A drone patrol management system and method based on edge-cloud collaboration

Abstract

This invention discloses an edge-cloud collaborative drone patrol management system and method, relating to the field of drone patrol management technology. The system includes: acquiring patrol area division information, historical drone patrol data, and drone equipment parameter information; analyzing self-interference and related interference from different areas on drone patrols; comprehensively judging the overall interference from different areas on drones; marking key areas; locating drones; identifying a first and a second drone to be observed; acquiring partial patrol routes of the two drones; creating buffer zones for partial patrol routes; determining whether mutual interference will occur between the two drones during patrols in key areas; selectively adjusting the drone patrol routes based on the judgment results; controlling the drones to detect targets according to the patrol routes; and coordinating with ground-based gimbals to drive away targets, thereby reducing the risk of mutual interference between drones in specific areas during patrols.

Description

Technical Field

[0001] This invention relates to the field of drone patrol management technology, specifically a drone patrol management system and method based on edge-cloud collaboration. Background Technology

[0002] To improve the efficiency of target monitoring, existing technologies typically utilize drones to patrol areas and identify targets. The drones' pods are equipped with high-definition cameras or infrared devices, which can capture target dynamics in real time. When coordinated with cloud platforms and ground-based gimbals, the cloud platform analyzes the collected target data and coordinates the drones and ground-based gimbals to drive away targets. The drones and ground-based gimbals can also communicate through self-organizing networks to ensure reliable real-time response and linkage. By achieving a closed loop of detection, identification, and drive-away through cloud-edge collaboration, the efficiency of target detection can be effectively improved.

[0003] However, if multiple drones patrol a relatively narrow area with a high number of targets at the same time, the drones are easily subject to excessive interference. Existing patrol methods do not have early warning mechanisms for this situation, or they only identify the narrow area without accurately analyzing the pattern of the number of targets, making it impossible to adjust the drones' patrol routes in a timely manner to reduce the probability of interference. Summary of the Invention

[0004] The purpose of this invention is to provide an edge-cloud collaborative drone patrol management system and method to solve the problems raised in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a drone patrol management method based on edge-cloud collaboration, the method comprising:

[0006] Step 1: Obtain patrol area division information, historical patrol data of the drone, and drone equipment parameter information;

[0007] Step 2: Combining patrol area division information with historical patrol data, analyze the self-interference and related interference caused by different areas to drone patrols, comprehensively judge the overall interference of different areas to drones, and mark the key areas.

[0008] Step 3: Locate the drones and identify the first and second drones to be observed;

[0009] Step 4: Obtain partial patrol routes of the two drones, create buffer zones for these routes, determine whether the two drones will interfere with each other when patrolling in key areas, and selectively adjust the patrol routes of the drones based on the determination results.

[0010] Step 5: Control the drone to detect targets along the patrol route and coordinate with the ground-based gimbal to drive away targets.

[0011] Preferably, step one includes: collecting the set of all patrol areas for target detection by the pre-defined UAVs as s={s1,s2,...sn}, where n represents the total number of patrol areas; obtaining information on the number of targets detected in each area in the past; and obtaining the flight speed of all UAVs and the pre-defined patrol routes of the UAVs.

[0012] Preferably, step two includes: according to Wi=(1 / si) / ∑ n i=1 (1 / si) Calculates the self-interference coefficient Wi caused by the limitation of a randomly selected area during drone patrol. The self-interference coefficient is used to determine the interference caused by the narrowness of the patrol area itself. si represents the area of ​​a randomly selected area. The number of targets detected by the drone during previous patrols within the corresponding area is retrieved and arranged in ascending order to obtain the target set H = {H1, H2, ... H}. m}, where m represents the number of times the drone has patrolled the corresponding area in the past, and the corresponding area refers to the i-th area. An outlier criterion value (IQR) is set. , Indicates to Round down. Indicates to Round up, and set the non-outlier range to [ , Each target is evaluated to determine whether it falls within the non-outlier range. Targets not falling within the non-outlier range are removed, resulting in a set of remaining targets h = {h1, h2, ... h}. k}, k represents the remaining number of target quantities, and the smoothing coefficient is set to μ, 0 < μ < 1. A target quantity prediction model is established: Fi = h k+1 =μ*h k +(1-μ)*P k The estimated number of targets Fi that the current drone will detect when it enters the corresponding patrol area is predicted, where P k Let {F1, F2, ..., Fi, ..., Fn} be the exponentially smoothed value of the number of k-th targets in set h. Then, the estimated number of targets detected by the current UAV when it enters n areas for patrol is obtained. The analysis method for the estimated target number in other areas is the same as that for the corresponding areas, based on the formula wi = (Fi) / ∑ n i=1(Fi) Analysis yields the relevant interference coefficient wi for the corresponding area to the drone patrol. The relevant interference coefficient is used to determine the interference caused by targets appearing in the area to the drone patrol. The more targets appearing in the area and the narrower the area itself, the higher the probability of the drone colliding with the target during patrol, that is, the greater the interference caused to the drone patrol. The comprehensive interference coefficient Wi+wi for the corresponding area to the drone patrol is obtained. The threshold of the comprehensive interference coefficient is set to G. Wi+wi and G are compared: if Wi+wi>G, the corresponding area is marked as a critical area; otherwise, it is not marked.

[0013] Considering that the probability of interference between drones and between drones and targets is low in relatively spacious areas with fewer targets, this invention analyzes the interference caused by the area itself of different regions on drones by analyzing self-interference coefficients, analyzes the pattern of target quantity in different regions by analyzing relevant interference coefficients, and then judges the interference caused by regions with different target occurrence patterns on drone training. Combining the analysis of these two interference situations, key regions are selected and marked, that is, regions where route analysis is required when drones appear. For other regions, that is, regions with a low probability of interference, route analysis is not performed. This improves the timeliness of predicting interference during drone patrols while reducing unnecessary data analysis. For the analysis of the pattern of target quantity in regions, historical target quantities in different regions are collected. Considering that outliers in the historical target quantities are not conducive to the prediction of target quantity and will greatly reduce the accuracy of the prediction, the interquartile range screening method and target quantity prediction model are combined to analyze the pattern of target quantity, which effectively improves the accuracy of target quantity prediction and the rationality of key region selection.

[0014] Preferably, step three includes: acquiring the drone's location in real time while the drone is patrolling; when it is located that one and only one drone appears in a random key area, the corresponding drone is recorded as the first drone to be observed, the time when the first drone to be observed enters the key area is recorded as T1, the patrol route of the first drone to be observed is retrieved, the length of the patrol route of the first drone to be observed in the corresponding key area is obtained as D, the flight speed of the first drone to be observed is retrieved as v, the patrol duration of the first drone to be observed in the corresponding key area is obtained as D / v, the patrol time period of the first drone to be observed in the corresponding key area is obtained as [T1, D / v+T1], the start patrol time of a random drone other than the first drone to be observed that needs to patrol the corresponding key area is retrieved as t1 and the flight speed is V, and the random drone other than the first drone to be observed that needs to patrol the corresponding key area is recorded as the second drone to be observed.

[0015] Preferably, step four includes: retrieving the route length L from the starting point of the second drone to be observed to the intersection point of the patrol route and the key area. The intersection point of the patrol route and the key area refers to the first intersection point of the route in the patrol route direction and the boundary of the key area, that is, the position point where the second drone to be observed enters the corresponding key area.预判 the time for the second drone to be observed to enter the key area as t1 + L / V, and determine whether t1 + L / V is within the time period [T1, D / v + T1]. If it is, it is determined that the second drone to be observed and the first drone to be observed will be in the same key area at the same time, and further determine whether the patrol route needs to be adjusted. If not, it is determined that the second drone to be observed and the first drone to be observed will not be in the same key area at the same time, and no patrol route adjustment is made.

[0016] If it is determined that the second drone to be observed and the first drone to be observed are in the same key area, the position where the first drone to be observed is located when the second drone to be observed enters the key area is obtained: the position point where the route length to the intersection point of the patrol route of the first drone to be observed and the key area is (t1 + L / V - T1) * v. Obtain the patrol route of the first drone to be observed after this position point and within the key area, denoted as route R1, and the route length of R1 is d1. Obtain the patrol route of the second drone to be observed within the key area, denoted as route R2, and the route length of R2 is d2. Create two buffer zones for route R1 and route R2 respectively. The buffer zone refers to the geometric area generated around a route, and the geometric area is a strip area with the route as the center line. According to Calculate the proximity F between route R1 and route R2. Among them, d 12 Refers to the length of route R2 falling into the buffer zone created for route R1, d 21 Refers to the length of route R1 falling into the buffer zone created for route R2. Set the proximity threshold as f. If F ≥ f, adjust the patrol route of the second drone to be observed so that the second drone to be observed does not enter the corresponding key area. If F < f, choose not to adjust the patrol route.

[0017] Preferably, step five includes: controlling the drone to patrol according to the adjusted patrol route, and using the drone for target detection: using the pod of the drone to complete the acquisition and recognition of the target image, and transmitting the collected target image to the cloud platform. The cloud platform generates a target驱赶 strategy. The cloud platform can integrate multi-source data such as the target activity pattern and meteorological information, and generate a target驱赶 strategy through a digital twin model, and control the drone and the ground-based pan-tilt to cooperate in target驱赶. The ground-based pan-tilt can perform target驱赶 through technologies such as lasers. The drone and the ground-based pan-tilt communicate through a self-organizing network, which is beneficial to ensuring real-time response and cooperation reliability.

[0018] By using a cloud-edge collaborative approach, target detection and repulsion are achieved. After the drone collects data, the cloud performs data analysis to generate target repulsion strategies, which are then coordinated between the drone and the ground-based gimbal to drive away the target, thus improving the effectiveness of target detection and repulsion.

[0019] A drone patrol management system based on edge-cloud collaboration includes: a work information collection module, an area marking and planning module, a patrol management module, and a target collaborative detection module;

[0020] The work information collection module is used to collect patrol area division information, historical patrol data of UAVs, and UAV equipment parameter information.

[0021] The area marking and planning module is used to combine patrol area division information with historical patrol data to analyze the self-interference and related interference of different areas to drone patrols, comprehensively judge the overall interference of different areas to drones, and mark key areas.

[0022] The patrol management module is used to locate the drone, identify the first and second drones to be observed, determine whether the two drones will cause mutual interference when patrolling in key areas, and selectively adjust the patrol route of the drones based on the judgment result.

[0023] The target collaborative detection module is used to control the UAV to detect targets according to the patrol route and to coordinate with the ground-based gimbal to drive away the targets.

[0024] Preferably, the work information collection module includes a patrol area information collection unit, a patrol information acquisition unit, and a device information acquisition unit;

[0025] The patrol area information collection unit is used to collect the area information of all patrol areas that the UAVs need to detect targets in. The patrol information acquisition unit is used to acquire the number of targets detected in each area in the past. The equipment information acquisition unit is used to acquire the flight speed information of all UAVs and the patrol route information of the UAVs that have been planned.

[0026] Preferably, the region marking planning module includes a self-interference analysis unit, a related interference analysis unit, and a region selection marking unit;

[0027] The self-interference analysis unit is used to analyze the self-interference coefficient of the UAV caused by the range limitation of each area during UAV patrol; the correlation interference analysis unit is used to establish a target quantity prediction model to predict the number of targets that will appear in different areas, and analyze the correlation interference coefficient of the targets in the area to the UAV patrol based on the prediction results; the area selection and marking unit is used to comprehensively analyze the comprehensive interference coefficient of different areas to the UAV patrol, set the comprehensive interference coefficient threshold, and selectively mark key areas based on the coefficient comparison results.

[0028] Preferably, the patrol management module includes a regional prediction unit and a patrol adjustment selection unit;

[0029] The same region prediction unit is used to locate the UAV in real time. When it is located that there is only one UAV in a random key area, the corresponding UAV is recorded as the first UAV to be observed, and another UAV that needs to patrol the corresponding key area is recorded as the second UAV to be observed. The patrol routes of the second UAV to be observed and the first UAV to be observed are analyzed to determine whether the two UAVs will be in the same key area at the same time.

[0030] The patrol adjustment selection unit is used to select whether to make route adjustments: if the two drones will not be in the same critical area at the same time, no patrol route adjustment is made; otherwise, it further determines whether route adjustment is needed: obtain part of the patrol routes of the two drones, create a buffer for part of the patrol routes, determine whether the two drones will cause mutual interference when patrolling in the critical area, and selectively adjust the patrol routes of the drones according to the judgment result.

[0031] The target collaborative detection module includes an information acquisition and identification unit, a data transmission unit, and an operation execution unit. The information acquisition and identification unit controls the UAV to patrol along the adjusted patrol route and uses the UAV's pod to acquire and identify target images. The data transmission unit transmits the acquired target images to the cloud platform. The operation execution unit generates a target driving-off strategy on the cloud platform and then controls the UAV and the ground-based gimbal to work together to drive off the target.

[0032] Compared with the prior art, the beneficial effects of the present invention are:

[0033] This invention considers that in relatively spacious areas with fewer targets, the probability of interference between drones and between drones and targets is low. Therefore, this invention analyzes the interference caused by the area itself of different regions on drones by analyzing self-interference coefficients, analyzes the pattern of the number of targets appearing in different regions by analyzing relevant interference coefficients, and then judges the interference caused by regions with different target appearance patterns on drone training. Combining the analysis of these two interference situations, key regions are selected and marked, that is, regions where route analysis is required when drones appear. For other regions, that is, regions with a low probability of interference, route analysis is not performed. This improves the timeliness of predicting interference during drone patrols while reducing unnecessary data analysis. For the analysis of the pattern of the number of targets appearing in a region, the historical number of targets appearing in different regions is collected. Considering that the presence of outliers in the historical number of targets is not conducive to the prediction of the number of targets and the accuracy of the prediction will be greatly reduced, the interquartile range screening method and the target number prediction model are combined to analyze the pattern of the number of targets, which effectively improves the accuracy of the target number prediction and the rationality of the selection of key regions.

[0034] After identifying key areas, early warnings are issued for situations where multiple drones appear simultaneously within these areas. The potential for mutual interference between drones is also predicted. Based on these predictions, patrol routes are selectively adjusted. For drones whose initial planned patrol routes all involve key areas and are predicted to appear simultaneously within the same key area, further adjustments are made. Considering the risk of collision and interference if two drones get too close in a narrow area, two buffer zones are established to analyze the proximity of routes rather than their overlap. This allows for a more accurate assessment of the potential collision risk between two drones, enabling selective adjustments to patrol routes and reducing the risk of mutual interference between drones when patrolling narrow areas with a large number of targets. Attached Figure Description

[0035] Figure 1 This is a schematic diagram of the structure of an edge-cloud collaborative drone patrol management system according to the present invention;

[0036] Figure 2 This is a flowchart illustrating a drone patrol management method based on edge-cloud collaboration according to the present invention. Detailed Implementation

[0037] 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.

[0038] Example 1: As Figure 2 As shown, this embodiment provides a drone patrol management method based on edge-cloud collaboration. The method includes: Step 1: Obtaining patrol area division information, historical patrol data of drones, and drone equipment parameter information: Collecting the area set of all patrol areas that the drones need to detect targets in, which has been divided into n patrol areas, s={s1,s2,...sn}, where n represents the total number of patrol areas; obtaining the number of targets detected in each area in the past; obtaining the flight speed of all drones and the pre-defined patrol route information of the drones;

[0039] Step Two: Combining patrol area division information with historical patrol data, analyze the self-interference and related interference caused by different areas to drone patrols, comprehensively judge the overall interference of different areas to drones, and mark the key areas: According to Wi=(1 / si) / ∑ n i=1 (1 / si) Calculate the self-interference coefficient Wi caused by a random area limitation on the drone during patrol, where si represents the area of ​​the random area. Retrieve the number of targets detected by the drone during previous patrols within the corresponding area, and arrange the target numbers in ascending order to obtain the target number set H={H1,H2,...H m}, where m represents the number of times the drone has patrolled the corresponding area in the past, and the corresponding area refers to the i-th area. An outlier criterion value (IQR) is set. , Indicates to Round down. Indicates to Round up, and set the non-outlier range to [ , Each target is evaluated to determine whether it falls within the non-outlier range. Targets not falling within the non-outlier range are removed, resulting in a set of remaining targets h = {h1, h2, ... h}. k}, k represents the remaining number of target quantities, and the smoothing coefficient is set to μ, 0 < μ < 1. A target quantity prediction model is established: Fi = h k+1 =μ*h k +(1-μ)*P kThe estimated number of targets Fi that the current drone will detect when it enters the corresponding patrol area is predicted, where P k Let {F1, F2, ..., Fi, ..., Fn} be the exponentially smoothed value of the number of k-th targets in set h. Then, the estimated number of targets detected by the current UAV when it enters n areas for patrol is obtained. The analysis method for the estimated target number in other areas is the same as that for the corresponding areas, based on the formula wi = (Fi) / ∑ n i=1 (Fi) Analysis yields the relevant interference coefficient wi for the corresponding area to the drone patrol, and the comprehensive interference coefficient Wi+wi for the corresponding area to the drone patrol is obtained. The threshold for the comprehensive interference coefficient is set to G. Wi+wi and G are compared: if Wi+wi>G, the corresponding area is marked as a critical area; otherwise, it is not marked.

[0040] Step 3: Locate the drones and identify the first and second drones to be observed: While the drones are patrolling, their locations are acquired in real time. When only one drone is located in a random key area, it is designated as the first drone to be observed. The time when the first drone enters the key area is recorded as T1. The patrol route of the first drone is retrieved, and the length of its patrol route in the corresponding key area is obtained as D. The flight speed of the first drone is retrieved as v, and the patrol duration of the first drone in the corresponding key area is obtained as D / v. The patrol time period of the first drone in the corresponding key area is [T1, D / v + T1]. The start patrol time and flight speed of a random drone (excluding the first drone to be observed) that needs to patrol the corresponding key area are retrieved as t1 and V, respectively. This random drone (excluding the first drone to be observed) that needs to patrol the corresponding key area is designated as the second drone to be observed.

[0041] Step 4: Obtain the partial patrol routes of two drones, create buffers for the partial patrol routes, and determine whether there will be mutual interference between the two drones during patrol in the critical area. Selectively adjust the patrol routes of the drones based on the judgment result: The length of the route from the starting point of the second drone to be observed to the intersection point of the patrol route and the critical area is L. The intersection point of the patrol route and the critical area refers to the first intersection point of the route in the patrol route direction and the boundary of the critical area, that is, the position point where the second drone to be observed enters the corresponding critical area. Predict the time when the second drone to be observed enters the critical area as t1 + L / V, and determine whether t1 + L / V is within the time period [T1, D / v + T1]. If it is, judge that the second drone to be observed and the first drone to be observed will be in the same critical area at the same time, and further judge whether the patrol route needs to be adjusted: Obtain the position of the first drone to be observed when the second drone to be observed enters the critical area: at the position point where the length of the route to the intersection point of the patrol route of the first drone to be observed and the critical area is (t1 + L / V - T1)*v. Obtain the patrol route of the first drone to be observed after this position point and within the critical area and record it as route R1, the length of route R1 is d1, obtain the patrol route of the second drone to be observed within the critical area and record it as route R2, the length of route R2 is d2. Create two buffers for route R1 and route R2 respectively. A buffer refers to a geometric area generated around a route. The geometric area is a strip area with the route as the center line. According to Calculate the proximity degree F of route R1 and route R2. Among them, d 12 refers to the length of route R2 falling into the buffer created for route R1, d 21 refers to the length of route R1 falling into the buffer created for route R2. Set the proximity degree threshold as f. If F ≥ f, adjust the patrol route of the second drone to be observed so that the second drone to be observed does not enter the corresponding critical area; if F < f, choose not to adjust the patrol route; if not, judge that the second drone to be observed and the first drone to be observed will not be in the same critical area at the same time, and do not adjust the patrol route;

[0042] For example: Obtain that the length of route R1 is d1 = 50 meters and the length of route R2 is d2 = 72 meters. Create two buffers for route R1 and route R2 respectively. The length of route R2 falling into the buffer created for route R1 is 38 meters, and the length of route R1 falling into the buffer created for route R2 is 30 meters. Calculate that the proximity degree F of route R1 and route R2 is 0.53. Set the proximity degree threshold as f = 0.50. F > f, adjust the patrol route of the second drone to be observed so that the second drone to be observed does not enter the corresponding critical area;

[0043] Step 5: Control the drone to detect targets along the patrol route and coordinate with the ground-based gimbal to drive away targets: Control the drone to patrol along the adjusted patrol route and use the drone to detect targets: Use the drone's pod to collect and identify target images, transmit the collected target images to the cloud platform, the cloud platform generates a target driving away strategy, and controls the drone and the ground-based gimbal to coordinate to drive away targets.

[0044] Example 2: Figure 1 As shown, this embodiment provides a drone patrol management system based on edge-cloud collaboration. It is implemented based on the patrol management method described in this embodiment and specifically includes: a work information collection module, a region marking and planning module, a patrol management module, and a target collaborative detection module. The work information collection module collects patrol area division information, historical patrol data of drones, and drone equipment parameter information. The region marking and planning module combines patrol area division information and historical patrol data to analyze the self-interference and related interference caused by different regions to drone patrols, comprehensively judges the overall interference of different regions to drones, and marks key areas. The patrol management module locates drones, identifies the first and second drones to be observed, determines whether the two drones will cause mutual interference when patrolling in key areas, and selectively adjusts the drone patrol route based on the judgment result. The target collaborative detection module controls the drone to detect targets according to the patrol route and coordinates with the ground-based gimbal to drive away targets.

[0045] The work information acquisition module includes a patrol area information acquisition unit, a patrol information acquisition unit, and an equipment information acquisition unit. The patrol area information acquisition unit is used to collect the area information of all patrol areas that the UAVs need to detect targets in. The patrol information acquisition unit is used to acquire the number of targets detected in each area in the past. The equipment information acquisition unit is used to acquire the flight speed information of all UAVs and the patrol route information of the UAVs that have been planned.

[0046] The area marking planning module includes a self-interference analysis unit, a related interference analysis unit, and an area selection marking unit. The self-interference analysis unit is used to analyze the self-interference coefficient of each area limitation on the drone during drone patrol. The related interference analysis unit is used to establish a target quantity prediction model to predict the number of targets that will appear in different areas, and analyze the related interference coefficient of targets in the area on the drone patrol based on the prediction results. The area selection marking unit is used to comprehensively analyze the comprehensive interference coefficient of different areas on the drone patrol, set the comprehensive interference coefficient threshold, and selectively mark key areas based on the coefficient comparison results.

[0047] The patrol management module includes a co-occurrence area prediction unit and a patrol adjustment selection unit. The co-occurrence area prediction unit is used to perform real-time positioning of UAVs. When it locates exactly one UAV in a random key area, the corresponding UAV is marked as the first UAV to be observed, and another random UAV that needs to patrol the corresponding key area is marked as the second UAV to be observed. The patrol routes of the second and first UAVs to be observed are analyzed to determine whether the two UAVs will be in the same key area at the same time. The patrol adjustment selection unit is used to select whether to adjust the route: if the two UAVs will not be in the same key area at the same time, no patrol route adjustment is made; otherwise, it further determines whether route adjustment is needed: it obtains part of the patrol routes of the two UAVs, creates a buffer for part of the patrol routes, determines whether the two UAVs will cause mutual interference when patrolling in the key area, and selectively adjusts the patrol routes of the UAVs based on the judgment result.

[0048] The target collaborative detection module includes an information acquisition and identification unit, a data transmission unit, and an operation execution unit. The information acquisition and identification unit controls the UAV to patrol along the adjusted patrol route and uses the UAV's pod to acquire and identify target images. The data transmission unit transmits the acquired target images to the cloud platform. The operation execution unit generates a target driving-off strategy on the cloud platform and then controls the UAV and ground-based gimbal to work together to drive off the target.

[0049] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A method for managing unmanned aerial vehicle (UAV) patrols based on edge-cloud collaboration, characterized in that: include: Step 1: Obtain patrol area division information, historical patrol data of the drone, and drone equipment parameter information; Step 2: Combining patrol area division information with historical patrol data, analyze the self-interference and related interference caused by different areas to drone patrols, comprehensively judge the overall interference of different areas to drones, and mark the key areas. Step 3: Locate the drones and identify the first and second drones to be observed; Step 4: Obtain partial patrol routes of the two drones, create buffer zones for these routes, determine whether the two drones will interfere with each other when patrolling in key areas, and selectively adjust the patrol routes of the drones based on the determination results. Step 5: Control the drone to detect targets according to the patrol route, and coordinate with the ground-based gimbal to drive away targets; Step 3 includes: acquiring the drone's location in real time while the drone is patrolling; when only one drone is located in a random key area, the corresponding drone is recorded as the first drone to be observed, the time when the first drone to be observed enters the key area is recorded as T1, the patrol route of the first drone to be observed is retrieved, the length of the patrol route of the first drone to be observed in the corresponding key area is obtained as D, the flight speed of the first drone to be observed is retrieved as v, the patrol duration of the first drone to be observed in the corresponding key area is obtained as D / v, the patrol time period of the first drone to be observed in the corresponding key area is obtained as [T1, D / v+T1], the start patrol time of a random drone other than the first drone to be observed that needs to patrol the corresponding key area is retrieved as t1, the flight speed is V, and the random drone other than the first drone to be observed that needs to patrol the corresponding key area is recorded as the second drone to be observed. Step four includes: retrieving the route length L from the starting point of the second drone to be observed to the intersection of the patrol route and the key area; predicting the time when the second drone to be observed enters the key area as t1+L / V; and determining whether t1+L / V is within the time period [T1, D / v+T1]. If it is, it is determined that the second drone to be observed will be in the same key area as the first drone to be observed, and further determining whether the patrol route needs to be adjusted; if it is not, it is determined that the second drone to be observed will not be in the same key area as the first drone to be observed, and no patrol route adjustment is made. If it is determined that the second drone to be observed will be in the same critical area as the first drone to be observed, the position of the first drone to be observed when the second drone to be observed enters the critical area is obtained as: the position point where the length of the route from the first drone to be observed to the intersection point of the patrol route and the critical area is (t1 + L / V - T1) * v. The patrol route of the first drone to be observed after this position point and within the critical area is obtained and denoted as route R1, and the length of route R1 is d1. The patrol route of the second drone to be observed within the critical area is obtained and denoted as route R2, and the length of route R2 is d2. Two buffers are created for route R1 and route R2 respectively. According to Calculate the proximity degree F of route R1 and route R2. Among them, d 12 Refers to the length of route R2 falling into the buffer created for route R1, d 21 Refers to the length of route R1 falling into the buffer created for route R2. Set the proximity degree threshold as f. If F ≥ f, adjust the patrol route of the second drone to be observed so that the second drone to be observed does not enter the corresponding critical area; if F < f, choose not to adjust the patrol route.

2. The UAV patrol management method based on edge-cloud collaboration according to claim 1, characterized in that: Step one includes: collecting the area set of all patrol areas that the drones need to detect targets in, which is s={s1,s2,...sn}, where n represents the total number of patrol areas; obtaining the number of targets detected in each area in the past; and obtaining the flight speed of all drones and the pre-defined patrol routes of the drones.

3. The UAV patrol management method based on edge-cloud collaboration according to claim 2, characterized in that: Step two includes: based on Wi=(1 / si) / ∑ n i=1 (1 / si) Calculate the self-interference coefficient Wi caused by a random area limitation on the drone during patrol, where si represents the area of ​​the random area. Retrieve the number of targets detected by the drone during previous patrols within the corresponding area, and arrange the target numbers in ascending order to obtain the target number set H={H1,H2,...H m }, where m represents the number of times the drone has patrolled the corresponding area in the past, and the outlier detection criterion value IQR is set. Set the non-outlier range as Each target is checked to see if it is outside the range. Targets outside the range are removed, resulting in the set of remaining targets after the removal process, which is h = {h1, h2, ... h}. k }, k represents the remaining number of target quantities, and the smoothing coefficient is set to μ, 0 < μ < 1. A target quantity prediction model is established: Fi = h k+1 =μ*h k +(1-μ)*P k The estimated number of targets Fi that the current drone will detect when it enters the corresponding patrol area is predicted, where P k Let be the exponentially smoothed value of the number of k-th targets in set h. Then, we obtain the estimated number of targets {F1, F2, ..., Fi, ..., Fn} that the current UAV will detect when patrolling n areas. This is calculated using the formula wi = (Fi) / ∑ n i=1 (Fi) Analysis yields the relevant interference coefficient wi for the corresponding area to the drone patrol, and the comprehensive interference coefficient Wi+wi for the corresponding area to the drone patrol is obtained. The threshold for the comprehensive interference coefficient is set to G. Wi+wi and G are compared: if Wi+wi>G, the corresponding area is marked as a critical area; otherwise, it is not marked.

4. The UAV patrol management method based on edge-cloud collaboration according to claim 1, characterized in that: Step five includes: controlling the drone to patrol along the adjusted patrol route and using the drone for target detection: using the drone's pod to collect and identify target images, transmitting the collected target images to the cloud platform, generating a target driving-away strategy on the cloud platform, and controlling the drone and ground-based gimbal to work together to drive away the target.

5. A drone patrol management system based on edge-cloud collaboration, applied to the drone patrol management method based on edge-cloud collaboration as described in claim 1, characterized in that: The system includes a work information collection module, an area marking and planning module, a patrol management module, and a target collaborative detection module; The work information collection module is used to collect patrol area division information, historical patrol data of UAVs, and UAV equipment parameter information. The area marking and planning module is used to combine patrol area division information with historical patrol data to analyze the self-interference and related interference of different areas to drone patrols, comprehensively judge the overall interference of different areas to drones, and mark key areas. The patrol management module is used to locate the drone, identify the first and second drones to be observed, determine whether the two drones will cause mutual interference when patrolling in key areas, and selectively adjust the patrol route of the drones based on the judgment result. The target collaborative detection module is used to control the UAV to detect targets according to the patrol route and to coordinate with the ground-based gimbal to drive away the targets.

6. The UAV patrol management system based on edge-cloud collaboration according to claim 5, characterized in that: The work information collection module includes a patrol area information collection unit, a patrol information acquisition unit, and a device information acquisition unit; The patrol area information collection unit is used to collect the area information of all patrol areas that the UAVs need to detect targets in. The patrol information acquisition unit is used to acquire the number of targets detected in each area in the past. The equipment information acquisition unit is used to acquire the flight speed information of all UAVs and the patrol route information of the UAVs that have been planned.

7. The UAV patrol management system based on edge-cloud collaboration according to claim 6, characterized in that: The region marking planning module includes a self-interference analysis unit, a related interference analysis unit, and a region selection marking unit; The self-interference analysis unit is used to analyze the self-interference coefficient of the UAV caused by the range limitation of each area during UAV patrol; the related interference analysis unit is used to establish a target quantity prediction model to predict the number of targets that will appear in different areas, and analyze the related interference coefficient of the targets in the area to the UAV patrol based on the prediction results. The region selection and marking unit is used to comprehensively analyze the comprehensive interference coefficient of different regions on UAV patrols, set a threshold for the comprehensive interference coefficient, and selectively mark key regions based on the coefficient comparison results.

8. The UAV patrol management system based on edge-cloud collaboration according to claim 7, characterized in that: The patrol management module includes a regional prediction unit and a patrol adjustment selection unit. The same region prediction unit is used to locate the UAV in real time. When it is located that there is only one UAV in a random key area, the corresponding UAV is recorded as the first UAV to be observed, and another UAV that needs to patrol the corresponding key area is recorded as the second UAV to be observed. The patrol routes of the second UAV to be observed and the first UAV to be observed are analyzed to determine whether the two UAVs will be in the same key area at the same time. The patrol adjustment selection unit is used to select whether to make route adjustments: if the two drones will not be in the same critical area at the same time, no patrol route adjustment is made; otherwise, it further determines whether route adjustment is needed: obtain part of the patrol routes of the two drones, create a buffer for part of the patrol routes, determine whether the two drones will cause mutual interference when patrolling in the critical area, and selectively adjust the patrol routes of the drones according to the judgment result. The target collaborative detection module includes an information acquisition and identification unit, a data transmission unit, and an operation execution unit. The information acquisition and identification unit controls the UAV to patrol along the adjusted patrol route and uses the UAV's pod to acquire and identify target images. The data transmission unit transmits the acquired target images to the cloud platform. The operation execution unit generates a target driving-off strategy on the cloud platform and then controls the UAV and the ground-based gimbal to work together to drive off the target.

Images