Unmanned aerial vehicle flight state detection method based on frequency spectrum detection technology

Abstract

The invention relates to the technical field of radio orientation, in particular to an unmanned aerial vehicle flight state detection method based on a frequency spectrum detection technology, which comprises the following steps: acquiring an initial time domain signal of an initial frequency spectrum signal of an unmanned aerial vehicle in a monitoring period in a flight process; segmenting the initial time domain signal according to the flight state so as to screen out each complex change time period; performing signal decomposition on the time domain sub-signals in each complex change time period to obtain a plurality of components; and determining an interference frequency tendency index of each component, determining an interference characteristic comprehensive index of each component according to the interference frequency tendency index of each component, the oscillation intensity characteristic value and the correlation coefficient, and reconstructing the frequency spectrum sub-signal corresponding to the time domain sub-signal to obtain a target frequency spectrum signal. According to the method, the data reference value of the target spectrum signal obtained through adaptive filtering is larger, and the accuracy of detection of the flight state of the unmanned aerial vehicle is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of radio orientation, in particular to a method for detecting flight state of unmanned aerial vehicle based on spectrum detection technology. BACKGROUND

[0002] In recent years, the unmanned aerial vehicle racing is developing rapidly, and the demand for high-precision and real-time detection of flight state is increasingly urgent. At present, the radio frequency detection technology based on spectrum analysis is one of the effective ways to achieve this goal.

[0003] However, the electromagnetic environment of the racing unmanned aerial vehicle competition site is extremely complex. The camera, image transmission equipment, remote controller and numerous audience electronic devices work at the same time, generating a large number of electromagnetic signals of different frequency bands and different types. These signals are superimposed with the radio frequency signals of the unmanned aerial vehicle in the limited space, forming serious co-frequency interference, adjacent frequency interference and multipath interference, which causes distortion and attenuation of the radio frequency signals. Such complex electromagnetic background noise greatly increases the difficulty of effectively separating and analyzing the real state of the unmanned aerial vehicle from the mixed spectrum, resulting in high misjudgment rate of the existing detection system and insufficient reliability of the obtained flight trajectory, attitude and other data. SUMMARY

[0004] In order to solve the technical problem of low reliability of spectrum signal for flight state detection due to interference of multiple source heterogeneous signals in complex electromagnetic environment, the purpose of the present application is to provide a method for detecting flight state of unmanned aerial vehicle based on spectrum detection technology, and the technical scheme adopted is as follows: One embodiment of the present application provides a method for detecting flight state of unmanned aerial vehicle based on spectrum detection technology, which comprises the following steps: Obtaining the initial spectrum signal of the unmanned aerial vehicle in the monitoring period during flight, and converting the initial spectrum signal into an initial time domain signal; Segmenting the initial time domain signal according to the flight state to screen out each complex change period; the complex change period is composed of at least multiple acceleration moments, deceleration moments, unstable acceleration moments and unstable deceleration moments which appear alternately; Respectively performing signal decomposition on the time domain sub-signal of each complex change period to obtain a plurality of components; determining the interference frequency tendency index of each component according to the frequency distribution set corresponding to the target energy set of each component; Obtaining the oscillation intensity characteristic value of each component and the correlation coefficient of each component with other components in the time domain sub-signal; determining the interference feature comprehensive index of each component according to the interference frequency tendency index, the oscillation intensity characteristic value and the correlation coefficient of each component; According to the interference feature comprehensive index of each component, the spectrum sub-signal corresponding to the time domain sub-signal is reconstructed to obtain a target spectrum signal.

[0005] Further, the initial time-domain signal is segmented according to the flight state to screen out each complex change period, comprising: obtaining an amplitude sequence according to the initial time-domain signal; determining each normal period according to the amplitude difference between adjacent time points in the amplitude sequence; the normal period includes a hovering period, an acceleration period and a deceleration period; In the monitoring period, the time range between any two normal periods is taken as a complex change period.

[0006] Further, the determination of each normal period according to the amplitude difference between adjacent time points in the amplitude sequence comprises: calculating the amplitude difference between the corresponding amplitudes of the previous time point and the next time point in the amplitude sequence, taking two time points with an absolute value of the amplitude difference less than or equal to an amplitude threshold as hovering time points, and taking more than two consecutive hovering time points as a hovering period; taking each time point in the monitoring period other than the hovering period as a pending time point, taking two pending time points with a positive amplitude difference as an acceleration time point, and taking two pending time points with a negative amplitude difference as a deceleration time point; determining each acceleration period and deceleration period according to each acceleration time point and each deceleration time point in the monitoring period.

[0007] Further, the determination of each acceleration period and deceleration period according to each acceleration time point and each deceleration time point in the monitoring period comprises: for each continuously distributed target time point period, obtaining an amplitude difference sequence corresponding to the period; the target time point is an acceleration time point or a deceleration time point; determining a stability attribute index of each target time point according to the difference between each two adjacent elements in the amplitude difference sequence; determining a target time point with a stability attribute index greater than a stability threshold as a stable target time point, and taking more than two consecutive stable target time points as a target period; the target period is an acceleration period or a deceleration period.

[0008] Further, the determination of the interference frequency tendency index of each component according to the frequency distribution set corresponding to the target energy set of each component comprises: determining the maximum energy corresponding to each component, and determining a target energy set based on the maximum energy; obtaining a frequency distribution set corresponding to the target energy set of the component, and determining the interference frequency tendency index of the component according to the similarity between the frequency distribution set and a reference frequency distribution set; the similarity and the interference frequency tendency index are negatively correlated.

[0009] Further, the obtaining of the oscillation intensity characteristic value of each component comprises: For each component respectively, the kurtosis of each wave peak in the time domain sub-signal of the component is calculated; the kurtosis is used to at least represent the state of the wave peak; The oscillation intensity characteristic value of the component is determined according to the kurtosis of all wave peaks corresponding to the component; the kurtosis is positively correlated with the oscillation intensity characteristic value.

[0010] Further, the determining of the interference characteristic comprehensive index of each component according to the interference frequency tendency index, the oscillation intensity characteristic value and the correlation coefficient of each component comprises: For each component respectively, the interference characteristic factor of the component is determined according to each correlation coefficient of the component; The product of the interference frequency tendency index, the oscillation intensity characteristic value and the interference characteristic factor of the component is calculated, and the product is taken as the interference characteristic comprehensive index of the component.

[0011] Further, the determining of the interference characteristic factor of the component according to each correlation coefficient of the component comprises: The absolute values of each correlation coefficient of the component are determined, and the average value of the absolute values of all correlation coefficients is calculated; The average value is negatively correlated to obtain a negatively correlated value as the interference characteristic factor of the component.

[0012] Further, the reconstructing of the frequency spectrum sub-signal corresponding to the time domain sub-signal according to the interference characteristic comprehensive index of each component of the time domain sub-signal to obtain the target frequency spectrum signal comprises: For each time domain sub-signal, the interference intensity of the frequency spectrum sub-signal corresponding to the time domain sub-signal is determined according to the interference characteristic comprehensive index of each component of the time domain sub-signal; The initial step length in the filtering algorithm is adjusted according to the interference intensity of the frequency spectrum sub-signal to determine the adaptive step length when the frequency spectrum sub-signal is filtered; The frequency spectrum sub-signal is filtered by using the filtering algorithm with the adaptive step length to obtain a new frequency spectrum sub-signal; Each new frequency spectrum sub-signal is used to replace the frequency spectrum sub-signal of the corresponding complex change period in the initial frequency spectrum signal to obtain the target frequency spectrum signal.

[0013] Further, after the target frequency spectrum signal is obtained, the method further comprises: The flight state detection of the unmanned aerial vehicle in the monitoring period is performed according to the target frequency spectrum signal.

[0014] The present application has the following beneficial effects: The application provides a UAV flight state detection method based on a spectrum detection technology, which realizes signal segmentation according to a flight state, is beneficial to adaptive noise removal of a spectrum sub-signal in a complex change period, is beneficial to enhancing the pertinence of interference suppression, realizes accurate noise removal in a key area, and makes the subsequent noise removal strategy have state adaptability; the interference feature comprehensive index of each component is determined, interference degree analysis and calculation are performed on the components, the fluctuation characteristics of the signal at different frequencies can be reflected, and the specific frequency and change characteristics of different types of interference can also be reflected; the interference feature comprehensive index comprehensively reflects the interference features from multiple different angles, is beneficial to obtaining a target spectrum signal with stronger noise removal reliability, that is, to a certain extent, filtering the influence of interference factors on the spectrum signal, making the spectrum features related to the UAV flight state clearly presented, and highlighting the key features of the UAV flight state; finally, the UAV flight state is detected through the target spectrum signal, which is helpful to more accurately identifying the UAV flight parameters, improving the accuracy of the UAV flight state detection, and providing strong support for the design and performance improvement of the UAV. BRIEF DESCRIPTION OF DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, and the advantages thereof, a brief introduction will be given to the drawings needed in the embodiments or the prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0016] Figure 1 A step flowchart of a UAV flight state detection method based on a spectrum detection technology provided by an embodiment of the present application; Figure 2 An implementation flowchart of step S2 in the embodiment of the present application; Figure 3 An implementation flowchart of determining the interference frequency tendency index of each component according to the frequency distribution set corresponding to the target energy set of each component in the embodiment of the present application; Figure 4 An implementation flowchart of step S5 in the embodiment of the present application. DETAILED DESCRIPTION

[0017] In order to further illustrate the technical means and effects adopted by the present application to achieve the predetermined invention purpose, the specific implementation, structure, features and effects of the technical solutions proposed by the present application are described in detail below in combination with the drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0019] In order to remove the interference of other factors in the acquired spectrum signal to some extent, so that the spectrum signal can better reflect the flight state of the unmanned aerial vehicle, an embodiment of the present application provides an unmanned aerial vehicle flight state detection method based on spectrum detection technology, as shown in Figure 1 The method comprises the following steps: S1, acquiring an initial spectrum signal of the unmanned aerial vehicle in a monitoring period during flight, and converting the initial spectrum signal into an initial time domain signal.

[0020] In order to detect the flight state of the unmanned aerial vehicle using the spectrum detection technology, the spectrum signal of the unmanned aerial vehicle needs to be acquired, including: First, according to the relevant technical data or product manual of the unmanned aerial vehicle, the monitoring frequency range is determined.

[0021] Second, according to the monitoring frequency range, an antenna capable of covering the response frequency range is selected. It should be noted that considering that the unmanned aerial vehicle may fly at different distances and directions, an antenna with appropriate gain and directivity needs to be selected.

[0022] Third, a spectrum analyzer with sufficient frequency coverage, high resolution, adjustable video bandwidth and appropriate scanning time range is selected to meet the analysis requirements of different signals.

[0023] Fourth, the selected antenna is connected to the spectrum analyzer through a radio frequency cable, and the related parameters of the spectrum analyzer are set to meet the signal acquisition requirements.

[0024] Fifth, based on the spectrum analyzer after the setting is completed, the signals in the determined monitoring frequency range are scanned to acquire the initial spectrum signal of the unmanned aerial vehicle in the monitoring period during flight.

[0025] Among them, the acquisition of the initial spectrum signal can be customized by the implementer according to the specific scene requirements and needs, which is not limited here. The length of the monitoring period should at least contain a complete state change period of the unmanned aerial vehicle, and the person skilled in the art can select and adjust it within the range of 10 seconds to 60 seconds according to the actual application scene, such as racing, aerial photography or logistics transportation.

[0026] Time domain analysis is more suitable for capturing the dynamic behavior and state transient of the unmanned aerial vehicle flight. In order to facilitate subsequent data analysis, the initial spectrum signal needs to be converted into an initial time domain signal.

[0027] In one embodiment, the initial frequency spectrum signal is converted into an initial time domain signal using a discrete Fourier transform.

[0028] The implementation process of the discrete Fourier transform is prior art and will not be described here.

[0029] S2, segmenting the initial time domain signal according to the flight state to screen out each complex change period.

[0030] Here, the complex change period is at least composed of multiple times of acceleration moments, deceleration moments, unstable acceleration moments, and unstable deceleration moments, such as unstable acceleration or deceleration, sudden maneuver, and specific mode electromagnetic interference.

[0031] For the complex change period, a wide-band electromagnetic noise may be generated due to a dramatic change in motor power, or a multi-path effect may be caused due to a rapid adjustment of the attitude, so the frequency spectrum sub-signal corresponding to the complex change period is the data with the most concentrated electromagnetic interference and the worst signal quality. The subsequent adaptive denoising of the frequency spectrum sub-signal of the complex change period is beneficial to enhancing the pertinence of interference suppression and realizing accurate denoising in the key area, so that the subsequent denoising strategy has state adaptivity.

[0032] The flight state of the unmanned aerial vehicle changes during flight, and the frequency spectrum signals under different flight states exhibit different characteristics in many aspects. For example, when flying in a hover state, the amplitude of the time domain signal fluctuates little and remains at a relatively stable level; when flying at an acceleration state, the amplitude significantly increases and the change rate increases obviously; when flying in a turning state, the power output of the motor is constantly adjusted to maintain balance and change direction, causing the amplitude of the related signal to fluctuate up and down. Therefore, the entire flight period can be preliminarily segmented according to the amplitude change of the collected time domain signal, that is, the time periods of different flight states are preliminarily determined.

[0033] Among them, the horizontal cruising, hovering, accelerating, decelerating, and turning are the basic flight states covering the basic actions of changing position, speed, and direction of the unmanned aerial vehicle in the air, and other complex flight states are composed and superimposed by these basic states. Since the racing unmanned aerial vehicle needs to frequently change speed, direction, and height during the competition, it rarely has the opportunity to perform horizontal cruising for a long time, and the turning state is also constantly changing direction and speed to achieve, so this embodiment takes the time periods corresponding to hovering, acceleration, and deceleration as normal periods, and determines each complex change period based on the normal periods.

[0034] As an exemplary embodiment, the above step S2 can be implemented by the steps shown in FIG. 2: Figure 2 S21, obtaining an amplitude sequence according to the initial time domain signal.

[0035] ​In one embodiment, based on the initial time domain signal, a direct time domain extraction method can be used to obtain the amplitude sequence corresponding to the initial time domain signal.

[0036] S22, according to the amplitude difference between the adjacent two time points in the amplitude sequence, determine each normal period.

[0037] Here, the normal period includes hovering period, acceleration period and deceleration period, that is, its flight process is predictable, and the generated frequency spectrum is continuous and smooth.

[0038] Since the frequency spectrum change of the above normal period is itself part of the normal flight state of the unmanned aerial vehicle, it is a signal that needs to be accurately measured. In order to retain the signal characteristics of the normal period, the noise does not need to be removed, so as to avoid the over-smoothing or distortion caused by strong denoising. For the complex change period, deep denoising is needed to strip off those abnormal and strong interference.

[0039] As an exemplary embodiment, the above step S22 can be realized by the following steps: First, calculate the amplitude difference between the amplitude corresponding to the latter time point and the former time point in the amplitude sequence, and take the two time points whose absolute value of the amplitude difference is less than or equal to the amplitude threshold as hovering time points, and take the continuous two or more hovering time points as hovering time period.

[0040] In one embodiment, the amplitude difference between the latter time point and the former time point in the adjacent two time points in the amplitude sequence is calculated as the amplitude difference, which is used to represent the change amplitude between the adjacent two amplitudes; the amplitude threshold is an empirical value of 3, when the absolute value of a certain amplitude difference is less than or equal to the amplitude threshold, it means that the change fluctuation of the amplitudes corresponding to the two adjacent time points is small and relatively stable, and the unmanned aerial vehicle is in hovering state, so the two adjacent time points can be taken as hovering time points; all hovering time points are obtained, and the continuous two or more hovering time points are taken as hovering time period.

[0041] The size of the amplitude threshold can be set by the implementer according to the specific unmanned aerial vehicle model and hovering flight standard, which is not limited here.

[0042] Second, take each time point in the monitoring period except the hovering time period as a to-be-determined time point, and take two to-be-determined time points with positive amplitude difference as acceleration time points and two to-be-determined time points with negative amplitude difference as deceleration time points.

[0043] In one embodiment, after obtaining each hovering period, the time instants other than the hovering periods in the monitoring period are regarded as pending time instants; for two adjacent pending time instants, if the amplitude difference corresponding to the two adjacent pending time instants is positive, it is determined that the two adjacent pending time instants are acceleration time instants, and if the amplitude difference corresponding to the two adjacent pending time instants is negative, it is determined that the two adjacent pending time instants are deceleration time instants.

[0044] In a third step, each acceleration period and deceleration period are determined according to each acceleration time instant and each deceleration time instant in the monitoring period.

[0045] Since the amplitude fluctuates up and down when the UAV is turning, and sometimes increases and sometimes decreases, it is also possible that two or more consecutive acceleration or deceleration time instants are part of the turning state, and therefore it is not possible to only divide the acceleration period and the deceleration period according to the continuity of the same type of time instants, and the stability of the amplitude change needs to be considered.

[0046] As an exemplary embodiment, determining each acceleration period and deceleration period comprises: In a first sub-step, for each continuously distributed target time instant period, the amplitude difference sequence corresponding to the period is obtained.

[0047] Here, the determination process of the acceleration time instant and the deceleration time instant is the same, in order to reduce unnecessary description, the acceleration time instant and the deceleration time instant are both referred to as target time instants. It is worth noting that when determining the target period, the time instants of the same attribute are analyzed.

[0048] During the acceleration or deceleration of the UAV, the motor power changes relatively stably, although there may be some small fluctuations, but the overall amplitude fluctuation frequency is not high, so the amplitude difference of adjacent target time instants should be close.

[0049] In one embodiment, for each continuously distributed target time instant period, the amplitude difference between every two adjacent target time instants in the period is obtained first to form an amplitude difference sequence.

[0050] In a second sub-step, according to the difference between each two adjacent elements in the amplitude difference sequence, the stability attribute index of each target time instant is determined.

[0051] In one embodiment, the absolute value of the difference between each two adjacent elements in the amplitude difference sequence is calculated to represent the difference between the two adjacent amplitude differences; all difference absolute values are processed by maximum-minimum value normalization to obtain normalized values limited between 0 and 1; and the normalized values obtained are processed in a negative correlation manner, i.e. taking the reciprocal, to obtain the stability attribute index corresponding to three target time instants, and thus the stability attribute index of each target time instant is obtained.

[0052] Of course, the exponential function with natural constant as base, i.e. exp(-) can also be used to realize the normalization processing of the negative correlation of data to obtain the stable attribute index.

[0053] In a third sub-step, the target time point with the stable attribute index greater than the stable threshold value is determined as a stable target time point, and more than two stable target time points are grouped to form a target period.

[0054] Here, the target period is an acceleration period or a deceleration period.

[0055] In an embodiment, the stable threshold value can be an empirical value of 0.6 or 0.7. If a certain stable attribute index is greater than the stable threshold value, it indicates that the amplitude difference of adjacent target time points is close, the amplitude fluctuation is small, and the acceleration or deceleration is stable. Therefore, each target time point corresponding to the calculation of the stable attribute index is regarded as a stable target time point, and more than two stable target time points are grouped to form a target period. The size of the stable threshold value can be set by the implementer according to the specific actual situation, which is not limited here.

[0056] Based on the above determination process of the target period, each acceleration period and deceleration period in the monitoring period can be obtained.

[0057] S23, in the monitoring period, the time range between any two normal periods is regarded as a complex change period.

[0058] In an embodiment, the time range between any two adjacent normal periods is regarded as a complex change period. Each complex change period is not a non-overlapping period.

[0059] Thus, all complex change periods in the monitoring period are obtained in the embodiment.

[0060] It should be noted that by screening out the complex change period for targeted processing, the processing efficiency is significantly improved. By only performing in-depth processing on the complex period with a small proportion, the calculation cost can be reduced, and the system has real-time processing capability.

[0061] After obtaining the complex change period, if there are some periods that do not belong to the normal period and do not belong to the complex change period, a backup processing strategy is started. The backup processing strategy includes but is not limited to: regarding these periods as complex change periods for processing, or recording the period data for subsequent model optimization. For isolated time points, they can be divided into any adjacent attribute period.

[0062] It also needs to be explained that the complex change period contains the most discriminative feature information of the UAV flight, if uniform and excessive filtering is performed on these periods, important state features may be smoothed, and through targeted processing, the key spectral features reflecting the real flight state are preserved to the maximum extent while removing interference, and a cleaner and more reliable data basis is provided for subsequent flight state analysis, that is, the target spectral signal is obtained, which is beneficial to reduce the misjudgment rate and omission rate of state recognition, and improve the accuracy and reliability of the entire detection system. Corresponding to the following steps S3 to S5.

[0063] S3, signal decomposition is performed on the time domain sub-signal of each complex change period respectively to obtain a plurality of components; and an interference frequency tendency index of each component is determined according to a frequency distribution set corresponding to a target energy set of each component.

[0064] In actual scenarios, there are many electronic devices and communication devices in the UAV racing field, such as wireless communication base stations, intercoms used by event organizers, wireless cameras on site, mobile phones carried by on-site audience and staff, etc., which will occupy certain frequency bands on the spectrum and emit signals. At this time, when using a spectrum detection device to collect data, the signals of electronic devices and communication devices are collected together with the spectrum signal of the target UAV, and when multiple UAVs are flying at the same time, the spectrum signals of other UAVs will also be superimposed on the collected data, and the signals of other devices will not only exist as independent signal components in the collected data, but also interfere with the spectrum signal of the target UAV. Therefore, the complex change period may be caused by the UAV constantly changing its flight state during flight, or it may be caused by the spectrum signal constantly changing due to various external disturbances during flight.

[0065] Each component obtained after signal decomposition can represent the oscillation mode with a specific time scale in the original signal, which reflects the fluctuation characteristics of the signal at different frequencies. For the initial spectrum signal of the monitoring period, some components after decomposition may be related to the flight state of the UAV, different flight states of the UAV will cause changes in the working state of other electronic systems, and then affect the signal quality. At the same time, it is also related to external environmental factors, such as electromagnetic interference of nearby electronic devices, environmental noise, etc., and different types of interference often have specific frequencies and change characteristics, which are reflected in the corresponding components. Therefore, the components of each complex change period need to be obtained to analyze the signal interference characteristics of different components.

[0066] Signal decomposition is performed on the time domain sub-signal of each complex change period respectively to obtain a plurality of components, including: In one embodiment, the principal component analysis method is used to decompose the time domain sub-signal of each complex change period into a plurality of components.

[0067] The implementation process of the principal component analysis method is prior art and will not be described here. In addition, each complex change period has a plurality of components corresponding thereto.

[0068] As an exemplary embodiment, the interference frequency tendency index of each component can be determined according to the frequency distribution set corresponding to the target energy set of each component, which can be achieved by Figure 3 The steps are shown as follows: S31, for each component, the maximum energy corresponding to the component is determined, and the target energy set is determined based on the maximum energy.

[0069] For the components related to the flight state of the unmanned aerial vehicle, the center frequency of the component is relatively stable and has a specific range, while the interference signal source is diverse, the component frequency range can be scattered in a wide frequency band, and sudden changes can occur, and the amplitude fluctuation will be larger.

[0070] In an embodiment, for each component of the same complex change period, the maximum energy is determined, and the maximum energy multiplied by a preset multiple is taken as the minimum value of the target energy set, and the maximum energy value is taken as the maximum value of the target energy set, so as to obtain the energy data between the two extreme values. Wherein, the preset multiple can be 0.2 or 0.3, and the implementer can set the size of the preset multiple according to the specific actual demand, which is not limited here.

[0071] S32, the frequency distribution set corresponding to the target energy set of the component is obtained, and the interference frequency tendency index of the component is determined according to the similarity degree between the frequency distribution set and the reference frequency distribution set.

[0072] Here, the similarity degree and the interference frequency tendency index are negatively correlated, that is, the greater the similarity degree, the smaller the interference frequency tendency index.

[0073] In an embodiment, the intersection between the frequency distribution set and the reference frequency distribution set is determined, the length of the intersection is taken as the similarity degree between the frequency distribution set and the reference frequency distribution set, and the similarity degree is negatively correlated to obtain a negative correlation value, such as an inverse, which is taken as the interference frequency tendency index of the corresponding component.

[0074] As an example, the calculation formula of the interference frequency tendency index of the component can be: ; In the formula, A represents the interference frequency tendency index of the component, B represents a non-zero constant, which can be 0.01, which is used to avoid the case that the denominator of the fraction is zero, C represents the intersection between the frequency distribution set and the reference frequency distribution set, that is, the similarity degree between the frequency distribution set and the reference frequency distribution set.

[0075] The reference frequency distribution set can be the intersection of the frequency distribution sets of all components. When the coincidence degree between the frequency distribution set of a component and the reference frequency distribution set is higher, it indicates that the frequency distribution of the component is more consistent with the frequency distributions of other components, the component is less affected by interference, and the possibility of the component being an interference component is smaller.

[0076] In another embodiment, based on the frequency distribution set and the reference frequency distribution set, the similarity before analyzing the frequency existence, amplitude distribution and energy concentration degree of the two sets is analyzed to determine indicators in multiple different dimensions. Comprehensive analysis of all indicators can obtain an interference frequency tendency index of a component with higher numerical accuracy. The calculation processes of the frequency existence, amplitude distribution and energy concentration degree are prior art and will not be described here.

[0077] Referring to the calculation process of the interference frequency tendency index of the single component, the interference frequency tendency index of each component can be obtained.

[0078] Thus, the interference frequency tendency index of each component is determined in the embodiment.

[0079] S4, obtaining the oscillation intensity characteristic value of each component and the correlation coefficient of the component with other components in the time domain sub-signal; and determining an interference characteristic comprehensive index of each component based on the interference frequency tendency index, the oscillation intensity characteristic value and the correlation coefficient of each component.

[0080] Here, the interference characteristic comprehensive index can comprehensively analyze the interference degree of a single component from multiple angles, which is beneficial to subsequent adaptive filtering processing.

[0081] As an exemplary embodiment, obtaining the oscillation intensity characteristic value of each component and the correlation coefficient of the component with other components in the time domain sub-signal includes: First, obtaining the oscillation intensity characteristic value of each component.

[0082] The interference signal generally appears instantaneously and randomly and is unstable. Therefore, in the time domain signal of each component, the kurtosis of each peak in the component is calculated to represent the state of each peak. The greater the kurtosis, the sharper the corresponding peak, and the more likely the corresponding peak is to appear suddenly and be an instantaneous interference.

[0083] First sub-step, for each component, the kurtosis of each peak in the time domain sub-signal of the component is calculated.

[0084] Here, the kurtosis is used to represent at least the state of the peak.

[0085] Second sub-step, determine the oscillation intensity characteristic value of the component according to the kurtosis of all the peaks corresponding to the component.

[0086] Here, the kurtosis is positively correlated with the oscillation intensity characteristic value, and the greater the oscillation intensity characteristic value, the more likely it is that the peak in the component is caused by the instantaneous interference, and the greater the possibility that the corresponding component is an interference component.

[0087] In one embodiment, the average or cumulative value of the kurtosis of all the peaks corresponding to the component is calculated as the oscillation intensity characteristic value of the corresponding component.

[0088] Wherein, the calculation process of the kurtosis of the peak in the component is prior art, which will not be described here.

[0089] Second step, obtain the correlation coefficient of each component and other components in the time domain sub-signal.

[0090] As an exemplary embodiment, according to the interference frequency tendency index, the oscillation intensity characteristic value and the correlation coefficient of each component, the interference feature comprehensive index of each component is determined, including: First step, for each component respectively, according to the correlation coefficient of the component, the interference feature factor of the component is determined.

[0091] Because each component corresponding to the signal is related to the motion of the unmanned aerial vehicle and the system characteristics, under normal circumstances, there may be a certain correlation between the components reflecting the flight state of the unmanned aerial vehicle. However, the correlation between the interference signal and other normal components is usually low due to its randomness and independence. Therefore, the smaller the correlation between a certain component and other components, the more likely it is that the corresponding component is an interference component.

[0092] First sub-step, determine the absolute value of each correlation coefficient of the component, and then calculate the average value of the absolute values of all the correlation coefficients.

[0093] In one embodiment, the correlation coefficient can be a Pearson correlation coefficient, whose value range is -1 to 1. Wherein, the calculation process of the Pearson correlation coefficient is prior art, which will not be described here.

[0094] Second sub-step, negatively correlate the average value, and obtain the negative correlation value as the interference feature factor of the component.

[0095] In one embodiment, the negative correlation processing can be realized by taking the inverse method or exp(-).

[0096] In another embodiment, the average value of the DTW distance between a single component and other components in the same signal is calculated as the interference feature factor of the component.

[0097] Wherein, the DTW distance mean is used to represent the correlation coefficient, the greater the DTW distance mean, the smaller the correlation, the less similar the component to other components, and the greater the possibility of the corresponding component being an interference component. The calculation process of the DTW distance is prior art, which will not be described here.

[0098] Secondly, the product of the interference frequency tendency index, the oscillation intensity characteristic value and the interference characteristic factor of the component is calculated, and the product is taken as the interference characteristic comprehensive index of the component.

[0099] For each component, the greater the interference frequency tendency index of the component and the greater the amplitude fluctuation of the component, the lower the correlation between the component and other components, and the more likely the corresponding component is an interference component, and the greater the corresponding interference characteristic comprehensive index.

[0100] As an example, the calculation formula of the interference characteristic comprehensive index of the component can be: In the formula, B represents the interference characteristic comprehensive index of the component, A represents the interference frequency tendency index of the component, j represents the oscillation intensity characteristic value of the component, and q represents the interference characteristic factor of the component.

[0101] Referring to the calculation process of the interference characteristic comprehensive index of the component, the interference characteristic comprehensive index of each component can be obtained.

[0102] Thus far, the interference characteristic comprehensive index of each component is determined.

[0103] S5, reconstructing the frequency spectrum sub-signal corresponding to the time domain sub-signal according to the interference characteristic comprehensive index of each component, to obtain a target frequency spectrum signal.

[0104] Here, the target frequency spectrum signal refers to the initial frequency spectrum signal after adaptive filtering processing.

[0105] As an exemplary embodiment, the above step S5 can be implemented by the steps shown in the following table: Figure 4 S51, for each time domain sub-signal, determining the interference intensity of the frequency spectrum sub-signal corresponding to the time domain sub-signal according to the interference characteristic comprehensive index of each component of the time domain sub-signal.

[0106] In one embodiment, the average of the interference characteristic comprehensive indexes of all components of the time domain sub-signal is calculated as the interference intensity of the frequency spectrum sub-signal corresponding to the corresponding time domain sub-signal.

[0107] ​It should be noted that the greater the interference intensity, the more the corresponding time period of the spectrum sub-signal is interfered by the outside world, and the lower the credibility of the corresponding spectrum sub-signal in presenting the flight state of the unmanned aerial vehicle; the smaller the interference intensity, the less the spectrum of the unmanned aerial vehicle is interfered by the outside world in the corresponding time period, and the more the flight state of the unmanned aerial vehicle in the corresponding time period can be truly reflected.

[0108] S52, adjusting the initial step length in the filtering algorithm according to the interference intensity of the spectrum sub-signal to determine the adaptive step length when the spectrum sub-signal is filtered.

[0109] In one embodiment, the initial step length is set to 0.01, and the interference intensity is normalized to the range of [-0.5, 0.5]; according to the normalized interference intensity, the initial step length is adjusted, and the adjusted step length is used as the adaptive step length when the spectrum sub-signal is filtered.

[0110] When the interference intensity is large, it means that the spectrum sub-signal is interfered more strongly, and the step length factor should be reduced. A smaller step length factor can slow down the convergence speed of the filter, but can improve the stability of the filtering and better suppress strong interference; when the interference intensity is small, it means that the spectrum sub-signal is interfered more weakly, and the step length factor can be appropriately increased to speed up the convergence speed of the filter, so that the filter can adapt to the change of the signal faster.

[0111] As an example, the calculation formula of the adaptive step length when the spectrum sub-signal is filtered can be: ; wherein S0 represents the initial step length, C represents the normalized interference intensity of the spectrum sub-signal, and S represents the adaptive step length when the spectrum sub-signal is filtered.

[0112] S53, filtering the spectrum sub-signal by using the filtering algorithm with the adaptive step length to obtain a new spectrum sub-signal.

[0113] In one embodiment, the implementation process of the filtering algorithm is a prior art, which will not be described here.

[0114] It should be noted that by adjusting the step length factor, the adaptive filter can adaptively remove the external interference according to the interference intensity, highlight the spectrum signal features related to the flight state of the unmanned aerial vehicle, and make the detected spectrum signal better reflect the flight state of the unmanned aerial vehicle.

[0115] S54, replacing the spectrum sub-signal of the corresponding complex change period in the initial spectrum signal with each new spectrum sub-signal to obtain a target spectrum signal.

[0116] Thus, the target spectrum signal of the monitoring time period is obtained in this embodiment.

[0117] S6, according to the target spectrum signal, the flight state of the UAV in the monitoring period is detected.

[0118] In one embodiment, the target spectrum signal is converted into a two-dimensional format suitable for convolutional neural network input, and the two-dimensional format target spectrum signal is input into the trained convolutional neural network model, and the model outputs the predicted flight state, so as to obtain more accurate UAV flight state information.

[0119] Among them, the convolutional neural network model for predicting the flight state can be trained using historical flight spectrum data after removing the interference and the corresponding flight state to ensure that the network can learn the feature mode of the spectrum signal under different flight states. The training process of the convolutional neural network model is prior art, which will not be described here.

[0120] So far, the embodiment has completed the UAV flight state detection with high accuracy.

[0121] In summary, the present application provides a UAV flight state detection method based on spectrum detection technology. The method adjusts the detected UAV spectrum signal through an adaptive filtering algorithm, can filter out the interference of some external factors in the spectrum detection process, makes the spectrum features related to the UAV flight state clearly presented, highlights the key features of the UAV flight state, and helps to more accurately identify the UAV flight parameters, and improves the accuracy of the UAV flight state detection.

[0122] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A method for detecting the flight state of a UAV based on spectrum detection technology, characterized in that, The method comprises the following steps: obtaining an initial frequency spectrum signal of a monitoring period of a UAV in flight and converting the initial frequency spectrum signal into an initial time domain signal; segmenting the initial time domain signal according to flight states to screen out each complex change period; the complex change period is composed of at least multiple acceleration time points, deceleration time points, unstable acceleration time points and unstable deceleration time points which are interspersed; performing signal decomposition on a time domain sub-signal of each complex change period to obtain several components; determining an interference frequency tendency index of each component according to a frequency distribution set corresponding to a target energy set of each component; obtaining an oscillation intensity characteristic value of each component and a correlation coefficient of each component with other components in the time domain sub-signal; determining an interference feature comprehensive index of each component according to the interference frequency tendency index, the oscillation intensity characteristic value and the correlation coefficient of each component; reconstructing a frequency spectrum sub-signal corresponding to the time domain sub-signal according to the interference feature comprehensive index of each component to obtain a target frequency spectrum signal.

2. The method of claim 1, wherein the method further comprises: The segmentation of the initial time domain signal according to flight states to screen out each complex change period comprises: obtaining an amplitude sequence from the initial time domain signal; determining each normal period according to amplitude difference conditions of adjacent time points in the amplitude sequence; the normal period includes a hovering period, an acceleration period and a deceleration period; in the monitoring period, regarding a time range between any two normal periods as a complex change period.

3. The method of claim 2, wherein the method further comprises: The determination of each normal period according to the amplitude difference conditions of adjacent time points in the amplitude sequence comprises: calculating an amplitude difference corresponding to a later time point and an earlier time point in the amplitude sequence, regarding two time points with an absolute value of the amplitude difference less than or equal to an amplitude threshold as hovering time points, and regarding continuous two or more hovering time points as a hovering period; regarding each time point other than the hovering time points in the monitoring period as a to-be-determined time point, regarding two to-be-determined time points with a positive amplitude difference as acceleration time points, and regarding two to-be-determined time points with a negative amplitude difference as deceleration time points; determining each acceleration period and each deceleration period according to each acceleration time point and each deceleration time point in the monitoring period.

4. The method of claim 3, wherein the method further comprises: The determination of each acceleration period and each deceleration period according to each acceleration time point and each deceleration time point in the monitoring period comprises: for each continuous distribution target time period, obtaining an amplitude difference sequence corresponding to the target time period; the target time is an acceleration time or a deceleration time; determining a stability attribute index of each target time according to difference conditions between each two adjacent elements in the amplitude difference sequence; regarding a target time with a stability attribute index greater than a stability threshold as a stable target time, and regarding continuous two or more stable target times as a target period; the target period is an acceleration period or a deceleration period.

5. The method of claim 1, wherein the method further comprises: The determination of an interference frequency tendency index of each component according to a frequency distribution set corresponding to a target energy set of each component comprises: for each component, determining a maximum energy corresponding to the component, and determining a target energy set based on the maximum energy; The frequency distribution set corresponding to the target energy set of the component is acquired, and an interference frequency tendency index of the component is determined according to a similarity degree between the frequency distribution set and a reference frequency distribution set; the similarity degree and the interference frequency tendency index are negatively correlated.

6. The method of claim 1, wherein the method further comprises: The oscillation intensity characteristic value of each component is acquired, including: For each component, the kurtosis of each wave peak in the time domain sub-signal of the component is calculated; the kurtosis is used at least to represent the state of the wave peak; The oscillation intensity characteristic value of the component is determined according to the kurtosis of all wave peaks corresponding to the component; the kurtosis and the oscillation intensity characteristic value are positively correlated.

7. The method of claim 1, wherein the method further comprises: The interference characteristic comprehensive index of each component is determined according to the interference frequency tendency index, the oscillation intensity characteristic value and the correlation coefficient of each component, including: For each component, an interference characteristic factor of the component is determined according to each correlation coefficient of the component; The product of the interference frequency tendency index, the oscillation intensity characteristic value and the interference characteristic factor of the component is calculated, and the product is taken as the interference characteristic comprehensive index of the component.

8. The method of claim 7, wherein the method further comprises: The interference characteristic factor of the component is determined according to each correlation coefficient of the component, including: The absolute values of each correlation coefficient of the component are determined, and the average value of the absolute values of all correlation coefficients is calculated; The average value is negatively correlated to obtain a negatively correlated value as the interference characteristic factor of the component.

9. The method of claim 1, wherein the method further comprises: The frequency spectrum sub-signal corresponding to the time domain sub-signal is reconstructed according to the interference characteristic comprehensive index of each component to obtain a target frequency spectrum signal, including: For each time domain sub-signal, the interference intensity of the frequency spectrum sub-signal corresponding to the time domain sub-signal is determined according to the interference characteristic comprehensive index of each component of the time domain sub-signal; The initial step length in the filtering algorithm is adjusted according to the interference intensity of the frequency spectrum sub-signal to determine an adaptive step length when the frequency spectrum sub-signal is filtered; The frequency spectrum sub-signal is filtered by using the filtering algorithm with the adaptive step length to obtain a new frequency spectrum sub-signal; Each new frequency spectrum sub-signal is used to replace the frequency spectrum sub-signal of the corresponding complex change period in the initial frequency spectrum signal to obtain the target frequency spectrum signal.

10. The method of claim 1, wherein the method further comprises: After the target frequency spectrum signal is obtained, further including: The flight state detection of the unmanned aerial vehicle in the monitoring period is performed according to the target frequency spectrum signal.

Images