Unmanned aerial vehicle landing site selection method based on nhpp risk field and multi-dimensional comprehensive evaluation

Abstract

The application discloses a kind of based on NHPP risk field and multi-dimensional comprehensive evaluation UAV take-off and landing site selection method, its method includes: S1, creates n1 candidate area in study area;S2, the time-space risk monitoring data of all candidate areas in study area are input into time-space nonhomogeneous Poisson process wind disaster risk model calculation to obtain the NHPP risk cumulative intensity corresponding to candidate area;S3, build the multi-dimensional comprehensive evaluation system including NHPP risk cumulative intensity, construction cost includes P1 dimension, weighted evaluation obtains the evaluation value of candidate area;Select the optimal candidate area as take-off and landing site selection area, or set decision threshold and output site selection decision result according to decision threshold.The NHPP risk field of all candidate areas is constructed by time-space nonhomogeneous Poisson process wind disaster risk model, effectively utilize the time-space data to construct NHPP risk field and combine multi-dimensional scientific and reasonable evaluation UAV take-off field candidate area, provide reliable technical support for low-altitude aviation site selection decision.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of low-altitude aviation decision-making, and in particular to a UAV landing site selection method based on an NHPP risk field and multi-dimensional comprehensive evaluation. BACKGROUND

[0002] In the field of low-altitude aviation, low-altitude aviation infrastructure and intelligent decision-making are the key to the development of low-altitude aviation. How to reasonably evaluate and select the landing site of a UAV is a technical problem that needs to be solved in the field of low-altitude aviation. US patent US11741702B2 discloses an automatic safe landing site selection for a UAV system, which proposes generating a depth map by obtaining two overlapping images, identifying flat areas and calculating a "landing area quality score" based on depth variance and semantic type, and selecting the area with the highest score as the landing site. Canadian patent CA2977945A1 discloses a scanning environment and tracking UAV system, which scans the candidate area in real time through LiDAR and camera, constructs an environment map, and scores and filters the areas that meet the flatness and obstacle threshold. At present, single-factor risk or threshold alarm is also used. Some existing technologies only monitor wind speed or obstacle density and decide whether to enable or suspend landing according to fixed thresholds, lacking analysis of the spatio-temporal aggregation and multi-dimensional trade-off of risks. The above three methods rely only on one-way monitoring or regular manual review, which cannot accurately and scientifically evaluate the selection of UAV landing sites. SUMMARY

[0003] The present application relates to the technical field of low-altitude aviation decision-making, and in particular to a UAV landing site selection method based on an NHPP risk field and multi-dimensional comprehensive evaluation.

[0004] The purpose of the present application is achieved by the following technical solutions:

[0005] A UAV landing site selection method based on an NHPP risk field and multi-dimensional comprehensive evaluation, the method comprising:

[0006] S1, creating n1 candidate areas in a study area;

[0007] S2, collecting spatio-temporal risk monitoring data of all candidate areas in the study area and inputting the data into a constructed spatio-temporal non-homogeneous Poisson process wind disaster risk model to construct a spatio-temporal dimension NHPP risk field, and calculating the NHPP risk cumulative intensity corresponding to the candidate areas;

[0008] S3, a multi-dimensional comprehensive evaluation system including NHPP risk cumulative intensity and construction cost is constructed, including P1 dimensions, weighted evaluation is performed to obtain the evaluation value of the candidate region; the optimal candidate region is selected as the take-off and landing site selection region, or a decision threshold is set and the site selection decision result is output according to the decision threshold.

[0009] In order to better realize the present application, if the study area is a non-mountainous area, the spatio-temporal risk monitoring data includes historical wind monitoring spatio-temporal sequence data, and the spatio-temporal non-homogeneous Poisson process wind disaster risk model constructs an NHPP risk field according to a spatio-temporal intensity function, and the spatio-temporal intensity function expression is as follows:

[0010] , wherein is the expected frequency of event occurrence at spatial position , time , is the risk impact of historical events on spatial position , time , is the additional contribution to risk when the wind speed exceeds the safety threshold, and the historical events are wind monitoring data record points.

[0011] Preferably, if the study area is a mountainous area, the spatio-temporal risk monitoring data includes historical wind monitoring spatio-temporal sequence data, and the spatio-temporal non-homogeneous Poisson process wind disaster risk model constructs an NHPP risk field according to a risk density function, and the risk density function expression is as follows:

[0012] , wherein is the risk density of spatial position , time in the mountainous area, is the position of the mountainous event point i at time , is the weight of the historical event point i, is the wind speed at time t, is the safety wind speed threshold, is the spatial influence diffusion scale.

[0013] Preferably, the risk impact is expressed as follows:

[0014] , wherein is the weight of the historical event point i, is the position of the historical event point i, is the distance decay rate of the historical event point i, is the influence coefficient of the wind speed exceeding the safety threshold on the risk;

[0015] The additional contribution The expression is as follows:

[0016] , wherein is the impact coefficient of risk when the wind speed exceeds the safety threshold, is the wind speed at time t, is the safety wind speed threshold.

[0017] Preferably, the candidate area The NHPP risk cumulative intensity in the interval [t0, T] The expression is as follows: .

[0018] Preferably, the evaluation value of the candidate area is obtained based on the weighted calculation of the dimension indicators in the multi-dimensional comprehensive evaluation system, and the expression is as follows: , wherein is the evaluation value of the dimension indicator , and is the weight of the dimension indicator .

[0019] Preferably, the NHPP risk cumulative intensity is verified and compared by the following method: selecting a sample site and deploying a weather station, a laser radar and a camera through a 5G cell and a LoRaWAN network at the sample site to monitor data and evaluate the observed risk , the predicted risk of the sample site is calculated through the spatio-temporal non-homogeneous Poisson process wind disaster risk model, the difference value is obtained by subtracting the predicted risk from the observed risk .

[0020] Preferably, the parameters of the spatio-temporal non-homogeneous Poisson process wind disaster risk model are adjusted based on the difference value , and the expression is as follows:

[0021] , wherein is the original parameter, is the adjusted parameter, is the risk sensitivity threshold, is the learning rate or the proportion factor.

[0022] Preferably, if the study area is a mountainous area and the rescue is in the mountainous area, the multi-dimensional comprehensive evaluation system further includes communication quality evaluation and ground rescue accessibility evaluation, and the communication quality evaluation and the ground rescue accessibility evaluation are evaluated by real-time feedback of detection data to the command center through edge wind field detection, air channel detection and video shooting by the unmanned aerial vehicle.

[0023] Preferably, the evaluation value of the candidate region is a positive evaluation value, and the greater the evaluation value of the candidate region, the better it represents; the site selection decision result includes three kinds of recommendation, conditional recommendation and non-recommendation.

[0024] Compared with the prior art, the present application has the following advantages and beneficial effects:

[0025] (1) The present application constructs the NHPP risk field of all candidate regions through the spatio-temporal non-homogeneous Poisson process wind disaster risk model, for the first time applying the spatio-temporal dimensional NHPP risk field to the evaluation of landing site selection, using a multi-dimensional comprehensive evaluation system for weighted evaluation and selecting the optimal candidate region as the landing site selection region, effectively utilizing the spatio-temporal data to construct the NHPP risk field and combining multi-dimensional scientific and reasonable evaluation of the unmanned aerial vehicle landing site candidate region, providing reliable technical support for low-altitude aviation site selection decision, and promoting the development of low-altitude intelligent decision technology.

[0026] (2) The present application obtains the difference value by the difference between the observed risk and the predicted risk of the sample point, adjusts the parameters of the spatio-temporal non-homogeneous Poisson process wind disaster risk model based on the difference value, realizes the closed-loop feedback and online correction parameter adjustment mechanism, forms a dynamic response mechanism through group cooperation and task feedback loop, and significantly improves the risk field evaluation accuracy (error ≤ 5%) and self-adaptive ability. BRIEF DESCRIPTION OF DRAWINGS

[0027] Figure 1 The method flowchart of the unmanned aerial vehicle landing site selection method of the present application. DETAILED DESCRIPTION

[0028] The present application will be further described in detail below in conjunction with the embodiments:

[0029] Example 1

[0030] As shown in the figure, an unmanned aerial vehicle landing site selection method based on NHPP risk field and multi-dimensional comprehensive evaluation, the method comprises: Figure 1

[0031] S1, creating n1 candidate regions in the study area.

[0032] S2, collecting spatio-temporal risk monitoring data of all candidate regions in the study area to input the constructed spatio-temporal non-homogeneous Poisson process wind disaster risk model to construct the NHPP risk field of spatio-temporal dimension, and calculating to obtain the corresponding NHPP risk cumulative intensity of the candidate region.

[0033] If the study area is a non-mountainous area, the spatio-temporal risk monitoring data contains historical wind monitoring spatio-temporal sequence data, and the spatio-temporal non-homogeneous Poisson process wind disaster risk model constructs the NHPP risk field according to the spatio-temporal intensity function, and the expression of the spatio-temporal intensity function is as follows:

[0034] , wherein​ In spatial location ,time Expected frequency of events For historical events in relation to spatial location ,time The risk impact decreases exponentially with increasing distance; For the additional contribution to risk when wind speed exceeds the safety threshold, historical events are recorded as wind monitoring data points. The risk impact is... The expression is as follows:

[0035] ,in Weights for historical event point i, For the location of historical event point i, Let i be the decay rate of historical event point i with distance. This represents the impact coefficient on risk when wind speed exceeds the safety threshold. Additional contribution is also included. The expression is as follows:

[0036] ,in This is the risk impact coefficient when wind speed exceeds the safety threshold. Let be the wind speed at time t. The safe wind speed threshold.

[0037] If the study area is mountainous, the spatiotemporal risk monitoring data includes historical wind monitoring spatiotemporal sequence data. The spatiotemporal nonhomogeneous Poisson process wind disaster risk model constructs the NHPP risk field according to the risk density function, the expression of which is as follows:

[0038] ,in Spatial location of mountainous areas ,time risk density, For event point i in the mountainous area, in time Location, Weights for historical event point i, Let be the wind speed at time t. For safe wind speed thresholds, This represents the scale of spatial impact diffusion.

[0039] Candidate region of the present invention The cumulative NHPP risk intensity in the interval [t0,T] The expression is as follows: If the study area is not a mountainous area, for If the study area is mountainous, for ;NHPP risk accumulation intensity reflects the spatio-temporal clustering trend and risk exposure.

[0040] In some embodiments, the NHPP risk accumulation intensity is verified and compared by the following method: sample sites are selected, and meteorological stations, laser radars, and cameras are deployed at the sample sites through 5G cellular and LoRaWAN networks to monitor data and evaluate observed risks The predicted risks of the sample sites are calculated through the spatio-temporal non-homogeneous Poisson process wind disaster risk model The predicted risks are subtracted from the observed risks to obtain differences (the sample point difference calculation, overestimation or underestimation of the actual wind disaster frequency). Based on the differences the parameters of the spatio-temporal non-homogeneous Poisson process wind disaster risk model are adjusted, and the expression is as follows:

[0041] , wherein is the original parameter, is the adjusted parameter, is the risk sensitivity threshold (controls the parameter update range and avoids overcorrection), is the learning rate or proportion factor (controls the correction speed and stability). The differences are used for closed-loop feedback and online optimization processing (adjusting the parameters ), reflecting the risk sensitivity adjustment after feedback on the observed data residual. The present application adjusts the parameters of the spatio-temporal non-homogeneous Poisson process wind disaster risk model based on the differences, realizes the closed-loop feedback and online correction parameter adjustment mechanism, forms a dynamic response mechanism through group cooperation and task feedback loop, and significantly improves the risk field evaluation accuracy (error ≤ 5%) and self-adaptive ability.

[0042] S3, a multi-dimensional comprehensive evaluation system containing NHPP risk accumulation intensity and construction cost containing P1 dimensions is constructed, and the evaluation value of the candidate area is obtained by weighted evaluation. The evaluation value of the candidate area in this embodiment is calculated based on the weighted calculation of the dimension indicators in the multi-dimensional comprehensive evaluation system, and the expression is as follows: , wherein is the evaluation value of the dimension indicator , and is the weight of the dimension indicator . In some embodiments, if the study area is a mountainous area and rescue is needed in the mountainous area, the multi-dimensional comprehensive evaluation system further includes two dimensions of communication quality evaluation and ground rescue accessibility evaluation. The communication quality evaluation and the ground rescue accessibility evaluation are evaluated by real-time feedback of detection data to the command center through edge wind field detection, air channel detection, and video shooting by the unmanned aerial vehicle.

[0043] The optimal candidate region is selected as the take-off and landing site selection region, or a decision threshold is set and a site selection decision result is output according to the decision threshold. In some embodiments, the evaluation value of the candidate region is a positive evaluation value, and the greater the evaluation value of the candidate region, the better it is; the site selection decision result includes three kinds of recommendation, conditional recommendation and non-recommendation.

[0044] Embodiment two

[0045] An unmanned aerial vehicle take-off and landing site selection method based on an NHPP risk field and multi-dimensional comprehensive evaluation, under the idea of the present application, the present embodiment is used for city logistics center take-off and landing site selection (a city A hopes to set up a multi-rotor unmanned aerial vehicle take-off and landing site near the logistics center in the southwest area for high-frequency distribution operation. The logistics center address is "City A, Wanshou Road Logistics Park", and "City A, Wanshou Road Logistics Park" is a non-mountainous area), and the method comprises:

[0046] S1, creating n1 candidate regions in the study area. Specifically, a circular region with a radius of 1.5 km is set around the "Wanshou Road Logistics Park", and a plurality of candidate regions are set. Through high-resolution remote sensing images or map data (such as OpenStreetMap or city construction BIM data), all vacant land, building roofs, parking lots, and open squares in the circular region are extracted as candidate points, and a spatial clustering algorithm is applied to merge the candidate points into a plurality of candidate regions.

[0047] S2, collecting the spatio-temporal risk monitoring data of all candidate regions in the study area to input the constructed spatio-temporal non-homogeneous Poisson process wind disaster risk model to construct the NHPP risk field in the spatio-temporal dimension, and calculating to obtain the corresponding NHPP risk cumulative intensity of the candidate region.

[0048] Since the Wanshou Road Logistics Park is the study area, the study area is a non-mountainous area, and the spatio-temporal risk monitoring data includes historical wind monitoring spatio-temporal sequence data. The spatio-temporal non-homogeneous Poisson process wind disaster risk model constructs the NHPP risk field according to the spatio-temporal intensity function, and the spatio-temporal intensity function expression is as follows:

[0049] , wherein is the expected frequency of the event occurring at the spatial position , time is the weight of the historical event point i, is the position of the historical event point i, is the distance decay rate of the historical event point i, is the influence coefficient of the risk when the wind speed exceeds the safety threshold, is the wind speed at time t, is the safety wind speed threshold, and the present embodiment​ 7m / s (i.e. the maximum take-off and landing wind speed for multi-rotor safety).

[0050] Candidate regions of the present application The cumulative intensity of NHPP risk in the interval [t0, T] is verified and compared as follows: sample sites are selected and meteorological stations, laser radars and cameras are deployed at the sample sites through 5G cellular and LoRaWAN networks to monitor and evaluate the observed risk The predicted risk of the sample site is calculated through the spatio-temporal non-homogeneous Poisson process wind disaster risk model The predicted risk is subtracted from the observed risk to obtain the difference (sample point difference calculation, overestimation or underestimation of the actual wind disaster frequency). Based on the difference , the parameters of the spatio-temporal non-homogeneous Poisson process wind disaster risk model are adjusted, and the expression is as follows:

[0051] , wherein is the original parameter, is the adjusted parameter, is the risk sensitivity threshold (controls the parameter update range and avoids overcorrection), is the learning rate or proportion factor (controls the correction speed and stability). The difference is used for closed-loop feedback and online optimization processing (adjusting the parameters ) to reflect the risk sensitivity adjustment after feedback on the observed data residual. The above method is an error-driven model parameter update method, which is a nonlinear residual feedback correction mechanism. The main logic is as follows: if the model prediction R pred deviates significantly from the observation R obs , i.e. |ΔR| is large, then the risk response of certain regions / time periods in the model is too high or too low; by amplifying or reducing the related parameters , the fitting accuracy of the model is improved; R th ensures that the correction effect has the same scale for risk fields of different scales; controls the correction "step length" to prevent oscillation or overfitting. For example: suppose the model parameters of a certain region are , the model predicts the risk R pred = 0.6, but the observation is R obs = 0.4, i.e. ΔR = 0.2. Let R th = 1.0, κ = 0.1, then , i.e.: slightly increase the parameters of this region to more accurately match the observed data.

[0052] S3, build a multi-dimensional comprehensive evaluation system including NHPP risk cumulative intensity and construction cost, and get the evaluation value of the candidate area by weighted evaluation. In the research area of Wanshou Road Logistics Park, the multi-dimensional comprehensive evaluation system includes the distance cost of the logistics center, terrain features (slope, shielding degree), and land accessibility or usability score in addition to NHPP risk cumulative intensity and construction cost. The evaluation value of the candidate area in this embodiment Based on the dimension index in the multi-dimensional comprehensive evaluation system, the normalization processing is first performed (the evaluation value of the candidate area is a positive evaluation value, and the larger the evaluation value, the better. In the normalization processing, the larger the normalized dimension index, the better, for example, the normalization formula of NHPP risk cumulative intensity is: , and the normalization formula of distance cost is: ), and the weighted processing is calculated to obtain the expression as follows: , wherein is the evaluation value of the dimension index , and is the weight of the dimension index . The optimal candidate area is selected as the take-off and landing site selection area.

[0053] Embodiment Three

[0054] A UAV take-off and landing site selection method based on NHPP risk field and multi-dimensional comprehensive evaluation, which is used for UAV take-off and landing site selection in mountain emergency rescue scenarios (in mountain sudden disasters such as earthquakes, landslides, and debris flows, traditional ground transportation often fails due to road interruption, and UAVs play an increasingly key role in high-efficiency and flexible deployment in emergency rescue. However, the complex terrain of mountains, variable weather, and frequent signal blind areas bring great challenges to the safety of UAV take-off and landing and task scheduling. Therefore, selecting the riskiest and most efficient take-off and landing site is a core technical problem in the emergency UAV system), which includes the following steps:

[0055] S1, create n1 candidate areas in the research area. Take a typical mountain area (take a certain earthquake-prone mountain area in Sichuan as an example) as the research area, build a three-dimensional digital elevation model (DEM), and select several candidate areas.

[0056] S2, collect the spatio-temporal risk monitoring data of all candidate areas in the research area, input the constructed spatio-temporal non-homogeneous Poisson process wind disaster risk model, build the NHPP risk field of the spatio-temporal dimension, and calculate the NHPP risk cumulative intensity corresponding to the candidate area.

[0057] ​​In this embodiment, an earthquake-prone mountainous area in Sichuan Province is used as the study area. The spatiotemporal risk monitoring data includes historical wind monitoring spatiotemporal sequence data. The spatiotemporal nonhomogeneous Poisson process wind disaster risk model is used to construct the NHPP risk field according to the risk density function. The expression of the risk density function is as follows:

[0058] ,in Spatial location of mountainous areas ,time risk density, For event point i in the mountainous area, in time Location, Weights for historical event point i, Let be the wind speed at time t. For safe wind speed thresholds, This represents the scale of spatial impact diffusion.

[0059] Candidate region of the present invention The cumulative NHPP risk intensity in the interval [t0,T] The expression is as follows: If the study area is not a mountainous area, for If the study area is mountainous, for .

[0060] S3. Construct a multi-dimensional comprehensive evaluation system including P1 dimensions, encompassing the cumulative intensity of NHPP risks and construction costs, and obtain the evaluation value of the candidate region through weighted evaluation. The evaluation value of the candidate region in this embodiment... The expression is derived from the weighted calculation of dimensional indicators in a multi-dimensional comprehensive evaluation system, as follows: ,in Dimensional indicators The evaluation value, Dimensional indicators The weighting is as follows. In this embodiment, the study area is mountainous and mountain rescue is conducted (for simplicity, construction costs are set to zero). The multi-dimensional comprehensive evaluation system also includes two dimensions: communication quality evaluation and ground rescue accessibility evaluation. Communication quality evaluation and ground rescue accessibility evaluation are conducted using drones for edge wind field detection, aerial channel detection, and video recording, with the detection data fed back to the command center in real time for evaluation. Evaluation value of candidate areas. The expression is: ,in For communication quality assessment results, Based on the results of the ground rescue accessibility assessment, The cumulative intensity of NHPP risk. respectively are the weights of NHPP risk cumulative intensity, communication quality evaluation and ground rescue accessibility evaluation, for example, 0.5, 0.3, 0.2 respectively. The optimal candidate area is selected as the take-off and landing site selection area, or the decision threshold is set and the site selection decision result is output according to the decision threshold. In a certain earthquake-prone mountainous area in Sichuan as a research area, after being evaluated by the method of the embodiment, 3 alternative sites are screened out, and through subsequent multi-machine exploration verification, a temporary landing pad in a valley bottom open land (low risk, good signal and accessible) is finally selected, which effectively supports the emergency delivery and wounded evacuation tasks.

[0061] The above merely describes preferred embodiments of the present application and is not intended to limit the present application, and any modification, equivalent replacement and improvement made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for selecting a landing site for a UAV based on an NHPP risk field and multi-dimensional comprehensive evaluation, characterized in that: The method comprises: S1, creating n1 candidate areas in the study area; S2, collecting the spatio-temporal risk monitoring data of all candidate areas in the study area to construct the NHPP risk field of the spatio-temporal dimension in the constructed spatio-temporal non-homogeneous Poisson process wind disaster risk model, if the study area is a non-mountainous area, the spatio-temporal risk monitoring data contains historical wind monitoring spatio-temporal sequence data, and the spatio-temporal non-homogeneous Poisson process wind disaster risk model constructs the NHPP risk field according to the spatio-temporal intensity function, and the expression of the spatio-temporal intensity function is as follows: wherein is the expected frequency of occurrence of an event at a spatial location , time , is the risk impact of historical events on spatial location , time , is the additional contribution to risk when wind speed exceeds a safety threshold, historical events are wind monitoring data records; if the study area is a mountainous area, the spatio-temporal risk monitoring data include historical wind monitoring spatio-temporal sequence data, and the spatio-temporal non-homogeneous Poisson process wind disaster risk model constructs an NHPP risk field according to a risk density function, and the expression of the risk density function is as follows: wherein is the spatial position of the mountainous area , time , risk density, is the position of the mountainous area event point i at time , is the historical event point i weight, is the wind speed at time t, is the safety wind speed threshold, is the spatial influence diffusion scale; the NHPP risk cumulative intensity corresponding to the candidate area is calculated to obtain; S3, constructing a multi-dimensional comprehensive evaluation system containing the NHPP risk cumulative intensity and construction cost, containing P1 dimensions, and obtaining the evaluation value of the candidate area through weighted evaluation; selecting the optimal candidate area as the take-off and landing site selection area, or setting a decision threshold and outputting the site selection decision result according to the decision threshold. 2.The UAV landing site selection method based on NHPP risk field and multi-dimensional comprehensive evaluation according to claim 1, characterized in that: The risk impact The expression is as follows: wherein is a weight for historical event point i, is a location for historical event point i, is a decay rate with distance for historical event point i, is a coefficient of influence on risk when wind speed exceeds a safety threshold; The additional contribution The expression is as follows: wherein is a coefficient of influence on the risk when the wind speed exceeds the safety threshold, is the wind speed at time t, is the safety wind speed threshold. 3.The UAV landing site selection method based on NHPP risk field and multi-dimensional comprehensive evaluation according to claim 1, characterized in that: The candidate region NHPP risk cumulative intensity in the interval [t0, T] The expression is as follows: .

4. The unmanned aerial vehicle landing site selection method based on NHPP risk field and multi-dimensional comprehensive evaluation according to claim 1, characterized in that: The evaluation value of the candidate region The evaluation value of the candidate region is calculated based on the weighted calculation of the dimension indicators in the multi-dimension comprehensive evaluation system, and the expression is as follows: Wherein is the evaluation value of the dimension indicator , is the weight of the dimension indicator .

5. The unmanned aerial vehicle landing site selection method based on NHPP risk field and multi-dimensional comprehensive evaluation according to claim 1, characterized in that: The cumulative intensity of the NHPP risk is verified and compared by the following method: sample sites are selected, and meteorological stations, laser radars, and cameras are deployed at the sample sites through 5G cellular and LoRaWAN networks to monitor data and evaluate observed risks The predicted risks of the sample sites are calculated through a spatiotemporal nonhomogeneous Poisson process wind disaster risk model The difference between the predicted risks and the observed risks is obtained .

6. The unmanned aerial vehicle landing site selection method based on NHPP risk field and multi-dimensional comprehensive evaluation according to claim 5, characterized in that: Based on the difference Adjusting the parameters of the spatio-temporal non-homogeneous Poisson process wind disaster risk model, the expression is as follows: wherein is the original parameter, is the adjusted parameter, is the risk sensitivity threshold, is the learning rate or scaling factor.

7. The unmanned aerial vehicle landing site selection method based on NHPP risk field and multi-dimensional comprehensive evaluation according to claim 1, characterized in that: If the study area is a mountainous area and rescue is needed, the multi-dimensional comprehensive evaluation system further includes two dimensions of communication quality evaluation and ground rescue accessibility evaluation, the communication quality evaluation and the ground rescue accessibility evaluation are detected by an unmanned aerial vehicle through edge wind field detection, air channel detection and video shooting, and the detection data is fed back to the command center in real time for evaluation. 8.The unmanned aerial vehicle landing site selection method based on NHPP risk field and multi-dimensional comprehensive evaluation according to claim 1, wherein: The evaluation value of the candidate area is a positive evaluation value, and the larger the evaluation value of the candidate area, the better; the site selection decision result includes three kinds of recommendation, conditional recommendation and non-recommendation.

Images