An artificial intelligence-based low-altitude safety monitoring system and method

Abstract

The application discloses a low-altitude safety monitoring system and method based on artificial intelligence, and relates to the technical field of low-altitude safety monitoring.The system comprises an airspace division storage module, a target collision event screening module, a deep division module, a collision risk index volatility rate analysis module and a suspicious collision node response monitoring module.The airspace division storage module is used for storing the flight data of low-altitude equipment in each low-altitude airspace in history according to airspace division.The target collision event screening module is used for screening target collision events caused by collision index differences.The deep division module is used for deep division based on a collision risk index for the purpose of generating a collision type set.The collision risk index volatility rate analysis module is used for calculating the collision risk index volatility rate of the corresponding target time length before collision in each collision type set.The suspicious collision node response monitoring module is used for response monitoring of the collision index of the corresponding collision type set based on suspicious collision node updating.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of low-altitude safety monitoring, and particularly relates to a low-altitude safety monitoring system and method based on artificial intelligence. BACKGROUND

[0002] In the existing low-altitude safety monitoring system, the collision indicators, such as minimum safety distance and relative speed threshold, of different types of low-altitude flight equipment (such as industrial-grade unmanned aerial vehicles, gliders, manned aircrafts and the like) are significantly different. For example, when the industrial-grade unmanned aerial vehicle and the glider fly in the same airspace, the minimum horizontal safety distance of the unmanned aerial vehicle from other equipment can be 100 meters, while the corresponding standard of the glider can be 500 meters. The incompatibility of such collision indicators can cause abnormal collision risks. The existing system cannot optimize the monitoring logic in a targeted manner, resulting in low collision warning accuracy and response lag of different types of equipment in the same airspace. In addition, the existing technology lacks dynamic analysis capability for collision risks. It is difficult to determine the critical risk index and risk fluctuation rate of different collision types based on historical collision data, to identify potential risk nodes caused by collision indicator differences in advance, and to dynamically update the collision indicators to adapt to complex scenarios in different airspaces. Ultimately, the intelligent and accurate level of low-altitude safety monitoring is insufficient. SUMMARY

[0003] The application aims to provide a low-altitude safety monitoring system and method based on artificial intelligence to solve the problems in the prior art.

[0004] To achieve the above-mentioned purpose, the application provides the following technical scheme: a low-altitude safety monitoring method based on artificial intelligence, the method comprising the following specific steps:

[0005] Step S100: storing the historical low-altitude equipment flight data in each low-altitude airspace according to airspace division; extracting collision events recorded in each low-altitude airspace, and screening target collision events caused by collision indicator differences;

[0006] Step S200: preliminarily distinguishing the target collision events in each low-altitude airspace, extracting the low-altitude equipment flight data recorded in each target collision event to analyze the collision risk index, and performing deep distinction based on the collision risk index to generate a collision type set;

[0007] Step S300: marking collision events different from the target collision events as control collision events, analyzing the critical collision risk index of the control collision events, and determining a collision response interval duration; taking the collision response interval duration as a target duration, and calculating the collision risk index fluctuation rate of the corresponding target duration before collision in each collision type set;

[0008] Step S400: Based on the collision risk index fluctuation rate, determine the suspicious collision nodes corresponding to each collision type set of each low-altitude airspace, and update the collision index of the corresponding collision type set based on the suspicious collision nodes for response monitoring.

[0009] Further, the target collision event caused by the difference in collision index is screened, including the following specific steps:

[0010] Extract the flight data of each low-altitude device in the collision event, including position trajectory data, motion parameters, device state performance data and environmental data; and obtain the collision index and standard index data of each low-altitude device in each collision event;

[0011] Mark the low-altitude device in each collision event that has a collision relationship as a collision device, and compare the standard index data of each collision device based on the collision index, when the standard index data of the corresponding collision device under the same collision index is different, output the collision event as the first collision event;

[0012] Based on the first collision event, search the flight data recorded by each collision device, mark the first collision event corresponding to the flight data that meets the safety flight standard in device state performance data and environmental data, and the position trajectory data and motion parameters that meet the flight target setting.

[0013] The analysis of the target collision event in this application is to extract the collision event corresponding to the collision risk caused by the difference in obstacle avoidance index when the industrial-grade unmanned aerial vehicle and the glider fly in the same airspace; thereby realizing intelligent and safe monitoring of different types of flight devices in the same airspace, and solving the problem of abnormal collision caused by the difference in obstacle avoidance index.

[0014] Further, step S200 includes the following specific steps:

[0015] Step S210: preliminary differentiation refers to differentiating based on the collision device of the target collision event, and the target collision event is different for different collision devices; obtain the standard index data P1 and the actual index data P2 of the corresponding collision index of each collision device at the moment of collision of the target collision event, and use the formula:

[0016] W=a1*(P 21 / P 11 )+a2*(P 22 / P 12 )+....+a n *(P 2n / P 1n );

[0017] Calculate the collision risk index W at the moment of collision of the collision device, wherein a1, a2,...an a1, a2,... an represent the weight of the first, second,... nth collision indicator, a1+a2+..+an=1, P n 1n Pn represents the standard indicator data under the nth collision indicator, P 2n Pn represents the actual indicator data under the nth collision indicator;

[0018] Step S220: The collision devices of the same type are divided into a target set; the collision devices of the same type refer to at least two types of low-altitude flight devices in the target collision event, and any target collision event is taken as an initial event, the types of low-altitude flight devices in the initial event are extracted, the types of low-altitude flight devices recorded in other target collision events are searched, and if the device types are all the same, the collision devices in the two target collision events are output as the collision devices of the same type;

[0019] Step S230: The collision risk indexes of the collision devices in the same target collision event are extracted, the risk difference K is calculated, K=W max -W min , wherein W max represents the maximum value of the collision risk indexes of all the collision devices in the same target collision event, and W min represents the minimum value of the collision risk indexes of all the collision devices in the same target collision event; based on the target set, the dispersion coefficient R of the risk difference corresponding to each target collision event in the target set is calculated, R={[∑(K1-K0) 2 ] / m} 1 / 2 ; wherein K1 represents the risk difference of each target collision event in the target set, K0 represents the average value of the risk difference in the same target set, and m represents the total number of target collision events in the same set;

[0020] Step S240: A dispersion coefficient threshold R0 is set, when R≥R0, the target collision events with different collision device names in the target set are independently generated into a collision type set; when R<R0, the target set is taken as a collision type set.

[0021] Further, step S300 includes the following specific steps:

[0022] Step S310: The collision events with different causes refer to the collision events not caused by the difference in collision indicators, the collision risk indexes of the control collision events are calculated based on step S200, the specific collision devices included in the target collision events in each collision type set are taken as the range, and the minimum value of the recorded collision risk indexes of all the same collision devices under the control collision events is determined as the critical collision risk index corresponding to the collision type set;

[0023] ​Step S320: Taking the critical collision risk index as the starting monitoring point of each control collision event in the same collision type set, the starting monitoring point refers to extracting and calculating the collision risk index of the flight data before the collision in each control collision event until the corresponding flight time when the collision risk index is determined to be the same or the difference is less than the difference threshold; output the maximum interval time length between the starting monitoring point and the collision time in the same collision type set as the collision response interval time length T1;

[0024] Step S330: Taking the collision time of each target collision event in the collision type set as the search point, search forward for the collision risk index of the target time length; calculate the first risk fluctuation index of each target collision event in the same collision type set using the formula: Z=(W 碰撞 -W 起始 ) / T1, wherein W 碰撞 represents the collision risk index corresponding to the collision time in the target time length, and W 起始 represents the collision risk index corresponding to the starting time in the target time length; obtain the average value Z0 of the first risk fluctuation index corresponding to the same collision type set as the collision risk index fluctuation rate of the corresponding target time length.

[0025] Analyzing the collision risk index fluctuation rate can dynamically find the data fluctuation change based on the collision risk index before the low-altitude flying device is about to collide from the data angle; and make early warning response for subsequent analysis of how to produce collision risk under the analysis of different types of low-altitude devices in the same domain.

[0026] Further, step S400 includes the following specific steps:

[0027] Step S410: Extract the output node in each collision type set within the target time length to obtain the flight data to calculate the collision risk index, calculate the first risk fluctuation index with the time length formed by adjacent output nodes, form the fluctuation index set A of the output nodes other than the starting node and the collision node, A=(Z 前 ,Z 后 ), Z 前 represents the first risk fluctuation index calculated by the interval time length between the output node and the previous adjacent node, and Z 后 represents the first risk fluctuation index calculated by the interval time length between the output node and the next adjacent node;

[0028] Step S420: When there is Z 前 <Z0 and Z 后When Z0, the output node responding to the first collision node is determined, all output nodes in the same collision type set are determined, and the output node closest to the starting monitoring node of the target time length distance of the first collision node is selected as the suspicious collision node; The purpose of analyzing the suspicious collision node is to determine the data node that can be used as an update collision index from the historical data record flight data, and in the real-time monitoring process, the determination of the collision node is based on the relationship between the above fluctuation index and the average value Z0;

[0029] Step S430: Extract the actual index data of each collision index in the same collision type set at the corresponding time of the suspicious collision node, and select the maximum limit value of the actual data to replace the standard index data of each collision index in the corresponding collision type set; The maximum limit value refers to the maximum difference value between the actual index data and the standard index data corresponding to the actual index data when the difference value between the actual index data and the standard index data of each target collision event in the same collision type set at the corresponding time of the suspicious collision node is the largest.

[0030] An artificial intelligence-based low-altitude safety monitoring system, the system comprising an airspace division storage module, a target collision event screening module, a deep division module, a collision risk index fluctuation rate analysis module, and a suspicious collision node response monitoring module;

[0031] The airspace division storage module is used to store the flight data of low-altitude equipment in each low-altitude airspace according to airspace division.

[0032] The target collision event screening module is used to extract the collision events recorded in each low-altitude airspace and screen the target collision events caused by collision index differences.

[0033] The deep division module is used for deep division based on the collision risk index for the purpose of generating a collision type set.

[0034] The collision risk index fluctuation rate analysis module is used to calculate the collision risk index fluctuation rate of the corresponding target time length before collision in each collision type set.

[0035] The suspicious collision node response monitoring module is used to determine the suspicious collision node corresponding to each collision type set in each low-altitude airspace, and to update the collision index of the corresponding collision type set based on the suspicious collision node for response monitoring.

[0036] Further, the deep division module comprises a preliminary division unit, a collision risk index calculation unit, a risk difference discrete coefficient analysis unit, and a collision type set output unit.

[0037] The preliminary division unit is used to divide based on the collision equipment of the target collision event,

[0038] The collision risk index calculation unit is used to calculate the collision risk index at the time of collision of the collision equipment.

[0039] The risk difference discrete coefficient analysis unit is configured to extract the collision risk indexes of the collision devices in the same target collision event, calculate the risk difference, and calculate the discrete coefficient of the risk difference corresponding to each target collision event in the target set;

[0040] The collision type set output unit is configured to, based on the relationship between the discrete coefficient and the threshold, independently generate a collision type set for the target collision events with different collision device names in the target set or take the target set as a collision type set.

[0041] Further, the collision risk index fluctuation rate analysis module includes a critical collision risk index determination unit, a collision response interval time length determination unit, and a first risk fluctuation index.

[0042] The critical collision risk index determination unit is configured to determine the minimum value of the recorded collision risk indexes of all the same collision devices in the contrast collision event as the critical collision risk index of the corresponding collision type set.

[0043] The collision response interval time length determination unit is configured to output the maximum interval time length between the starting monitoring point and the collision time in the same collision type set as the collision response interval time length.

[0044] The first risk fluctuation index is configured to take the collision time of each target collision event in the collision type set as a search point, search forward for the collision risk indexes within the target time length, calculate the first risk fluctuation index of each target collision event in the same collision type set, and obtain the average value Z0 of the first risk fluctuation indexes corresponding to the same collision type set as the collision risk index fluctuation rate of the corresponding target time length.

[0045] Further, the suspicious collision node response monitoring module includes a fluctuation index array generation unit, a first collision node analysis unit, and an update response unit.

[0046] The fluctuation index array generation unit is configured to extract the output nodes in each collision type set for obtaining the flight data to calculate the collision risk indexes within the target time length, calculate the first risk fluctuation index based on the time length formed by the adjacent output nodes, and form the fluctuation index array of the output nodes other than the starting node and the collision node.

[0047] The first collision node analysis unit is configured to determine the first collision node based on the data of the fluctuation index array, judge all the output nodes in the same collision type set, and select the output node closest to the starting monitoring node in the target time length where the first collision node is located as the suspicious collision node.

[0048] The updating response unit is used for extracting actual index data of each collision index in the same collision type set corresponding to the suspicious collision node at the moment, and selecting the maximum limit value of the actual data to replace the standard index data of each collision index in the corresponding collision type set.

[0049] Compared with the prior art, the present application has the following advantages:

[0050] By screening the "target collision event", the present application excludes factors such as device self-failure and environmental interference, and focuses only on collisions caused by different collision index standards, thereby achieving accurate positioning of the collision reasons of different types of devices in the same airspace, and solving the problem of "ambiguous collision reasons and difficulty in targeted prevention and control" in the existing system.

[0051] By collision risk index calculation and dispersion coefficient analysis, the collision event is divided into different "collision type sets", so that the system can develop differentiated monitoring strategies for the collision risks of different device combinations such as industrial-grade unmanned aerial vehicles and gliding parachutes, and helicopters and consumer-grade unmanned aerial vehicles, thereby avoiding "false positives / misses caused by unified standard monitoring";

[0052] By analyzing the critical collision risk index, the collision risk index fluctuation rate and the suspicious collision node, the system can extract the risk fluctuation characteristics before the collision from the historical data, and update the standard index based on the actual index data of the suspicious collision node, such as expanding the minimum safety distance threshold, thereby achieving dynamic adaptation of the collision index and solving the problem of "standard solidification leading to incompatible monitoring in the same domain"; by calculating the collision response interval length and the risk index fluctuation rate, the system can identify the risk fluctuation trend within the "target time length" before the collision, locate the suspicious collision node in advance and trigger the response monitoring, thereby upgrading the traditional "post-tracing" to "pre-warning", and significantly improving the initiative and foresight of low-altitude safety monitoring.

[0053] The present application is suitable for complex airspace scenarios and improves the intelligent level of monitoring the same domain of multiple types of devices; for different airspaces, such as urban dense areas and mountainous areas, and different device combinations, such as unmanned aerial vehicles and hot air balloons, manned aircraft and power parachutes, the system can achieve adaptive monitoring through the updated collision index, thereby solving the intelligent monitoring problem caused by "diverse device types and fragmented standards", and providing technical support for the safe development of low-altitude economy. BRIEF DESCRIPTION OF DRAWINGS

[0054] Figure 1 The figure is a structural schematic diagram of the low-altitude safety monitoring method based on artificial intelligence. DETAILED DESCRIPTION

[0055] All other embodiments obtained by those of ordinary skill in the art without creative labor based on the embodiments of the present application fall within the scope of protection of the present application.

[0056] Embodiment: As shown in the figure, the present application provides a low-altitude safety monitoring method based on artificial intelligence, which comprises the following specific steps: Figure 1

[0057] Step S100: Store the flight data of low-altitude equipment in each low-altitude airspace according to airspace division; extract collision events recorded in each low-altitude airspace, and screen target collision events caused by differences in collision indicators;

[0058] Step S200: Preliminarily distinguish target collision events under each low-altitude airspace, extract low-altitude equipment flight data recorded in each target collision event to analyze collision risk indexes, and perform deep distinction based on the collision risk indexes for the purpose of generating a collision type set;

[0059] Step S300: Mark collision events different from the cause of the target collision event as control collision events, analyze the critical collision risk index of the control collision event, and determine the collision response interval duration; take the collision response interval duration as the target duration, and calculate the collision risk index fluctuation rate of the corresponding target duration before collision in each collision type set;

[0060] Step S400: Based on the collision risk index fluctuation rate, determine the suspicious collision node corresponding to each collision type set in each low-altitude airspace, and update the collision indicators of the corresponding collision type set based on the suspicious collision node for response monitoring.

[0061] Screening target collision events caused by differences in collision indicators includes the following specific steps:

[0062] Extract the flight data of each low-altitude equipment in the collision event, which includes position trajectory data, motion parameters, equipment state performance data and environmental data; the low-altitude equipment flight data stored in the present application is obtained by flight test in a safe and compliant airspace; and obtain the collision indicators and standard indicator data of each low-altitude equipment in each collision event;

[0063] Mark each low-altitude equipment in each collision event that has a collision relationship as a collision equipment, and compare the standard indicator data of each collision equipment based on the collision indicators; when the standard indicator data of the corresponding collision equipment under the same collision indicator is different, output the collision event as a first collision event;

[0064] ​Based on the first collision event, the flight data of each collision device record is searched, and the target collision event is marked as the first collision event in which the device state performance data and the environment data in the flight data meet the safety flight standard, and the position trajectory data and the motion parameters meet the flight target setting.

[0065] The analysis of the target collision event in the application is to extract the collision event corresponding to the collision risk caused by the difference in obstacle avoidance indicators when the industrial unmanned aerial vehicle and the glider fly in the same airspace, so as to realize intelligent and safe monitoring of different types of flight devices in the same airspace and solve the problem of abnormal collision caused by the difference in obstacle avoidance indicators.

[0066] The target collision event is determined after excluding the collision events caused by the problems of the flight device itself, such as the speed, acceleration, historical trajectory and predicted trajectory in the position trajectory data; in the monitoring process, the speed and acceleration meet the speed set by the flight target, the deviation of the historical trajectory and the predicted trajectory also meets the set standard, and the device state performance data includes the power in the hardware and the motor monitoring in the software system, which all meet the flight standard, but the collision occurs, and the collision indicators of the corresponding collision device also exist differences, such as the minimum horizontal safety distance, the standard indicator data of the unmanned aerial vehicle is: ≥ 300 meters with manned aircraft; ≥ 100 meters with other unmanned aerial vehicles; the standard indicator data of the glider is: ≥ 500 meters with other devices (regardless of type); at this time, it is indicated that the collision is caused by the difference in collision indicators. For example, the distance between the unmanned aerial vehicle and the glider is 400 meters, which meets the collision indicator of the unmanned aerial vehicle, but does not meet the collision indicator of the glider, so the collision is caused.

[0067] Step S200 includes the following specific steps:

[0068] Step S210: preliminary classification means classifying based on the collision device of the target collision event, and the collision device is different for different target collision events; the standard indicator data P1 and the actual indicator data P2 of the corresponding collision indicator of each collision device at the time of the target collision event are obtained, and the formula is used:

[0069] W = a1*(P 21 / P 11 )+a2*(P 22 / P 12 )+....+a n *(P 2n / P 1n );

[0070] The collision risk index W of the collision device at the time of the collision is calculated, wherein a1, a2,... a n represent the weights of the corresponding first, second,... n collision indicators, and a1+a2+..+an =1, P 1n represents the standard index data under the nth collision index, P 2n represents the actual index data under the nth collision index;

[0071] As shown in the examples: when an industrial-grade unmanned aerial vehicle flies in urban airspace, the safety distance is 300 meters (threshold value 500 meters, weight 40%), and the relative speed is 100 km / h (threshold value 80 km / h, weight 30%); then W=0.4x(300 / 500)+0.3x(100 / 80)=0.24+0.375=0.615;

[0072] Step S220: classifying the collision devices of the same type into a target set; the collision devices of the same type refer to at least two types of low-altitude flying devices in a target collision event, taking any target collision event as an initial event, extracting the type of low-altitude flying device in the initial event, searching for the type of low-altitude flying device recorded in other target collision events, and if the device types are all the same, outputting that the collision devices in the two target collision events are collision devices of the same type;

[0073] For example, low-altitude devices are classified into manned aerial vehicles, unmanned aerial vehicles, etc.; manned aerial vehicles include helicopters, paragliders, etc., and unmanned aerial vehicles include industrial-grade unmanned aerial vehicles, consumer-grade unmanned aerial vehicles, etc.; the collision devices in collision event 1 are paragliders and industrial-grade unmanned aerial vehicles, and the collision devices in collision event 2 are helicopters and consumer-grade unmanned aerial vehicles, so that, after type analysis, collision event 1 and collision event 2 can be classified into a target set;

[0074] Step S230: extracting the collision risk index of each collision device under the same target collision event, calculating the risk difference K, K=W max -W min , wherein W max represents the maximum value of the collision risk index of all collision devices under the same target collision event, and W min represents the minimum value of the collision risk index of all collision devices under the same target collision event; based on the target set, the dispersion coefficient R of the risk difference of each target collision event in the target set is calculated, R={[∑(K1-K0) 2 ] / m} 1 / 2 ; wherein K1 represents the risk difference of each target collision event in the target set, K0 represents the average value of the risk difference in the same target set, and m represents the total number of target collision events in the same set;

[0075] Step S240: set a discrete coefficient threshold R0, when R≥R0, independently generate a collision type set for the target collision events with different device names in the target set; when R<R0, take the target set as a collision type set.

[0076] Step S300 includes the following specific steps:

[0077] Step S310: the collision events with different causes refer to the collision events not caused by the collision index difference, based on the collision risk index of the contrast collision event calculated in step S200, taking the specific collision device contained in each target collision event in each collision type set as the range, determining the minimum value of the collision risk index recorded by all the same collision devices under the contrast collision event as the critical collision risk index of the corresponding collision type set;

[0078] Step S320: take the critical collision risk index as the starting monitoring point of each contrast collision event in the same collision type set, the starting monitoring point refers to extracting and calculating the collision risk index of the flight data before the collision in each contrast collision event until the corresponding flight time when the critical collision risk index is determined or the difference is less than the difference threshold; output the maximum interval time length between the starting monitoring point and the collision time in the same collision type set as the collision response interval time length T1;

[0079] Step S330: take the collision time of each target collision event in the collision type set as the search point, search backward for the collision risk index of the target time length; calculate the first risk fluctuation index of each target collision event in the same collision type set by using the formula: Z=(W 碰撞 -W 起始 ) / T1, wherein W 碰撞 represents the collision risk index corresponding to the collision time in the target time length, W 起始 represents the collision risk index corresponding to the starting time in the target time length; take the average value Z0 of the first risk fluctuation index corresponding to the same collision type set as the collision risk index fluctuation rate of the corresponding target time length.

[0080] Analyzing the collision risk index fluctuation rate can dynamically find the data fluctuation change based on the collision risk index before the low-altitude flying device is about to collide from the data angle; and make early warning response for subsequent analysis of how to produce collision risk under the analysis of different types of low-altitude devices in the same domain.

[0081] Step S400 includes the following specific steps:

[0082] Step S410: Extract the output nodes in each collision type set within the target time length to calculate the flight data collision risk index, and calculate the first risk fluctuation index with adjacent output nodes, form the fluctuation index set A of the output nodes except the starting node and the collision node, A=(Z 前 后 ), Z 前 represents the first risk fluctuation index calculated by the interval time length between the output node and the previous adjacent node, and Z 后 represents the first risk fluctuation index calculated by the interval time length between the output node and the next adjacent node.

[0083] Step S420: When Z 前 <Z0 and Z 后 ≥Z0, respond to the output node as the first collision node, judge all output nodes in the same collision type set, and select the output node closest to the starting monitoring node in the target time length where the first collision node is located as the suspicious collision node; The purpose of analyzing the suspicious collision node is to determine the data node that can be used as the updated collision index from the historical data record flight data, and in the real-time monitoring process, the determination of the collision node is based on the relationship between the above fluctuation index set and the average value Z0;

[0084] Step S430: Extract the actual index data of each collision index based on the same collision type set at the corresponding time of the suspicious collision node, and select the maximum limit value of the actual data to replace the standard index data of each collision index in the corresponding collision type set; The maximum limit value refers to the actual index data corresponding to the maximum difference between the actual index data and the standard index data of all target collision events in the same collision type set at the corresponding suspicious collision node time.

[0085] In this application, the collision indexes corresponding to different types of low-altitude equipment are the same, and the specific standard index data is different.

[0086] For example, based on the collision index of the minimum horizontal safety distance, determine the same collision type set including target collision event 1 and target collision event 2.

[0087] The actual index data of the unmanned aerial vehicle recorded by the target collision event 1 at the suspicious collision node is 450m away from the manned aircraft, and the actual index data of the glider is 600m away from the unmanned aerial vehicle.

[0088] The actual index data of the consumer unmanned aerial vehicle recorded by the target collision event 2 at the suspicious collision node is 520m away from the helicopter, and the actual index data of the helicopter is 650m away from the consumer unmanned aerial vehicle.

[0089] ​The maximum difference between the standard index data of the corresponding type of device collision indicator and the actual index data corresponding to the target collision event 2 is the actual index data corresponding to the target collision event 2; therefore, the original standard index data of the unmanned aerial vehicle is: ≥ 300 meters from manned aircraft; the standard index data of the manned aircraft is: ≥ 500 meters from other devices (regardless of type); the update is: the standard index data of the unmanned aerial vehicle is: ≥ 520 meters from manned aircraft; the standard index data of the manned aircraft is: ≥ 650 meters from other devices (regardless of type).

[0090] An artificial intelligence-based low-altitude safety monitoring system, the system comprising an airspace division storage module, a target collision event screening module, a deep division module, a collision risk index volatility rate analysis module, and a suspicious collision node response monitoring module;

[0091] The airspace division storage module is used to store the flight data of low-altitude devices in each low-altitude airspace according to airspace division.

[0092] The target collision event screening module is used to extract collision events recorded in each low-altitude airspace and screen target collision events caused by collision indicator differences.

[0093] The deep division module is used to perform deep division based on the collision risk index for the purpose of generating a collision type set.

[0094] The collision risk index volatility rate analysis module is used to calculate the collision risk index volatility rate of the corresponding target duration before collision in each collision type set.

[0095] The suspicious collision node response monitoring module is used to determine the suspicious collision nodes corresponding to each collision type set in each low-altitude airspace, and to update the collision indicators of the corresponding collision type set based on the suspicious collision nodes for response monitoring.

[0096] The deep division module comprises a preliminary division unit, a collision risk index calculation unit, a risk difference discrete coefficient analysis unit, and a collision type set output unit.

[0097] The preliminary division unit is used to divide based on the collision devices of the target collision event,

[0098] The collision risk index calculation unit is used to calculate the collision risk index at the moment of collision of the collision device.

[0099] The risk difference discrete coefficient analysis unit is used to extract the collision risk indices of each collision device under the same target collision event, calculate the risk difference, and calculate the discrete coefficient of the risk difference corresponding to each target collision event in the target set.

[0100] The collision type set output unit is used to independently generate a collision type set for target collision events with different collision device names in the target set or to treat the target set as a collision type set based on the relationship between the discrete coefficient and the threshold.

[0101] The collision risk index volatility analysis module includes a critical collision risk index determination unit, a collision response interval duration determination unit, and a first risk volatility index;

[0102] The critical collision risk index determination unit is used to determine the minimum value of the collision risk index recorded by all the same collision devices under the control collision event as the critical collision risk index of the corresponding collision type set;

[0103] The collision response interval duration determination unit is used to output the maximum interval duration between the starting monitoring point and the collision occurrence moment in the same collision type set as the collision response interval duration;

[0104] The first risk fluctuation index is used to use the collision moment of each target collision event in the collision type set as the retrieval point to search forward for the collision risk index of the target duration; calculate the first risk fluctuation index of each target collision event in the same collision type set, and obtain the average value Z0 of the first risk fluctuation index corresponding to the same collision type set as the collision risk index volatility of the corresponding target duration.

[0105] The suspicious collision node response monitoring module includes a fluctuation index group generation unit, a first collision node analysis unit and an update response unit;

[0106] The fluctuation index group generating unit is used to extract the output nodes in each collision type set that obtain flight data within the target duration to calculate the collision risk index, calculate the first risk fluctuation index based on the duration formed by adjacent output nodes, and form a fluctuation index group for the output nodes other than the starting node and the collision node;

[0107] The first collision node analysis unit is used to determine the first collision node based on the data of the fluctuation index group, judge all output nodes in the same collision type set, and select the output node whose target time distance of the first collision node is closest to the starting monitoring node as the suspicious collision node;

[0108] The update response unit is used to extract the actual indicator data of each collision indicator in the same collision type set at the corresponding moment of the suspicious collision node, select the maximum value of the actual data to replace and update the standard indicator data of each collision indicator in the corresponding collision type set.

[0109] Finally, it should be noted that the above only describes the preferred embodiments of the present application and is not intended to limit the present application. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art will appreciate that the technical solutions described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalent ones. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A low-altitude safety monitoring method based on artificial intelligence, characterized by: The method comprises the following specific steps: Step S100: storing the historically stored low-altitude equipment flight data in each low-altitude airspace separately according to the airspace division; extracting the collision events recorded in each low-altitude airspace, and screening the target collision events caused by the difference in collision indicators; Step S200: Preliminarily distinguish target collision events in each low-altitude airspace, then extract the low-altitude equipment flight data recorded for each target collision event to analyze the collision risk index, and perform in-depth differentiation based on the collision risk index to generate a collision type set; Obtain the standard index data P1 and actual index data P2 of each collision device corresponding to the collision index when the target collision event occurs, using the formula: W=a1*(P 21 / P 11 )+a2*(P 22 / P 12 )+....+a n *(P 2n / P 1n ); Calculate the collision risk index W at the moment of collision of the collision equipment, where a1, a2, ...a n Indicates the weights corresponding to the 1st, 2nd, ...nth collision indicators, a1+a2+..+a n =1,P 1n Indicates the standard index data under the nth collision index, P 2n Indicates the actual indicator data under the nth collision indicator; Step S300: Marking collision events with different causes from the target collision event as control collision events, analyzing the critical collision risk index of the control collision event, and determining the collision response interval duration; using the collision response interval duration as the target duration, calculating the collision risk index volatility corresponding to the target duration before collision in each collision type set; Taking the collision time of each target collision event in the collision type set as the retrieval point, search forward for the collision risk index of the target duration; using the formula: Z=(W 碰撞 -W 起始 ) / T1, calculate the first risk fluctuation index of each target collision event in the same collision type set, where W 碰撞 represents the collision risk index corresponding to the collision moment in the target duration, W 起始 represents the collision risk index corresponding to the starting time in the target duration, T1 is the collision response interval duration; the average value Z0 of the first risk volatility index corresponding to the same collision type set is obtained as the collision risk index volatility of the corresponding target duration; Step S400: Based on the collision risk index volatility, determine the suspicious collision nodes corresponding to each collision type set in each low-altitude airspace, and update the collision indicators of the corresponding collision type set based on the suspicious collision nodes for response monitoring; The step S400 includes the following specific steps: Step S410: Extract the output nodes of each collision type set that obtain flight data within the target duration to calculate the collision risk index, calculate the first risk fluctuation index based on the duration formed by adjacent output nodes, and form a fluctuation index group A of the output nodes other than the starting node and the collision node, A=(Z 前 ,Z 后 ), Z 前 Represents the first risk fluctuation index calculated based on the time interval between the output node and the previous adjacent node, Z 后 The first risk fluctuation index calculated by the time interval between the output node and the next adjacent node; Step S420: When there is Z 前 <Z0 and Z 后 ≥Z0, in response to the output node being the first collision node, judge all output nodes in the same collision type set, and select the output node that is the closest to the starting monitoring node in the target time duration where the first collision node is located as the suspicious collision node; Step S430: extract the actual indicator data of each collision indicator in the same collision type set at the corresponding moment of the suspicious collision node, select the maximum value of the actual data to replace and update the standard indicator data of each collision indicator in the corresponding collision type set; the maximum value refers to the actual indicator data corresponding to the maximum difference between all actual indicator data and the standard indicator data obtained through each target collision event in the same collision type set at the corresponding suspicious collision node moment.

2. The low-altitude safety monitoring method based on artificial intelligence according to claim 1, characterized in that: The screening of target collision events caused by differences in collision indicators includes the following specific steps: Extracting flight data of each low-altitude device involved in a collision event, including position trajectory data, motion parameters, device status performance data, and environmental data; and obtaining collision indicators and standard indicator data of each low-altitude device involved in each collision event; Marking the low-altitude equipment that has a collision relationship in each collision event as a collision equipment, comparing the standard indicator data of each collision equipment based on the collision index, and if the standard indicator data corresponding to different collision equipment under the same collision index are different, outputting the collision event as the first collision event; Based on the first collision event, the flight data recorded by each collision device is searched, and the first collision event corresponding to the device status performance data and environmental data in the flight data meeting the safe flight standards and the position trajectory data and motion parameters meeting the flight target setting is marked as a target collision event.

3. The low-altitude safety monitoring method based on artificial intelligence according to claim 1, characterized in that: The step S200 includes the following specific steps: Step S210: The preliminary differentiation refers to differentiation based on the collision device of the target collision event, and different collision devices have different target collision events; Step S220: Divide the collision devices of the same type into a target set; the collision devices of the same type refer to that there are at least two types of low-altitude flight devices in the target collision event. Taking any target collision event as the initial event, extract the types of low-altitude flight devices in the initial event, and retrieve the types of low-altitude flight devices recorded in other target collision events. If all the device types are the same, then the collision devices in the two target collision events are output as the collision devices of the same type. Step S230: Extract the collision risk index of each collision device under the same target collision event and calculate the risk difference K, K=W max -W min , where W max Indicates the maximum collision risk index of all collision devices under the same target collision event, W min Indicates the minimum collision risk index of all collision devices under the same target collision event; Based on the target set, the discrete coefficient R of the risk difference corresponding to the collision event of each target in the target set is calculated, R={[∑(K1-K0) 2 ] / m} 1 / 2 ; Where K1 represents the risk difference of each target collision event in the target set, K0 represents the average risk difference in the same target set, and m represents the total number of target collision events in the same set; Step S240: Set the discrete coefficient threshold R0. When R≥R0, independently generate a collision type set for the target collision events with different collision device names in the target set. When R<R0, use the target set as a collision type set.

4. The low-altitude safety monitoring method based on artificial intelligence according to claim 3 is characterized in that: The step S300 includes the following specific steps: Step S310: The collision events with different causes refer to the collision events not caused by the difference in collision indicators. Based on step S200, calculate the collision risk index of the control collision events. Taking the specific collision devices included in the target collision events in each collision type set as the scope, determine the minimum value of the collision risk index recorded for all the same collision devices under the control collision events as the critical collision risk index corresponding to the collision type set. Step S320: Use the critical collision risk index as the starting monitoring point for each control collision event in the same collision type set. The starting monitoring point refers to extracting and calculating the collision risk index for the flight data before the collision in each control collision event until the flight moment corresponding to the same critical collision risk index or a difference less than the difference threshold is determined; output the maximum interval duration between the starting monitoring point and the collision moment in the same collision type set as the collision response interval duration T1.

5. An artificial intelligence-based low-altitude safety monitoring system, such as using an artificial intelligence-based low-altitude safety monitoring method according to any one of claims 1 to 4, characterized in that: The system includes an airspace division and storage module, a target collision event screening module, a depth differentiation module, a collision risk index volatility analysis module, and a suspicious collision node response monitoring module; The airspace division and storage module is used to store the flight data of low-altitude devices in each historical low-altitude airspace separately according to airspace division; The target collision event screening module is used to extract the collision events recorded in each low-altitude airspace and screen the target collision events caused by the difference in collision indicators; The depth differentiation module is used for depth differentiation aiming at generating a collision type set based on the collision risk index; The collision risk index volatility analysis module is used to calculate the collision risk index volatility corresponding to the target duration before the collision in each collision type set; The suspicious collision node response monitoring module is used to determine the suspicious collision nodes corresponding to each collision type set in each low-altitude airspace and perform response monitoring based on the suspicious collision nodes to update the collision indicators of the corresponding collision type sets.

6. The artificial intelligence-based low-altitude safety monitoring system according to claim 5, characterized in that: The depth differentiation module includes a preliminary differentiation unit, a collision risk index calculation unit, a risk difference discrete coefficient analysis unit, and a collision type set output unit; The preliminary differentiation unit is used to differentiate based on the collision devices of the target collision events; The collision risk index calculation unit is used to calculate the collision risk index at the moment of collision of the collision devices. The risk difference dispersion coefficient analysis unit is used to extract the collision risk index of each collision device under the same target collision event, calculate the risk difference, and calculate the dispersion coefficient of the risk difference corresponding to each target collision event in the target set; The collision type set output unit is used to independently generate a collision type set for target collision events with different collision device names in the target set or to treat the target set as a collision type set based on the relationship between the discrete coefficient and the threshold.

7. The artificial intelligence-based low-altitude safety monitoring system according to claim 6, characterized in that: The collision risk index volatility analysis module includes a critical collision risk index determination unit, a collision response interval duration determination unit, and a first risk volatility index; The critical collision risk index determination unit is used to determine the minimum value of the collision risk index recorded by all the same collision devices under the control collision event as the critical collision risk index corresponding to the collision type set; The collision response interval duration determination unit is used to output the maximum interval duration between the starting monitoring point and the collision occurrence moment in the same collision type set as the collision response interval duration; The first risk fluctuation index is used to use the collision moment of each target collision event in the collision type set as the retrieval point to search forward for the collision risk index of the target duration; calculate the first risk fluctuation index of each target collision event in the same collision type set, and obtain the average value Z0 of the first risk fluctuation index corresponding to the same collision type set as the collision risk index volatility of the corresponding target duration.

8. The artificial intelligence-based low-altitude safety monitoring system according to claim 6, characterized in that: The suspicious collision node response monitoring module includes a fluctuation index group generation unit, a first collision node analysis unit and an update response unit; The fluctuation index group generating unit is used to extract the output nodes in each collision type set that obtain flight data within the target duration to calculate the collision risk index, calculate the first risk fluctuation index based on the duration formed by adjacent output nodes, and form a fluctuation index group for the output nodes other than the starting node and the collision node; The first collision node analysis unit is used to determine the first collision node based on the data of the fluctuation index group, judge all output nodes in the same collision type set, and select the output node whose target time distance of the first collision node is closest to the starting monitoring node as the suspicious collision node; The update response unit is used to extract actual indicator data of each collision indicator in the same collision type set at the corresponding moment of the suspicious collision node, select the maximum value of the actual data to replace and update the standard indicator data of each collision indicator in the corresponding collision type set.

Images