CN-121981533-A - Low-altitude landing field bidirectional safety verification and early warning method based on data feature fusion

CN121981533ACN 121981533 ACN121981533 ACN 121981533ACN-121981533-A

Abstract

The application belongs to the technical field of low-altitude economy, and provides a low-altitude landing field bidirectional safety verification and early warning method based on data feature fusion, which comprises the following steps: through collecting heterogeneous environmental parameters of multisource and landing guarantee data and preprocessing, a consistency data set is formed, cross check logic of the environmental parameters on the landing guarantee data is respectively constructed aiming at four core scenes of microclimate, road surface, airspace and equipment environment, when data distortion is found through check, the corresponding environment is further subjected to anti-evidence check by utilizing the landing guarantee data, bidirectional verification of the environment and the data is realized, finally, credibility judgment is carried out based on consistency indexes, and a five-stage grading response strategy is executed by a dynamic airspace management system. The application can obviously improve the data credibility, hidden danger identification accuracy and operation response efficiency, and provides reliable technical support for low-altitude take-off and landing occasion regularization and safe operation.

Inventors

XU YI
XU XIAOWEI
XU SHU
GUO XUZHOU
LIU SIXIAN
WANG FULIN
ZHANG ZHENKUN
YANG PEI

Assignees

南京熊猫电子股份有限公司

Dates

Publication Date: 20260505
Application Date: 20260108

Claims (10)

1. The low-altitude take-off and landing field bidirectional safety verification and early warning method based on data feature fusion is characterized by comprising the following steps of: s1, acquiring multi-source heterogeneous environment parameters and lifting guarantee data, and preprocessing the environment parameters and the lifting guarantee data to generate a consistency data set; S2, respectively constructing cross checking logic of environmental parameters of four risk scenes on the lifting security data based on the consistency data set, wherein the four risk scenes comprise a microclimate scene, a road surface scene, an airspace scene and an equipment environment scene; s3, based on the cross check logic of the four risk scenes, cross check is carried out on the lifting and falling guarantee data through corresponding environment parameters in each risk scene, and when the cross check result is data distortion, anti-evidence check is carried out on the corresponding environment parameters through the lifting and falling guarantee data, and an anti-evidence check result is generated to identify environment abnormality; and S4, calculating consistency indexes of the low-altitude landing field environment and landing guarantee data according to the cross check result and the anti-evidence check result in each risk scene, performing credibility judgment on the low-altitude landing field operation state, and triggering corresponding risk early warning and grading response according to the credibility judgment result.
2. The method of claim 1, wherein the classifying preprocessing of the multisource heterogeneous environmental parameters and the landing assurance data in S1 to generate a consistency dataset specifically comprises: denoising the environment parameters and the numerical data in the lifting guarantee data; Sequentially carrying out Gaussian denoising and contour extraction on the image data in the environment parameters and the lifting guarantee data; and mapping the environment parameters and the lifting guarantee data to a unified time axis based on a Network Time Protocol (NTP) timestamp, filling the missing data by adopting linear interpolation, and generating the consistency data set.
3. The method of claim 1, wherein when the risk scenario is a microclimate scenario, the S2 is executed to construct cross-checking logic of environmental parameters of four risk scenarios on the take-off and landing security data based on the consistency data set, respectively, and specifically includes: extracting environmental parameter wind speed from the consistent dataset Airplane attitude angle based on take-off and landing guarantee data ; Based on low-altitude landing field historical data, constructing by adopting least square fitting method The linear association model is specifically expressed as: ; Wherein, the Is the wind speed coefficient of the attitude angle, Is a reference attitude angle; Inputting the current wind speed acquired in real time into the device Outputting a corresponding predicted value of the attitude angle of the airplane by using the linear correlation model ; According to the predicted value of the attitude angle of the airplane And real-time measured attitude angle actual value The deviation rate is determined, and is specifically expressed as: ; Wherein, the Is the deviation rate; If the deviation rate is And if the data distortion is larger than 5%, judging that the attitude angle of the unmanned aerial vehicle has data distortion.
4. The method of claim 1, wherein when the risk scenario is a road surface scenario, the S2 is configured to construct cross-checking logic of environmental parameters of four risk scenarios on the take-off and landing security data based on the consistency data set, respectively, and specifically includes: Extracting environmental parameters of the road surface water accumulation depth H and the road surface image water accumulation depth D in the lifting guarantee data from the consistency data set; In the water accumulation state, if H is more than D× (1+5%), judging that the data distortion exists in a sensor for acquiring the road surface water accumulation depth H, and correcting the road surface water accumulation depth H by using the road surface image water accumulation depth D.
5. The method of claim 1, wherein when the risk scenario is a airspace scenario, the S2 is configured to construct cross-checking logic of environmental parameters of four risk scenarios on landing assurance data based on the consistency data set, respectively, and specifically includes: extracting environmental parameter radar target data T_radar and lifting guarantee data target panoramic images T_vision from the consistency data set; reflection area based on the radar target data Target speed The outline characteristics of the target panoramic image are used for judging the type of the target through a preset decision tree model; if the radar target data indicates that a target exists in a preset distance range and the panoramic optical image data does not identify a corresponding target, judging that the panoramic image of the target has data distortion; The preset decision tree model specifically comprises the following steps: And (3) primary judgment: By radar reflection area For determining the index, the data is derived from the actual measurement of the low-altitude blind-complement radar, and the determination logic is as follows =0.1 As a critical threshold, dividing the airspace target into two categories, namely a small-reflection-area target and a medium-large-reflection-area target; And (3) secondary judgment: branch 1-small reflective area target, i.e. when < 0.1: at a target speed In order to judge the index, the data is acquired by the cooperation of the radar and the image transmission system, and the judgment logic is as follows, small interference targets are distinguished through a speed interval, and the small interference targets comprise birds and stray targets; Branch 2, medium-large reflective area target, i.e. when And (3) when the temperature is more than or equal to 0.1: Taking the optical image contour regularity as a judging index, wherein data is derived from a target image acquired by an optical camera, and calculating the regularity after contour extraction by a Canny operator, wherein the judging logic is as follows, based on morphological characteristics of the optical data, distinguishing an artificially manufactured target from a natural/stray target; and (3) three-stage judgment: Branch 1-1 small reflection area target classification: sub-threshold 1: the determination result is that the bird flies; sub-threshold 2: <5 or 15, Judging that the result is a stray target; branch 2-1: medium-large reflective area target classification: sub-determination 1, contour rule and linkage target speed The judgment result is unmanned aerial vehicle; and 2, sub-judging that the contour is irregular, wherein the judging result is an unknown target, and manual rechecking is required to be started.
6. The method of claim 1, wherein when the risk scenario is a device environment scenario, the S2 is configured to construct cross-checking logic of environmental parameters of four risk scenarios on the take-off and landing security data based on the consistency data set, respectively, and specifically includes: extracting environmental parameter shelter temperature T and lifting guarantee data monitoring radar signal intensity S from the consistency data set; based on the pearson correlation coefficient, the correlation between the shelter temperature T and the monitoring radar signal intensity S is analyzed, and the calculation formula of the pearson correlation coefficient r is as follows: ; Wherein, the Is that Is used to determine the covariance of (1), Is that Is set in the standard deviation of (2), Is that Is set in the standard deviation of (2), In order to obtain the number of samples, And Respectively is And Average value of (2); when the radar system is operating normally, the temperature variation of the shelter And (2) and All are lower than 0.2 when the temperature change of the shelter is low And actually monitors radar signal strength Monitoring radar signal strength with theory during normal operation of radar system If the deviation of the signal intensity of the monitoring radar does not exceed the preset deviation value, judging that the data distortion exists in the signal intensity of the monitoring radar.
7. The method of any one of claims 3 to 6, wherein the performing a countercheck on the corresponding environmental parameter using the take-off and landing security data in S3, generating a countercheck result to identify an environmental anomaly, specifically includes: In microclimate scene, extracting visual navigation recognition rate in take-off and landing guarantee data And with visibility in environmental parameters Check the countercheck when 1000M When the fog is present, the local fog is reflected; Extracting landing gear pressure F in the take-off and landing guarantee data in a road surface scene, and performing anti-evidence verification with the road surface flatness, wherein when a sensor for acquiring the road surface flatness is normal and the increment of F in a preset time range is larger than a preset increment value, local bulges exist on the anti-evidence road surface; In an airspace scene, extracting a picture transmission image T_picture in the take-off and landing guarantee data, and performing anti-evidence verification with radar blind area information, wherein when the radar does not find an abnormal obstacle and the T_picture identifies the abnormal obstacle, the radar blind area is verified; Extracting communication time delay L in the take-off and landing guarantee data in an equipment environment scene, and performing anti-evidence verification with the electromagnetic interference intensity; After the local light fog, the local bulge, the radar blind area and the electromagnetic interference environment abnormality are detected and confirmed in the field, the environment parameter database is updated, and hidden danger levels corresponding to the environment abnormalities are marked.
8. The method of claim 7, wherein the calculating, in S4, a consistency index of the low-altitude landing field environment and landing guarantee data according to the cross check result and the anti-evidence check result in each risk scene, and the performing the trusted determination on the low-altitude landing field operation state specifically includes: For four scenes of the microclimate, the road surface, the airspace and the equipment environment, respectively calculating corresponding quantitative consistency indexes based on the results of cross check and anti-evidence check, wherein the quantitative consistency indexes comprise the deviation rate calculated by the microclimate and the equipment environment scene Calculating consistency coefficient K of road surface scene and calculating target fitness of airspace scene Deviation rate of Consistency coefficient K, target fitness The calculation process of (a) is as follows: ; ; ; wherein the deviation rate The normal threshold value of the (C) is less than 5%, the normal threshold value of the consistency coefficient K is more than 0.8, and the target consistency degree is achieved Is greater than 85%; If the quantized consistency indexes of all scenes meet the corresponding normal thresholds, judging that the low-altitude landing field environment and landing guarantee data are reliable, and allowing landing operation; if the consistency index of any scene does not meet the normal threshold, triggering an exception handling closed-loop flow, wherein the exception handling closed-loop flow comprises: data plane processing, performing at least one operation of sensor calibration, data recovery and model reconstruction; processing an environment layer, and executing at least one operation of field hidden trouble shooting and safeguard upgrading; and after the processing operation is finished, re-executing the flow from data acquisition to consistency calculation and judgment to verify until the quantized consistency indexes of all scenes meet the normal threshold value.
9. The method of claim 8, wherein the risk early warning and grading response in S4 specifically comprises: If the quantized consistency indexes of all scenes meet the normal threshold, triggering a first-level normal response, wherein the linkage action is the allowable take-off and landing, and the UTM system opens an airspace; If the difference value between the quantized consistency index of one scene and the corresponding normal threshold value is smaller than a second preset threshold value and does not exceed the normal threshold value, triggering a secondary early warning response, wherein the linkage action is to prompt operation and maintenance personnel to pay attention to, and the UTM system strengthens airspace monitoring; If the quantized consistency index of one scene exceeds the normal threshold, triggering three-level deceleration response, wherein the linkage action is that the take-off and landing speed of the aircraft is reduced, and the UTM system reduces the airspace range; if the quantitative consistency indexes of two scenes exceed the normal threshold, triggering a four-level avoidance response, wherein the linkage action is to suspend take-off and landing, and the UTM system guides surrounding aircraft to avoid and check hidden danger; If the quantitative consistency indexes of three or more scenes exceed the normal threshold, triggering five-level emergency response, triggering an emergency landing program by linkage action, closing an airspace by a UTM system, and starting an emergency rescue plan.
10. The utility model provides a low altitude take-off and landing field two-way safety verification and early warning system based on data feature fuses which characterized in that includes: The multi-source heterogeneous data fusion module is used for collecting multi-source heterogeneous environment parameters and lifting guarantee data, and carrying out classification pretreatment on the environment parameters and the lifting guarantee data to generate a consistency data set; The scene cross check logic construction module is used for respectively constructing cross check logic of environmental parameters of four risk scenes on the taking-off and landing guarantee data based on the consistency data set, wherein the four risk scenes comprise a microclimate scene, a road surface scene, an airspace scene and an equipment environment scene; the environment-data bidirectional verification execution module is used for executing cross verification on the lifting and falling guarantee data through corresponding environment parameters in each risk scene based on the cross verification logic of the four risk scenes, and when the cross verification result is data distortion, performing anti-verification on the corresponding environment parameters by utilizing the lifting and falling guarantee data to generate an anti-verification result so as to identify environment abnormality; The credibility judgment and grading response module is used for calculating consistency indexes of the low-altitude landing field environment and landing guarantee data according to the cross check result and the anti-evidence check result in each risk scene, carrying out credibility judgment on the low-altitude landing field operation state, and triggering corresponding risk early warning and grading response according to the credibility judgment result.

Description

Low-altitude landing field bidirectional safety verification and early warning method based on data feature fusion Technical Field The application belongs to the technical field of low-altitude economy, and particularly relates to a low-altitude landing field bidirectional safety verification and early warning method and system based on data feature fusion. Background With the large-scale development of low-altitude economy, unmanned aerial vehicles and vertical take-off and landing aircrafts are increasingly frequently applied to scenes such as logistics distribution, emergency disaster relief and the like. The low-altitude landing field is used as a key node for connecting aerial operation and ground guarantee, the operation environment has the characteristics of distributed dispersion, small scale and complex peripheral conditions, and safety control faces multiple challenges such as microclimate mutation, abnormal road surface state, airspace stray target burst, electromagnetic interference and the like. At present, the safety verification method for the low-altitude landing field mainly relies on a single sensor or a limited data source to carry out environment monitoring, and carries out one-way verification and alarm on collected data by setting a threshold value, so that a conventional technical path for judging the credibility of the data by using environment parameters is formed. However, the method has the problems of single data source, unidirectional check logic and insufficient scene suitability, which leads to data distortion and missed judgment of environmental hidden danger and restricts safe operation. Disclosure of Invention The embodiment of the application provides a low-altitude landing field bidirectional safety verification and early warning method and system based on data feature fusion, which can solve the problems of data distortion and environment hidden danger missed judgment and constraint on safe operation caused by the limitations of single data source, unidirectional verification logic and insufficient scene suitability. The embodiment of the application provides a low-altitude landing-off and landing-on-off bidirectional safety verification and early warning method based on data feature fusion, which comprises the steps of S1 collecting multi-source heterogeneous environment parameters and landing-off guarantee data, classifying and preprocessing the environment parameters and the landing-off guarantee data to generate a consistency data set, S2 respectively constructing cross check logic of the environment parameters of four risk scenes on the landing-off guarantee data based on the consistency data set, wherein the four risk scenes comprise microclimate scenes, road surface scenes, airspace scenes and equipment environment scenes, S3 executing cross check on the landing-off guarantee data through corresponding environment parameters in each risk scene based on the cross check logic of the four risk scenes, and when the cross check result is data distortion, performing anti-evidence check on the corresponding environment parameters by utilizing the landing-off guarantee data to generate an anti-evidence check result so as to identify the environment abnormality, S4 calculating the consistency index of the landing-off and landing-off environment and landing-off guarantee data according to the cross check result and the anti-evidence check result under each risk scene, and triggering the reliable and early warning response and classification judgment result according to the cross check result under each risk scene. In a possible implementation manner of the first aspect, the classifying preprocessing in S1 is performed on the multisource heterogeneous environmental parameter and the landing guarantee data to generate a consistency data set, which specifically includes: denoising the environment parameters and the numerical data in the lifting guarantee data; sequentially carrying out Gaussian denoising and contour extraction on image data in the environmental parameters and the lifting guarantee data; Based on Network Time Protocol (NTP) time stamps, mapping the environment parameters and the lifting guarantee data to a unified time axis, filling the missing data by adopting linear interpolation, and generating a consistency data set. Optionally, in another possible implementation manner of the first aspect, when the risk scenario is a microclimate scenario, S2 above, based on the consistency dataset, cross checking logic of environmental parameters of four risk scenarios on the take-off and landing security data is respectively constructed, and specifically includes: Extracting environmental parameter wind speed from consistent data set Airplane attitude angle based on take-off and landing guarantee data; Based on low-altitude landing field historical data, constructing by adopting least square fitting methodThe linear association model is specifically expressed as: Wherein