A method, system, device and storage medium for modeling camera layout

Abstract

The application belongs to the technical field of monitoring security and protection, and specifically provides a modelized camera layout method, system, device and storage medium, the method comprising the steps of: acquiring three-dimensional space data of a target scene, and constructing a three-dimensional scene model; training a machine learning model based on preset layout optimization target parameters, a dynamic constraint condition set and a training data set; inputting the three-dimensional scene model and real-time acquired security requirement parameters into the machine learning model, and outputting an optimal camera layout scheme meeting the layout optimization target parameters through the machine learning model; and dynamically iteratively optimizing the position, angle and coordination parameters of the camera through the machine learning model according to real-time acquired environmental change parameters of the target scene, and outputting an updated camera layout scheme. Through the collaborative optimization of three-dimensional scene modeling and the machine learning model, the traditional experience-driven camera layout is converted into a data-driven intelligent decision, and the objectivity and adaptability of the layout are significantly improved.

Description

Technical Field

[0001] This invention belongs to the field of surveillance and security technology, specifically relating to a model-based camera layout method, system, device, and storage medium. Background Technology

[0002] With increasing security awareness, surveillance systems have become a crucial component of security work and are widely used. However, most existing video surveillance layout designs are based on the experience of engineers or simple measurements and estimates. This manual approach is prone to significant subjective bias, making it difficult to assess the rationality of camera placement. There is no unified objective standard to evaluate the quality of camera layouts, resulting in inconsistent quality and the potential for blind spots.

[0003] To address these issues, the present invention provides a modeling camera layout method, system, device, and storage medium. Summary of the Invention

[0004] To overcome the shortcomings of the prior art, the present invention provides a modeling camera layout method, system, device and storage medium to solve the problems in the prior art.

[0005] One embodiment of the present invention provides a modeling camera layout method, comprising the following steps:

[0006] S10. Obtain the three-dimensional spatial data of the target scene and construct a three-dimensional scene model that includes static object structures and dynamic environmental parameters.

[0007] S20. Based on the preset layout optimization target parameters, dynamic constraint condition set and training dataset, train the machine learning model to obtain the trained machine learning model, which is used to associate the mapping relationship between camera layout parameters and scene features.

[0008] S30. Input the three-dimensional scene model and the security requirement parameters acquired in real time into the trained machine learning model, and output the optimal camera layout scheme that meets the layout optimization target parameters through the machine learning model.

[0009] S40. Based on the environmental change parameters of the target scene acquired in real time, the position, angle and collaborative parameters of the camera are dynamically iteratively optimized through the machine learning model, and an updated camera layout scheme is output.

[0010] In one embodiment, after step S40, the following step is further included:

[0011] S50. During the execution of the camera layout scheme, when a camera in the target scene is detected to have triggered a preset monitoring failure judgment rule, a monitoring failure signal is generated and the location area of ​​the failed camera is determined.

[0012] S60. Based on the location area, activate the backup camera that meets the preset backup device activation rules;

[0013] S70. Dynamically adjust the monitoring parameters of the backup camera to cover the failed area, and update the collaborative parameter set in the camera layout scheme.

[0014] In one embodiment, in step S60, the preset backup device activation rule includes at least one of the following:

[0015] If the straight-line distance or path planning distance between the backup camera and the failed area is less than a preset distance threshold, the nearest available device will be selected first.

[0016] The current workload of the backup camera is lower than a preset threshold, and the workload includes at least one of CPU utilization, network bandwidth utilization, and remaining storage space.

[0017] The field of view of the backup camera is unobstructed and has unobstructed line-of-sight coverage with the failed area in the 3D scene model.

[0018] In one embodiment, step S10 specifically includes the following steps:

[0019] S11. Acquire three-dimensional spatial data of the target scene through a multimodal acquisition module. The three-dimensional spatial data includes geometric topology data, texture image data, and dynamic environmental parameters.

[0020] S12. Based on the geometric topology data and texture image data, extract the spatial structure of buildings, fixed equipment and functional areas, and use a semantic analysis model to label the access priority and safety risk level of the static objects;

[0021] S13. In real time, acquire the illumination distribution characteristics, surface reflection characteristics and moving target trajectory in the dynamic environment parameters, perform behavioral semantic classification on the moving target, and generate a security response priority label that is dynamically associated with the three-dimensional spatial coordinates.

[0022] S14. Based on the illumination distribution characteristics and surface reflection characteristics, the imaging attenuation area within the camera's field of view is predicted using a light field reconstruction model and labeled as light field interference characteristics.

[0023] S15. By integrating the spatial structure, semantic annotation, security response priority label, and light field interference features of the static objects, a multi-dimensional three-dimensional scene model containing geometric constraints, semantic rules, and light field interference features is constructed.

[0024] In one embodiment, in step S20, the preset layout optimization target parameter includes at least one of the following:

[0025] Coverage optimization objective: Using the security risk level in the scene semantic attributes as weight, maximize the number of sub-regions that are effectively covered within the monitoring area. The effective coverage must satisfy that the light field interference characteristics in the three-dimensional scene model are below a preset threshold.

[0026] Cost optimization objective: Minimize the overall cost of camera deployment by using the number of devices, energy consumption, and deployment complexity as joint optimization variables. The deployment complexity is dynamically evaluated based on the obstacle distribution in the 3D scene model.

[0027] Redundancy optimization objective: Within an area where the safety risk level is at a preset threshold, maximize the number of sub-regions that are covered by the overlapping fields of view of multiple cameras.

[0028] In one embodiment, in step S20, the set of dynamic constraints includes at least one of the following:

[0029] Field of view integrity constraint: The unobstructed rate between the camera's field of view coverage path and the obstacle area in the 3D scene model must reach a preset threshold. The obstacle area is dynamically verified based on real-time updated semantic attributes.

[0030] Timeliness constraint: Based on the scene activity density in the dynamic environment parameters, the maximum generation delay threshold of the layout scheme is adaptively adjusted to ensure that the response time of high-density scenes meets the preset safety timeliness.

[0031] Cooperative deployment constraints: The deployment distance between cameras must meet the signal compatibility threshold and the physical security threshold.

[0032] In one embodiment, step S20 specifically includes the following steps:

[0033] S21. Based on the preset layout optimization target parameters, dynamic constraint set, and training dataset, generate a multi-scenario training dataset, wherein each training sample includes:

[0034] Input features: geometric topology, dynamic environmental parameters, and security requirement parameters extracted from historical 3D scene models;

[0035] Output label: The set of optimal camera layout parameters that satisfy the layout optimization objective parameters and conform to the set of dynamic constraints;

[0036] S22. Construct a multi-objective optimization neural network model, the model including:

[0037] Coverage optimization module: Calculates the coverage effectiveness of camera locations on the monitored area based on security risk level weights;

[0038] Cost optimization module: Assess the complexity of equipment deployment based on obstacle distribution density and generate a layout scheme under cost constraints;

[0039] Redundancy optimization module: Calculates the redundancy of overlapping coverage of multiple camera views within an area where the security risk level reaches a preset threshold;

[0040] S23. Input the multi-scenario training dataset into the multi-objective optimization neural network model, with the coverage optimization objective, cost optimization objective, and redundancy optimization objective as optimization terms, and the dynamic constraint condition set as hard constraint terms, and train the model parameters together through the constraint optimization algorithm;

[0041] S24. Verify whether the layout scheme output by the trained model in the test scenario simultaneously satisfies all dynamic constraints. If there is a constraint violation, adjust the weight coefficients of the layout optimization objective parameters and retrain.

[0042] This application also relates to a modeled camera layout system, including:

[0043] The data acquisition module is used to acquire three-dimensional spatial data of the target scene and construct a three-dimensional scene model that includes static object structures and dynamic environmental parameters.

[0044] The model training module is used to train a machine learning model based on preset layout optimization target parameters, dynamic constraint condition set and training dataset, and obtain a trained machine learning model, which is used to associate the mapping relationship between camera layout parameters and scene features.

[0045] The solution output module is used to input the three-dimensional scene model and the security requirement parameters acquired in real time into the trained machine learning model, and output the optimal camera layout solution that meets the layout optimization target parameters through the machine learning model.

[0046] The iterative optimization module is used to dynamically iteratively optimize the position, angle, and collaborative parameters of the camera based on the environmental change parameters of the target scene acquired in real time, and output an updated camera layout scheme.

[0047] This application also relates to a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described modeled camera layout method.

[0048] This application also relates to a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described modeled camera layout method.

[0049] The model-based camera layout method, system, device, and storage medium provided in the above embodiments have the following beneficial effects:

[0050] 1. By collaboratively optimizing 3D scene modeling and machine learning models, traditional experience-driven camera placement is transformed into data-driven intelligent decision-making, significantly improving the objectivity and adaptability of the placement. Based on the precise quantification of static structure and dynamic environmental parameters using 3D scene models, and combined with machine learning models to construct a mapping relationship between scene features and placement parameters, subjective biases from human experience are eliminated. Driven by both real-time security requirement parameters and environmental change parameters, dynamic iterative optimization of camera position, angle, and collaborative strategies is achieved, effectively reducing blind spots and improving response speed to sudden risks. Simultaneously, relying on training datasets and automated placement generation mechanisms, redundant equipment deployment and manual debugging costs are reduced, providing a high-precision, low-cost standardized solution for security monitoring systems.

[0051] 2. In one embodiment, a failure detection and dynamic adjustment mechanism solves the problem of monitoring blind spots caused by equipment malfunctions in traditional layouts. It identifies camera failure events in real time and locates the faulty area, activating backup equipment for compensation coverage to ensure monitoring continuity; it optimizes multi-camera task allocation by dynamically updating the collaborative parameter set to avoid resource redundancy; and it relies on an automated response mechanism to shorten abnormal interruption time from minutes to seconds, significantly improving security reliability in high-density or high-risk scenarios.

[0052] 3. In one embodiment, multimodal data fusion and light field enhancement modeling address the shortcomings of traditional 3D scene models in analyzing complex environments. Multimodal sensing synchronously acquires geometric and texture data, and semantic analysis is used to accurately label passage priorities and security risk areas, improving the accuracy of static modeling. Dynamic light field reconstruction quantifies light reflection and moving target interference, predicts monitoring blind spots, and adaptively adjusts anti-interference parameters. A multi-dimensional model integrating geometric, semantic, and physical interference assessments provides holographic decision-making support for complex scenes, enabling the scientific deployment of security layouts in highly dynamic environments and significantly enhancing the system's environmental adaptability and anti-interference capabilities.

[0053] 4. In one embodiment, a joint training mechanism combining multi-objective collaborative optimization and dynamic constraints addresses the insufficient optimization capabilities of traditional layout models under multi-objective conflicts and complex physical rules. The multi-objective optimization neural network module learns hierarchically, covering decision rules for efficiency, cost control, and redundancy security, and generates adaptive layout strategies by combining historical scene data and real-time environmental parameters. The constraint optimization algorithm rigorously verifies the visibility, timeliness, and device coordination rules of the layout scheme, ensuring that the scheme strictly conforms to dynamic constraints. Through incremental training and weight adjustment mechanisms, the model's generalization ability and response speed in highly dynamic scenarios are significantly improved, providing intelligent layout decision support that balances efficiency and reliability for complex security scenarios. Attached Figure Description

[0054] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0055] Figure 1 A flowchart illustrating a modeled camera layout method provided in an embodiment of the present invention;

[0056] Figure 2 Another flowchart of a modeling camera layout method provided in an embodiment of the present invention;

[0057] Figure 3 A flowchart illustrating step S10 of a modeling camera layout method provided in an embodiment of the present invention;

[0058] Figure 4 A flowchart illustrating step S20 of a modeling camera layout method provided in an embodiment of the present invention;

[0059] Figure 5 This is a schematic block diagram of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0060] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0061] It should be noted that if the embodiments of the present invention involve directional indications (such as up, down, left, right, front, back, etc.), the directional indications are only used to explain the relative positional relationship and movement of the components in a specific posture. If the specific posture changes, the directional indications will also change accordingly.

[0062] Furthermore, if the embodiments of this invention involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined with "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, the use of "and / or" or "and / or" throughout the text includes three parallel solutions. For example, "A and / or B" includes solution A, solution B, or a solution where both A and B are satisfied simultaneously. Furthermore, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by this invention.

[0063] Reference Figure 1 One embodiment of the present invention provides a modeling camera layout method, comprising the following steps:

[0064] S10. Obtain the three-dimensional spatial data of the target scene and construct a three-dimensional scene model that includes static object structures and dynamic environmental parameters.

[0065] S20. Based on the preset layout optimization target parameters, dynamic constraint condition set and training dataset, train the machine learning model to obtain the trained machine learning model, which is used to associate the mapping relationship between camera layout parameters and scene features.

[0066] S30. Input the three-dimensional scene model and the security requirement parameters acquired in real time into the trained machine learning model, and output the optimal camera layout scheme that meets the layout optimization target parameters through the machine learning model.

[0067] S40. Based on the environmental change parameters of the target scene acquired in real time, the position, angle and collaborative parameters of the camera are dynamically iteratively optimized through the machine learning model, and an updated camera layout scheme is output.

[0068] As described in step S10 above, the three-dimensional spatial data includes the geometric topology data of the target scene (such as building outlines and obstacle coordinates), texture image data (such as surface materials and color information), and dynamic environmental parameters (such as light intensity, pedestrian flow trajectory, temperature, and humidity). Data is collected synchronously using multimodal sensing devices (such as LiDAR and RGB-D cameras), and a three-dimensional scene model is constructed using point cloud registration and semantic segmentation algorithms. This model includes spatial representations of static object structures (such as walls and equipment locations) and real-time mapping of dynamic environmental parameters, providing a digital environmental foundation for layout optimization.

[0069] As described in step S20 above, the preset layout optimization target parameters are defined as maximizing coverage, minimizing equipment cost, and redundant monitoring in high-risk areas; the dynamic constraint set includes line-of-sight constraints, response latency constraints, and equipment anti-interference spacing. Based on the training dataset (historical scene samples and corresponding optimal layout schemes), a deep reinforcement learning framework is used to train a machine learning model, establishing a mapping relationship between scene features (such as security level distribution and obstacle density) and layout parameters (camera coordinates, tilt angle, and focal length). A multi-objective optimization algorithm ensures that the layout scheme simultaneously meets physical feasibility and security requirements.

[0070] As described in step S30 above, the 3D scene model (including static structure and dynamic parameter annotations) and security requirement parameters (such as key monitoring areas, risk level weights, budget costs, etc. in the target scene) are input into the trained model. The model outputs the optimal camera layout scheme through iterative calculation, including device type, quantity, spatial coordinates, and collaborative strategy (such as master-slave tracking logic), ensuring that the scheme meets the preset optimization objectives and dynamic constraints.

[0071] As described in step S40 above, environmental change parameters include the coordinates of newly added obstacles, sudden changes in pedestrian density, and shifts in lighting conditions. Real-time data drives a machine learning model for dynamic iterative optimization: adjusting camera positions to avoid obstructions, switching angles to expand coverage, updating the collaborative parameter set (e.g., enabling backup device linkage), and outputting the updated layout parameter set to achieve continuous adaptation.

[0072] In this embodiment, the traditional experience-driven camera layout is transformed into data-driven intelligent decision-making through the collaborative optimization of 3D scene modeling and machine learning models, significantly improving the objectivity and adaptability of the layout: Based on the 3D scene model, static structure and dynamic environmental parameters are accurately quantified, and the mapping relationship between scene features and layout parameters is constructed by combining machine learning models, eliminating the subjective bias of human experience; Through the dual drive of real-time security demand parameters and environmental change parameters, the dynamic iterative optimization of camera position, angle and collaborative strategy is realized, effectively reducing monitoring blind spots and improving the response speed to sudden risks; At the same time, relying on the training dataset and automated layout generation mechanism, the deployment of redundant equipment and manual debugging costs are reduced, providing a high-precision, low-cost standardized solution for security monitoring systems.

[0073] In one feasible embodiment, it is assumed that:

[0074] A shopping mall needs to deploy a surveillance system covering the jewelry area (high-risk), fire exits (critical access points), and the atrium (a high-traffic area).

[0075] Step S10:

[0076] The building structure is scanned by LiDAR, and the texture of the shelves is captured by an RGB-D camera to build a 3D model with centimeter-level precision.

[0077] Mark static structures: such as jewelry counters (security level 9) and fire doors (access priority A);

[0078] Integrated dynamic parameters: peak reflectivity of the glass curtain wall at midday (1200 lux), pedestrian flow trajectory during off-peak hours (2 people / 10㎡).

[0079] Step S20:

[0080] Preset goals: 100% coverage of the jewelry area, total cost ≤ ¥500,000, and redundant monitoring rate ≥ 90%;

[0081] Constraints: Line-of-sight ratio ≥ 98%, response latency ≤ 1.5 seconds;

[0082] Model training: Input several relevant historical layout samples of shopping malls, and output the scene feature-layout parameter mapping relationship (e.g., "high reflective area → wide-angle anti-glare camera").

[0083] Step S30:

[0084] Input real-time requirements: During the event, the risk weight of the jewelry area will be increased to the highest level;

[0085] Model output scheme:

[0086] Jewelry area: 3 4K cameras (redundant coverage + face tracking), with a 25° tilt angle to avoid glare;

[0087] Fire exit: Two-way cameras facing each other to ensure no blind spots;

[0088] Central courtyard: One PTZ camera is deployed for every 100 square meters, reducing the total number of devices by 22%.

[0089] Step S40:

[0090] Sudden change in scenario 1: A temporary promotional booth obstructs the fire escape camera;

[0091] Occlusion is detected in real time using point cloud. Within 0.5 seconds, the blocked camera is turned off, a backup device is activated, and the angle of the adjacent camera is adjusted to fill in the blind spot.

[0092] Sudden Scene 2: Heavy rain causes natural light to disappear;

[0093] The light sensor triggers the model to switch all cameras to infrared mode and simultaneously increases the gain parameters to ensure image clarity.

[0094] Reference Figure 2 In one embodiment, after step S40, the following step is further included:

[0095] S50. During the execution of the camera layout scheme, when a camera in the target scene is detected to have triggered a preset monitoring failure judgment rule, a monitoring failure signal is generated and the location area of ​​the failed camera is determined.

[0096] S60. Based on the location area, activate the backup camera that meets the preset backup device activation rules;

[0097] S70. Dynamically adjust the monitoring parameters of the backup camera to cover the failed area, and update the collaborative parameter set in the camera layout scheme.

[0098] As described in step S50 above, the monitoring failure determination rule is defined as being triggered by any of the following conditions:

[0099] The camera remains offline for ≥10 seconds;

[0100] The effective frame rate of the video stream is ≤5fps (normally it should be ≥15fps);

[0101] The coverage loss rate in critical areas is ≥30% (e.g., a sudden increase in blind spots).

[0102] Through real-time heartbeat detection and video quality analysis engine, a monitoring failure signal is generated when a failure is determined, and the location area of ​​the failed camera (such as coordinate range and associated coverage responsibility area) is determined based on the coordinate mapping of the 3D scene model.

[0103] As described in step S60 above, the standby device activation rule includes three constraints:

[0104] Line-of-sight coverage constraint: There is no physical obstruction between the backup camera and the failed area (line-of-sight rate ≥ 95%);

[0105] Load balancing constraint: The current load of the standby equipment is ≤70% (to avoid overload);

[0106] Path reachability constraint: If it is a mobile device, the path distance to the failure area must be ≤50 meters and there must be no dynamic obstacles.

[0107] The optimal backup camera is selected based on rules, and the activation command is sent in real time through the device control bus.

[0108] As described in step S70 above:

[0109] (1) Dynamically adjusting monitoring parameters includes:

[0110] Adjust the focus and tilt angle of the backup camera to cover the failed area;

[0111] Switch to wide-angle mode to expand the area of ​​responsibility;

[0112] Enable digital zoom to enhance detail capture.

[0113] (2) Update the cooperative parameter set:

[0114] Reassign master-slave tracking relationships (e.g., transferring the tracking target of a failed camera to a backup device);

[0115] Modify alarm linkage rules (e.g., backup equipment takes over access control permissions);

[0116] Generate a new device topology diagram and synchronize it to the security platform.

[0117] In this embodiment, the problem of monitoring blind spots caused by equipment failure in traditional layouts is solved by failure detection and dynamic adjustment mechanism: real-time identification of camera failure events and location of fault areas, activation of backup equipment to compensate coverage, and ensuring monitoring continuity; optimization of multi-camera task allocation by dynamically updating collaborative parameter set to avoid resource redundancy; and by relying on automated response mechanism, the abnormal interruption time is shortened from minutes to seconds, significantly improving the security reliability of high-density or high-risk scenarios.

[0118] In one feasible embodiment, it is assumed that:

[0119] A fixed camera in the jewelry section of a shopping mall failed due to a circuit malfunction. The area where the camera failed was the high-risk area of ​​the cashier (coordinates X1-Y1).

[0120] Step S50 is executed (failure detection and location):

[0121] The system detected that the camera had been offline for 12 seconds, the video stream was interrupted, and a monitoring failure signal was triggered.

[0122] The 3D model locates the failure area: a radius of 5 meters around the cashier (security level 9).

[0123] Step S60 is executed (standby device activation), filtering candidate standby devices:

[0124] Fixed backup camera A: line-of-sight coverage is 99%, but the load has reached 85% (exceeding the threshold);

[0125] Patrol Robot B: Load 40%, line-of-sight 100%, distance to failure zone 20 meters;

[0126] Ceiling-mounted PTZ camera C: 60% load, 90% line-of-sight, 8 meters from the failure zone.

[0127] Activation decision: Select patrol robot B (satisfies the triple rule and the path is reachable).

[0128] Step S70 execution (dynamic adjustment and collaborative update):

[0129] (1) Parameter adjustment:

[0130] Patrol robot B switches to 10x zoom mode and lowers its tilt angle by 15° to cover the cashier counter.

[0131] Enable thermal imaging to enhance low-light details (jewelry case glass reflection interference).

[0132] (2) Collaborative updates:

[0133] The face tracking task of the original malfunctioning camera was transferred to robot B;

[0134] Modify alarm rules: Abnormal transactions at the checkout will trigger Robot B to automatically approach and record video;

[0135] Update the layout plan: Mark the failed equipment as pending maintenance, and integrate robot B into the fixed monitoring network.

[0136] In one embodiment, in step S60, the preset backup device activation rule includes at least one of the following:

[0137] 1) The straight-line distance or path planning distance between the backup camera and the failed area is less than the preset distance threshold, and the nearest available device is selected first;

[0138] 2) The current workload of the backup camera is lower than a preset threshold, and the workload includes at least one of CPU utilization, network bandwidth utilization, and remaining storage space;

[0139] 3) The field of view of the backup camera has unobstructed line-of-sight coverage with the failed area in the three-dimensional scene model.

[0140] In this embodiment:

[0141] (1) Path planning distance refers to the shortest feasible path length (including bypassing obstacles) from the backup camera to the failed area in the 3D scene model. The preset distance threshold is set to 50 meters (configurable). The distance priority strategy calculates the path distance of each backup device using the Dijkstra algorithm and selects the device with the minimum value to activate. If the straight-line distance is ≤10 meters and there is no obstruction, the straight-line distance is used directly for determination.

[0142] (2) Workload quantification indicators include:

[0143] CPU utilization (threshold ≤ 80%);

[0144] Network bandwidth utilization (threshold ≤ 70%);

[0145] Remaining storage space (threshold ≥ 20%).

[0146] The preset load threshold can be dynamically adjusted (e.g., relaxed to 90% during peak hours). The load assessment engine collects device status data in real time and selects only standby devices whose all indicators are below the threshold.

[0147] (3) Unobstructed line-of-sight coverage must meet the following requirements:

[0148] There are no physical obstacles (walls, large equipment) between the center line of the backup camera's field of view and the area where the camera failed;

[0149] Visibility ≥ 95% (verified by ray collision detection of 3D model).

[0150] If there is a semi-transparent medium such as glass, the light transmittance must be additionally checked to be ≥85% (to avoid blurry images).

[0151] In one feasible embodiment, it is assumed that:

[0152] A fixed security camera in the jewelry section of a shopping mall has failed due to a circuit malfunction. The area affected by the failure is the high-risk area near the checkout counter (coordinates X1-Y1). Assume the following are potential backup cameras:

[0153] equipment Path distance CPU utilization Viewing rate Ceiling camera A 6 meters 75％ 100％ Patrol Robot B 8 meters 40％ 98％ Pillar camera C 4 meters 90％ 100％

[0154] The rule execution logic is as follows:

[0155] Distance priority (rule 1):

[0156] The pillar camera C has the shortest path distance (4 meters), but its CPU utilization is 90%, exceeding the threshold (preset 80%), so it is excluded.

[0157] Ceiling camera A (6 meters) and robot B (8 meters) will proceed to the next round of screening.

[0158] Load verification (rule 2):

[0159] Ceiling-mounted camera A has a load of 75% ≤ 80%, which meets the requirements.

[0160] Robot B met the 40% load requirement; both passed.

[0161] Line-of-sight coverage verification (Rule 3):

[0162] Ceiling camera A: X-ray detection shows 100% line-of-sight (no obstruction);

[0163] Robot B: 98% visibility (due to 2% of the area being blocked by a temporary promotional booth), excluded.

[0164] Finally, ceiling camera A is activated.

[0165] In this embodiment, multi-dimensional backup device activation rules are used to address the issues of misselection and inefficiency under a single decision dimension: the line-of-sight coverage rule avoids secondary blind spots caused by wall obstruction, the load balancing rule prevents cascading failures caused by hardware overload, and the path planning distance rule adapts to complex building structures, reducing the cost of ineffective deployment; at the same time, low-load, close-range, and line-of-sight backup devices are prioritized to extend the lifespan of hardware and reduce maintenance costs, forming an intelligent fault-tolerant mechanism that balances efficiency and economy.

[0166] Reference Figure 3 In one embodiment, step S10 specifically includes the following steps:

[0167] S11. Acquire three-dimensional spatial data of the target scene through a multimodal acquisition module. The three-dimensional spatial data includes geometric topology data, texture image data, and dynamic environmental parameters.

[0168] S12. Based on the geometric topology data and texture image data, extract the spatial structure of buildings, fixed equipment and functional areas, and use a semantic analysis model to label the access priority and safety risk level of the static objects;

[0169] S13. In real time, acquire the illumination distribution characteristics, surface reflection characteristics and moving target trajectory in the dynamic environment parameters, perform behavioral semantic classification on the moving target, and generate a security response priority label that is dynamically associated with the three-dimensional spatial coordinates.

[0170] S14. Based on the illumination distribution characteristics and surface reflection characteristics, the imaging attenuation area within the camera's field of view is predicted using a light field reconstruction model and labeled as light field interference characteristics.

[0171] S15. By integrating the spatial structure, semantic annotation, security response priority label, and light field interference features of the static objects, a multi-dimensional three-dimensional scene model containing geometric constraints, semantic rules, and light field interference features is constructed.

[0172] As described in step S11 above, the multimodal acquisition module includes a lidar (to acquire geometric topology point clouds), an RGB-D camera (to capture texture images), and an environmental sensor (to monitor temperature, humidity, and light intensity).

[0173] Three-dimensional spatial data includes:

[0174] Geometric topology data: building outline, obstacle coordinates, and spatial connectivity (such as door / window opening / closing status);

[0175] Texture image data: Surface material properties (such as glass reflectivity, wall roughness);

[0176] Dynamic environmental parameters: real-time pedestrian density, illumination distribution map, and trajectory of moving targets (people / vehicles).

[0177] As described in step S12 above, spatial structures (such as wall boundaries and shelf locations) and functional areas (cashier counters and fire exits) are extracted from geometric and texture data through point cloud segmentation and image semantic segmentation algorithms.

[0178] The semantic analysis model (based on a hybrid network of ResNet-50+CRF) annotates static objects as follows:

[0179] Access priority: Fire exits are level A (highest), and shelving clearance is level C (restricted access);

[0180] Safety risk level: Jewelry counters are level 9 (highest risk), and ordinary shops are level 3.

[0181] As described in step S13 above:

[0182] (1) Dynamic environmental parameter processing:

[0183] Illumination distribution characteristics: A 0.5m × 0.5m gridded illumination map is generated using a light intensity sensor array;

[0184] Surface reflection characteristics: Based on the material library, reflectivity parameters are matched (e.g., 85% reflectivity for glass curtain walls);

[0185] Moving target trajectory: Track the trajectory of people / vehicles using the YOLOv7+DeepSORT algorithm (target tracking algorithm).

[0186] (2) Behavioral semantic classification:

[0187] Normal behavior (walking in a straight line, staying still for less than 30 seconds) is marked as low priority;

[0188] Abnormal behaviors (loitering, reversing) are marked as high priority and security response priority labels (such as "abnormal aggregation - high risk") are generated and dynamically bound to the three-dimensional coordinates.

[0189] As described in step S14 above, the light field reconstruction model (based on NeRF neural radiation field) simulates the transmission path of light in the scene and predicts the imaging attenuation area within the camera's field of view by combining surface reflection characteristics:

[0190] Highly reflective areas: Areas where the glass curtain wall causes overexposure;

[0191] Low light attenuation area: The area of ​​blurred imaging caused by shadow occlusion.

[0192] The attenuation area is marked as a light field interference feature (such as coordinate range + interference intensity value) to avoid or enable anti-interference mode during layout optimization.

[0193] As described in step S15 above, multi-dimensional model fusion:

[0194] Geometric constraints: Building structure limits the camera installation location;

[0195] Semantic rule: Mandatory redundant coverage in high-risk areas;

[0196] Light field interference characteristics: Mark the imaging attenuation areas that need to be avoided or compensated.

[0197] By aligning the temporal and spatial dimensions and weighting features (such as a 60% weighting for security level), a unified 3D scene representation is generated to drive subsequent layout decisions.

[0198] In this embodiment, multimodal data fusion and light field enhancement modeling address the shortcomings of traditional 3D scene models in analyzing complex environments: multimodal sensing synchronously collects geometric and texture data, and semantic analysis is used to accurately label passage priorities and security risk areas, improving the accuracy of static modeling; dynamic light field reconstruction quantifies light reflection and moving target interference, predicts monitoring blind spots, and adaptively adjusts anti-interference parameters; a multi-dimensional model integrating geometric, semantic, and physical interference assessments provides holographic decision-making basis for complex scenes, enabling the scientific deployment of security layout in highly dynamic environments and significantly enhancing the system's environmental adaptability and anti-interference capabilities.

[0199] In one feasible embodiment, suppose a shopping mall needs to construct a 3D scene model for security monitoring, covering the jewelry area, atrium, and fire exits.

[0200] Step S11 (Data Acquisition):

[0201] LiDAR scanning: Obtaining the coordinates of shelves and columns with centimeter-level accuracy;

[0202] Shot with an RGB-D camera: Glass curtain wall (85% reflectivity), marble floor (10% reflectivity).

[0203] Environmental sensor: monitors the peak intensity of natural sunlight at midday (1500 lux).

[0204] Step S12 is executed (static structure annotation):

[0205] Semantic model output:

[0206] Jewelry counter (security level 9, access priority B);

[0207] Fire lane (safety level 7, access priority A);

[0208] The glass railing in the atrium (85% reflectivity, marked as a source of reflective interference).

[0209] Step S13 is executed (dynamic parameter processing):

[0210] Light distribution: The atrium lighting diagram shows that the reflectivity of the glass curtain wall on the west side reaches 1300 lux;

[0211] Moving targets: Three people were detected loitering in the jewelry area for more than 2 minutes. They were marked as "abnormal gathering - high risk" and their coordinates were bound to the jewelry cabinet (X3-Y3).

[0212] Security label generation: The jewelry cabinet is associated with the "high risk + abnormal gathering" label, and its weight value is increased to 0.9.

[0213] Step S14 is performed (light field interference prediction):

[0214] The NeRF model predicts that a 2m wide strong reflective band will form on the west side of the curtain wall between 14:00 and 15:00 (interference intensity 0.8).

[0215] This is marked as a light field interference feature; it is recommended to avoid it or use a polarizing filter during layout.

[0216] Step S15 is executed (model fusion):

[0217] Fusion results:

[0218] Geometric constraints: Suspended equipment is prohibited from being installed in fire lanes;

[0219] Semantic rule: The jewelry display case needs to be equipped with 2 cameras (for redundant coverage);

[0220] Light field interference: The camera tilt angle in the west curtain wall area must be ≥30°.

[0221] In one embodiment, in step S20, the preset layout optimization target parameter includes at least one of the following:

[0222] Coverage optimization objective: Using the security risk level in the scene semantic attributes as weight, maximize the number of sub-regions that are effectively covered within the monitoring area. The effective coverage must satisfy that the light field interference characteristics in the three-dimensional scene model are below a preset threshold.

[0223] Cost optimization objective: Minimize the overall cost of camera deployment by using the number of devices, energy consumption, and deployment complexity as joint optimization variables. The deployment complexity is dynamically evaluated based on the obstacle distribution in the 3D scene model.

[0224] Redundancy optimization objective: Within an area where the safety risk level is at a preset threshold, maximize the number of sub-regions that are covered by the overlapping fields of view of multiple cameras.

[0225] In this embodiment:

[0226] (1) Coverage optimization objective:

[0227] Security risk level weighting: High-risk areas (such as the jewelry area) are weighted at 1.0, medium-risk areas (cashier counter) at 0.6, and low-risk areas (rest area) at 0.3;

[0228] For effective coverage to be determined, the following conditions must be met simultaneously:

[0229] The monitoring field of view completely covers the sub-area;

[0230] The light field interference characteristic value is ≤0.3 (preset threshold, corresponding to an image sharpness of ≥90%).

[0231] Optimization logic: With the goal of maximizing the number of weighted sub-regions covered, priority is given to ensuring interference-free coverage in high-risk areas.

[0232] (2) Cost optimization objective:

[0233] Joint optimization variables:

[0234] Quantity of equipment (cost coefficient per unit: 1.0);

[0235] Energy consumption (cost factor of 0.2 per watt-hour);

[0236] Deployment complexity (based on dynamic calculation of obstacle density):

[0237] Complexity = Number of obstacles / 10㎡ × Installation height coefficient (ceiling = 1.0, column = 1.5);

[0238] The optimization objective is to minimize the total cost function.

[0239] Total cost = Number of equipment × 1.0 + Total energy consumption × 0.2 + Total complexity × 0.5.

[0240] (3) Redundancy optimization objective:

[0241] Preset risk threshold: Security level ≥ 8 (e.g., jewelry area, vault);

[0242] Overlap coverage requirement: At least two cameras must have an overlap rate of ≥30% in their field of view;

[0243] Optimization logic: Maximize the proportion of sub-regions that meet the overlap requirements within high-risk areas.

[0244] In one feasible embodiment, it is assumed that the shopping mall needs to optimize the monitoring layout, focusing on covering the jewelry area (risk level 9), fire exits (risk level 7), and general shelving area (risk level 3).

[0245] (1) Execution of coverage optimization objectives:

[0246] Input data:

[0247] The light field interference value in the jewelry area is 0.4 (caused by glass reflection), and needs to be reduced to ≤0.3;

[0248] The interference value of the fire escape route is 0.1 (meets the requirements).

[0249] Optimization measures:

[0250] Jewelry area: Deploy wide-angle anti-glare cameras, installed at a 25° angle to avoid reflections, reducing the interference value to 0.2;

[0251] General shelf area: Due to the low risk weight, the number of cameras will be reduced by 20%, and resources will be allocated to high-risk areas.

[0252] result:

[0253] The number of sub-regions effectively covered has increased to 98% (compared to 85% in the original plan);

[0254] High-risk areas are 100% covered.

[0255] (2) Implementation of cost optimization objectives:

[0256] Variable calculation:

[0257]

[0258]

[0259] Optimization strategy:

[0260] Equipment quantity: Replace the three standard cameras in the shelving area with one 360° panoramic camera;

[0261] Energy consumption: Low-light mode is activated in high-risk areas (power consumption reduced by 40%);

[0262] Complexity: Fire escape routes were replaced with wall-mounted brackets (complexity increased from 0.3 to 0.1).

[0263] result:

[0264] Total cost reduced by 35% (number of devices -22%, energy consumption -30%, complexity -40%).

[0265] (3) Redundancy optimization target execution:

[0266] High-risk area: Jewelry area (risk level 9 > threshold 8);

[0267] Overlapping coverage requirements:

[0268] Deploy 3 cameras (A / B / C) with a field-of-view overlap rate of ≥40%;

[0269] Real-time verification: A and B overlap by 45%, and B and C overlap by 50%, both of which meet the standards.

[0270] Dynamic adjustment:

[0271] The addition of the exhibition booth reduced the overlap rate of camera B's field of view to 25%.

[0272] The system automatically adjusted the angle of camera C to restore the overlap rate to 38%.

[0273] result:

[0274] The high-risk area overlaps with 100% of the coverage (compared to only 70% in traditional schemes).

[0275] In this embodiment, the collaborative design of multi-dimensional layout optimization target parameters solves the problems of uneven coverage and resource waste caused by single-target optimization in traditional camera layout: the coverage optimization target with security level as the weight prioritizes the monitoring efficiency of high-risk areas, and avoids areas with image quality degradation by combining light field interference threshold, eliminating the risk of missed detection in key areas due to human experience; through joint cost optimization of equipment quantity, energy consumption and deployment complexity, security efficiency and economic benefits are balanced, and redundant equipment deployment is reduced; at the same time, redundant coverage is forced in core areas where the security level reaches the threshold, forming multiple monitoring defenses, which significantly improves the security reliability of high-risk scenarios such as bank vaults and hazardous chemical warehouses.

[0276] In one embodiment, in step S20, the set of dynamic constraints includes at least one of the following:

[0277] Field of view integrity constraint: The unobstructed rate between the camera's field of view coverage path and the obstacle area in the 3D scene model must reach a preset threshold. The obstacle area is dynamically verified based on real-time updated semantic attributes.

[0278] Timeliness constraint: Based on the scene activity density in the dynamic environment parameters, the maximum generation delay threshold of the layout scheme is adaptively adjusted to ensure that the response time of high-density scenes meets the preset safety timeliness.

[0279] Cooperative deployment constraints: The deployment distance between cameras must meet the signal compatibility threshold and the physical security threshold.

[0280] In this embodiment:

[0281] (1) Visual field integrity constraint:

[0282] Unobstructed coverage rate preset threshold: ≥95% (i.e., the percentage of the effective coverage area within the camera's field of view);

[0283] Dynamic verification of obstacle areas: Based on the semantic attributes in the real-time 3D scene model (such as marking newly added temporary booths as "movable obstacles"), the proportion of intersection points between the view path and obstacles is calculated through the ray collision detection algorithm. If the occlusion rate is >5%, the constraint is determined to be unsatisfied.

[0284] Verification mechanism: Scan for scene changes every 5 seconds and trigger a real-time recalculation of the view ratio.

[0285] (2) Time constraints:

[0286] Scene activity density classification:

[0287] Low density (≤3 people / 10㎡): Layout generation delay threshold ≤5 seconds;

[0288] Medium density (4-6 people / 10㎡): Delay threshold ≤ 3 seconds;

[0289] High density (≥7 people / 10㎡): Delay threshold ≤1 second (preset safety time limit);

[0290] Adaptive adjustment logic: Density areas are divided in real time using a heatmap of pedestrian flow, and differentiated delay thresholds are assigned to different areas.

[0291] (3) Cooperative deployment constraints:

[0292] Signal compatibility threshold:

[0293] Cameras operating on the same frequency band should be spaced at least 3 meters apart (to prevent Wi-Fi / Bluetooth interference).

[0294] Spacing between devices operating in different frequency bands ≥ 1 meter;

[0295] Physical security threshold:

[0296] Fixed camera spacing ≥ device width × 1.5 (to prevent mechanical collision);

[0297] The vertical distance between mobile and fixed equipment shall be ≥0.5 meters (to prevent suspension collisions).

[0298] In one feasible embodiment, assuming the shopping mall is in the midst of an event and the pedestrian density in the atrium reaches "8 people / 10㎡" (high risk), the layout needs to be adjusted in real time.

[0299] (1) Execution of vision integrity constraints:

[0300] Scene change: A temporary promotional booth (coordinates X4-Y4) will be set up to block the original camera C's field of view;

[0301] Dynamic verification:

[0302] X-ray collision detection showed that the occlusion rate increased to 12% (>5% threshold);

[0303] The semantic attribute marker for the booth is "temporary obstacle";

[0304] Response measures:

[0305] The camera's tilt angle was automatically lowered by 20°, reducing the obstruction rate to 4%.

[0306] Update the 3D model obstacle coordinate library.

[0307] (2) Execution of time-sensitive constraints:

[0308] Density monitoring:

[0309] Atrium (8 people / 10㎡) → Delay threshold ≤ 1 second;

[0310] Clothing area (5 people / 10㎡) → Delay threshold ≤ 3 seconds;

[0311] Layout generation:

[0312] A sudden gathering incident occurs in the atrium; the model outputs a solution within 0.8 seconds:

[0313] Activate 2 patrol robots as reinforcements;

[0314] Adjust four wide-angle cameras to track the flow of people.

[0315] (3) Coordinated deployment constraint execution:

[0316] New device conflict:

[0317] equipment coordinate frequency band Physical dimensions Fixed camera D (X5,Y5) 2.4GHz 0.2m × 0.2m Patrol Robot E (X5.2, Y5.3) 5GHz 0.8m × 0.6m

[0318] Constraint verification:

[0319] Signal compatibility: Horizontal spacing of 0.3 meters (<1 meter threshold) → collision

[0320] Physical safety: Vertical spacing of 0.4 meters (<0.5 meter threshold) → conflict

[0321] Dynamic adjustment:

[0322] Move robot E's path westward by 1.2 meters to meet the spacing requirements (1.5 meters horizontally and 0.6 meters vertically).

[0323] Switch robot E to the 5GHz high-frequency channel to eliminate signal interference.

[0324] In this embodiment, the real-time adaptation mechanism of dynamic constraint condition set overcomes the problem of insufficient scene compatibility caused by the rigidity of traditional layout rules: the field of view integrity constraint is based on the dynamic updating of obstacles with semantic annotation to ensure that the unobstructed rate of the camera's field of view path is greater than or equal to the preset threshold, eliminating monitoring blind spots caused by temporary obstacles (such as construction fences); the timeliness constraint adaptively adjusts the response delay according to the scene activity density, increasing the layout generation speed during peak traffic hours by more than 50%, meeting the emergency response needs of sudden events; the collaborative deployment constraint solves the risk of signal interference and mechanical collision in multi-device deployment through dual verification of signal compatibility and physical safety threshold, reducing equipment failure rate and extending hardware life.

[0325] Reference Figure 4 In one embodiment, step S20 specifically includes the following steps:

[0326] S21. Based on the preset layout optimization target parameters, dynamic constraint set, and training dataset, generate a multi-scenario training dataset, wherein each training sample includes:

[0327] Input features: geometric topology, dynamic environmental parameters, and security requirement parameters extracted from historical 3D scene models;

[0328] Output label: The set of optimal camera layout parameters that satisfy the layout optimization objective parameters and conform to the set of dynamic constraints;

[0329] S22. Construct a multi-objective optimization neural network model, the model including:

[0330] Coverage optimization module: Calculates the coverage effectiveness of camera locations on the monitored area based on security risk level weights;

[0331] Cost optimization module: Assess the complexity of equipment deployment based on obstacle distribution density and generate a layout scheme under cost constraints;

[0332] Redundancy optimization module: Calculates the redundancy of overlapping coverage of multiple camera views within an area where the security risk level reaches a preset threshold;

[0333] S23. Input the multi-scenario training dataset into the multi-objective optimization neural network model, with the coverage optimization objective, cost optimization objective, and redundancy optimization objective as optimization terms, and the dynamic constraint condition set as hard constraint terms, and train the model parameters together through the constraint optimization algorithm;

[0334] S24. Verify whether the layout scheme output by the trained model in the test scenario simultaneously satisfies all dynamic constraints. If there is a constraint violation, adjust the weight coefficients of the layout optimization objective parameters and retrain.

[0335] As described in step S21 above, the logic for constructing the multi-scenario training dataset is as follows:

[0336] Input features:

[0337] Geometric topology: point cloud of building outline, spatial coordinates of obstacles;

[0338] Dynamic environmental parameters: illumination distribution matrix, pedestrian flow heat map;

[0339] Security requirements parameters: Risk level distribution map (e.g., jewelry area = level 9).

[0340] Output tags:

[0341] Optimal layout parameter set: camera coordinates (x, y, z), pitch angle (α), focal length (f);

[0342] Verification criteria: Simultaneously meet the following requirements: coverage ≥ 95%, cost ≤ budget limit, and redundancy ≥ 2 in high-risk areas.

[0343] As described in step S22 above, the multi-objective optimization neural network model architecture is as follows:

[0344] (1) Coverage optimization module (graph convolutional network):

[0345] Input: Risk level weight matrix (e.g., jewelry area weight 1.0);

[0346] Output: Coverage effectiveness score (calculation formula: ∑(sub-region risk weight × coverage status)).

[0347] (2) Cost optimization module (fully connected network):

[0348] Input: Obstacle density map + installation location code;

[0349] Output: Deployment complexity score (Formula: 0.5 × number of devices + 0.3 × complexity).

[0350] (3) Redundancy optimization module (attention mechanism network):

[0351] Input: Coordinate range of the high-risk area;

[0352] Output: Redundancy score (calculation formula: overlapping area / total area of ​​high-risk area).

[0353] As described in step S23 above, the joint training mechanism is as follows:

[0354] Optimization objective: Maximize the total utility function.

[0355] U = 0.6 × coverage efficiency + 0.3 × (1 / cost) + 0.1 × redundancy.

[0356] Hard constraint terms (injected via penalty function):

[0357] Visibility rate <95% → Penalty +1000;

[0358] Equipment spacing < safety threshold → penalty +500;

[0359] Training algorithm: The constrained NSGA-II multi-objective evolutionary algorithm is used, with 1000 iterations.

[0360] As described in step S24 above, the constraint violation handling process is as follows:

[0361] Test scenario: Simulating the addition of new obstacles causing the visibility to drop to 90%;

[0362] Constraint detection: The visibility rate of the model output solution does not meet the standard (<95%);

[0363] Weight adjustment:

[0364] The original coverage effectiveness weight was 0.6, which has been increased to 0.7.

[0365] Retrain until the test set constraint compliance rate is ≥99%.

[0366] In this embodiment, a joint training mechanism combining multi-objective collaborative optimization and dynamic constraints addresses the insufficient optimization capabilities of traditional layout models under multi-objective conflicts and complex physical rules. The multi-objective optimization neural network module learns hierarchically to cover decision rules for efficiency, cost control, and redundancy security, and generates adaptive layout strategies by combining historical scene data and real-time environmental parameters. The constraint optimization algorithm forcibly verifies the visibility, timeliness, and device coordination rules of the layout scheme, ensuring that the scheme strictly conforms to dynamic constraints. Through incremental training and weight adjustment mechanisms, the generalization and response speed of the model in highly dynamic scenarios are significantly improved, providing intelligent layout decision support that balances efficiency and reliability for complex security scenarios.

[0367] In one feasible embodiment, it is assumed that during an event, the pedestrian density in the mall atrium surges (12 people / 10㎡), requiring urgent optimization of the layout.

[0368] Step S21 is executed (training data generation):

[0369] Input feature extraction:

[0370] Geometric topology: coordinates of the atrium pillars, temporary stage boundaries;

[0371] Dynamic parameters: Peak area of ​​pedestrian flow heat (coordinates X6-Y6);

[0372] Security requirements: The risk level of the stage area has been raised to level 8.

[0373] Output tag generation:

[0374] The best historical solution: deploy 8 cameras (tilt angle 15°-30°), with a coverage efficiency of 92%.

[0375] Step S22 execution (model building):

[0376] Optimization module Input data Output Decision Example Coverage optimization Stage area risk weight: 0.9 Add 2 wide-angle cameras Cost optimization Column density: 2.5 columns per 10 square meters Switch to wall-mounted brackets (reduces complexity by 30%). Redundancy optimization Stage area coordinate range The field of view overlap rate has been increased to 45%.

[0377] Step S23 (Joint Training) Execution:

[0378] Initial solution flaws:

[0379] Equipment spacing 1.2 meters (<1.5 meters safety threshold) → Penalty +500;

[0380] The stage area coverage efficiency is only 85% (<95% target).

[0381] Optimized solution:

[0382] Adjust the camera position, increasing the spacing to 1.6 meters;

[0383] An additional anti-obstruction camera has been added to the stage area, achieving a coverage efficiency of 96%.

[0384] Step S24 is executed (constraint verification and tuning):

[0385] Test scenario: Temporary installation of a light fixture (addition of obstructions);

[0386] Constraint violation: Line-of-sight rate dropped to 88%;

[0387] Weight adjustment:

[0388] Coverage effectiveness weighting changed from 0.6 to 0.75.

[0389] After retraining, the line-of-sight accuracy recovered to 96%.

[0390] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0391] In one embodiment, a modeled camera layout system is provided, which corresponds to the modeled camera layout method described in the above embodiments. The modeled camera layout system includes:

[0392] The data acquisition module is used to acquire three-dimensional spatial data of the target scene and construct a three-dimensional scene model that includes static object structures and dynamic environmental parameters.

[0393] The model training module is used to train a machine learning model based on preset layout optimization target parameters, dynamic constraint condition set and training dataset, and obtain a trained machine learning model, which is used to associate the mapping relationship between camera layout parameters and scene features.

[0394] The solution output module is used to input the three-dimensional scene model and the security requirement parameters acquired in real time into the trained machine learning model, and output the optimal camera layout solution that meets the layout optimization target parameters through the machine learning model.

[0395] The iterative optimization module is used to dynamically iteratively optimize the position, angle, and collaborative parameters of the camera based on the environmental change parameters of the target scene acquired in real time, and output an updated camera layout scheme.

[0396] For specific limitations regarding a model-based camera layout system, please refer to the limitations of a model-based camera layout method described above, which will not be repeated here. Each module in the aforementioned model-based camera layout system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0397] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows. Figure 5 The computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database is used for data storage, data processing, and data analysis. The network interface is used for communication with external terminals via a network connection. When executed by the processor, the computer program implements a modeled camera layout method.

[0398] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement a modeled camera layout method.

[0399] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements a modeled camera layout method.

[0400] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0401] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0402] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method of modeling camera placement, the method comprising: Comprise the following steps: S10, acquire the three-dimensional space data of the target scene, construct the three-dimensional scene model containing static object structure and dynamic environment parameter; Wherein, specifically comprising the following steps: S11, acquire the three-dimensional space data of the target scene through the multi-modal acquisition module, the three-dimensional space data includes geometric topology data, texture image data and dynamic environment parameter; S12, based on the geometric topology data and texture image data, extract the spatial structure of building, fixed equipment and functional area, and label the static object's traffic priority and safety risk level through semantic analysis model; S13, real-time acquisition of the light distribution characteristics, surface reflection characteristics and moving target trajectory in the dynamic environment parameter, behavior semantic classification is carried out to the moving target, and security response priority label dynamically associated with three-dimensional space coordinate is generated; S14, based on the light distribution characteristics and surface reflection characteristics, the imaging attenuation area in the camera field of view is predicted through the light field reconstruction model, and is labeled as light field interference feature; S15, fusion the spatial structure, semantic labeling, security response priority label and light field interference feature of the static object, construct the multi-dimensional three-dimensional scene model containing geometric constraint, semantic rule and light field interference feature; S20, based on the preset layout optimization target parameter, dynamic constraint condition set and training data set, train the machine learning model, obtain the trained machine learning model, and be used for the mapping relationship between the camera layout parameter and the scene feature; S30, input the three-dimensional scene model and the real-time acquisition security demand parameter into the trained machine learning model, and output the optimal camera layout scheme meeting the layout optimization target parameter through the machine learning model; S40, according to the environmental change parameter of the target scene acquired in real time, the position, angle and coordination parameter of the camera are dynamically iteratively optimized through the machine learning model, and the updated camera layout scheme is output.

2. The method of claim 1, wherein, After the step S40, further comprising the following steps: S50, in the camera layout scheme execution process, when the camera in the target scene triggers the preset monitoring failure determination rule, a monitoring failure signal is generated and the position area of the failure camera is determined; S60, according to the position area, the standby camera meeting the preset standby equipment activation rule is activated; S70, dynamically adjust the monitoring parameter of the standby camera to cover the failure area, and update the coordination parameter set in the camera layout scheme.

3. The method of claim 2, wherein, In the step S60, the preset standby equipment activation rule comprises at least one of the following: The straight line distance or path planning distance between the standby camera and the failure area is less than the preset distance threshold, and the nearest available equipment is preferentially selected; The current work load of the standby camera is lower than the preset threshold, and the work load includes at least one of CPU utilization, network bandwidth occupancy and storage space remaining amount; The field of view angle of the standby camera exists unobstructed visibility coverage with the failure area in the three-dimensional scene model.

4. The method of claim 1, wherein, In the step S20, the preset layout optimization target parameter comprises at least one of the following: The coverage optimization objective is to maximize the number of sub-regions in the monitoring area that are effectively covered, with the safety risk level in the semantic attribute of the scene as the weight, and the effective coverage satisfying that the light field interference feature in the three-dimensional scene model is lower than a preset threshold; The cost optimization objective is to minimize the comprehensive cost of camera deployment, with the number of devices, energy consumption, and deployment complexity as the joint optimization variables, and the deployment complexity being dynamically evaluated based on the obstacle distribution in the three-dimensional scene model; The redundancy optimization objective is to maximize the number of sub-regions that are overlapped and covered by multiple camera fields of view in the area where the safety risk level is a preset threshold.

5. The method of claim 4, wherein, In the step S20, the set of dynamic constraint conditions includes at least one of: The field of view integrity constraint requires that the unobstruction rate between the camera field of view coverage path and the obstacle region in the three-dimensional scene model reaches a preset threshold, and the obstacle region is dynamically verified based on the real-time updated semantic attribute; The timeliness constraint adaptively adjusts the maximum generation delay threshold of the layout scheme according to the scene activity density in the dynamic environment parameter, to ensure that the response time of high-density scenes meets the preset safety timeliness; The cooperative deployment constraint requires that the deployment distance between cameras meets the signal compatibility threshold and the physical safety threshold.

6. The method of claim 5, wherein, In the step S20, the method specifically includes the following steps: S21, generating a multi-scene training data set according to preset layout optimization objective parameters, a set of dynamic constraint conditions, and a training data set, wherein each training sample includes: Input features: geometric topology structure, dynamic environment parameters, and security demand parameters extracted from historical three-dimensional scene models; Output labels: an optimal camera layout parameter set that meets the layout optimization objective parameters and conforms to the set of dynamic constraint conditions; S22, constructing a multi-objective optimization neural network model, which includes: A coverage optimization module: calculating the coverage efficiency of the camera position on the monitoring area based on the safety risk level weight; A cost optimization module: evaluating the device deployment complexity according to the obstacle distribution density to generate a layout scheme under the cost constraint; A redundancy optimization module: calculating the redundancy of multiple camera field of view overlap in the area where the safety risk level reaches a preset threshold; S23, inputting the multi-scene training data set into the multi-objective optimization neural network model, taking the coverage optimization objective, the cost optimization objective, and the redundancy optimization objective as optimization items, and taking the set of dynamic constraint conditions as hard constraint items, and training the model parameters through a constraint optimization algorithm; S24, verifying whether the layout scheme output by the trained model in the test scene meets all the dynamic constraint conditions, and if there is a constraint violation, adjusting the weight coefficient of the layout optimization objective parameters and retraining.

7. A modelled camera layout system for implementing the steps of a modelled camera layout method as claimed in any one of claims 1 to 6, characterized by The method includes: A data acquisition module for acquiring three-dimensional spatial data of a target scene and constructing a three-dimensional scene model containing static object structures and dynamic environment parameters; A model training module for training a machine learning model based on preset layout optimization objective parameters, a set of dynamic constraint conditions, and a training data set to obtain a trained machine learning model for associating the mapping relationship between camera layout parameters and scene features. A scheme output module is configured to input the three-dimensional scene model and real-time acquired security requirement parameters into a trained machine learning model, and output an optimal camera layout scheme meeting a layout optimization target parameter through the machine learning model. An iterative optimization module is configured to dynamically and iteratively optimize the position, angle and coordination parameters of the camera through the machine learning model according to real-time acquired environmental change parameters of the target scene, and output an updated camera layout scheme.

8. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The processor executes the computer program to implement the steps of the model camera layout method according to any one of claims 1-6.

9. A computer-readable storage medium storing a computer program, the computer program comprising instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 8. The computer program is executed by the processor to implement the steps of the model camera layout method according to any one of claims 1-6.

Images