Unmanned aerial vehicle public safety intelligent inspection early warning method and system

Abstract

The invention relates to the technical field of unmanned aerial vehicles, in particular to an unmanned aerial vehicle public safety intelligent inspection early warning method and system. The method comprises the following steps: collecting inspection data based on training parameters, creating an early warning model, inputting preprocessed data into the early warning model, generating an inspection route of a target unmanned aerial vehicle through the early warning model, judging whether an abnormal condition exists in combination with multi-source data on the inspection route, and giving out early warning based on the severity of the abnormal condition. And finally, inputting the real-time data into the trained early warning model, and judging whether the real-time inspection route has a real-time abnormal condition, the abnormal degrees of the key targets are divided according to the abnormal values, and the abnormal degrees of the inspection route are calculated by combining the abnormal degrees of all the key targets. Therefore, the reliability of abnormity judgment of the inspection road is improved, the abnormity can be found, the danger level can be quantified, and a differentiated early warning mode is executed according to the danger level, so that a manager can deal with the most critical event preferentially.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) technology, and more specifically, to a method and system for intelligent inspection and early warning of public safety using UAVs. Background Technology

[0002] Drone inspection is a technology that uses drones equipped with various sensors to conduct automated inspections, data collection and analysis of target areas or facilities. It is profoundly changing the traditional inspection modes in many fields such as power grids, transportation, environmental protection, and urban management by improving efficiency, ensuring safety and realizing intelligence.

[0003] Chinese Patent Publication No. CN118692166A discloses a drone inspection system and method, relating to the field of drone inspection technology. The system includes a tower area division module, a monitoring data determination module, an inspection equipment screening module, an inspection equipment analysis module, a drone screening module, a tower status analysis module, and an early warning terminal. It divides the area into normal zones and key zones, then screens the corresponding inspection equipment for each zone, predicts the inspection time and energy consumption of the equipment for each zone, and selects the appropriate drones for inspection. After the drone inspection is completed, the system assesses the status of the power tower to ensure its safety. However, existing technologies do not consider key targets along the inspection route, leading to potential errors in judging abnormal conditions along the user's inspection route. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for intelligent inspection and early warning of public safety using unmanned aerial vehicles (UAVs), in order to solve the problems mentioned in the background art.

[0005] To address the aforementioned technical problems, one objective of this invention is to provide a method for intelligent inspection and early warning of public safety using unmanned aerial vehicles (UAVs), comprising: Set the inspection parameters for the target drone, collect inspection data based on the training parameters, and preprocess the inspection data to obtain multiple preprocessed data. Create an early warning model; The preprocessed data is input into the early warning model, which generates the inspection route of the target UAV. The model combines multi-source data on the inspection route to determine whether there are any abnormalities and issues an early warning based on the severity of the abnormalities, thus obtaining the trained early warning model. Real-time data from drones is collected and input into a trained early warning model to determine if any real-time anomalies exist along the inspection route.

[0006] Preferably, the process of setting the inspection parameters of the target UAV, collecting inspection data based on training parameters, and preprocessing the inspection data to obtain multiple preprocessed data includes the following steps: Set up a drone database; For each target UAV, corresponding data acquisition parameters are set; these parameters include the inspection area, key targets, and inspection altitude. Inspection data for each target UAV is collected based on the acquisition parameters, and the collected inspection data is input into the UAV database; the inspection data includes image data and information data; Randomly select inspection parameters for a target drone from the drone inspection database; Determine if there is any missing data in the inspection parameters of the target drone; If there are missing data in the inspection parameters of the target drone, the missing parameters are filled in based on the mean. Return the inspection parameters of a target drone randomly selected from the drone inspection database, and continue until the inspection parameters of all target drones in the drone database have been selected, resulting in multiple preprocessed data.

[0007] Preferably, the step of inputting preprocessed data into the early warning model, generating an inspection route for the target UAV through the early warning model, determining whether there are any anomalies by combining multi-source data on the inspection route, and issuing an early warning based on the severity of the anomalies, thereby obtaining the trained early warning model, includes the following steps: All preprocessed data are divided into training and test sets according to a random ratio; The training set is input into the early warning model, and the early warning model creates a corresponding inspection route for each target drone. The preprocessed data is then used to determine whether there are any abnormalities on the inspection route. If there are abnormalities on the inspection route, the severity of the abnormality is determined, and the degree of abnormality of the inspection route is calculated based on the degree of abnormality of all key targets on the inspection route. The corresponding early warning method is executed based on the severity to obtain the trained early warning model. Input the test set into the trained early warning model to verify whether the early warning model has been trained successfully.

[0008] Preferably, the step of inputting the training set into the early warning model, creating a corresponding inspection route for each target UAV through the early warning model, and determining whether there are any abnormalities on the inspection route based on the preprocessed data includes the following steps: Preprocessed data of a target UAV is randomly selected from the training set; The inspection route and key targets of the target UAV are obtained from the preprocessed data, and the key targets are marked on the inspection route. For each key objective, calculate the outlier for each key objective separately; Return the preprocessed data of a target drone randomly selected from the training set, until the preprocessed data of all target drones in the training set have been selected, and obtain the anomaly level corresponding to each target drone.

[0009] Preferably, if an anomaly exists on the inspection route, the severity of the anomaly is determined, and the anomaly severity of the inspection route is calculated based on the anomaly severity of all key targets on the inspection route. Based on the severity, a corresponding early warning method is executed to obtain the trained early warning model, including the following steps: Set multiple levels of anomaly and the corresponding anomaly value for each level of anomaly; Each key objective is assigned an anomaly level based on outliers; The degree of abnormality of each inspection route is calculated based on the degree of abnormality of all key targets along that route. Set an anomaly threshold and determine whether the anomaly level of the inspection route is greater than or equal to the anomaly threshold. If the degree of abnormality of the inspection route is greater than or equal to the abnormality threshold, the inspection route is recorded as an abnormal route.

[0010] Preferably, the step of collecting real-time data from the drone and inputting the real-time data into the trained early warning model to determine whether there are any real-time anomalies on the real-time inspection route includes the following steps: Collect real-time inspection data from drones; the real-time inspection data includes real-time image data and real-time information data; The real-time inspection data is input into the trained early warning model, and the training early warning model is used to calculate the degree of anomaly of the real-time key targets and the degree of anomaly of the real-time inspection route of the real-time UAV. The degree of abnormality of the real-time inspection route is used to determine whether an anomaly has occurred on the real-time inspection route.

[0011] Preferably, determining whether an anomaly has occurred in the real-time inspection route based on the degree of anomaly includes the following steps: Cluster analysis is performed on the real-time inspection routes to obtain the inspection routes corresponding to the real-time inspection routes; the inspection routes corresponding to the real-time inspection routes are recorded as the target routes. Set route anomaly thresholds; Calculate the difference between the anomaly level of the real-time inspection route and the anomaly level of the target route, and determine whether the difference between the anomaly level of the real-time inspection route and the anomaly level of the target route is greater than or equal to the route anomaly threshold. If the difference between the abnormality level of the real-time inspection route and the abnormality level of the target route is greater than or equal to the route abnormality threshold, then the real-time inspection route is a real-time abnormal route.

[0012] On the other hand, this application also provides an intelligent inspection and early warning system for public safety using unmanned aerial vehicles (UAVs), applied to the intelligent inspection and early warning method for public safety using UAVs as described in any of the preceding statements. The system includes a target UAV, a data acquisition component, and an early warning component. The data acquisition component is installed on the target UAV and collects inspection data generated by the target UAV during the inspection process. The early warning component is communicatively connected to the data acquisition component. All data information collected by the data acquisition component is input to the early warning component. The early warning component combines multi-source data along the inspection route to determine whether there are any abnormalities and issues an early warning based on the severity of the abnormality.

[0013] Preferably, the early warning component includes a processor and a storage unit, wherein the storage unit stores inspection data and the processor executes the UAV public safety intelligent inspection and early warning method.

[0014] Preferably, the acquisition component includes an image acquisition module and an information acquisition module. The image acquisition module acquires inspection images of the inspection route, and the information acquisition module acquires inspection information of the inspection route.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: By setting inspection parameters for the target drone, collecting inspection data based on training parameters, and preprocessing the inspection data to obtain multiple preprocessed data sets, an early warning model is created. The preprocessed data is then input into the early warning model, which generates the inspection route for the target drone. The model combines multi-source data along the inspection route to determine if any anomalies exist and issues an early warning based on the severity of the anomalies, resulting in a trained early warning model. Finally, real-time data from the drone is collected and input into the trained early warning model to determine if any real-time anomalies exist along the inspection route. This application classifies the degree of anomaly of key targets based on outliers and calculates the degree of anomaly of the inspection route by combining the anomalies of all key targets. This improves the reliability of anomaly detection for inspection routes, enabling the identification of anomalies and quantification of their hazard levels. Differentiated early warning methods are then implemented, allowing managers to prioritize the most critical events and significantly optimizing emergency response procedures. Attached Figure Description

[0016] Figure 1 A flowchart illustrating the intelligent patrol and early warning method for public safety using drones; Figure 2 A schematic diagram of the framework of an unmanned aerial vehicle (UAV) intelligent inspection and early warning system for public safety. Reference numerals: 100, Target UAV; 200, Acquisition Components; 201, Image Acquisition Module; 202. Information acquisition module; 300. Early warning component; 301. Processor; 302. Storage unit. Detailed Implementation

[0017] 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 some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Example 1

[0018] like Figure 1 As shown, one of the objectives of this invention is to provide a method for intelligent inspection and early warning of public safety using unmanned aerial vehicles (UAVs), comprising: S100: Set the inspection parameters of the target UAV, collect inspection data based on the training parameters, and preprocess the inspection data to obtain multiple preprocessed data. S200, create an early warning model; S300 inputs preprocessed data into the early warning model, generates the inspection route of the target UAV through the early warning model, and judges whether there are any abnormalities by combining the multi-source data on the inspection route, and issues an early warning based on the severity of the abnormality, thus obtaining the trained early warning model. The S400 collects real-time data from the drone and inputs it into the trained early warning model to determine whether there are any real-time anomalies along the inspection route.

[0019] It should be noted that by setting the inspection parameters of the target UAV, collecting inspection data based on the training parameters, and preprocessing the inspection data to obtain multiple preprocessed data, an early warning model is created. The preprocessed data is then input into the early warning model, which generates the inspection route of the target UAV. The model combines multi-source data on the inspection route to determine whether there are any anomalies and issues an early warning based on the severity of the anomalies, resulting in a trained early warning model. Finally, real-time data from the UAV is collected and input into the trained early warning model to determine whether there are any real-time anomalies on the real-time inspection route. This application classifies the degree of anomaly of key targets based on outliers and calculates the degree of anomaly of the inspection route by combining the degree of anomalies of all key targets. This improves the reliability of anomaly judgment on inspection routes, enables the detection of anomalies, quantifies their danger level, and implements differentiated early warning methods accordingly. This allows managers to prioritize the handling of the most critical events and greatly optimizes the emergency response process.

[0020] In one embodiment of this application, S100 includes: S110, set up the drone database; S120, set corresponding acquisition parameters for each target UAV; the acquisition parameters include the inspection area, key targets, and inspection altitude; S130: Collect inspection data for each target UAV based on the acquisition parameters, and input the collected inspection data into the UAV database; the inspection data includes image data and information data; S140, randomly select inspection parameters for a target drone from the drone inspection database; S150, determine whether there is missing data in the inspection parameters of the target drone; S160, If there are missing data in the inspection parameters of the target UAV, fill in the missing parameters based on the mean. S170 returns the inspection parameters of a target drone randomly selected from the drone inspection database, until the inspection parameters of all target drones in the drone database have been selected, resulting in multiple preprocessed data.

[0021] It should be noted that by setting collection parameters and preprocessing inspection data, the accuracy and efficiency of inspections can be improved. The early warning model trained in this way can intelligently identify and warn of abnormal public safety situations, thereby improving the response speed to emergencies.

[0022] First, the drone's inspection parameters are set according to different inspection scenarios, such as inspection area, key targets, and inspection altitude. Then, the drone collects data according to these parameters and preprocesses the collected image, environmental, and audio data. Next, this preprocessed data is used to train an early warning model, enabling it to identify abnormal situations such as fires and gatherings of people, and issue corresponding warnings based on the severity of the anomaly.

[0023] By setting up a unified UAV database and collection parameters, the standardization and consistency of data sources were ensured. In particular, the strategy of using mean imputation for missing data is simple and efficient, and can ensure the integrity of the dataset without introducing excessive bias. This lays a solid data foundation for the stable training and accurate prediction of the subsequent early warning model.

[0024] In one embodiment of this application, S300 includes: S310, divide all preprocessed data into training and test sets according to a random ratio; S320 inputs the training set into the early warning model, which then creates a corresponding inspection route for each target UAV and uses preprocessed data to determine whether there are any anomalies on the inspection route. S330, If there are abnormalities on the inspection route, the severity of the abnormality is determined, and the degree of abnormality of the inspection route is calculated based on the degree of abnormality of all key targets on the inspection route. The corresponding early warning method is executed based on the severity to obtain the trained early warning model. S340: Input the test set into the trained early warning model to verify whether the trained early warning model has been successfully trained.

[0025] It should be noted that when dividing the training set and the test set, the proportion of the training set should be greater than that of the test set to ensure that there are enough training samples in the training set.

[0026] After the training set is divided, the training samples in the training set are input into the early warning model. The early warning model creates inspection routes and identifies abnormal situations, such as analyzing images and environmental data to determine if there is a fire. If an anomaly is detected, the system will execute the corresponding early warning method according to the severity of the anomaly, thereby effectively identifying abnormal situations on the inspection route and issuing early warnings based on the severity of the anomaly, improving the accuracy and practicality of the early warning. By integrating a series of complex operations such as "generating routes", "judging anomalies", "assessing severity" and "executing early warnings" into the model training process, the model can directly learn the end-to-end mapping relationship from raw data to the final early warning decision, realizing the automation and intelligence of the decision-making process.

[0027] After obtaining the trained early warning model, the training samples in the test set are input into the trained early warning model, and the response time / accuracy is used as the criterion for judging whether the early warning model has been trained successfully.

[0028] In one embodiment of this application, S320 includes: S321, randomly select preprocessed data of a target UAV from the training set; S322: Obtain the inspection route and key targets of the target UAV from the preprocessed data, and mark the key targets on the inspection route; S323, For each key target, calculate the outlier for each key target separately; S324 returns the preprocessed data of a target drone randomly selected from the training set, until the preprocessed data of all target drones in the training set have been selected, and obtains the anomaly level corresponding to each target drone.

[0029] It should be noted that outliers for key targets are calculated using Formula 1. Formula 1; in, Before calculation, each sub-data needs to be normalized to unify the contribution of each sub-data to the outlier, making the outlier more reliable. The outlier is actually a value used to describe the possible abnormal situation of the key target. That is, the larger the outlier of the key target, the more likely it is to have an abnormal situation. This quantifies the probability of the abnormal situation, so that users can more intuitively understand the probability of the key target having an abnormal situation.

[0030] By breaking down the assessment of the entire inspection route into the calculation of outliers for each "key target" along the route, this "divide and conquer" analysis method not only determines whether there is a problem with the entire route, but also pinpoints the specific location of the problem (i.e., which key target is abnormal). This provides extremely precise information guidance for subsequent emergency response and greatly shortens the response time.

[0031] In one embodiment of this application, S330 includes: S331, set multiple anomaly levels and the anomaly value corresponding to each anomaly level; S332, set the degree of abnormality for each key target based on outliers; S333, calculate the degree of abnormality of the inspection route based on the degree of abnormality of all key targets along each inspection route; S334, Set an anomaly threshold and determine whether the anomaly level of the inspection route is greater than or equal to the anomaly threshold; S335, if the degree of abnormality of the inspection route is greater than or equal to the degree of abnormality threshold, then the inspection route is recorded as an abnormal route.

[0032] It should be noted that since the larger the outlier of a key target, the more likely it is to experience an anomaly, the larger the outlier of a key target, the more severe the anomaly. Once the anomaly of all key targets on the inspection route is obtained, the anomaly of the inspection route can be calculated using Formula 2. Formula 2; in, Since the inspection route is derived based on the anomaly levels of all key targets along it, the anomaly level of the inspection route also reflects the probability of abnormal situations occurring on that route. This, combined with anomaly level thresholds, allows for a judgment on whether the inspection route is normal. Generally, the higher the anomaly level threshold, the easier it is to judge the inspection route; conversely, the lower the threshold, the stricter the judgment. When the early warning model identifies an anomaly on the inspection route, the system determines whether to issue an early warning based on the preset anomaly level threshold. If the anomaly level exceeds the threshold, the system marks the inspection route as abnormal and triggers the corresponding early warning mechanism.

[0033] By establishing a quantitative and hierarchical anomaly assessment system and setting clear anomaly levels and corresponding anomaly values, the system is able to digitally measure abstract "abnormal situations" for the first time. Based on this, by calculating the overall anomaly level of the entire inspection route and comparing it with preset thresholds, a leap from "qualitative judgment" to "quantitative decision-making" has been achieved. This makes the system's judgment results more objective.

[0034] In one embodiment of this application, S400 includes: S410, collects real-time inspection data from the drone; the real-time inspection data includes real-time image data and real-time information data; S420 inputs real-time inspection data into the trained early warning model, and calculates the anomaly level of the real-time key targets and the anomaly level of the real-time inspection route of the real-time UAV through the trained early warning model. S430 determines whether an anomaly has occurred on the real-time inspection route based on the degree of anomaly in the real-time inspection route.

[0035] It should be noted that by inputting real-time inspection data into the trained early warning model, the degree of anomaly of the real-time inspection route can be obtained. In order to improve the reliability of the anomaly degree, the real-time inspection route is further judged by combining the degree of anomaly of the real-time inspection route, so as to promptly detect and deal with abnormal situations and improve the level of public safety protection.

[0036] By successfully applying the well-trained intelligent model to real-world business scenarios, a leap from theory to practice has been achieved. Combining real-time collected multimodal data with the trained model enables immediate diagnosis of the current inspection status.

[0037] In one embodiment of this application, S430 includes: S431, perform cluster analysis on the real-time inspection route to obtain the inspection route corresponding to the real-time inspection route; record the inspection route corresponding to the real-time inspection route as the target route. S432, Set route anomaly threshold; S433, calculate the difference between the abnormality level of the real-time inspection route and the abnormality level of the target route, and determine whether the difference between the abnormality level of the real-time inspection route and the abnormality level of the target route is greater than or equal to the route abnormality threshold. S434, if the difference between the abnormality level of the real-time inspection route and the abnormality level of the target route is greater than or equal to the route abnormality threshold, then the real-time inspection route is a real-time abnormal route.

[0038] It's worth noting that the system innovatively incorporates cluster analysis and a comparison mechanism with the "target route." Even if individual real-time data may contain noise or transient interference, comparing the degree of anomaly with historical normal patterns (target route) effectively filters out sporadic fluctuations and identifies truly abnormal and persistent risk trends. This significantly reduces the false alarm rate and improves the reliability of the system's early warnings.

[0039] like Figure 2 As shown, in one embodiment of this application, a drone public safety intelligent inspection and early warning system is also provided, applied to the drone public safety intelligent inspection and early warning method described in any of the foregoing. The system includes a target drone 100, a data acquisition component 200, and an early warning component 300. The data acquisition component 200 is disposed on the target drone 100 and collects inspection data generated by the target drone 100 during the inspection process. The early warning component 300 is communicatively connected to the data acquisition component 200. All data information collected by the data acquisition component 200 is input to the early warning component 300. The early warning component 300 combines multi-source data along the inspection route to determine whether there are any abnormalities and issues an early warning based on the severity of the abnormality.

[0040] It should be noted that the organic integration of software algorithms and physical devices forms a complete and deployable solution. The system clearly defines the collaborative working architecture of "target UAV 100 - data acquisition component 200 - early warning component 300", highlighting the data flow and processing, reflecting the systematic nature and completeness of the invention, meeting the requirements of technical solution feasibility, and achieving high-precision identification of various public safety anomalies by combining multi-source data fusion technology.

[0041] In one embodiment of this application, the early warning component 300 includes a processor 301 and a storage unit 302. The storage unit 302 stores inspection data, and the processor 301 executes the UAV public safety intelligent inspection and early warning method.

[0042] It should be noted that the processor 301 in this application uses an embedded AI chip (such as NVIDIA Jetson Xavier NX), which has powerful parallel computing capabilities and supports real-time image recognition and data fusion analysis. The storage unit 302 is equipped with ≥128GB of high-speed flash memory, which is used to cache the collected images, audio, environmental data and recognition results, and supports local data storage and cloud synchronization.

[0043] In one embodiment of this application, the acquisition component 200 includes an image acquisition module 201 and an information acquisition module 202. The image acquisition module 201 acquires inspection images of the inspection route, and the information acquisition module 202 acquires inspection information of the inspection route.

[0044] It should be noted that by integrating the image acquisition module 201 and the information acquisition module 202, comprehensive image and information data along the inspection route can be collected, improving the comprehensiveness and accuracy of abnormal situation identification. The image acquisition module 201 of this application consists of a high-definition infrared dual-light camera and a panoramic camera. The high-definition camera (resolution ≥ 4K) is used for detailed shooting in visible light scenes during the day, the infrared camera (detection distance ≥ 500m) is used for thermal imaging acquisition in nighttime or low visibility (rain, fog, dense smoke) environments, and the panoramic camera (field of view ≥ 180°) is used for large-area scene coverage. The camera supports 360° rotation and 10x optical zoom, and can accurately capture target details.

[0045] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for intelligent inspection and early warning of public safety using unmanned aerial vehicles (UAVs), characterized in that, include: Set the inspection parameters for the target drone, collect inspection data based on the training parameters, and preprocess the inspection data to obtain multiple preprocessed data. Create an early warning model; The preprocessed data is input into the early warning model, which generates the inspection route of the target UAV. The model combines multi-source data on the inspection route to determine whether there are any abnormalities and issues an early warning based on the severity of the abnormalities, thus obtaining the trained early warning model. Real-time data from drones is collected and input into a trained early warning model to determine if any real-time anomalies exist along the inspection route.

2. The unmanned aerial vehicle (UAV) intelligent inspection and early warning method for public safety according to claim 1, characterized in that: The process of setting the inspection parameters for the target UAV, collecting inspection data based on training parameters, and preprocessing the inspection data to obtain multiple preprocessed data sets includes the following steps: Set up a drone database; For each target UAV, corresponding data acquisition parameters are set; these parameters include the inspection area, key targets, and inspection altitude. Inspection data for each target UAV is collected based on the acquisition parameters, and the collected inspection data is input into the UAV database; the inspection data includes image data and information data; Randomly select inspection parameters for a target drone from the drone inspection database; Determine if there is any missing data in the inspection parameters of the target drone; If there are missing data in the inspection parameters of the target drone, the missing parameters are filled in based on the mean. Return the inspection parameters of a target drone randomly selected from the drone inspection database, and continue until the inspection parameters of all target drones in the drone database have been selected, resulting in multiple preprocessed data.

3. The intelligent inspection and early warning method for public safety using unmanned aerial vehicles (UAVs) according to claim 2, characterized in that: The process of inputting preprocessed data into the early warning model, generating an inspection route for the target UAV through the early warning model, determining whether there are any anomalies by combining multi-source data on the inspection route, and issuing an early warning based on the severity of the anomalies, to obtain the trained early warning model, includes the following steps: All preprocessed data are divided into training and test sets according to a random ratio; The training set is input into the early warning model, and the early warning model creates a corresponding inspection route for each target drone. The preprocessed data is then used to determine whether there are any abnormalities on the inspection route. If there are abnormalities on the inspection route, the severity of the abnormality is determined, and the degree of abnormality of the inspection route is calculated based on the degree of abnormality of all key targets on the inspection route. The corresponding early warning method is executed based on the severity to obtain the trained early warning model. Input the test set into the trained early warning model to verify whether the early warning model has been trained successfully.

4. The unmanned aerial vehicle (UAV) intelligent inspection and early warning method for public safety according to claim 3, characterized in that: The process of inputting the training set into the early warning model, creating a corresponding inspection route for each target UAV through the early warning model, and determining whether there are any anomalies on the inspection route based on the preprocessed data includes the following steps: Preprocessed data of a target UAV is randomly selected from the training set; The inspection route and key targets of the target UAV are obtained from the preprocessed data, and the key targets are marked on the inspection route. For each key objective, calculate the outlier for each key objective separately; Return the preprocessed data of a target drone randomly selected from the training set, until the preprocessed data of all target drones in the training set have been selected, and obtain the anomaly level corresponding to each target drone.

5. The unmanned aerial vehicle (UAV) intelligent inspection and early warning method for public safety according to claim 4, characterized in that: If an anomaly is found on the inspection route, the severity of the anomaly is determined, and the anomaly severity of the inspection route is calculated based on the anomaly severity of all key targets on the inspection route. The corresponding early warning method is then executed based on the severity, resulting in a trained early warning model. This process includes the following steps: Set multiple levels of anomaly and the corresponding anomaly value for each level of anomaly; Each key objective is assigned an anomaly level based on outliers; The degree of abnormality of each inspection route is calculated based on the degree of abnormality of all key targets along that route. Set an anomaly threshold and determine whether the anomaly level of the inspection route is greater than or equal to the anomaly threshold. If the degree of abnormality of the inspection route is greater than or equal to the abnormality threshold, the inspection route is recorded as an abnormal route.

6. The unmanned aerial vehicle (UAV) intelligent inspection and early warning method for public safety according to claim 5, characterized in that: The process of collecting real-time data from the drone and inputting it into the trained early warning model to determine whether there are any real-time anomalies along the inspection route includes the following steps: Collect real-time inspection data from drones; The real-time inspection data includes real-time image data and real-time information data; The real-time inspection data is input into the trained early warning model, and the training early warning model is used to calculate the degree of anomaly of the real-time key targets and the degree of anomaly of the real-time inspection route of the real-time UAV. The degree of abnormality of the real-time inspection route is used to determine whether an anomaly has occurred on the real-time inspection route.

7. The unmanned aerial vehicle (UAV) intelligent inspection and early warning method for public safety according to claim 6, characterized in that: The process of determining whether a real-time inspection route has experienced an anomaly, based on the degree of anomaly in the real-time inspection route, includes the following steps: Cluster analysis is performed on the real-time inspection routes to obtain the inspection routes corresponding to the real-time inspection routes. Record the inspection route corresponding to the real-time inspection route as the target route; Set route anomaly thresholds; Calculate the difference between the anomaly level of the real-time inspection route and the anomaly level of the target route, and determine whether the difference between the anomaly level of the real-time inspection route and the anomaly level of the target route is greater than or equal to the route anomaly threshold. If the difference between the abnormality level of the real-time inspection route and the abnormality level of the target route is greater than or equal to the route abnormality threshold, then the real-time inspection route is a real-time abnormal route.

8. A drone-based intelligent inspection and early warning system for public safety, applied to any one of claims 1 to 7, characterized in that, include: Target drone; Data acquisition components; The data acquisition component is installed on the target drone, and the data acquisition component is used to collect the inspection data generated by the target drone during the inspection process. Early warning components; The early warning component is communicatively connected to the data acquisition component. All data information collected by the data acquisition component is input to the early warning component. The early warning component combines multi-source data on the inspection route to determine whether there is an abnormality and issues an early warning based on the severity of the abnormality.

9. The unmanned aerial vehicle (UAV) public safety intelligent inspection and early warning system according to claim 8, characterized in that, The early warning component includes a processor and a storage unit. The storage unit stores inspection data, and the processor executes the UAV public safety intelligent inspection and early warning method.

10. The unmanned aerial vehicle (UAV) public safety intelligent patrol and early warning system according to claim 9, characterized in that, The acquisition component includes an image acquisition module and an information acquisition module. The image acquisition module acquires inspection images of the inspection route, and the information acquisition module acquires inspection information of the inspection route.

Images