Baseband signal processing method suitable for GNSS-R height measurement of unmanned aerial vehicle with complex earth surface

Abstract

The invention discloses a baseband signal processing method suitable for GNSS-R height measurement of a complex surface type unmanned aerial vehicle, and the method comprises the steps: adaptively generating a signal collection strategy and a signal integration strategy according to at least one of the surface features of a mirror reflection point, the motion state of the unmanned aerial vehicle, and a height measurement task index; acquiring a GNSS-R height measurement direct signal and a reflection signal based on the matched signal acquisition strategy; performing signal-to-noise ratio adjustment on the collected GNSS-R height measurement direct signal and reflection signal based on the matched signal integration strategy; calculating a path difference between the GNSS-R height measurement direct signal and the reflection signal after the signal-to-noise ratio adjustment, and taking the path difference as an initial GNSS-R height measurement observed quantity; and carrying out system error and random error correction on the initial GNSS-R height measurement observed quantity to obtain an effective GNSS-R height measurement observed quantity.

Description

Technical Field

[0001] The present application relates to the field of automatic identification technology, and particularly relates to a baseband signal processing method for GNSS-R altimetry of unmanned aerial vehicles applicable to complex surface types. Background Art

[0002] The Global Navigation Satellite System Reflective Signal (GNSS R) technology realizes elevation measurement by analyzing the time delay difference between the direct signal and the reflected signal. However, when this technology is applied to dynamic and complex scenarios, it faces a series of severe challenges.

[0003] In existing research, it mainly focuses on static or low-dynamic scenarios. When it comes to the variable motion states of unmanned aerial vehicles and complex surface features, there are many problems in determining the observable quantities.

[0004] For multi-path land surface environments, the non-specular scattering caused by urban buildings, vegetation canopies, etc. makes the main peak of the Delay-Doppler Map (DDM) submerged by the sidelobes, which directly leads to the failure of the traditional leading-edge tracking algorithm in determining the observable quantities, and it is impossible to accurately obtain the required signal time delay difference, thereby affecting the accuracy of the observable quantities in elevation measurement.

[0005] In addition, the error sources of unmanned aerial vehicle altimetry are complex and strongly coupled. The ionospheric delay can reach the meter level in low-earth orbit satellite-borne scenarios. If the hardware delay difference between receiver channels is not calibrated, it will introduce systematic errors, seriously affecting the accuracy of the observable quantities. Moreover, most of the existing error compensation models are based on static or idealized assumptions, lacking the joint modeling of dynamic errors, and it is difficult to accurately determine effective observable quantities in complex scenarios.

[0006] At the same time, the experimental verification data mostly come from shore-based or low-dynamic airborne platforms, lacking the support of measured data with high mobility and complex landforms, resulting in doubts about the generalization ability of the algorithm in complex scenarios, and it is impossible to accurately determine the high-precision GNSS R altimetry observable quantities applicable to complex landforms.

[0007] In summary, how to obtain high-precision GNSS R altimetry observable quantities and solve the problem of determining observable quantities in high-precision unmanned aerial vehicle GNSS R ground clearance measurement in complex scenarios. Summary of the Invention

[0008] The embodiments of the present application provide a technical solution with high recognition accuracy for smoking behavior.

[0009] Specifically, a baseband signal processing method for GNSS-R altimetry of unmanned aerial vehicles applicable to complex surface types includes:

[0010] Generate a signal acquisition strategy and a signal integration strategy adaptively according to at least one of the surface features to which the specular reflection point belongs, the motion state of the unmanned aerial vehicle, and the altimetry task index;

[0011] Collect GNSS-R altimetry direct signals and reflected signals based on the matched signal acquisition strategy;

[0012] Adjust the signal-to-noise ratio of the collected GNSS-R altimetry direct signals and reflected signals based on the matched signal integration strategy;

[0013] Calculate the path difference between the GNSS-R altimetry direct signal and the reflected signal after signal-to-noise ratio adjustment as the initial GNSS-R altimetry observation;

[0014] Correct the systematic error and random error of the initial GNSS-R altimetry observation to obtain an effective GNSS-R altimetry observation.

[0015] The technical solution provided by the embodiment of the present application has at least the following beneficial effects:

[0016] 1. The present application adaptively generates a signal acquisition strategy and a signal integration strategy according to the surface features to which the specular reflection point belongs, which can better cope with non-specular scattering caused by urban buildings, vegetation canopies, etc., avoid the situation where the main peak of the delay-Doppler map (DDM) is submerged by side lobes, and then accurately obtain the signal delay difference, improve the accuracy of the observation in altitude measurement, and solve the defects of traditional methods in a multipath land surface environment. 2. The present application can effectively process the problems caused by the loss of carrier phase coherence and the distortion of the code phase waveform by adjusting the signal-to-noise ratio of the collected GNSS-R altimetry direct signals and reflected signals, accurately calculate the path difference between the direct signal and the reflected signal, and ensure the accuracy of the observation. 3. The present application generates a strategy adaptively based on factors such as the surface features to which the specular reflection point belongs, which can accurately reflect the actual signal characteristics under different landforms, provide a reliable basis for determining the observation, and solve the problems of the existing model in the mixed landform transition zone. 4. The altimetry error sources of unmanned aerial vehicles are complex and strongly coupled. Ionospheric delay and hardware delay differences between receiver channels will introduce systematic errors that affect the accuracy of the observation. The present application corrects the systematic error and random error of the initial GNSS-R altimetry observation, which can effectively reduce the influence of error sources, improve the accuracy of the observation, overcome the defects of the existing error compensation model based on static or idealized assumptions, and achieve effective processing of dynamic errors. 5. The present application adaptively generates a strategy according to different surface features, the motion state of the unmanned aerial vehicle, and the altimetry task index, which can adapt to complex scenarios, does not rely on the measured data of a specific platform, effectively enhances the generalization ability of the algorithm in complex scenarios, and accurately determines the high-precision GNSS-R altimetry observation applicable to complex landforms. Description of the Drawings

[0017] Figure 1 The flowchart of the baseband signal processing method for GNSS-R altimetry of unmanned aerial vehicles applicable to complex surface types provided for the implementation of this application. Specific implementation manners

[0018] As Figure 1 As shown, a baseband signal processing method for GNSS-R altimetry of unmanned aerial vehicles applicable to complex surface types is provided, including: adaptively generating a signal acquisition strategy and a signal integration strategy according to at least one of the surface characteristics to which the specular reflection point belongs, the motion state of the unmanned aerial vehicle, and the altimetry task index; collecting GNSS-R altimetry direct signals and reflected signals based on the matched signal acquisition strategy; performing signal-to-noise ratio adjustment on the collected GNSS-R altimetry direct signals and reflected signals based on the matched signal integration strategy; calculating the path difference between the GNSS-R altimetry direct signals and reflected signals after signal-to-noise ratio adjustment as the initial GNSS-R altimetry observation quantity; and correcting the systematic error and random error of the initial GNSS-R altimetry observation quantity to obtain an effective GNSS-R altimetry observation quantity.

[0019] Optionally, adaptively generating a signal acquisition strategy and a signal integration strategy according to at least one of the surface characteristics to which the specular reflection point belongs, the motion state of the unmanned aerial vehicle, and the altimetry task index includes: generating an environmental feature vector according to at least one of the surface characteristics to which the specular reflection point belongs, the motion state of the unmanned aerial vehicle, and the altimetry task index; and adaptively generating a signal acquisition strategy and a signal integration strategy according to the environmental feature vector.

[0020] Optionally, the method includes: extracting features from the surface image to determine the surface characteristics to which the specular reflection point belongs; and determining the motion state of the unmanned aerial vehicle based on the inertial navigation data on the unmanned aerial vehicle.

[0021] When generating the environmental feature vector, use a vision sensor to identify the surface roughness to determine the surface characteristics to which the specular reflection point belongs; determine its motion state based on the inertial navigation data on the unmanned aerial vehicle; and determine the altimetry task index in combination with the altimetry task requirements. Combine this information to generate an environmental feature vector. For example, when performing land surface altimetry, an increase in soil moisture will increase the intensity of the reflected signal, and an increase in roughness or vegetation coverage will increase the scattered signal and weaken the intensity of the specular reflection signal; this surface feature information is an important basis for generating the environmental feature vector.

[0022] Preferably, in a specific application scenario, for example, the generation of the environmental feature vector is achieved through the following specific processing process:

[0023] (1) Control the vision sensor to collect RGB images of the area around the specular reflection point at a fixed frequency (e.g., 5 frames per second). Use Gaussian filtering on the original images to remove noise, and enhance the image contrast through histogram equalization. Apply the Canny edge detection algorithm to extract the image edge contours, and calculate the fractal dimension of the contours to characterize the roughness. Calculate the texture feature parameters (such as contrast, entropy, inverse difference moment) of the image based on the gray-level co-occurrence matrix. Compare the roughness and texture feature parameters with a pre-established surface type feature library (including the feature threshold ranges of typical surfaces such as grassland, sandy land, and cement ground), so as to determine the surface feature to which the specular reflection point belongs, and encode it to obtain the surface feature vector.

[0024] (2) Read the three-axis data of the accelerometer and gyroscope in the unmanned aerial vehicle inertial navigation system; use the quaternion method to integrate the gyroscope data, update the quaternion attitude parameters in real time, and convert them into Euler angles (pitch angle, roll angle, yaw angle); after denoising the accelerometer data, combine the attitude information to convert it to the geographic coordinate system, and calculate the speed and position coordinates of the aircraft through double integration; finally, integrate the Euler angles, speed, and position information to determine the motion state of the unmanned aerial vehicle, and encode it to obtain the motion state vector.

[0025] (3) Parse the XML configuration file of the altimetry task, and extract the task parameters (such as the longitude and latitude range of the target area, the expected altimetry accuracy ±0.1m, and the task completion time of 30 minutes); perform one-hot encoding on non-numerical parameters (such as the task priority "high / medium / low") to convert them into numerical vectors, and perform normalization processing on numerical parameters (such as mapping the altimetry accuracy requirement to the interval [0,1], 0.1m accuracy corresponds to 0.8, and 0.05m accuracy corresponds to 1); splice all the processed parameters to obtain the altimetry task indicators, and vectorize them to obtain the altimetry task indicator vector.

[0026] (4) Splice the surface feature encoding vector obtained in step (1) (such as grassland encoded as [1,0,0], sandy land encoded as [0,1,0]), the motion state vector obtained in step (2) (including 9 parameters of Euler angles, speed, and position), and the altimetry task indicator vector obtained in step (3) in a fixed order to finally generate a 12-dimensional environmental feature vector containing surface, motion, and task information.

[0027] In particular, in a specific application scenario, the specific processing process for generating the environmental feature vector is as follows:

[0028] I. Environmental Feature Quantization Model

[0029] 1. Surface Feature Vector (S)

[0030] S = [S1, S2, S3, S4, S5], where S1 (roughness index): The contour is extracted by Canny edge detection, and the fractal dimension D is calculated f and normalized.

[0031] S2 (vegetation coverage rate): Green pixels are segmented in the HSV space (H ∈ [40, 80], S ≥ 0.2, V ≥ 0.2),

[0032]

[0033] S3 (soil moisture): Based on the reflectance of the multispectral red band (Red) and the near-infrared band (NIR)

[0034]

[0035] S4 (specular reflection intensity): The integral ratio of the power spectral density of the reflected signal to the direct signal

[0036]

[0037] S5 (surface type coding): Grassland = 0.1, Sandy land = 0.3, Cement ground = 0.7, City = 0.9, Mountain = 1.0

[0038] 2. Motion state vector (M)

[0039] M = [v, a, θ, ω, GPS std , INS bias

[0040] v: Flight speed (m / s), calculated by inertial navigation integration. a: Acceleration (m / s 2 ), from the original accelerometer data. θ: Pitch angle (rad), solved by the quaternion method. ω: Angular velocity (rad / s), real-time output of the gyroscope. GPS std : Standard deviation of GPS positioning error, estimated by Kalman filter. INS bias : Inertial navigation zero bias, obtained by Allan variance analysis.

[0041] 3. Mission index vector (T)

[0042] T = [T1, T2, T3, T4]

[0043] T1 (accuracy requirement): Exponential mapping normalization

[0044]

[0045] T2 (real-time performance): Ratio of the remaining time to the used time

[0046]

[0047] T3 (Power Consumption Constraint): Ratio of Remaining Power Consumption to Maximum Power Consumption

[0048]

[0049] T4 (Multipath Factor): Peak-to-Average Ratio of Autocorrelation Function

[0050] ACF peak Represents the peak value of the autocorrelation function, ACF rms Represents the root mean square value of the autocorrelation function

[0051] II. Frequency Point Selection Decision Model

[0052] 1. Signal Quality Evaluation Function:

[0053] Dynamic Weight Allocation:

[0054] Dynamic Weight Allocation: (w1: SNR, w2: Multipath Resistance, w3: Dynamic Adaptability, w4: Precision Potential) 2. Single-Frequency Signal Quality Index

[0055] B1C / L1 Frequency Point:

[0056] B2a / L5 Frequency Point:

[0057] 3. Dual-Frequency Signal Advantage Function

[0058] Eliminate Ionospheric Residual Error I residual , Enhance Multipath Robustness.

[0059] 4. Wideband Signal Advantage Function

[0060] The larger the bandwidth B, the better the measurement accuracy and real-time performance.

[0061] III. Adaptive Decision Algorithm

[0062] 1. Frequency Point Selection Rule

[0063] Selection Strategy = arg max f∈{单频,双频,宽带} Q(f)

[0064] Single-Frequency Priority Scenario:

[0065] Flat Ground (S5 < 0.3) + Low Speed (v < v threshold ) → B1C / L1

[0066] Flat Ground + High Speed (v ≥ vthreshold ) → B2a / L5

[0067] Dual - frequency priority scenario: Complex terrain (urban / mountainous, S5 ≥ 0.3) or medium - precision tasks (T1 ∈ [T min , T max )

[0068] Broadband priority scenario: High - precision tasks (T1 > T max ) → B1 I / B1C or B2a / B2b

[0069] 2. Integration time calculation

[0070] Coherent integration time:}]

[0071] The higher the carrier - to - noise ratio C / N0 and the lower the speed / roughness, the longer the integration time

[0072] Non - coherent integration time:

[0073] The lower the reflection intensity S4 and the higher the speed / roughness, the longer the integration time

[0074] In summary, the environmental feature quantization model transforms surface, motion, and task - related information into numerical vectors that can be processed by a computer through multi - source heterogeneous data fusion. Its core lies in constructing a mapping relationship between the physical environment and mathematical representation: By mathematical methods such as fractal dimension and spectral reflectance, physical properties such as surface roughness and vegetation coverage are quantified into numerical values, providing a basis for subsequent signal feature analysis. For example, using Canny edge detection and fractal dimension to calculate surface roughness, mathematically based on the self - similarity principle of image edges, the complex surface contour is transformed into a comparable numerical index. Based on inertial navigation system data, through algorithms such as quaternion method and Kalman filter, raw data such as acceleration and angular velocity are transformed into accurate motion parameters. The quaternion method solves the gimbal lock problem existing in Euler angles, and the Kalman filter effectively suppresses sensor noise, improving the accuracy and stability of motion parameters. Through methods such as exponential mapping and ratio calculation, abstract tasks such as measurement accuracy requirements and real - time requirements are transformed into numerical indexes, providing a quantitative basis for subsequent strategy decision - making.

[0075] When it comes to the frequency point selection decision model, the frequency point selection decision model constructs a signal quality evaluation system based on multi-index weighted fusion. Its essence is to quantify the applicability of different frequency points in a specific environment through mathematical modeling: weighted summation of key indicators such as signal-to-noise ratio (SNR), multipath resistance, dynamic adaptability, and precision potential, and the weights are dynamically adjusted according to task requirements. This multi-index fusion method can comprehensively consider various characteristics of signals in different environments and avoid the one-sidedness of single-index decision-making. Mathematical models are established for different frequency point characteristics. For example, the influence of roughness and speed on SNR is described by an exponential decay function, and the advantages of high-frequency signals in multipath resistance and dynamic adaptability are reflected by quadratic terms.

[0076] Furthermore, the adaptive decision algorithm realizes the dynamic selection of the optimal signal acquisition strategy by comparing the quality scores of different frequency points. By defining clear thresholds and conditions, the quality scores are combined with environmental characteristics and task indicators to form logical judgment rules. For example, single-frequency signals are selected according to the surface type coding and speed threshold, and dual-frequency or broadband signals are selected according to the precision requirement threshold. Based on parameters such as carrier-to-noise ratio and reflection intensity, the optimal coherent and non-coherent integration times are calculated mathematically in this application, balancing the contradiction between SNR improvement and signal processing delay, and ensuring the best signal processing effect in different environments.

[0077] In traditional technologies, fixed frequency points or simple empirical rules are usually used to select frequency points, lacking the dynamic adaptation ability to the environment and task requirements. For example, traditional methods may uniformly use a certain frequency point in all scenarios without considering the influence of factors such as surface type and aircraft speed on signal quality. Generally, a fixed integration time is used and cannot be dynamically adjusted according to signal strength and environment. This method is prone to insufficient SNR or excessive processing delay when the signal is weak or the environment is complex. It relies on a single sensor or simple threshold judgment and lacks multi-source data fusion and precise quantitative analysis. For example, only visually judging the surface type cannot accurately obtain key parameters such as roughness and vegetation coverage rate.

[0078] It is difficult to ensure performance stability in different environments with traditional technologies, while this solution can perceive environmental changes in real time and automatically select the optimal frequency point and integration time through multi-source data fusion and dynamic weight adjustment. For example, in an urban environment, the system can automatically switch to dual-frequency signals according to the multipath factor T4 and the high-precision measurement requirement T1, effectively suppressing multipath interference and improving the measurement accuracy. Traditional methods are greatly affected by ionospheric errors and multipath effects, while this solution can effectively eliminate the residual ionospheric error I residual by using the dual-frequency signal model and the broadband signal advantage function, and improve the resolution by using high-bandwidth signals. For example, in a high-precision task scenario (T1 > T 1,max) Selecting a broadband signal (such as B1I / B1C or B2a / B2b) can improve the height measurement accuracy by more than 30%. Traditional fixed strategies may cause resource waste or insufficient performance. In this solution, through power consumption constraint T3 and dynamic calculation of the integration time, on the premise of meeting the task requirements, the resource consumption of signal processing is optimized. For example, in a scenario with strong signals, the integration time is shortened to reduce the amount of calculation and power consumption. The performance of traditional technologies drops significantly in complex terrain or high-speed movement scenarios. In this solution, by considering factors such as surface roughness S1 and flight speed v, a targeted signal quality model is established. For example, in a mountainous environment, the weight of the anti-multipath algorithm is adjusted according to the roughness and multipath factor, reducing the signal loss lock rate of the system by more than 50% in complex scenarios.

[0079] Optionally, according to at least one of the surface characteristics to which the specular reflection point belongs, the motion state of the unmanned aerial vehicle, and the height measurement task index, a signal acquisition strategy and a signal integration strategy are adaptively generated, including: if it is determined according to the surface characteristics to which the specular reflection point belongs that the unmanned aerial vehicle is on flat ground, and, if it is determined according to the motion state of the unmanned aerial vehicle that the flight speed of the unmanned aerial vehicle is less than the set speed threshold, then the first single-frequency signal height measurement is adaptively selected to generate a signal acquisition strategy; if it is determined according to the surface characteristics to which the specular reflection point belongs that the unmanned aerial vehicle is on flat ground, and, if it is determined according to the motion state of the unmanned aerial vehicle that the flight speed of the unmanned aerial vehicle is greater than or equal to the set speed threshold, then the second single-frequency signal height measurement is adaptively selected to generate a signal acquisition strategy; if it is determined according to the surface characteristics to which the specular reflection point belongs that the unmanned aerial vehicle is in a city or mountainous area, then the dual-frequency signal height measurement is adaptively selected to generate a signal acquisition strategy; if the height measurement task index is greater than the set first task index threshold but less than the set second task index threshold, then the dual-frequency signal height measurement is adaptively selected to generate a signal acquisition strategy; if the height measurement task index is greater than the set second task index threshold, then the broadband signal height measurement is adaptively selected to generate a signal acquisition strategy.

[0080] Optionally, the first single-frequency signal height measurement is based on the B1C / L1 frequency point, the second single-frequency signal height measurement is based on the B2a / L5 frequency point; the dual-frequency signal height measurement is based on the B1C / L1 frequency point and the B2a / L5 frequency point; the broadband signal height measurement is based on the B1I / B1C or B2a / B2b frequency point.

[0081] Therefore, when generating the above signal acquisition strategy, for the first single-frequency signal altitude measurement selection: if it is determined that the unmanned aerial vehicle is on flat ground according to the surface characteristics of the specular reflection point, and it is determined that its flight speed is less than the set speed threshold based on the motion state of the unmanned aerial vehicle, considering that in such a relatively simple environment, the B1C / L1 frequency band has a certain applicability, therefore, adaptively select the first single-frequency signal altitude measurement based on the B1C / L1 frequency band to generate the signal acquisition strategy. For the second single-frequency signal altitude measurement selection: if it is determined that the unmanned aerial vehicle is on flat ground according to the surface characteristics of the specular reflection point, and it is determined that its flight speed is greater than or equal to the set speed threshold based on the motion state of the unmanned aerial vehicle, since the B2a / L5 signal has a longer wavelength, smaller free-space attenuation, better anti-multipath effect, and a high symbol rate, and has better satellite tracking performance when the aircraft is moving at high speed, so adaptively select the second single-frequency signal altitude measurement based on the B2a / L5 frequency band to generate the signal acquisition strategy. For the dual-frequency signal altitude measurement selection: if it is determined that the unmanned aerial vehicle is in a complex environment such as a city or mountain according to the surface characteristics of the specular reflection point, or the altitude measurement task index is greater than the set first task index threshold but less than the set second task index threshold, considering that the dual-frequency altitude measurement based on the B1C / L1 frequency band and the B2a / L5 frequency band can basically eliminate the ionospheric delay error, has strong anti-interference ability, and can also distinguish between direct signals and reflected signals, and performs well in areas with obvious multipath effects, so adaptively select the dual-frequency signal altitude measurement to generate the signal acquisition strategy. For the wideband signal altitude measurement selection: if the altitude measurement task index is greater than the set second task index threshold, at this time, the requirements for altitude measurement accuracy and real-time performance are extremely high, and the wideband signal based on the B1I / B1C or B2a / B2b frequency band has a larger bandwidth, which can bring better tracking performance and higher altitude measurement accuracy. At the same time, it can also form subcarrier phase observations and exhibit better positioning performance than pseudo-code, so at this time, adaptively select the wideband signal altitude measurement to generate the signal acquisition strategy.

[0082] In the GNSS-R altitude measurement application of unmanned aerial vehicles, challenges such as complex surface environments (such as flat ground, cities, mountains), variable motion states of the aircraft (changes in speed and acceleration), and diverse altitude measurement task requirements (different accuracy and real-time requirements) are faced. A single signal acquisition and integration strategy cannot meet the requirements of altitude measurement accuracy and efficiency in different scenarios. Therefore, when designing the algorithm, it is necessary to comprehensively consider the three key factors: the surface characteristics of the specular reflection point, the motion state of the unmanned aerial vehicle, and the altitude measurement task index. Through multi-source information fusion and intelligent decision-making, the adaptive selection of signal acquisition strategies and signal integration strategies is realized. Specifically, it is necessary to establish an environmental perception model to quantify the surface characteristics, construct a motion state estimation model to obtain the dynamic parameters of the aircraft, formulate a task requirement evaluation model to clarify the task index, and finally establish a decision-making model based on this information to output the optimal signal strategy to ensure high-precision altitude measurement in various complex scenarios.

[0083] To this end, particularly in a specific application scenario, the specific implementation of generating a signal acquisition strategy is as follows:

[0084] II. Algorithm implementation logic

[0085] (I) Environmental perception model

[0086] 1. Quantification of surface roughness in this application

[0087]

[0088] In the GNSS-R altimetry scenario, surface roughness affects the signal reflection characteristics. Rough surfaces scatter signals, while smooth surfaces are prone to specular reflection, which affects the altimetry accuracy. In this application, through multi-scale LoG operators ( corresponding to small, medium, and large scales, σ1 = 1.0, σ2 = 1.5, σ3 = 2.0), the surface image I(x j , y j ) is convolved to obtain texture features at different scales. w i is the scale weight (determined by PCA, w1 = 0.2, w2 = 0.5, w3 = 0.3). The medium-scale feature has a greater impact on signal reflection and has a higher weight. N i is the number of effective pixels at each scale, measuring the degree of texture change. α i is the variance correction coefficient (α1 = 0.1, α2 = 0.3, α3 = 0.5), which is used to compensate for texture uniformity, ensure accurate roughness evaluation, and provide a basis for subsequent surface type judgment.

[0089] 2. Probability discrimination of surface types in this application

[0090]

[0091] Different surface types (S t ∈ {flat, urban, mountain}) have different effects on the propagation of GNSS signals. Accurate discrimination is the key to selecting a signal strategy. In this application, based on logistic regression, F = [R rough , E d , C v , H s , D spec is a high-dimensional feature vector, including surface roughness R rough , edge density E d (high edge density in urban construction areas), color variance C v (large color variance in vegetation areas), texture entropy H s , specular reflection intensity D spec , etc. w t and b tThey are the logistic regression parameters obtained through training, which determine the weights of each feature. Calculate the cosine similarity between the actual feature vector and the reference feature vector. γ = 0.2 is the similarity influence factor to enhance the recognition of typical surface features, obtain the probability of the surface type, and provide a basis for signal strategy selection.

[0092] (2) Motion state estimation model

[0093] 1. Federated Kalman filter in this application

[0094]

[0095] During flight, a single sensor (GNSS, IMU, optical flow) has limitations. Federated Kalman filter is used to fuse multi-sensor data to achieve accurate state estimation. i = 1, 2, 3 correspond to different sensors. They are the state estimation vectors of each sub-filter. It is the state transition matrix (designed according to the sensor characteristics. For example, GNSS considers the earth's curvature, and IMU is based on Newtonian mechanics). It is the Kalman gain matrix, weighing the prediction and measurement errors. β i It is the information distribution coefficient (calculated based on the reciprocal of the covariance matrix and introducing the credibility factor γ i Adjusted. For example, GNSS: 0.2, IMU: 0.5, optical flow: 0.3), dynamically adjust the weights of each sensor in state estimation, and obtain accurate aircraft motion state parameters (position, velocity, acceleration, etc.) for subsequent signal strategy decision-making.

[0096] 2. Dynamic adjustment of speed threshold in this application

[0097] v thresh = v0·exp(-λ·R rough ) + Δv wind

[0098] To adapt to the signal reception requirements in different environments, the speed threshold needs to be dynamically adjusted. v0 = 10m / s is the reference threshold, λ = 0.5 is the roughness influence coefficient. The rougher the surface, the larger R rough is, and the lower the speed threshold v thresh , because the signal reflection on the rough surface is complex, and it is easy to lose signals during high-speed flight. Δv wind = u·d is the projection of the wind speed vector u and the flight direction d, which compensates for the influence of the wind field on the aircraft speed in real time, ensures the rationality of the speed threshold, and provides an accurate threshold for signal selection based on speed in the signal acquisition strategy.

[0099] (3) Task requirement assessment model

[0100] 1. Quantification of high-precision measurement requirement in this application

[0101]

[0102] Different altimetry tasks have different precision requirements, and this application quantifies the precision requirements. It is the basic precision requirement, reflecting the ratio of the task requirement to the system capability. It is the altitude correction term, where δ = 0.2 is the correction coefficient. As the flight altitude Altitude increases, the signal propagation distance is long and the intensity is weak, resulting in a decrease in altimetry precision. The precision requirement is corrected through a sine function. It is the target size correction term, where α = 0.3 is the coefficient. Small targets require higher precision measurement, while large targets have relatively lower precision requirements. The altimetry precision requirement index T1 is comprehensively obtained for signal strategy selection.

[0103] 2. Quantify the real-time requirement of this application

[0104]

[0105] The real-time requirement is related to the motion state of the aircraft, and this application quantifies this requirement. It is the basic term of the update rate. The exponential function makes the index change sensitive at low update rates and gentle at high update rates. It is the velocity deviation term, v opt = 15 m / s is the optimal sampling velocity. When the aircraft velocity deviates, the real-time requirement increases to ensure accurate tracking under high-speed motion. It is the acceleration influence term, where β = 0.4 is the coefficient. When accelerating or decelerating violently, a higher update rate is required to capture the motion changes, and the real-time requirement index T2 is obtained to assist in signal strategy decision-making.

[0106] (IV) Signal strategy decision model

[0107]

[0108] Based on the comprehensive evaluation results of environmental perception, motion state, and task requirements, the signal acquisition strategy S is decided. When it is determined that the terrain is flat (P(flat) > 0.8) and the velocity is less than the threshold (v < v thresh ), the first single-frequency signal of the B1C / L1 frequency band is selected, which is suitable for simple environments; when the velocity is greater than or equal to the threshold (v ≥ v thresh ), the second single-frequency signal of the B2a / L5 frequency band is selected, and its characteristics such as wavelength and symbol rate are suitable for high-speed scenarios. When it is determined that the area is urban or mountainous (P(urban) > 0.5 ∨ P(mountain) > 0.5), or the altimetry precision requirement is within a certain range (T 1min < T1 ≤ T 1max), select the dual-frequency signal (B1C / L1 + B2a / L5), which can eliminate ionospheric errors and resist interference. When the high-precision measurement requirement is extremely high (T1 > T 1max ), select the broadband signal (B1 I / B1C or B2a / B2b) and use the large bandwidth to achieve high-precision height measurement.

[0109] Optionally, according to at least one of the surface characteristics to which the specular reflection point belongs, the motion state of the unmanned aerial vehicle, and the height measurement task index, adaptively generate a signal acquisition strategy and a signal integration strategy, including: adaptively matching the coherent integration and non-coherent integration durations of the signal according to at least one of the surface characteristics to which the specular reflection point belongs, the motion state of the unmanned aerial vehicle, and the height measurement task index to generate a signal integration strategy.

[0110] Optionally, according to at least one of the surface characteristics to which the specular reflection point belongs, the motion state of the unmanned aerial vehicle, and the height measurement task index, adaptively match the coherent integration and non-coherent integration durations of the signal to generate a signal integration strategy, including: if it is determined according to the surface characteristics to which the specular reflection point belongs that the unmanned aerial vehicle is on flat ground or in the mountains, and it is determined according to the motion state of the unmanned aerial vehicle that the flight speed of the unmanned aerial vehicle is less than the set speed threshold, adaptively match the coherent integration duration of the signal to be greater than the set coherent integration duration threshold and the non-coherent integration duration to be greater than the set non-coherent integration duration threshold to generate a signal integration strategy; if it is determined according to the surface characteristics to which the specular reflection point belongs that the unmanned aerial vehicle is on flat ground or in the mountains, and it is determined according to the motion state of the unmanned aerial vehicle that the flight speed of the unmanned aerial vehicle is greater than or equal to the set speed threshold, adaptively match the coherent integration duration of the signal to be less than the set coherent integration duration threshold and the non-coherent integration duration to be less than the set non-coherent integration duration threshold to generate a signal integration strategy; if it is determined according to the surface characteristics to which the specular reflection point belongs that the unmanned aerial vehicle is in the city, and it is determined according to the motion state of the unmanned aerial vehicle that the flight speed of the unmanned aerial vehicle is less than the set speed threshold, adaptively match the coherent integration duration of the signal to be greater than the set coherent integration duration threshold and the non-coherent integration duration to be greater than the set non-coherent integration duration threshold to generate a signal integration strategy; if it is determined according to the surface characteristics to which the specular reflection point belongs that the unmanned aerial vehicle is in the city, and it is determined according to the motion state of the unmanned aerial vehicle that the flight speed of the unmanned aerial vehicle is greater than the set speed threshold, adaptively match the coherent integration duration of the signal to be less than the set coherent integration duration threshold and the non-coherent integration duration to be less than the set non-coherent integration duration threshold to generate a signal integration strategy.

[0111] Therefore, in specific applications, for example, when on flat ground or in mountainous areas and the flight speed is less than the set speed threshold, the received signal is relatively stable. To improve the height measurement accuracy, the adaptive matching signal coherent integration duration is greater than the set coherent integration duration threshold, and the non-coherent integration duration is greater than the set non-coherent integration duration threshold; when on flat ground or in mountainous areas but the flight speed is greater than or equal to the set speed threshold, to reduce the influence of the Doppler effect and phase change on coherent integration, the adaptive matching signal coherent integration duration is less than the set coherent integration duration threshold, and the non-coherent integration duration is less than the set non-coherent integration duration threshold. Similarly, in an urban environment, the integration duration is also matched similarly according to the flight speed. This is because increasing the coherent integration duration can improve the signal-to-noise ratio of the signal, but it is limited by the navigation data bit length and will introduce frequency errors; non-coherent accumulation is not affected by data jumps and can integrate and accumulate for a long time, but long-term non-coherent integration will introduce square loss, so it is necessary to comprehensively consider multiple factors to select the appropriate integration duration.

[0112] Optionally, based on the matched signal integration strategy, the signal-to-noise ratio of the collected GNSS-R height measurement direct signal and reflected signal is adjusted, including: based on the matched signal integration strategy, by increasing or shortening the signal coherent integration and non-coherent integration durations, to adjust the signal-to-noise ratio of the collected GNSS-R height measurement direct signal and reflected signal.

[0113] In summary, based on the matched signal integration strategy, the signal-to-noise ratio of the collected GNSS R height measurement direct signal and reflected signal is adjusted by changing the signal coherent integration and non-coherent integration durations. When the reflected signal is strong, appropriately shorten the coherent integration and non-coherent integration times to reduce the signal processing time; when the reflected signal is weak, increase the coherent integration time and non-coherent time to obtain a higher signal-to-noise ratio.

[0114] In the unmanned aerial vehicle GNSS R height measurement system, the signal-to-noise ratio of the direct signal and the reflected signal directly affects the accuracy and reliability of the height measurement result. Facing complex and variable environments (such as multipath interference in cities and signal occlusion in mountainous areas), as well as different intensities of direct / reflected signal inputs, it is necessary to dynamically adjust the coherent integration and non-coherent integration durations to optimize the signal-to-noise ratio. The detailed principle is described as follows:

[0115] I. Construct a joint model of signal energy accumulation and noise growth

[0116]

[0117] It can be seen that this application starts from the interaction relationship between signal energy and noise energy to construct a signal-to-noise ratio adjustment model. The numerator represents signal energy accumulation, through the coherent integration energy E s-c and the non-coherent integration energy E s-i respectively multiplied by the corresponding number of integration times Nc , N i After superposition, it reflects the enhancement effect of the integration process on the signal; the denominator represents the noise energy accumulation, considering the coherent integration noise E n-c and the non - coherent integration noise E n-i 's contribution, and at the same time introduces a multipath effect correction term In the GNSS R altimetry scenario, the multipath effect (MPR) is widespread, especially in complex terrains such as cities and mountains. This correction term can adjust the noise influence weight according to the intensity of the multipath effect. Using this application of the present invention, it is possible to comprehensively quantify the dual effects of the integration duration change on the signal and the noise, accurately guide the signal - to - noise ratio optimization strategy, and meet the altimetry requirements in different environments.

[0118] In the above application of the present invention: SNR adj : The target signal - to - noise ratio after adjusting the integration duration, which is the core index for measuring the altimetry signal quality and directly determines the accuracy of the altimetry result. In high - precision altimetry tasks, it is necessary to ensure that SNR adj reaches a relatively high level to reduce measurement errors. E s-c : The signal energy of a single coherent integration, which is related to factors such as signal transmission power, propagation path loss, and antenna gain. In a flat environment, the signal propagation loss is relatively small, and E s-c is relatively high; while in areas blocked by high - rise buildings in the city or mountains, the signal undergoes multiple reflections and scatterings, and E s-c is significantly reduced. N c : The number of coherent integrations, which is related to the coherent integration duration T c ( T c-unit is the basic duration of a single coherent integration). Coherent integration improves the signal - to - noise ratio by accumulating in - phase signal energy, but limited by Doppler frequency shift and code - phase change, too many integration times will lead to phase divergence and reduce the signal accumulation efficiency. E s-i : The signal energy of a single non - coherent integration, which is the energy gain after processing the results of multiple coherent integrations. Non - coherent integration eliminates phase uncertainty through square summation and plays an important role when the signal phase is unstable, but it also introduces the accumulation of noise square terms. N i : The number of non - coherent integrations, which is related to the non - coherent integration duration T i ( T i-unit is the basic duration of a single non - coherent integration). Increasing the number of non - coherent integrations can further enhance the signal energy, but at the same time it will amplify the noise, and it needs to be coordinated with coherent integration for adjustment. E n-c : The noise energy of a single coherent integration, mainly from receiver thermal noise, environmental electromagnetic interference, etc. In areas with complex electromagnetic environments (such as industrial areas, near communication base stations), E n-c will increase significantly, affecting the signal quality. E n-i:The noise energy of single non - coherent integration. Due to the square operation of non - coherent integration, the noise energy grows faster than the signal energy, so the number of non - coherent integration times needs to be carefully controlled. ρ: The multi - path effect sensitivity coefficient, which reflects the influence degree of multi - path effect on the signal - to - noise ratio and is dynamically adjusted through historical data and machine learning algorithms. In urban canyon areas with severe multi - path effects, ρ takes a larger value, emphasizing the suppression of multi - path noise; in open areas with weak multi - path effects, ρ takes a smaller value. MPR: The real - time multi - path effect intensity index, which is calculated through signal delay power spectrum analysis. The larger the value, the more severe the multi - path effect. For example, in urban streets with many high - rise buildings, MPR may reach a high value, resulting in signal quality deterioration. MPR thres : The multi - path effect intensity threshold, which serves as a benchmark for judging whether the multi - path effect is severe. When MPR > MPR thres : It is necessary to strengthen the suppression of noise and adjust the integration strategy.

[0119] II. Constructing the integration duration based on dynamic weights

[0120]

[0121] Thus, based on the difference between the input signal - to - noise ratio SNR in and the target signal - to - noise ratio SNR target , as well as the multi - path effect intensity MPR, this application dynamically adjusts the coherent integration duration T c and the non - coherent integration duration T i . For T c , when SNR in is lower than SNR target , the integration duration is increased through the exponential term to accumulate signal energy; meanwhile, the stronger the multi - path effect (the larger MPR is), the shorter T c is made to avoid the aggravation of phase divergence. For T i , when the signal is weak, the integration duration is increased through to enhance the signal, while when the multi - path effect is strong, T i is appropriately increased to suppress noise. This application comprehensively considers signal strength and environmental interference, realizes the precise adaptive adjustment of the integration duration, maximizes the signal - to - noise ratio in different scenarios, and meets the requirements of the height measurement task.

[0122] In the above application: T c : The adjusted coherent integration duration, which is a key parameter for controlling signal phase accumulation and energy enhancement, directly affecting the signal - to - noise ratio improvement effect and height measurement accuracy. T c-base : The coherent integration duration reference value, which is set according to device performance, signal frequency band characteristics, and general height measurement scenarios, serving as a basic reference for integration duration adjustment. ω c: Coherent integration duration adjustment exponent, which determines the influence degree of the difference between the input signal-to-noise ratio and the target signal-to-noise ratio on T c In high-precision altimetry tasks with extremely high requirements for signal-to-noise ratio, ω c takes a larger value, making T c more sensitive to signal intensity changes and quickly increasing the integration duration to improve signal quality. δ c : Inhibition exponent of multipath effect on coherent integration duration, which reflects the negative impact degree of multipath effect on coherent integration. In scenarios with complex multipath effects, δ c takes a larger value, and T c is shortened in a timely manner to avoid phase ambiguity and ensure signal processing effect. T i : Adjusted non-coherent integration duration, which is used for secondary energy accumulation and noise suppression of coherent integration results and affects the final optimization effect of signal-to-noise ratio. T i-base : Non-coherent integration duration reference value, which is set based on device performance and common altimetry conditions and provides a starting basis for non-coherent integration duration adjustment. ω i : Non-coherent integration duration adjustment exponent, which determines the action intensity of the difference between the input signal-to-noise ratio and the target signal-to-noise ratio on T i In scenarios where the signal is weak and the noise tolerance is high, ω i takes a larger value, allowing more non-coherent integration duration to be increased to enhance the signal. δ i : Promotion exponent of multipath effect on non-coherent integration duration. Considering the advantage of non-coherent integration in suppressing multipath noise, when the multipath effect is obvious, T i is appropriately increased through this exponent to improve the noise suppression ability. MPR max : Maximum value of multipath effect intensity, which is used as the upper limit of normalization processing and is used to measure the relative level of the current multipath effect intensity in extreme cases.

[0123] Optionally, systematic errors and random errors of the initial GNSS-R altimetry observations are corrected to obtain effective GNSS-R altimetry observations, including: based on the constructed atmospheric delay error correction model and random error correction model, systematic error correction and random error correction are respectively performed on the initial GNSS-R altimetry observations to obtain effective GNSS-R altimetry observations.

[0124] Optionally, based on the constructed atmospheric delay error correction model and random error correction model, systematic error correction and random error correction are respectively performed on the initial GNSS-R altimetry observations to obtain effective GNSS-R altimetry observations, including: based on the tropospheric error correction model and multi-frequency observation combination model, systematic errors in the initial GNSS-R altimetry observations are corrected.

[0125] To this end, in a specific application scenario, for shore-based and UAV-borne GNSS R altimetry, a large amount of observation data is collected to analyze the relationship between tropospheric refraction and temperature, humidity, pressure parameters, GNSS frequencies, and the height of the observation platform, and a high-precision tropospheric refraction error correction model is established. In the case of spaceborne applications, the low-altitude platform model is extended and optimized to establish a tropospheric correction model suitable for spaceborne applications. At the same time, the ionospheric influence is eliminated by using multi-frequency combinations or existing ionospheric correction models. Based on these tropospheric error correction models and multi-frequency observation combination models, the systematic errors in the initial GNSS R altimetry observations are corrected. In specific implementation, the exponential decay characteristic of atmospheric density with height is considered, which matches the height of the UAV flight environment (usually in the lower troposphere). The introduction of real-time meteorological parameters (temperature, humidity, pressure) in this application enables the model to adapt to different regions and weather conditions and improves the correction accuracy. In addition, for the characteristics of satellite high orbits, this application compensates for the deficiencies of traditional models in describing the delay of the upper atmosphere by introducing an orbital height correction term and an elevation angle non-linear correction term, which is particularly suitable for spaceborne GNSS-R systems. Furthermore, in the ionospheric model, the first-order ionospheric influence is completely eliminated based on the dual-frequency combination, which is the basis for high-precision positioning. In shore-based and spaceborne applications, dual-frequency receivers have become the standard configuration. Based on the triple-frequency combination, by combining wide-lane and narrow-lane combinations, both the ambiguity resolution speed and positioning accuracy are considered, which is particularly suitable for dynamic measurement scenarios. For example, when the UAV is rapidly maneuvering, the triple-frequency combination can fix the integer ambiguity faster and improve the positioning reliability.

[0126] In particular, in the GNSS-R altimetry system, the systematic errors mainly come from atmospheric delays (troposphere and ionosphere) and multipath effects. The model and its principle are as follows:

[0127] I. Tropospheric error correction model

[0128] 1. Layered mapping function model (applicable to low-altitude platforms)

[0129]

[0130] where: Δρ tro : Tropospheric delay error (unit: meter), which is one of the main systematic errors affecting the propagation time of GNSS signals. C h and C w : Zenith delay coefficients of the dry and wet components of the troposphere (unit: meter), respectively, obtained by fitting the Global Pressure and Temperature model (GPT) or regional meteorological data. In practical applications, the shore-based station can use the interpolation of local meteorological station data, and the UAV can obtain it in real time through the on-board micro-meteorological sensor. f(h): Height attenuation function, where h is the height of the observation platform, h0 is the reference height (usually taken as the sea level), and H is the atmospheric scale height (about 8500 m). This function reflects the exponential decrease characteristic of atmospheric density with height. m h m(θ), m w m(θ): They are the mapping functions of the dry component and the wet component respectively, and θ is the satellite elevation angle. Commonly used mapping functions include NMF (Niell Mapping Function) and GMF (Global Mapping Function) to convert the zenith delay to the delay at any elevation angle. The delay increases significantly at low elevation angles (close to the horizon). P, T: The real-time barometric pressure (hPa) and temperature (K) at the observation point, which are obtained in real time by the airborne platform through sensors and can be interpolated from meteorological station data for the shore-based platform. P0, T0: Standard barometric pressure (1013.25 hPa) and standard temperature (288.15 K). α, β: Barometric pressure and temperature correction coefficients, obtained by regression from historical data (typical values are α = 0.0026 and β = 0.0009), used to adjust the influence of meteorological condition changes on the delay.

[0131] Thus, it can be seen that the tropospheric delay is directly related to the atmospheric density on the signal propagation path. The dry component accounts for about 90% of the total delay and mainly depends on the barometric pressure; the wet component is small (about 10%) but varies significantly and is closely related to the water vapor content. The altitude attenuation function f(h) reflects the exponential decrease of atmospheric density with height and is particularly suitable for describing the environmental characteristics of the airborne platform (usually flying in the lower troposphere). The mapping function m(θ) takes into account the change of the signal path length with the elevation angle, and the delay increases significantly at low elevation angles. The barometric pressure and temperature correction terms and enable the model to adapt to different meteorological conditions and improve the correction accuracy.

[0132] 2. Extended Model for Spaceborne Platform

[0133]

[0134] where: h orbit : The satellite orbital height (such as about 20,200 km for GPS). γ: Orbital height correction coefficient, reflecting the additional delay of the upper atmosphere on the signal (typical value γ = 0.001). δ: Elevation angle non-linearity correction coefficient, considering the high elevation angle observation characteristics of the spaceborne receiver (typical value δ = 0.005).

[0135] Thus, it can be seen that for the spaceborne GNSS-R system, the traditional model has large errors because it does not consider the characteristics of the upper atmosphere. The extended model introduces the orbital height correction term and the elevation angle non-linearity correction term (1 + δ·sin 2θ), which more accurately describes the signal propagation path from space to the ground. For example, when the satellite approaches the horizon (low elevation angle), the signal passes through a longer tropospheric path, and the sin 2 θ term significantly increases the delay correction amount.

[0136] II. Ionospheric Error Correction Model

[0137] 1. Dual-Frequency Ionosphere-Free Combination

[0138]

[0139] Where: ρ IF : Ionosphere-free combination pseudorange observation value (unit: meter). ρ1, ρ2: Original pseudorange observation values of frequencies f1 and f2 respectively. f1, f2: Two different frequencies of GNSS signals (such as for GPS, L1 = 1575.42 MHz, L2 = 1227.60 MHz; for Beidou, B1 = 1561.098 MHz, B2 = 1207.14 MHz).

[0140] It can be seen that the ionospheric delay is inversely proportional to the square of the signal frequency, that is Through the dual-frequency linear combination, the first-order ionospheric effect can be completely eliminated. This combination is widely used in high-precision positioning, especially suitable for shore-based and spaceborne GNSS-R systems. For an airborne UAV platform, if only single-frequency signals (such as B1C / L1) are received, corrections need to be made in combination with the Global Ionospheric Map (GIM).

[0141] 2. Triple-Frequency Optimization Combination (Taking Beidou Triple-Frequency as an Example)

[0142]

[0143] Where: ρ iono-free : Ionosphere-free combination pseudorange. φ1, φ2: Carrier phase observation values of different frequencies. φ wide-lane : Wide-lane combination phase (wavelength is about 10 meters), used for rapid ambiguity resolution. φ narrow-lane : Narrow-lane combination phase (wavelength is about 10 centimeters), used for high-precision positioning.

[0144] It can be seen that the triple-frequency combination provides richer observation information. The wide-lane combination has a longer wavelength, reducing the difficulty of integer ambiguity resolution; the narrow-lane combination retains high measurement accuracy. This combination is especially suitable for airborne UAV dynamic measurements, which can significantly improve the initialization speed and positioning reliability. For example, in an urban environment, quickly resolving the ambiguity helps reduce the problem of signal loss due to signal blockage.

[0145] In summary, assuming the initial GNSS-R altimetry observation value is P obs , the observation value after systematic error correction is Psys-corr , then:

[0146] For shore-based and UAV-borne (low-altitude platform)

[0147] 1. First, perform tropospheric error correction:

[0148] Calculate the tropospheric delay error Δρ tro , according to the stratified mapping function model:

[0149]

[0150] Obtain the preliminarily corrected observable P tro-corr1 = P obs - Δρ tro

[0151] 2. Then, perform ionospheric error correction (assuming the use of dual-frequency ionosphere-free combination, if it is triple-frequency, correct according to the triple-frequency model):

[0152] Calculate the ionosphere-free combination pseudorange observation value where ρ1 and ρ2 are the original pseudorange observation values of frequencies f1 and f2 respectively.

[0153] The finally system-error-corrected observable P sys-corr = ρ IF - (ρ IF - P tro-corr1 )(Here, the results of tropospheric correction and ionospheric correction are fused)

[0154] For spaceborne platforms

[0155] 1. First, perform tropospheric error correction:

[0156] Calculate the tropospheric delay error where Δρ tro is first calculated according to the low-altitude platform stratified mapping function model and then substituted into this formula.

[0157] Obtain the preliminarily corrected observable P tro-corr2 = P obs - Δρ space

[0158] 2. Then, perform ionospheric error correction (assuming the use of dual-frequency ionosphere-free combination, if it is triple-frequency, correct according to the triple-frequency model):

[0159] Calculate the ionosphere-free combination pseudorange observation value where ρ1 and ρ2 are the original pseudorange observation values of frequencies f1 and f2 respectively.

[0160] The finally system-error-corrected observable P sys-corr = ρ IF-(ρ IF -P tro-corr2 (Here, the results after tropospheric correction are fused with the ionospheric correction results)

[0161] In summary, the present application comprehensively considers tropospheric and ionospheric error corrections, reflecting the processing methods of observables in the system error correction process for different platforms (low-altitude and spaceborne).

[0162] Alternatively, in another application scenario, based on the constructed atmospheric delay error correction model and random error correction model, systematic error correction and random error correction are respectively performed on the initial GNSS-R altimetry observables to obtain effective GNSS-R altimetry observables, including: calling the constructed tropospheric model and inputting environmental parameters (temperature, humidity, pressure parameters, GNSS frequency points, observation platform height data) to correct the tropospheric refraction error of the initial path difference observation value to obtain a preliminarily corrected path difference observation value; performing epoch-by-epoch difference operation on the preliminarily corrected path difference observation value and the theoretical path difference value calculated based on the geometric positions of the satellite and the aircraft to obtain a tropospheric residual error sequence; extracting the residual error values of each epoch in the tropospheric residual error sequence, correlating the corresponding environmental parameters (wet temperature, pressure parameters, GNSS-R frequency points, observation platform height, satellite elevation angle), and performing statistical analysis to quantify the correlation degree between the residual error and each environmental parameter and obtain an error-parameter correlation feature set accordingly; based on the error-parameter correlation feature set, constructing a tropospheric residual error correction model with environmental parameters as input variables and residual errors as output variables; inputting the preliminarily corrected path difference observation value into the tropospheric residual error correction model to calculate the predicted residual error value, and subtracting the predicted residual error value from the preliminarily corrected path difference observation value; determining multi-frequency observables to generate ionosphere-free combined observables; and eliminating the ionospheric delay error in the preliminarily corrected path difference observation value based on the ionosphere-free combined observables.

[0163] Particularly, in another specific application scenario, the specific implementation processes of the above steps are as follows:

[0164] I. Tropospheric refraction error correction

[0165] (1) Calculation of the initial value of tropospheric delay considering multi-factor coupling

[0166] Traditional tropospheric models only simply correlate temperature, humidity, pressure with delay, which is greatly expanded in this technology. The GNSS frequency point is introduced. Since the propagation characteristics of signals at different frequency points in the troposphere have slight differences, the frequency point influence factor f is obtained by fitting a large amount of experimental data factor . For the observation platform height, it is no longer limited to simple linear correction, and a non-linear functional relationship is constructed. Based on the improved Saastamoinen model, the calculation of the zenith tropospheric delay ZTD is as follows:

[0167] ZTD = ZHD + ZWD + ΔZ freq +ΔZ height

[0168]

[0169]

[0170] ΔZ freq = α × (f - f0) × ZHD

[0171]

[0172] Wherein, P s is the surface air pressure (hPa), φ is the station latitude (radians), H s is the station altitude (km), k3 is the water vapor refraction constant (22.1K / hPa), T m is the average atmospheric temperature (K), and in this application, T m = 70.2 + 0.72 × T s is calculated, T s is the surface temperature (K), e s is the surface water vapor pressure (hPa), and is calculated from the relative humidity RH: α and β are coefficients obtained by regression analysis of a large amount of data, f is the current GNSS frequency point, and f0 is the reference frequency point.

[0173] In terms of the mapping function, the traditional single fixed form is abandoned, and an adaptive mapping function is adopted. According to the height H of the observation platform s and the satellite elevation angle θ, the mapping function parameters a and b are dynamically adjusted:

[0174]

[0175] a = 0.0011 + 0.0001 × H s

[0176] b = 0.002 + 0.0002 × H s

[0177] Where R e is the average radius of the earth (6371 km).

[0178] The calculation of the preliminary corrected path difference observation value P corr1 is:

[0179] P corr1 = P obs - M(θ) × ZTD Where P obs is the initial path difference observation value.

[0180] (2) In-depth mining of residual errors and construction of multivariate models

[0181] After obtaining the preliminary corrected path difference observation value P corr1 , perform epoch-by-epoch difference operation with the theoretical path difference P theory calculated strictly based on the geometric positions of satellites and aircraft. The theoretical path difference P theory is determined based on a high-precision orbit model and the satellite-receiver relative position relationship:

[0182]

[0183] where (X sat , Y sat , Z sat ) is the position of the satellite in the ECEF coordinate system, and (X rec , Y rec , Z rec ) is the position of the receiver in the ECEF coordinate system.

[0184] Calculate the residual error ΔP epoch by epoch:

[0185] ΔP i = P corr1,i - P theory,i (i = 1, 2,..., n)

[0186] For the residual error sequence {ΔP1, ΔP2,..., ΔP n}, deeply analyze its relationship with various environmental parameters. Not only consider the conventional wet temperature, pressure, GNSS-R frequency points, observation platform height, satellite elevation angle, but also innovatively incorporate meteorological micro-environment parameters such as wind speed and wind direction. Use a method combining principal component analysis (PCA) and partial least squares regression (PLSR) to quantify the correlation degree between the residual error and each environmental parameter.

[0187] The constructed multiple regression model is:

[0188] ΔP = β0 + β1×T + β2×P + β3×e + β4×f + β5×H + β6×θ + β7×v w + β8×d w + ∈

[0189] where T is the temperature (°C), P is the pressure (hPa), e is the water vapor pressure (hPa), f is the GNSS frequency (GHz), H is the observation platform height (km), θ is the satellite elevation angle (radians), v w is the wind speed (m / s), d w is the wind direction (degrees), β0, β1,..., β8 are regression coefficients to be estimated, and ∈ is the random error term.

[0190] To solve the problem of multicollinearity, this application adopts the Ridge Regression method and determines the optimal regularization parameter λ through leave-one-out cross-validation. The finally obtained tropospheric residual error correction model is as follows:

[0191]

[0192] where are the regression coefficients estimated by ridge regression.

[0193] The quadratic correction is applied to the predicted value of the residual error to obtain the observed value P of the quadratic correction path difference corr2 :

[0194] P corr2 = P corr1 - ΔP pred

[0195] II. Elimination of ionospheric delay error

[0196] (I) Optimization algorithm for multi-frequency combination

[0197] For the elimination of ionospheric delay error, the traditional simple dual-frequency combination mode is not applicable. Considering the multi-frequency signal characteristics of different satellite systems (such as GPS, Beidou, GLONASS, Galileo), a generalized multi-frequency combination model is constructed. Taking the L1, L2, L5 frequency points of GPS and the B1I, B2I, B3I frequency points of Beidou as examples, the ionosphere-free combination observed value P IF :

[0198] where P ij is the pseudorange observed value of the j-th frequency point of the i-th satellite system, n is the number of satellite systems, m is the number of frequency points participating in the combination for each satellite system, and w ij is the combination weight.

[0199] The weight w ij is determined by minimizing the ionospheric delay residual and the noise amplification effect. Based on a large amount of measured data, the genetic algorithm (GA) is used for global optimization search. The objective function is: where is the variance of the ionospheric delay residual, is the noise variance, and λ is the weight coefficient for balancing the ionospheric delay elimination effect and the noise amplification. Its optimal value is determined through multiple experiments.

[0200] (II) Combined tropospheric correction result

[0201] The ionospheric delay error elimination and the tropospheric refraction error correction result are deeply fused. After obtaining the ionosphere-free combination observed value PIF Afterwards, considering that the residual error after tropospheric correction still affects the final result, a joint optimization model is constructed. The covariance matrix C of the tropospheric residual error is introduced. tropo and the covariance matrix C of the ionospheric residual error iono , and joint estimation is carried out by the weighted least squares (WLS) method:

[0202]

[0203] where is the effective GNSS-R altimetry observation after joint correction of ionospheric and tropospheric errors.

[0204] In summary, in the tropospheric refraction error correction of this application, the limitation of only considering a few factors in the traditional way is broken through, and multi-dimensional complex factors such as GNSS frequency points, observation platform height, wind speed and direction are innovatively integrated to construct a highly comprehensive error correction model. Through a large amount of experimental data and advanced data analysis methods, the complex relationships between various factors and tropospheric delay and residual error are determined, greatly improving the accuracy and adaptability of tropospheric error correction.

[0205] In addition, when dealing with tropospheric residual error, various advanced algorithms such as principal component analysis, partial least squares regression, and ridge regression are used to deeply explore the potential correlations between residual error and various environmental parameters. The model parameters are optimized by means such as cross-validation to construct an accurate residual error correction model, effectively reducing the systematic error.

[0206] Here, for the elimination of ionospheric delay error, a generalized multi-frequency combination model is proposed, and the combination weights are optimized by using a genetic algorithm, significantly improving the elimination effect of ionospheric delay. At the same time, the ionospheric error elimination and the tropospheric correction result are innovatively jointly optimized, considering the mutual influence between different error sources, further improving the accuracy of GNSS-R altimetry observations.

[0207] Optionally, based on the constructed atmospheric delay error correction model and random error correction model, systematic error correction and random error correction are respectively performed on the initial GNSS-R altimetry observations to obtain effective GNSS-R altimetry observations, including:

[0208] Evaluating the accuracy of GNSS-R code delay observations based on correlation waves, and evaluating the accuracy of GNSS-R phase delay observations according to the continuity of phases to obtain the random error of the observations;

[0209] Based on the random error of the observations, satellite elevation angle, and the phase continuity of GNSS-R altimetry direct signal and reflected signal, a dynamic accuracy evaluation model that can flexibly select different types of GNSS-R altimetry observations is constructed;

[0210] Evaluate the reliability of different types of GNSS-R altimetry observables, and assign different reference weights to the corresponding types of GNSS-R altimetry observables based on the robust variance component;

[0211] Based on the output precision evaluation value of the dynamic precision evaluation model and the variances of different types of GNSS-R altimetry observables, perform elastic inversion operations on the constructed initial random model through the reference weights of different types of GNSS-R altimetry observables to obtain a random error correction model.

[0212] Therefore, in the process of random error correction, through precision evaluation and acquisition of the random error of the observed value, perform precision evaluation on the GNSS R-code delay observable by analyzing the relevant waveform quality, and perform precision evaluation on the GNSS R-phase delay observable according to the phase continuity to obtain the random error of the observed value. Then, in the constructed dynamic precision evaluation model, combine the random error of the observed value, the satellite elevation angle, and the phase continuity of the direct and reflected signals of GNSS-R altimetry to construct a dynamic precision evaluation model that can flexibly select different types of GNSS R altimetry observables. For the reference weight assignment, evaluate the reliability of different types of GNSS R altimetry observables and assign different reference weights to them based on the robust variance component. Finally, generate a random error correction model: based on the precision evaluation value output by the dynamic precision evaluation model and the variances of different types of GNSS R altimetry observables, combined with the reference weight, perform elastic inversion operations on the constructed initial random model to obtain a random error correction model. Finally, based on the constructed atmospheric delay error correction model (troposphere and ionosphere related models) and the random error correction model, perform systematic error correction and random error correction on the initial GNSS R altimetry observables respectively to obtain effective GNSSR altimetry observables.

[0213] In summary, in the GNSS-R altimetry system, random error correction is mainly achieved through dynamic precision evaluation and robust estimation. The following takes an application in a specific application scenario as an example, and combines the present application that reflects the algorithm principle to illustrate the principle of random error correction.

[0214] I. Acquisition of the random error of the observed value

[0215] 1. Precision evaluation of the code delay based on the correlation wave

[0216]

[0217] Where: σ code : Standard deviation of code delay measurement (unit: meter), c: speed of light (about 299,792,458 m / s), σ corr: Full width at half maximum of the correlation function (unit: chip), SNR: Signal-to-noise ratio, reflecting the signal quality, B: Receiver bandwidth (unit: Hz), determining the signal resolution.

[0218] It can be seen that for the physical modeling of the code delay measurement error of the GNSS receiver. The receiver matches the received signal with the local code through the correlation function, and noise will cause the peak of the correlation function to broaden (σ corr ), directly affecting the accuracy of code delay measurement. In this application, the speed of light c converts the chip error into a distance error; the signal-to-noise ratio SNR is inversely proportional to the noise power, reflecting the damage of noise to the quality of the correlation function - when the environmental electromagnetic interference increases and causes the SNR to decrease, the code delay error will increase significantly; the receiver bandwidth B determines the signal resolution, and a higher B can provide a sharper correlation function and reduce the error.

[0219] 2. Phase Delay Accuracy Evaluation Based on Phase Continuity

[0220]

[0221] Where: σ phase : Standard deviation of phase delay measurement (unit: meter), λ: Signal wavelength (for example, the L1 frequency corresponds to approximately 0.19 meters), T coh : Coherent integration time (unit: second), reflecting the phase tracking stability.

[0222] It can be seen that in this application, the carrier phase measurement accuracy depends on the signal wavelength λ. The longer the wavelength, the smaller the inherent measurement error; the coherent integration time T coh suppresses noise by accumulating signal energy, but is limited by the Doppler frequency shift and the risk of signal loss of lock. The in this application reflects the balance between noise suppression and signal stability. In the scenario of high-speed flight of the unmanned aerial vehicle, an excessive T coh will cause phase loss of lock due to the Doppler frequency shift, and this application can accurately calculate the phase measurement error based on this, providing a basis for subsequent error correction.

[0223] II. Dynamic Accuracy Evaluation Model

[0224] 1. Comprehensive Accuracy Evaluation Index

[0225]

[0226] Where: σ est : Estimated accuracy of the observed value (unit: meter), σ0: Reference accuracy, depending on the receiver hardware characteristics, θ: Satellite elevation angle, low elevation angle observations are more severely affected by multipath and the atmosphere, Δφ: Phase change amount between adjacent epochs, reflecting phase continuity, φ thres : Phase jump threshold, exceeding the threshold is regarded as loss of lock, σ ref: Reference code delay accuracy, used for normalization. α, β, γ: Weight coefficients, calibrated through historical data.

[0227] It can be seen that this application provides a non-linear evaluation framework that fuses multiple factors. Among them, exp(-α·sinθ) is based on the principle of geometric optics and describes the influence of the satellite elevation angle θ on the signal propagation path - low-elevation signals are more affected by atmospheric refraction and multipath effects, and the exponential decay model is used to quantify the error growth trend; The signal quality is judged by using phase continuity. When the phase change Δφ between adjacent epochs exceeds the threshold φ thres it indicates that there may be a phase jump, and the error estimate value is improved through a linear amplification mechanism; Then the code delay accuracy is associated with the carrier phase accuracy to achieve the weight balance of the two types of observations. In the urban canyon scenario, this model can simultaneously consider the phase jumps caused by low elevation angles and signal blockages, as well as the code measurement errors caused by multipath, and comprehensively evaluate the accuracy of the observations.

[0228] 2. Elastic weight function

[0229]

[0230] Where: w i : The weight of the i-th observation value, σ i : The estimated accuracy of the i-th observation value, σ min 、σ max : The upper and lower limit thresholds of accuracy, p: Exponential parameter, controlling the weight decay rate.

[0231] It can be seen that this application provides a soft decision-making mechanism based on fuzzy mathematics. Compared with the traditional hard threshold method (directly eliminating low-quality observation values), the elastic weight function realizes the smooth transition of the observation value weights through the power function . When the accuracy σ of the observation value i is in the critical range, the weight does not suddenly drop to 0, but decays according to the power according to the error degree, which not only suppresses the negative impact of low-quality data, but also retains some available information, improving the data utilization rate and the robustness of the model.

[0232] III. Robust variance component estimation

[0233] 1. Robust weight function

[0234]

[0235] Where: u: Standardized residual, v i:Residual of the i-th observation value, k: Threshold parameter (typical value 2.5 - 3.0), controlling the robust range, ρ(u): Huber loss function, suppressing the influence of gross errors, w(u): Corresponding weight function, reducing the weight when the residual exceeds the threshold.

[0236] Thus, in Gaussian noise, the Huber loss function ensures the estimation efficiency; while when there are gross errors (non-Gaussian noise) in the data, the loss function turns into linear growth after the residual |u| exceeds the threshold k, avoiding the excessive influence of individual outliers on the estimation result. By adjusting the threshold k, a flexible trade-off can be made between the estimation efficiency and robustness. Compared with the traditional least squares method, it can effectively resist the gross error interference of 10% - 20% of the data volume and ensure the stability and reliability of the height measurement result.

[0237] 2. Variance component estimation

[0238]

[0239] Where: Variance estimation of the i-th type of observation value, v i : Residual vector of the i-th type of observation value, P i : Weight matrix of the i-th type of observation value, r i : Redundant observation number of the i-th type of observation value, τ i : Robust factor, reflecting the reliability of the observation value, Variance of unit weight.

[0240] Thus, in this application, the variance of the observation value is estimated through the residual vector v i , weight matrix P i and redundant observation number r i and the robust factor τ is introduced for adaptive adjustment according to the residual distribution. When there are gross errors in a certain type of observation value, τ i increases, causing the estimated variance i to overestimate the true variance, thereby reducing the weight of this type of observation value in the adjustment calculation and achieving the dynamic balance of the weights of different types of observation values (such as code phase, carrier phase), and improving the overall solution accuracy.

[0241] IV. Elastic inversion stochastic model

[0242] 1. Stochastic model update

[0243]

[0244] Where: ∑ k : Stochastic model covariance matrix at the k-th moment, α: Learning rate (between 0 and 1), controlling the update rate, w i : Weight of the i-th observation value, Variance estimation of the i-th observation value.

[0245] It can be seen that this application is based on the current stochastic model ∑ k and controls the update step size through the learning rate α, incorporating the variance information of new observation values into the model. In the scenario where the UAV enters the urban environment from an open area, this mechanism can quickly sense the change in signal quality. By adjusting the magnitude of α, on the premise of ensuring stability, the stochastic model can adapt to the new environment within seconds and optimize the weight allocation of observation values.

[0246] 2. Adaptive filtering gain

[0247]

[0248] Where: K k : Kalman filtering gain matrix, Prediction error covariance matrix, H k : Observation matrix, ∑ k : Updated stochastic model covariance matrix.

[0249] It can be seen that considering that traditional Kalman filtering relies on a fixed system noise model, while this application adjusts the filtering gain K k through the dynamically updated stochastic model covariance matrix ∑ k . When encountering non-Gaussian noise (such as sudden electromagnetic interference) or abnormal observations (such as loss of lock caused by signal occlusion), ∑ k reflects the error change in real time, enabling K k to automatically reduce the weight of affected observation values, enhancing the adaptability of the filtering algorithm to complex environments and ensuring the continuity and accuracy of the height measurement results.

[0250] In summary, let the initial GNSS-R height measurement observation be P obs , and the observation after random error correction be P rand-corr :

[0251] Based on the code delay accuracy assessment of the correlation wave, the code delay measurement standard deviation σ code :

[0252]

[0253] Based on the phase delay accuracy assessment of phase continuity, the phase delay measurement standard deviation σ phase :

[0254]

[0255] Calculate the comprehensive accuracy evaluation index σ est :

[0256]

[0257] Calculate the weight w of the i-th observation according to the elastic weight function i :

[0258] where σ i is the estimated accuracy of the i-th observation (σ est can be used here).

[0259] Calculate the robust weight function w(u):

[0260] where v i is the residual of the i-th observation, and ρ(u) is the Huber loss function:

[0261]

[0262] Calculate the variance estimate of the i-th type of observation

[0263]

[0264] Update the random model to obtain the updated random model covariance matrix ∑ k+1 :

[0265]

[0266] Calculate the adaptive filtering gain matrix K k :

[0267]

[0268] Through the above calculations, the finally observed quantity P after random error correction rand-corr can be expressed as:

[0269] P rand-corr = P obs - ΔP rand

[0270] The specific adjustment amount ΔP rand involves the comprehensive adjustment of the observed values according to weights, variance estimates, etc. In actual calculations, P obs can be adjusted by the filtering algorithm (such as the relevant calculations of Kalman filtering) combined with the parameters obtained from the above calculations to achieve random error correction. That is, for the initial observed quantity P obs , using the weight w i , variance estimate and the random model covariance matrix ∑ k+1 , filtering gain matrix K ketc., adjust the observation quantity according to the calculation logic of the filtering algorithm, so as to obtain P after random error correction rand-corr .

[0271] Let the adjustment amount of the random error based on the above calculation be ΔP rand , and the present application derives its mathematical formula in the following way:

[0272] 1. Estimate the relevant parameters of the adjustment amount based on robust variance component estimation and weight calculation

[0273] According to the robust variance component estimation, the variance estimation of the i-th type of observation value is Combined with the processing of the observation value residual v i by the robust weight function w(u), and the observation value weight w i , the weighted residual quantity related to the observation value can be obtained.

[0274] For the i-th observation value, the residual influence quantity after considering the weight and variance is Here, w i reflects the reliability weight of this observation value, and reflects the influence degree of the residual on the result under its variance level.

[0275] 2. Calculate the adjustment amount by combining random model update and filtering gain

[0276] In the elastic inversion random model, the random model is updated to obtain the covariance matrix ∑ k+1 , and the adaptive filtering gain matrix is

[0277] Taking all observation values into consideration, the adjustment amount ΔP of the random error based on the above calculation rand can be expressed as:

[0278]

[0279] where n is the total number of observation values. This formula comprehensively considers factors such as the weight, residual, variance, and filtering gain of the observation values, and reflects the adjustment effect on the random error in the initial GNSS-R altimetry observation quantity. Through this adjustment amount, combined with the initial observation quantity P obs , the finally obtained observation quantity P rand-corr = P obs - ΔP rand .

[0280] In summary, the present application is a process that starts from obtaining random error-related parameters, constructs an evaluation model, performs robust estimation, and updates the random model, and finally realizes the correction of the random error of the initial GNSS-R altimetry observation. Among them, the Huber loss function ρ(u) is used to suppress the influence of gross errors and plays a key role in calculating the robust weight function w(u), thereby affecting the variance component estimation and the subsequent update of the random model and the correction of the observations.

[0281] The above description is only for the embodiments of the present application and is not intended to limit the present application.

Claims

1. A baseband signal processing method applicable to GNSS-R altimetry of unmanned aerial vehicles for complex surface types, characterized in that, Including: Adaptive generate a signal acquisition strategy and a signal integration strategy according to at least one of the surface features to which the specular reflection point belongs, the motion state of the unmanned aerial vehicle, and the altimetry task index; Collect GNSS-R altimetry direct signals and reflection signals based on the matched signal acquisition strategy; Adjust the signal-to-noise ratio of the collected GNSS-R altimetry direct signals and reflection signals based on the matched signal integration strategy; Calculate the path difference between the GNSS-R altimetry direct signals and reflection signals after signal-to-noise ratio adjustment as the initial GNSS-R altimetry observation; Correct the systematic error and random error of the initial GNSS-R altimetry observation to obtain an effective GNSS-R altimetry observation.

2. The method according to claim 1, characterized in that, Adaptive generate a signal acquisition strategy and a signal integration strategy according to at least one of the surface features to which the specular reflection point belongs, the motion state of the unmanned aerial vehicle, and the altimetry task index, including: Generate an environmental feature vector according to at least one of the surface features to which the specular reflection point belongs, the motion state of the unmanned aerial vehicle, and the altimetry task index; Adaptive generate a signal acquisition strategy and a signal integration strategy according to the environmental feature vector.

3. The method according to claim 2, characterized in that, The method described above includes: Extract features from the surface image to determine the surface features to which the specular reflection point belongs; Determine the motion state of the unmanned aerial vehicle based on the inertial navigation data on the unmanned aerial vehicle.

4. The method according to claim 1, wherein Adaptive generate a signal acquisition strategy and a signal integration strategy according to at least one of the surface features to which the specular reflection point belongs, the motion state of the unmanned aerial vehicle, and the altimetry task index, including: If it is determined that the unmanned aerial vehicle is on flat ground according to the surface features to which the specular reflection point belongs, and it is determined that the flight speed of the unmanned aerial vehicle is less than the set speed threshold according to the motion state of the unmanned aerial vehicle, then adaptively select the first single-frequency signal altimetry to generate a signal acquisition strategy; If it is determined that the unmanned aerial vehicle is on flat ground according to the surface features to which the specular reflection point belongs, and it is determined that the flight speed of the unmanned aerial vehicle is greater than or equal to the set speed threshold according to the motion state of the unmanned aerial vehicle, then adaptively select the second single-frequency signal altimetry to generate a signal acquisition strategy; If it is determined that the unmanned aerial vehicle is in a city or mountain area according to the surface features to which the specular reflection point belongs, then adaptively select dual-frequency signal altimetry to generate a signal acquisition strategy; If the altimetry task index is greater than the set first task index threshold but less than the set second task index threshold, then adaptively select dual-frequency signal altimetry to generate a signal acquisition strategy; If the altimetry task index is greater than the set second task index threshold, then adaptively select wideband signal altimetry to generate a signal acquisition strategy.

5. The method according to claim 1, characterized in that, The first single-frequency signal altimetry is based on the B1C / L1 frequency point, the second single-frequency signal altimetry is based on the B2a / L5 frequency point; the dual-frequency signal altimetry is based on the B1C / L1 frequency point and the B2a / L5 frequency point; the wideband signal altimetry is based on the B1 I / B1C or B2a / B2b frequency point.

6. The method according to claim 1, characterized in that, Adaptive generate a signal acquisition strategy and a signal integration strategy according to at least one of the surface features to which the specular reflection point belongs, the motion state of the unmanned aerial vehicle, and the altimetry task index, including: Adaptive matching of the signal coherent integration and non-coherent integration durations is performed according to at least one of the surface features to which the specular reflection point belongs, the motion state of the unmanned aerial vehicle, and the altimetry task index to generate a signal integration strategy.

7. The method according to claim 6, wherein Adaptive matching of the signal coherent integration and non-coherent integration durations is performed according to at least one of the surface features to which the specular reflection point belongs, the motion state of the unmanned aerial vehicle, and the altimetry task index to generate a signal integration strategy, including: If it is determined according to the surface features to which the specular reflection point belongs that the unmanned aerial vehicle is in a flat area or a mountainous area, and it is determined according to the motion state of the unmanned aerial vehicle that the flight speed of the unmanned aerial vehicle is less than the set speed threshold, adaptively match the signal coherent integration duration to be greater than the set coherent integration duration threshold and the non-coherent integration duration to be greater than the set non-coherent integration duration threshold to generate a signal integration strategy; If it is determined according to the surface features to which the specular reflection point belongs that the unmanned aerial vehicle is in a flat area or a mountainous area, and it is determined according to the motion state of the unmanned aerial vehicle that the flight speed of the unmanned aerial vehicle is greater than or equal to the set speed threshold, adaptively match the signal coherent integration duration to be less than the set coherent integration duration threshold and the non-coherent integration duration to be less than the set non-coherent integration duration threshold to generate a signal integration strategy; If it is determined according to the surface features to which the specular reflection point belongs that the unmanned aerial vehicle is in a city, and it is determined according to the motion state of the unmanned aerial vehicle that the flight speed of the unmanned aerial vehicle is less than the set speed threshold, adaptively match the signal coherent integration duration to be greater than the set coherent integration duration threshold and the non-coherent integration duration to be greater than the set non-coherent integration duration threshold to generate a signal integration strategy; If it is determined according to the surface features to which the specular reflection point belongs that the unmanned aerial vehicle is in a city, and it is determined according to the motion state of the unmanned aerial vehicle that the flight speed of the unmanned aerial vehicle is greater than the set speed threshold, adaptively match the signal coherent integration duration to be less than the set coherent integration duration threshold and the non-coherent integration duration to be less than the set non-coherent integration duration threshold to generate a signal integration strategy.

8. The method according to claim 1, wherein Based on the matched signal integration strategy, the signal-to-noise ratio of the collected GNSS-R altimetry direct signal and reflected signal is adjusted, including: Based on the matched signal integration strategy, the signal-to-noise ratio of the collected GNSS-R altimetry direct signal and reflected signal is adjusted by increasing or shortening the signal coherent integration and non-coherent integration durations.

9. The method according to claim 1, characterized in that The initial GNSS-R altimetry observation is corrected for systematic errors and random errors to obtain an effective GNSS-R altimetry observation, including: Based on the constructed atmospheric delay error correction model and random error correction model, the initial GNSS-R altimetry observation is respectively corrected for systematic errors and random errors to obtain an effective GNSS-R altimetry observation.

10. The method according to claim 1, wherein Based on the constructed atmospheric delay error correction model and random error correction model, the initial GNSS-R altimetry observation is respectively corrected for systematic errors and random errors to obtain an effective GNSS-R altimetry observation, including: Based on the tropospheric error correction model and the multi-frequency observation combination model, the systematic errors in the initial GNSS-R altimetry observation are corrected.

Images