Intelligent management and control method and system for smart park

Abstract

The invention relates to the technical field of equipment management and control, in particular to an intelligent management and control method and system for a smart park, and proposes the following scheme: dividing a region into regular grids based on spatial data and equipment data, and generating an inspection path in combination with dynamic weights. And the inspection equipment collects equipment operation data and environmental parameters according to the path, and transmits the inspection data to the processing equipment in real time. And the processing equipment accurately identifies equipment operation abnormity, environment parameter abnormity or security risks through time sequence analysis and an abnormity detection model, and links the management and control equipment to process an abnormal area. The method improves the park inspection efficiency and the exception handling capability, and is suitable for industrial park scenes.

Description

Technical Field

[0001] The present invention relates to the field of equipment control technology, and in particular to an intelligent control method and system for a smart park. Background Art

[0002] With the rapid development of industrialization and informatization, demand for smart campus applications is increasing. Smart campuses integrate advanced IoT, artificial intelligence, and automation technologies to enable real-time monitoring and intelligent management of equipment within the campus. However, existing inspection and control methods still have many technical deficiencies when dealing with complex campus environments, hindering further improvements in smart campus management efficiency and security.

[0003] The existing technologies all have the problems raised in this background technology: low inspection efficiency and lack of dynamic path planning. To solve the above problems, this application designs an intelligent management and control method and system for smart parks. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to address the shortcomings of existing technologies and provide an intelligent management and control method and system for smart parks. Based on spatial data and equipment data, the area is divided into regular grids, and inspection paths are generated in combination with dynamic weights. Inspection equipment collects equipment operating data and environmental parameters along the path and transmits the inspection data to processing equipment in real time. The processing equipment uses time series analysis and anomaly detection models to accurately identify equipment operation anomalies, environmental parameter anomalies, or security risks, and links the management and control equipment to process abnormal areas. This improves the park's inspection efficiency and anomaly handling capabilities, and is suitable for industrial park scenarios.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] An intelligent management and control method for a smart park, wherein the smart park includes multiple smart workshops, each of which is equipped with an inspection device for inspection. The inspection device is in communication with a processing device, and the intelligent management and control method is executed by the processing device. The intelligent management and control method includes:

[0007] Generate an inspection path based on the spatial data and equipment data of the three-dimensional model of the smart workshop to be controlled, wherein the inspection path determines a regular grid through the spatial data and assigns grid weights to the regular grid through the equipment data;

[0008] Sending the inspection path to the inspection device, wherein the inspection device inspects the inspection path, obtains inspection data, and sends the inspection data to the processing device;

[0009] Receive the inspection data, detect whether there is any abnormal situation, and control the corresponding control equipment to control the area where the abnormal situation is located.

[0010] Generating the inspection path includes:

[0011] Dividing the smart workshop into regular grids according to the spatial data, and assigning grid weights according to the equipment data;

[0012] Setting an inspection target set, and mapping the inspection target set to the regular grid;

[0013] Generate a minimum circumscribed polygon covering the inspection target set, and generate multiple parallel inspection paths at fixed intervals according to the main axis direction of the minimum circumscribed polygon;

[0014] Detecting the parallel inspection paths to obtain a set of feasible paths;

[0015] The feasible path set is evaluated to obtain an inspection path.

[0016] Detecting the parallel inspection path includes:

[0017] Performing obstacle detection on the parallel inspection path;

[0018] If the path overlaps with the obstacle grid, the A* algorithm is used to replan the path in the obstacle grid area, reconnecting the starting and ending points of the path after bypassing the obstacle. The A* algorithm generates a local obstacle avoidance path based on the environmental constraint parameters in the device data.

[0019] If there is no barrier grid, the path is added to the feasible path set, and the traversal of the paths is repeated until all paths are added to the feasible path set.

[0020] The set of feasible paths is evaluated, including:

[0021] Establishing a multi-objective evaluation function, wherein the objectives include grid weight, path length, number of turns, coverage repetition rate and energy consumption;

[0022] Calculating a score for each path in the set of feasible paths according to the multi-objective evaluation function;

[0023] The path corresponding to the highest score is used as the inspection path.

[0024] When the inspection device is inspecting on the inspection path, the inspection device further includes:

[0025] When the inspection device detects an obstacle that does not belong to the obstacle grid, it updates the regular grid and adjusts the inspection path of the current inspection device;

[0026] When the inspection equipment detects detection data that deviates from the normal level, the equipment priority is reordered, the grid weight is updated, and the next inspection path of the inspection equipment is adjusted.

[0027] The detecting whether there is an abnormality includes:

[0028] Constructing a time series of the equipment operation data based on the equipment operation data in the multiple inspection data transmitted by the inspection equipment;

[0029] Calculating a time window based on the inspection time interval of the inspection equipment and the number of inspection path adjustments, and segmenting the time series based on the time window to obtain multiple window sequences;

[0030] Calculate the characteristic index of each window sequence and determine whether the characteristic index is abnormal;

[0031] If an anomaly exists, the characteristic indicator is input into a preset anomaly detection model, and the anomaly is output through the anomaly detection model.

[0032] The control device corresponding to the control controls the area where the abnormal situation occurs, including:

[0033] Displaying the abnormality type corresponding to the abnormal situation in the three-dimensional model;

[0034] Determine control measures based on the abnormality type, transmit the control measures to the control equipment, control the area where the abnormality occurs, and collect processing data during the control process;

[0035] After the abnormal situation is processed, the 3D model is updated based on the processed data.

[0036] The intelligent management and control method further includes constructing a three-dimensional model of the smart workshop, wherein constructing the three-dimensional model of the smart workshop includes:

[0037] The image sensor carried by the drone collects multi-angle images of the smart workshop, and the multi-angle images are stitched together using a feature matching algorithm to generate a continuous image;

[0038] Based on the continuous images, three-dimensional point cloud data of the smart workshop is generated using structured light technology, and the three-dimensional point cloud data is processed to extract spatial information;

[0039] Inputting the continuous image into a preset recognition network, and outputting device information through the recognition network;

[0040] A three-dimensional model is constructed according to the space information and the device information.

[0041] An intelligent management and control system for a smart park, comprising a three-dimensional modeling module, an inspection planning module, and an abnormality response module;

[0042] The three-dimensional modeling module is used to build a three-dimensional model of the smart workshop;

[0043] The inspection planning module is configured to generate an inspection path based on the spatial data and equipment data of the three-dimensional model, wherein the inspection path is a regular grid determined by the spatial data and a grid weight is assigned to the regular grid by the equipment data;

[0044] The abnormal response module is used to receive inspection data, detect whether there is an abnormal situation, and control the corresponding control equipment to control the area where the abnormal situation occurs.

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] The present invention constructs a three-dimensional model of a smart park and combines it with dynamic path planning, time series analysis, and deep learning anomaly detection models to achieve a significant improvement in inspection efficiency, precise anomaly detection, and rapid response of multi-system linkage. It solves the problems of inefficient inspection, delayed anomaly processing, and incomplete modeling information in the existing technology, can meet the intelligent management needs in complex park scenarios, and improve the safety and operational efficiency of the park. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Other features, objects and advantages of the present invention will become more apparent upon reading the detailed description of non-limiting embodiments made with reference to the following drawings:

[0048] Figure 1 This is a flow chart of an intelligent management and control method for a smart park according to embodiment 1 of the present invention;

[0049] Figure 2 This is a schematic diagram of an intelligent management and control system for a smart park according to Example 2 of the present invention;

[0050] Figure 3 This is a flowchart of the steps for constructing a three-dimensional model of a smart workshop in Example 3 of the present invention;

[0051] Figure 4 This is a diagram illustrating the network structure of embodiment 3 of the present invention. DETAILED DESCRIPTION

[0052] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0053] Example 1

[0054] See also Figure 1 The present invention provides an embodiment of an intelligent management and control method for a smart park. The specific steps of the intelligent management and control method are as follows:

[0055] S1: Generate an inspection route based on the spatial data and equipment data of the 3D model of the smart workshop to be controlled;

[0056] In this embodiment, processing devices (such as industrial servers or edge computing gateways) divide the smart workshop into regular grids. Weights are assigned to the grids based on the device's inspection priority and environmental constraints, and an optimization algorithm is used to generate inspection paths that cover the inspection targets. During the path generation process, the A* algorithm is used to plan around obstacles in the grid, ensuring the feasibility and safety of the path.

[0057] S2: Send the inspection route to the inspection equipment and collect inspection data;

[0058] In this embodiment, the inspection equipment can be a ground inspection robot, a drone, or a fixed-track inspection device, with the specific choice determined by the needs of the smart workshop. Ground inspection robots are suitable for complex terrain and are equipped with infrared thermal imaging sensors and gas detection modules; drones are suitable for inspecting large areas and high altitudes and are equipped with high-definition cameras and multispectral sensors; and fixed-track inspection devices are suitable for high-precision, repetitive inspections. The inspection equipment collects equipment operating data and environmental parameters along the route and transmits this data in real time to processing equipment via wireless communication networks (such as 5G or Wi-Fi).

[0059] S3: The inspection device sends the inspection data to the processing device;

[0060] S4: Detect whether there is any abnormality;

[0061] In this embodiment, after receiving inspection data, the processing device preprocesses it using time series analysis techniques and analyzes it using a deep learning-based anomaly detection model. This anomaly detection model combines time series feature extraction, an attention mechanism, and a classification and decision module to accurately identify equipment operational anomalies, environmental parameter anomalies, and security risks. Detection results are directly fed back into the 3D model, highlighting abnormal areas and annotating the anomaly type, impact range, and recommended actions.

[0062] S5: Control the corresponding control equipment to control the area where the abnormal situation occurs.

[0063] In this embodiment, the processing equipment is linked to the corresponding control equipment to perform intelligent processing. The control equipment includes environmental control equipment (such as ventilation systems, exhaust devices and temperature control equipment), equipment control systems (such as industrial equipment shutdown or mode switching modules) and security systems (such as patrol robots, smart cameras and alarm systems). For example, when a gas leak is detected, the ventilation equipment is linked to dilute the harmful gas and the environmental status in the three-dimensional model is dynamically updated; when a device failure is detected, the device is directly controlled to shut down and maintenance personnel are notified; when a security risk is discovered, the patrol robot is activated for real-time monitoring and alarm.

[0064] Specifically, equipment inspection and exception handling within smart parks usually rely on traditional manual operations or relatively rudimentary path planning methods. Although these methods can achieve equipment inspection to a certain extent, they face significant technical limitations in complex industrial park environments. For example, during the inspection process, traditional methods make it difficult to dynamically adjust the inspection priorities of multiple devices. This is especially true when the equipment is densely distributed, the environment is complex, or abnormal situations occur. Inspection equipment is prone to path duplication, resource waste, and inspection omissions. In addition, anomaly detection relies on fixed rule thresholds and lacks intelligent analysis capabilities, making it difficult to accurately respond to abnormal situations in dynamic environments. Control measures often rely on manual decision-making, which is slow to respond and can easily lead to the spread of anomalies and potential safety hazards in the park.

[0065] In this embodiment, the processing device uses spatial data and equipment data based on a three-dimensional model of the industrial park to generate a regular grid and assign dynamic weights to it. This ensures that path planning not only covers inspection targets but also takes into account equipment priority and obstacle bypassing capabilities. For example, if there are multiple high-priority devices in a certain area of the park that require inspection, the system can prioritize the shortest path for these high-priority devices while bypassing areas with obstacles, thus avoiding wasted paths and inefficient operation of the inspection equipment. The dynamic and intelligent nature of this path planning allows the inspection equipment to flexibly adapt to the needs of different areas and tasks, significantly improving inspection efficiency.

[0066] Furthermore, during the inspection process, the data collected by the inspection equipment will be transmitted to the processing equipment via wireless communication, and the processing equipment will use time series analysis and deep learning models to analyze the data in real time. For example, when the inspection equipment detects that the operating temperature of a certain device in an industrial workshop is abnormal, the system can use the time series analysis model to determine whether the device has a persistent overheating risk, and combine it with the deep learning model to extract abnormal features and accurately identify the cause of overheating (such as excessive circuit load or poor environmental heat dissipation). At the same time, the system will automatically generate linkage management and control strategies based on the type of abnormality, such as notifying related equipment to shut down or switch operating modes, and starting the ventilation system to improve the environmental conditions around the equipment. This automation and intelligence of exception handling avoids the response lag problem caused by relying on manual decision-making in traditional methods, and greatly shortens the exception handling time.

[0067] For example, suppose there are 10 key equipment in the workshop, 2 of which are located in a narrow area and surrounded by obstacles. When planning the path, the traditional inspection method may cause the inspection path to be lengthy or equipment to be missed due to the complexity of the obstacle area, or even fail to bypass obstacles during the inspection process. However, this embodiment uses a regular grid divided by a three-dimensional model and dynamic weight allocation to prioritize the generation of high-priority paths for key equipment in narrow areas, and uses the A* algorithm to accurately bypass obstacles. In addition, when the inspection equipment detects that a certain device has abnormal vibration, the system can adjust the priority of the device in real time, set it as the primary target of the next inspection, and update the inspection path to ensure that the abnormal device is detected and processed first. Through this method, not only the inspection efficiency is improved, but also the priority response capability of abnormal equipment is guaranteed. Especially in scenarios where the equipment is densely distributed and the environment is complex, the effect is particularly significant.

[0068] Preferably, in this embodiment, the inspection equipment can be a small wheeled robot used to monitor the operating status of public facilities within the community (such as abnormal elevator vibrations, streetlight failures, etc.). When an abnormal firefighting facility status is detected, the processing equipment can link with the firefighting system to initiate an early warning and push real-time alarm information to the community management center via a 3D model, thereby improving community safety.

[0069] Preferably, in this embodiment, the inspection equipment can be an autonomous mobile platform equipped with high-precision sensors to monitor equipment operating parameters and air quality within the medical campus. If an abnormal pressure in the oxygen supply equipment is detected, the processing equipment can activate a backup device to take over the oxygen supply and provide real-time notification of the equipment status to medical staff via a 3D model, ensuring continuity of medical services.

[0070] The specific steps of S1 are as follows:

[0071] S1.1: Divide the smart workshop into regular grids based on the spatial data, and assign grid weights based on the equipment data;

[0072] Specifically, based on the 3D model of the smart workshop, spatial data (including workshop dimensions, equipment distribution, and obstacle locations within the workshop) is extracted and divided into regular grids. The size of the regular grids is set based on the sensor coverage of the inspection equipment, ensuring that each grid is fully covered by the inspection equipment's sensors. A higher grid resolution improves path planning accuracy, but also increases the computational complexity. Therefore, a moderate grid size is selected to balance inspection accuracy and computational efficiency.

[0073] Furthermore, each grid is assigned a dynamic weight, determined by device data, including its priority, required inspection frequency, and the environmental complexity of its location. Grids containing high-priority devices receive a higher weight, while grids containing obstacles receive a zero weight, marking them as impassable. This ensures that important areas are prioritized during path planning while preventing obstacles from impacting the path, resulting in efficient and accurate path planning.

[0074] S1.2: Setting an inspection target set, and mapping the inspection target set to the regular grid;

[0075] Specifically, the inspection target set consists of a collection of devices that need to be inspected, including information such as the device's location coordinates, priority, and inspection requirements. When mapping the inspection target set to a regular grid, the grid number of each inspection target is determined by comparing the device's three-dimensional coordinates with the grid's location coordinates. During the inspection target mapping process, considering that the target device may be close to the obstacle boundary or overlap, the spatial position of the device center point is used as the basis for grid mapping, and the adjacent grids are appropriately expanded and marked according to the coverage range of the device. This ensures that the inspection path can cover all target devices during planning and avoids devices being missed or blocked by obstacles.

[0076] S1.3: Generate a minimum circumscribed polygon covering the inspection target set, and generate multiple parallel inspection paths at fixed intervals according to the main axis direction of the minimum circumscribed polygon;

[0077] Specifically, a minimum circumscribed polygon (MBP) can cover the entire workshop area with the smallest possible boundary area, thereby determining the optimal straight-line coverage direction and avoiding complex path intersections and extra work. A rotating caliper algorithm is used to find the minimum-area circumscribed polygon, and the polygon's principal axis is selected as the primary operating direction for inspection equipment, reducing unnecessary turns.

[0078] Furthermore, by calculating the boundary points of the grid where the inspection target set is located, a minimum circumscribed polygon is generated using the minimum circumscribed polygon algorithm. This polygon can cover all inspection targets while minimizing the inspection range to reduce the path length. Based on the main axis direction of the polygon, parallel inspection paths are generated by calculating the spacing. The size of the spacing matches the sensing coverage width of the inspection equipment to ensure that the area between the paths is fully covered. During the path generation process, the linearity of the path is prioritized to reduce the number of device turns and improve inspection efficiency.

[0079] The calculation formula for fixed spacing is:

[0080]

[0081] Where d represents the fixed spacing, Δw represents the overlap margin of the coverage range, W represents the sensing coverage width of the inspection device, A represents the area of the minimum circumscribed polygon, which is used to optimize the total coverage efficiency of the path, N represents the number of target grids that need to be inspected, which is related to the distribution density of the inspection target set, and θ represents the angle between the main axis direction and the parallel path, which is used to adjust the coverage direction of the path to reduce overlap or omission caused by oblique movement.

[0082] S1.4: Detect the parallel inspection paths to obtain a set of feasible paths;

[0083] Specifically, by comparing the overlap between each grid in the path and the obstacle grid, the path segments that overlap with the obstacle are marked. If the path segment is completely within the obstacle grid, it is determined that the path segment is infeasible and is replanned. An alternative path is generated to bypass the obstacle while keeping the starting and ending points of the alternative path consistent with the original path. This path detection and correction method can ensure that the inspection equipment will not be interrupted by obstacles when performing its tasks, while ensuring the continuity and optimality of the path;

[0084] S1.5: Evaluate the set of feasible paths to obtain an inspection path;

[0085] Specifically, paths with shorter lengths, higher coverage, and fewer turns receive higher scores. A greater cumulative score for each path passing through high-weighted grid cells indicates more comprehensive coverage of high-priority devices, further increasing the score. Based on the scoring results, the path with the highest overall score is selected as the final inspection route. This evaluation method effectively balances inspection efficiency and coverage priority, while ensuring the feasibility and energy efficiency of the inspection route. The final inspection route is sent by the processing device to the inspection device for execution.

[0086] The specific steps of S1.4 are as follows:

[0087] S1.4.1: Perform obstacle detection on the parallel inspection path;

[0088] Specifically, obstacle detection aims to identify obstacles along the inspection path that could impede the normal movement of inspection equipment. By processing the spatial data from the 3D model, the system uses obstacle perception algorithms to identify areas marked as obstacles within a regular grid. These obstacles can include fixed obstacles (such as equipment foundations and pillars) and dynamic obstacles (such as mobile equipment and temporarily stored items).

[0089] Furthermore, the processing equipment dynamically updates the obstacle positions in the 3D model using real-time environmental data collected by the inspection equipment (such as location information collected by lidar, ultrasonic sensors, and depth cameras). The obstacle information is then mapped to the corresponding grids in the parallel inspection path. Obstacle detection uses a grid-based feature matching method, specifically comparing each grid on the inspection path with the grids marking the obstacle to determine if the path overlaps with an obstacle.

[0090] S1.4.2: If the path overlaps with the obstacle grid, re-plan the path within the obstacle grid area using the A* algorithm, reconnecting the starting and ending points of the path after bypassing the obstacle. The A* algorithm generates a local obstacle avoidance path based on environmental constraints.

[0091] Specifically, the A* algorithm is a heuristic search algorithm, the core of which is to evaluate the quality of nodes by calculating the cost function of the path;

[0092] In this embodiment, the A* algorithm is optimized based on the specific environmental restriction parameters of the industrial park to adapt to the complex scenarios of local obstacle avoidance. The environmental restriction parameters include the mobility characteristics of the inspection equipment (such as turning radius, maximum climbing ability), the nature of the obstacles (such as the moving speed and direction of dynamic obstacles), and the passability of the environmental area (such as the reduction of the pass weight of high-risk areas). During the path search process, the A* algorithm dynamically adjusts the estimated cost from the current node to the target node in the cost function, giving priority to paths with high safety and low energy consumption. For example, when the inspection equipment needs to bypass an obstacle, the algorithm will generate a smooth path that meets the turning restrictions based on the minimum turning radius of the equipment, while avoiding high-risk areas.

[0093] Furthermore, after replanning the path, the processing device seamlessly connects the starting and ending points of the local obstacle avoidance path to the original path, ensuring the overall continuity of the inspection path. This system can quickly handle obstacles in complex scenarios, avoiding global path replanning and significantly reducing computing time and resource consumption. Furthermore, optimizing the path in conjunction with environmental constraint parameters can improve the practical feasibility of the path, especially in scenarios with dense or dynamically changing obstacles, providing extremely high adaptability and flexibility.

[0094] S1.4.3: If no obstacle grid exists, add the path to the set of feasible paths and repeat the traversal until all paths are added to the set of feasible paths;

[0095] Specifically, for parallel inspection paths with no detected obstacles, the processing device directly marks them as feasible paths and adds them to the set of feasible paths. To ensure efficient and consistent path coverage, the coverage rate and number of turns are pre-verified when adding a path to ensure that path planning does not result in redundant inspections or excessive equipment consumption. For example, when adding a path to the set of feasible paths, the system calculates the degree to which each path's actual coverage matches the grid distribution and excludes paths with excessively high duplication coverage.

[0096] The specific steps of S1.5 are as follows:

[0097] S1.5.1: Establish a multi-objective evaluation function, where the objectives include grid weight, path length, number of turns, coverage repetition rate, and energy consumption;

[0098] Specifically, a multi-objective evaluation function was established to comprehensively evaluate each path in the set of feasible paths. Its core purpose was to balance path quality and inspection efficiency under various inspection requirements, thereby selecting the optimal inspection path. Grid weights are determined by both device priority and environmental constraint parameters. High-priority devices or those in special environments receive higher grid weights, ensuring that paths prioritize coverage of these critical areas. Path length directly reflects the time it takes for inspection equipment to complete its tasks; shorter paths can reduce inspection time. The number of turns measures the smoothness of the path; excessive turns can reduce the efficiency of inspection equipment and even affect its lifespan. The coverage repetition rate is used to constrain redundant coverage of the path to avoid wasted resources. Energy consumption is estimated based on the motion characteristics of the inspection equipment and the complexity of the path to ensure rational energy use.

[0099] Furthermore, for the calculation of path length, the system will dynamically track the movement trajectory of the inspection equipment, accumulate the spatial distance of each path segment, and use the distribution of grid weights in the path as an adjustment coefficient to ensure that the path can take into account both efficiency and priority. In addition, for the calculation of the number of turns, by extracting the direction change points between the path nodes, the number of direction switches in the entire path is counted, and additional weight penalties are given to sharp turning areas to encourage the generation of smoother paths. The calculation of coverage repetition rate depends on the area coverage data in the three-dimensional model. By comparing the path coverage area with the uncovered area in real time, it ensures that repeated inspections are avoided first during path evaluation. In terms of energy consumption, by introducing the speed, motion inertia and load models of the inspection equipment, the actual energy consumption of different paths in complex environments is dynamically estimated.

[0100] S1.5.2: Calculate a score for each path in the set of feasible paths according to the multi-objective evaluation function;

[0101] S1.5.3: The path with the highest score is used as the inspection path;

[0102] Preferably, when there are multiple paths with similar scores, the importance of priority targets (such as equipment weight or priority coverage of abnormal areas) is further enhanced through weight distribution adjustment to ensure that the final path better meets the task requirements. The purpose of this is to dynamically adjust the path selection strategy under various constraints to cope with different inspection scenarios, such as prioritizing inspections of high-risk equipment areas in industrial workshops, or avoiding long stays in areas with complex environments. Through this path optimization mechanism, both inspection efficiency and resource utilization can be improved, while ensuring the accuracy and reliability of inspection tasks.

[0103] S2 also includes adjusting the inspection path. The specific steps are as follows:

[0104] S2.1: When the inspection device detects an obstacle that does not belong to the obstacle grid, it updates the regular grid, generates an adjusted obstacle avoidance path, and adjusts the inspection path of the current inspection device;

[0105] Specifically, when the inspection equipment detects an unmarked obstacle in the regular grid through built-in sensors (such as lidar, ultrasonic sensors or visual sensors) during the execution of the inspection task, the regular grid will be dynamically updated in real time, the grid where the obstacle is located will be marked as an inaccessible area, and the weights of the surrounding grids will be readjusted to reflect the impact range of the obstacle on the path.

[0106] Furthermore, the inspection equipment recalculates the adjusted obstacle avoidance path by processing the path planning algorithm embedded in the equipment. When calculating the path, the A* algorithm gives priority to paths with shorter path lengths and passing through low-risk areas, while assigning higher costs to paths close to obstacles or high-risk areas, to prevent the inspection equipment from entering dangerous areas or causing lengthy paths during the obstacle avoidance process. In addition, to ensure the smoothness of the path, the algorithm will perform secondary optimization on the newly generated path to reduce the frequent turning or path deviation of the inspection equipment. When goods suddenly pile up in the workshop or temporary mobile equipment appears, the inspection equipment can quickly adjust the path to bypass the obstacle, which not only improves the inspection efficiency, but also avoids collisions or stagnation of the equipment due to path planning errors.

[0107] S2.2: When an inspection device detects inspection data that deviates from the normal level, the device priority is reordered based on the device location, type, and severity of the inspection data, and the grid weight is updated to adjust the inspection path of the next inspection device.

[0108] Specifically, during the data collection process of the inspection equipment, if it is detected that the operating parameters of the equipment deviate from the normal range (such as temperature exceeding the standard, abnormal vibration or gas leakage, etc.), the processing equipment will dynamically adjust its priority according to the severity of the deviation data, the type of equipment and the location of the equipment in the park, and update the weight of the rule grid.

[0109] Furthermore, the priority adjustment logic is based on a pre-set device weight model, which integrates the device's failure risk, inspection frequency requirements, and the persistence of abnormal conditions. For example, if a device's temperature deviates from the normal range and continues to rise, the priority adjustment logic sets that device as the primary inspection target, increases the grid weight of its location, and reduces the weight of grids containing non-critical devices. The updated grid weights are used as input, combined with the inspection device's sensing range and motion characteristics, to replan the next inspection route, prioritizing inspection of critical devices.

[0110] In this embodiment, the device weight model can be expressed as:

[0111] Q=Q R +Q F +Q S ,

[0112] Among them, Q represents the device weight, Q R represents the failure risk weight of the equipment, Q F Indicates the inspection frequency weight of the device, Q S Indicates the abnormal state weight;

[0113] Specifically, the failure risk weight primarily reflects the criticality of the equipment and the likelihood of failure. A base risk value is set based on the equipment's historical operating data, maintenance records, and equipment category. For example, critical production equipment (such as high-pressure pumps and main controllers) has a higher risk value, while auxiliary equipment (such as auxiliary monitoring instruments) has a lower risk value. When equipment operating data (such as temperature, vibration, and noise) deviates from the normal range, the failure risk weight is dynamically increased, with the extent of the increase depending on the degree of deviation.

[0114] Specifically, the inspection frequency weight reflects the device's periodic need for inspections. Each device is assigned an inspection cycle, determined by the device type, operating environment, and maintenance recommendations. For example, high-load devices may require daily inspections, while low-risk devices may require weekly inspections. The device's last inspection time is recorded, and the weight is dynamically adjusted based on the time interval. If a device exceeds the set inspection cycle, its weight is significantly increased, ensuring that it is prioritized for the next inspection.

[0115] Specifically, the abnormal status weight reflects the current operating status of the device and its impact on priority. It monitors the device's operating data (such as temperature, gas leakage concentration, and vibration frequency) in real time. When abnormal data is detected, the abnormal status weight is calculated based on the type and severity of the abnormality. The abnormal status weight is also adjusted based on the duration of the abnormality. Devices with persistent abnormalities are assigned a higher weight, while short-term or transient abnormalities may be assigned a lower weight.

[0116] In this embodiment, if the inspection equipment frequently interrupts the task to re-plan the path during the inspection process, it will cause the processing equipment to frequently calculate the new path, increasing the computing burden of the processing equipment. At the same time, the motion inertia and hardware limitations of the inspection equipment (such as the stability and energy consumption distribution of the navigation module) also determine that frequent path adjustments may cause the equipment to have a lower operating efficiency and even increase hardware wear. Therefore, applying the priority adjustment results to the next inspection can effectively maintain the consistency and stability of the current inspection task and avoid negative effects on the overall efficiency of the system due to excessive dynamic adjustments. In addition, a slight increase or short-term fluctuation in temperature may be a normal fluctuation of the equipment rather than a serious failure. Directly interrupting the current task for processing may result in unnecessary waste of resources. By applying the adjustment results to the next inspection, the processing equipment can use the time window to further verify the abnormal data (for example, through the time series analysis model to determine the persistence and severity of the abnormality), reducing the possibility of misjudgment.

[0117] The specific steps of S4 are as follows:

[0118] S4.1: Construct a time series of the equipment operation data based on the equipment operation data in the plurality of inspection data transmitted by the inspection equipment;

[0119] Specifically, the system organizes and stores the operational data (such as temperature, vibration intensity, and gas concentration) transmitted by inspection equipment in chronological order, forming multidimensional time series. Each time series contains the operational parameters of the equipment at different points in time. This structured representation of the equipment's dynamic changes provides complete historical data support for subsequent anomaly analysis.

[0120] S4.2: Calculating a time window based on the inspection time interval of the inspection equipment and the number of inspection path adjustments, and segmenting the time series based on the time window to obtain multiple window sequences;

[0121] Specifically, during inspection tasks, the time intervals and path adjustments of inspection equipment directly affect the continuity and integrity of data collection. To ensure the accuracy of data analysis results, the time window size suitable for the current inspection conditions is calculated based on the fixed time intervals and dynamic path adjustment times of the inspection equipment during inspections. The selection of the time window will take into account the sampling frequency and the response time of abnormal events. For example, for high-frequency vibration detection of equipment, a shorter time window can be used to capture subtle changes; for temperature or gas concentration monitoring, the time window can be appropriately expanded to cover more samples. This dynamic window design can adapt to the characteristics of different equipment and environments, avoiding data fragmentation caused by too small a window and reducing the loss of details caused by too large a window, thereby ensuring that the abnormal analysis results are more accurate and efficient.

[0122] S4.3: Calculate the characteristic index of each window sequence and determine whether the characteristic index is abnormal;

[0123] Specifically, within each time window, a variety of characteristic indicators are extracted, including mean, standard deviation, rate of change, autocorrelation coefficient, etc. The specific extracted features are dynamically adjusted according to the operating characteristics of the equipment. For example, for vibration data, the rate of change and standard deviation can reflect whether the equipment has abnormal vibration; for temperature data, a continuous upward trend may be a sign of equipment overheating; for gas concentration, a sudden concentration peak may indicate a gas leak. The calculation of these characteristic indicators is based on the data points within the time window and can capture short-term and long-term dynamic characteristic changes. The system makes a preliminary judgment on these indicators through preset thresholds and statistical rules. If certain characteristic indicators exceed the normal range, they are marked as potential anomalies.

[0124] S4.4: If an anomaly exists, the characteristic indicator is input into a preset anomaly detection model, and the anomaly detection model outputs the anomaly.

[0125] In this embodiment, the anomaly detection model uses a deep learning-based long short-term memory (LSTM) network architecture combined with an attention mechanism to fully leverage the dynamic characteristics of time series. The LSTM network can memorize long-term trends in device status while capturing drastic short-term changes, enabling accurate identification of anomalies. The attention mechanism assigns higher weights to key time points, such as a sudden increase in device temperature or a sudden spike in gas concentration, allowing the model to focus more computing resources on important features.

[0126] For example, during inspections of high-temperature equipment, the LSTM model can identify abnormal trends in equipment temperature by analyzing the temperature time series of the equipment. For example, when a piece of equipment's temperature gradually rises due to long-term operation, traditional static threshold judgments may not be able to promptly identify overtemperature risks. However, the LSTM model can capture the slowly rising temperature trend and, through an attention mechanism, assign higher weights to key time points (such as the point where the temperature begins to rise significantly), generating timely anomaly alerts.

[0127] For example, in a chemical park, certain equipment may leak hazardous gases due to sealing issues. Real-time gas concentration data collected by inspection equipment is fed into the anomaly detection model as a time series. When a gas leak occurs, the LSTM model can identify abnormal fluctuations in gas concentration, such as a rapid increase in concentration over a short period of time. The attention mechanism focuses resources on this abnormal change, further confirming whether the leak is persistent and outputting a leak level judgment based on different gas types. Compared to traditional fixed concentration threshold detection, this approach significantly reduces false alarms caused by accidental factors (such as environmental noise or sensor errors) and is more sensitive to sudden gas leaks.

[0128] In this embodiment, a convolutional neural network (CNN) can also be used to process multi-dimensional data with spatial correlation, such as equipment monitoring videos or infrared imaging data. By extracting image features (such as edges, textures, and local structural changes), the CNN model can accurately determine whether there are abnormalities during equipment operation. For example, in thermal imaging images collected by inspection equipment, the CNN can quickly locate abnormal components by identifying local abnormal areas of temperature distribution on the equipment surface (such as overheating areas or uneven heat dissipation points).

[0129] The specific steps of S5 are as follows:

[0130] S5.1: Determine control measures based on the abnormality type, transmit the control measures to the control device, control the area where the abnormality occurs, and collect processing data during the control process;

[0131] S5.2: After the abnormal situation is processed, update the three-dimensional model based on the processed data.

[0132] Example 2

[0133] See also Figure 2 , the present invention provides an embodiment: an intelligent management and control system for a smart park, the system comprising a three-dimensional modeling module, an inspection planning module and an abnormality response module;

[0134] The three-dimensional modeling module is used to construct a three-dimensional model of the industrial park;

[0135] The inspection planning module is used to plan the inspection path of the inspection equipment through an optimization algorithm based on the spatial information and equipment information of the three-dimensional model;

[0136] The abnormal response module is used to control the area where the abnormal situation is located through the three-dimensional model linkage with the campus security when an abnormal situation is detected.

[0137] Example 3

[0138] See also Figure 3 The present invention provides an embodiment: an intelligent management and control method for a smart park, further comprising constructing a three-dimensional model of a smart workshop, the specific steps of which are as follows:

[0139] S01: Use the image sensor carried by the drone to collect multi-angle images of the smart workshop, and use the feature matching algorithm to stitch the multi-angle images to generate a continuous image;

[0140] In this step, drones are used as mobile acquisition platforms, equipped with high-resolution image sensors, to capture full-coverage aerial images of the target area from various heights and angles. This multi-angle image acquisition ensures the integrity of the target area and captures complex structures without omission, making it particularly suitable for scenarios such as smart workshops with densely packed equipment and complex environments. The collected multi-angle images are then used to extract key feature points from each image using a feature matching algorithm (such as the ORB algorithm or the SIFT algorithm). These images are then aligned and integrated using a stitching algorithm to generate a consistent and continuous panoramic image.

[0141] S02: Generate three-dimensional point cloud data of the smart workshop using structured light technology based on the continuous images, process the three-dimensional point cloud data, and extract spatial information;

[0142] Specifically, the system projects multiple grating patterns across the workshop and combines them with the depth information of the image to reconstruct the 3D spatial structure of the target area. The resulting 3D point cloud data undergoes filtering (such as voxel filtering and boundary noise reduction) to extract key spatial information, including the equipment's geometric characteristics, location distribution, and relative relationship to the surrounding environment.

[0143] S03: inputting the continuous image into a preset recognition network, and outputting device information through the recognition network;

[0144] In this step, continuous images are fed into a deep learning-based recognition network to extract classification information and location features of equipment within the workshop. This recognition network, which uses a residual network as its backbone, expands the feature dimensions of the input images through bilinear interpolation. Deep features are used to extract semantic information about the equipment, while shallow features are used to recover edge details. Ultimately, the device recognition layer generates semantic segmentation results for the equipment area, and a classifier outputs the category and specifications of each device.

[0145] S04: Constructing a three-dimensional model according to the space information and the device information.

[0146] Specifically, by fusing the 3D point cloud data generated by structured light with the device information output by the recognition network, a complete 3D model of the smart workshop is constructed. This 3D model contains the spatial location, geometric features, and semantic attribute information of the equipment, and is stored in a hierarchical structure on the industrial server to facilitate subsequent inspection route planning and anomaly detection.

[0147] See also Figure 4 , an embodiment of the present invention identifies a network structure diagram, wherein the identification network includes:

[0148] The residual layer is used to expand the plane dimension of the input parameters through bilinear interpolation, and uses the residual network as the backbone network to extract deep features and shallow features, wherein the deep features are used to extract semantic information and the shallow features are used to restore boundary detail information;

[0149] The input image is preprocessed using bilinear interpolation to expand its planar dimensions, enabling the feature extraction process to better preserve detailed information within the input image. Furthermore, a residual network (ResNet) is used as the backbone architecture to extract multi-layer features. Deep features utilize multiple layers of convolution and nonlinear activation functions to extract global semantic information from the image. These features are primarily used to identify the overall distribution patterns and shape characteristics of the workshop area. Shallow features, through fewer convolution operations, preserve image boundaries and texture information, enabling the restoration of detailed workshop area boundaries.

[0150] The device identification layer is used to fuse the deep features and shallow features to generate a feature map, process the feature map through a convolution kernel and an attention mechanism to extract regional semantic information, calculate label weights based on the regional semantic information, classify the label weights through a Softmax classifier, and determine the category information of the device;

[0151] Deep and shallow features are combined through pixel-by-pixel weighted addition to generate a feature map that captures global semantic information while preserving boundary details. Multiple sets of convolution kernels are applied to the fused feature map to further extract spatial correlation information. Convolution kernel sizes vary, including 1×1, 3×3, and 5×5, to capture spatial features at varying scales. Channel-wise and spatial-wise attention mechanisms are used to dynamically adjust the response values of important regions within the feature map, focusing on key features within the device region and minimizing interference from non-target areas.

[0152] The output layer is used to segment the device area based on the classification results of the device identification layer, perform pixel-level statistics on the device area, calculate the device coverage area and distribution characteristics, and generate device information.

[0153] Although the embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and are not to be construed as limitations on the present invention. A person skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. An intelligent management and control method for a smart park, wherein the smart park includes multiple smart workshops, each of which is equipped with an inspection device for inspection. The inspection device is in communication with a processing device, and the intelligent management and control method is executed by the processing device, characterized in that: The intelligent management and control method includes: Generate an inspection path based on the spatial data and equipment data of the three-dimensional model of the smart workshop to be controlled, wherein the inspection path determines a regular grid through the spatial data and assigns grid weights to the regular grid through the equipment data; Sending the inspection path to the inspection device, wherein the inspection device inspects the inspection path, obtains inspection data, and sends the inspection data to the processing device; Receive the inspection data, detect whether there is any abnormal situation, and control the corresponding control equipment to control the area where the abnormal situation is located.

2. The intelligent management and control method for a smart park according to claim 1, characterized in that: Generating the inspection path includes: Dividing the smart workshop into regular grids according to the spatial data, and assigning grid weights according to the equipment data; Setting an inspection target set, and mapping the inspection target set to the regular grid; Generate a minimum circumscribed polygon covering the inspection target set, and generate multiple parallel inspection paths at fixed intervals according to the main axis direction of the minimum circumscribed polygon; Detecting the parallel inspection paths to obtain a set of feasible paths; The feasible path set is evaluated to obtain an inspection path.

3. The intelligent management and control method for a smart park according to claim 2, characterized in that: Detecting the parallel inspection path includes: Performing obstacle detection on the parallel inspection path; If the path overlaps with the obstacle grid, the A* algorithm is used to replan the path in the obstacle grid area, reconnecting the starting and ending points of the path after bypassing the obstacle. The A* algorithm generates a local obstacle avoidance path based on the environmental constraint parameters in the device data. If there is no barrier grid, the path is added to the feasible path set, and the traversal of the paths is repeated until all paths are added to the feasible path set.

4. The intelligent management and control method for a smart park according to claim 2, characterized in that: The set of feasible paths is evaluated, including: Establishing a multi-objective evaluation function, wherein the objectives include grid weight, path length, number of turns, coverage repetition rate and energy consumption; Calculating a score for each path in the set of feasible paths according to the multi-objective evaluation function; The path corresponding to the highest score is used as the inspection path.

5. The intelligent management and control method for a smart park according to claim 1, characterized in that: When the inspection device is inspecting on the inspection path, the inspection device further includes: When the inspection device detects an obstacle that does not belong to the obstacle grid, it updates the regular grid and adjusts the inspection path of the current inspection device; When the inspection equipment detects detection data that deviates from the normal level, the equipment priority is reordered, the grid weight is updated, and the next inspection path of the inspection equipment is adjusted.

6. The intelligent management and control method for a smart park according to claim 1, characterized in that: The detecting whether there is an abnormality includes: Constructing a time series of the equipment operation data based on the equipment operation data in the multiple inspection data transmitted by the inspection equipment; Calculating a time window based on the inspection time interval of the inspection equipment and the number of inspection path adjustments, and segmenting the time series based on the time window to obtain multiple window sequences; Calculate the characteristic index of each window sequence and determine whether the characteristic index is abnormal; If an anomaly exists, the characteristic indicator is input into a preset anomaly detection model, and the anomaly is output through the anomaly detection model.

7. The intelligent management and control method for a smart park according to claim 6, characterized in that: The control device corresponding to the control controls the area where the abnormal situation occurs, including: Displaying the abnormality type corresponding to the abnormal situation in the three-dimensional model; Determine control measures based on the abnormality type, transmit the control measures to the control equipment, control the area where the abnormality occurs, and collect processing data during the control process; After the abnormal situation is processed, the 3D model is updated based on the processed data.

8. The intelligent management and control method for a smart park according to claim 1, characterized in that: The intelligent management and control method further includes constructing a three-dimensional model of the smart workshop, wherein constructing the three-dimensional model of the smart workshop includes: The image sensor carried by the drone collects multi-angle images of the smart workshop, and the multi-angle images are stitched together using a feature matching algorithm to generate a continuous image; Based on the continuous images, three-dimensional point cloud data of the smart workshop is generated using structured light technology, and the three-dimensional point cloud data is processed to extract spatial information; Inputting the continuous image into a preset recognition network, and outputting device information through the recognition network; A three-dimensional model is constructed according to the space information and the device information.

9. An intelligent management and control system for a smart park, implemented based on the intelligent management and control method for a smart park according to any one of claims 1 to 8, characterized in that: The system includes a three-dimensional modeling module, an inspection planning module and an abnormal response module; The three-dimensional modeling module is used to build a three-dimensional model of the smart workshop; The inspection planning module is configured to generate an inspection path based on the spatial data and equipment data of the three-dimensional model, wherein the inspection path is a regular grid determined by the spatial data and a grid weight is assigned to the regular grid by the equipment data; The abnormal response module is used to receive inspection data, detect whether there is an abnormal situation, and control the corresponding control equipment to control the area where the abnormal situation occurs.

Images