidate points into the second candidate point set.

5. The method for selecting locations for public video surveillance based on a large inference model as described in claim 1, characterized in that: Step 3 includes the following steps: Step 3.1: Based on the rules of "circles, blocks, grids, lines, and points", determine the level of each candidate point in the second candidate point set and mark it with a level label to realize the identification level; Step 3.2: Based on the hierarchical labels, query and assign the corresponding weight coefficients from the point type-dimension weight mapping table to realize the dynamic weight adjustment; Step 3.3: Calculate the strategic value of the candidate points based on the strategic value calculation formula to obtain the strategic value Sv of the candidate points; Step 3.4: Calculate the risk level of the candidate locations based on the risk level calculation formula to obtain the risk level Rr; Step 3.5: Calculate the functional criticality of the candidate points based on the aforementioned functional criticality calculation formula to obtain the functional criticality Fc; Step 3.6: Based on the comprehensive importance score calculation formula and the weight coefficients corresponding to the candidate points, calculate the comprehensive importance score of the candidate points to obtain the comprehensive importance score S.

6. The method for selecting locations for public video surveillance based on a large inference model as described in claim 1, characterized in that: Step 4 includes the following steps: Step 4.1: Based on the comprehensive importance score S of each candidate point, perform an initial sorting from high to low according to the comprehensive importance score S; Step 4.2: Based on the list obtained from the initial sorting, for multiple batches of construction at the same location in the list, the comprehensive importance score S of each subsequent batch is reduced by a fixed amount to form a construction priority list, where the fixed amount is 10%. Step 4.3: Based on the basic data, assign a standard name, latitude and longitude, and recommended equipment type and quantity to each candidate point in the construction priority list.

7. A public video surveillance site selection system based on a large inference model, used to implement the public video surveillance site selection method based on a large inference model as described in any one of claims 1-6, characterized in that, include: A data access module, used to import the basic data; The reasoning model includes a rule base and knowledge base module, a point generation unit, a spatial analysis unit, a scoring calculation unit, and a scheme generation unit. The rule base and knowledge base module receives and stores the basic data. The point generation unit generates candidate points based on the "circle, block, grid, line, point" rules and removes duplicates from the candidate points based on the established video surveillance point table. The spatial analysis unit performs spatial analysis based on the basic data. The scoring calculation unit calculates the strategic value Sv, risk level Rr, functional criticality Fc, and comprehensive importance score S of the candidate points based on the point type-dimensional weight mapping table, the comprehensive importance score calculation formula, the strategic value calculation formula, the risk level calculation formula, and the functional criticality calculation formula. The scheme generation unit sorts all the candidate points according to the comprehensive importance score S and generates a planning scheme based on the basic data. The output and visualization module outputs a planning scheme, a visualization map, and a data analysis report based on the basic data and the scheme generation unit.

Images