An unmanned aerial vehicle-based overhead water surface apparent spectrum multi-probe synchronous observation system

Abstract

The application discloses a kind of based on unmanned aerial vehicle's suspended water surface apparent spectrum multi-probe synchronous observation system, it is related to unmanned aerial vehicle technical field, the system includes: water radiance probe, for measuring water surface spectral radiation signal;Downlink irradiance probe, for measuring incident to water surface spectral radiance;First micro-drive mechanism is connected to water radiance probe, for based on received first control instruction, drive the receiving field of view of water radiance probe first periodic motion is carried out according to preset trajectory;Data processing unit, for first control instruction as first reference input signal, signal analysis is carried out to the water surface spectral radiation signal synchronously collected, separates out the spectral signal component modulated by first periodic motion;Based on the first field of view center pointing of each sampling time and spectral signal component, reconstructs the spatial distribution information of the apparent optical characteristics of water in measurement region.

Description

Technical Field

[0001] This application relates to the field of unmanned aerial vehicle (UAV) technology, and more specifically to the field of UAV-mounted observation equipment. In particular, it relates to a UAV-based suspended multi-probe synchronous observation system for apparent spectral data of water bodies. Background Technology

[0002] In fields such as water resource management, water environment monitoring, and water ecological protection, the accurate acquisition of the apparent spectral characteristics of water bodies is of paramount importance. Observation and analysis of the apparent spectral information of water bodies can provide a scientific basis for water pollution early warning and water environment quality assessment, playing an irreplaceable role in ensuring the sustainable use of water resources and the stability of the ecological environment.

[0003] Currently, observation techniques for the apparent spectra of water bodies have made some progress, but there are still significant shortcomings in practical applications. On the one hand, traditional methods for observing the apparent spectra of water bodies mostly rely on ground-based fixed-point sampling or fixed observation platforms. These methods are difficult to achieve efficient synchronous observation of large areas of water, resulting in low observation efficiency. Especially when facing complex terrain or vast water bodies, it is difficult to comprehensively obtain the overall distribution of the surface optical properties of water bodies.

[0004] There is currently no effective solution to the above problems. Summary of the Invention

[0005] This application provides a UAV-based suspended water body apparent spectrum multi-probe synchronous observation system to solve the above-mentioned technical problems.

[0006] This application provides a suspended multi-probe synchronous observation system for apparent spectra of water bodies based on unmanned aerial vehicles (UAVs), including:

[0007] A water radiance probe, mounted on a drone platform, is used to measure the spectral radiation signal of the water surface;

[0008] A downlink irradiance probe, mounted on the UAV platform, is used to measure the spectral irradiance incident on the water surface;

[0009] A first micro-motion mechanism is connected to the water radiance probe and is used to drive the receiving field of view of the water radiance probe to perform a first periodic movement along a preset trajectory based on the received first control command.

[0010] The data processing unit is used to determine the orientation of the first field of view center of the water radiance probe at each sampling time based on the pose data of the UAV platform collected synchronously and the first control command; to perform signal analysis on the synchronously collected water surface spectral radiation signal using the first control command as the first reference input signal, and to separate the spectral signal component modulated by the first periodic motion; and to reconstruct the spatial distribution information of the surface optical properties of the water body within the measurement area based on the orientation of the first field of view center and the spectral signal component at each sampling time.

[0011] Based on the embodiments provided in this application, by integrating the water radiance probe and the downlink irradiance probe onto a UAV platform, the flexible mobility of the UAV can overcome the spatial limitations of ground-based fixed-point sampling or fixed observation platforms, easily covering complex terrain or vast water areas. This enables the synchronous acquisition of apparent spectral information of water bodies within a large measurement area, significantly improving the efficiency and coverage of water body surface spectral characteristics observation. It effectively solves the shortcomings of traditional observation methods that make it difficult to comprehensively obtain the overall distribution of water body surface spectral characteristics. The first micro-motion mechanism can drive the water radiance probe's receiving field of view to perform a first periodic movement along a preset trajectory. Simultaneously, the data processing unit can accurately determine the center direction of the first field of view of the water radiance probe at each sampling moment based on the synchronously acquired UAV pose data and the first control command. Furthermore, the data processing unit uses the first control command as the first reference input signal to analyze the synchronously acquired water surface spectral radiation signal, separating the spectral signal component modulated by the first periodic movement. Then, it combines the center direction of the field of view at each sampling moment with the spectral signal component to reconstruct the spatial distribution information of water body surface spectral characteristics. This process can effectively reduce the interference of external environmental factors on spectral data acquisition, ensure accurate matching between the acquired spectral data and the corresponding observation location and field of view, significantly improve the accuracy and reliability of surface spectral characteristic data of water bodies, and thus ensure the scientific nature of subsequent analysis results and meet the actual needs of water environment monitoring. Attached Figure Description

[0012] The accompanying drawings, which are included to provide a further understanding of embodiments of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0013] Figure 1 This is a structural diagram of an optional UAV-based suspended water body apparent spectrum multi-probe synchronous observation system according to an embodiment of this application;

[0014] Figure 2 This is a flowchart of an optional adaptive weighted interpolation algorithm according to an embodiment of this application.

[0015] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0017] According to one aspect of the embodiments of this application, such as Figure 1 As shown, this application provides a suspended multi-probe synchronous observation system for apparent spectra of water bodies based on unmanned aerial vehicles (UAVs), comprising:

[0018] A water radiance probe 101 is mounted on an unmanned aerial vehicle platform 102 to measure the spectral radiation signal of the water surface.

[0019] Downlink irradiance probe 103 is mounted on UAV platform 102 and is used to measure spectral irradiance incident on the water surface;

[0020] The first micro-motion mechanism 104 is connected to the water radiance probe and is used to drive the receiving field of view of the water radiance probe to perform a first periodic movement according to a preset trajectory based on the received first control command.

[0021] It's important to explain that a preset trajectory refers to a pre-defined, regular path pattern. In this system, this could be a circle, an infinity symbol shape (a type of Lissajous shape), or a set of parallel lines scanning back and forth. The purpose of these trajectories is to allow the probe's line of sight to perform regular, repeated scans within a small circular or square area (e.g., the size of a dinner plate), thereby covering multiple points within the area instead of constantly focusing on a single point.

[0022] Periodic motion refers to the scanning action described above being repeated continuously in a fixed pattern. Its repetition speed (frequency) is extremely fast, far exceeding the speed of the drone's own wobbling. The amplitude (displacement) of its back-and-forth movement is extremely small, typically between a few tenths of a millimeter and a few millimeters. This rapidly repeating scanning cycle must be matched with the exposure time of the spectral camera to ensure that the probe can complete a sufficient number of scans within a single exposure time, thereby capturing the complete motion pattern.

[0023] The data processing unit 105 is used to determine the orientation of the first field of view center of the water radiance probe at each sampling moment based on the pose data of the synchronously acquired UAV platform and the first control command; to perform signal analysis on the synchronously acquired water surface spectral radiation signal using the first control command as the first reference input signal, and to separate the spectral signal component modulated by the first periodic motion; and to reconstruct the spatial distribution information of the surface optical properties of the water body within the measurement area based on the orientation of the first field of view center and the spectral signal component at each sampling moment.

[0024] It needs to be explained that the first control command is the “command” sent by the brain (synchronous control unit) to the muscle (micro-motion mechanism), which describes how the muscle is to move (e.g., “move one micrometer to the right now”).

[0025] The "command" itself is used as a reference signal because there is a strict causal relationship and synchronization between it and the light signal actually received by the probe, which is "contaminated" by motion. It should be understood that changes in the light signal strictly follow this "command" in time. Therefore, this "command" becomes the "key" to extracting useful information from the mixed signal. Traditional methods, lacking this actively issued, known "command," cannot perform this efficient signal separation.

[0026] For example, the aforementioned "command" (reference signal) is input into this mixing console. The console continuously adjusts its parameters, striving to generate a noise signal that best matches the "command." It then subtracts this self-generated noise signal from the original mixed signal (main input). Since the "command" is synchronized with the changes in the light signal caused by motion and is independent of other noise, the subtraction leaves behind a clean, effective light signal modulated by our active motion. The specification should include a block diagram of the filter and describe the initialization settings of its parameters (such as filter length and learning step size) and the iterative convergence process to demonstrate its feasibility.

[0027] It should be noted that reconstruction refers to using the separated effective optical signals and the precise spatial location of each signal point, through computer algorithms, like piecing together... Figure 1 Similarly, a "map" is reassembled to display the optical properties of every point within the entire scanned area. This process includes: a) creating a regular pixel grid for the entire scanned area; b) assigning or interpolating the signal value of each sampled point to the pixels of the grid based on its geographical location; and c) finally generating a complete image. The information displayed on this "map" can be the radiance above water or more advanced information such as remotely sensed reflectance.

[0028] refer to Figure 1The present application relates to the field of observation equipment mounted on UAVs for a suspended multi-probe synchronous observation system of apparent spectra of water bodies.

[0029] It is important to explain that the fundamental difference between this embodiment and existing technologies lies in the introduction of a completely new methodology ("active modulation and demodulation") to solve the core problem. Existing technologies focus on "robustness" and "compensation," aiming to maintain probe stability as much as possible (e.g., through gimbals or stabilizing pan-tilt units) or to retroactively remove "bad" data caused by platform jitter. This is a passive and reactive approach; data quality is highly susceptible to environmental disturbances, and the information dimension is singular (only a single spectral value).

[0030] The approach of this embodiment is "utilization" and "enhancement." Specifically, it includes: actively introducing a known disturbance, that is, actively and consciously driving the probe's field of view to perform a preset, periodic movement through a micro-motion mechanism. This is not for stabilization, but to create a known and controllable scanning pattern. The core algorithm of the data processing unit is not a simple averaging or filtering, but uses the "control command" (i.e., the known signal driving the micro-motion mechanism) as a reference signal. Using this reference signal, the acquired mixed spectral signal is analyzed (for example, using techniques similar to coherent detection or adaptive filtering), actively separating the spectral signal components modulated by this known movement. The ultimate goal is not to obtain a more stable "point" spectrum, but to generate a small patch of "spatial distribution information of spectral intensity within the measurement area." This is a completely new dimension of information that traditional methods simply cannot obtain.

[0031] This embodiment cleverly addresses the challenge of maintaining absolute stability on unmanned aerial vehicle (UAV) platforms, rather than combating it, by transforming it into a new measurement dimension. The resulting spatial distribution information can be used to analyze the aggregation state and spatial structure of algal blooms, oil films, and other phenomena on the water surface, which is completely impossible with traditional single-point measurements, opening a new analytical window for water quality remote sensing.

[0032] In other words, this embodiment describes a novel method and system for implementing the method that uses the control signal of the probe to perform periodic motion by actively controlling it and then uses this motion as a reference to demodulate the spectral spatial distribution information. Existing technologies focus on measuring a point more "statically," while this embodiment focuses on measuring a surface "dynamically." This fundamental difference in concept constitutes its essential characteristic.

[0033] Furthermore, the UAV-based suspended water body apparent spectral multi-probe synchronous observation system also includes:

[0034] The pose sensing unit is used to collect attitude data of the UAV platform.

[0035] The synchronous control unit is used to generate a first control command and send the first control command to the first micro-motion mechanism, and to synchronously trigger the data acquisition of the pose sensing unit and the spectral measurement of the water radiance probe and the downlink irradiance probe.

[0036] It should be noted that this system employs a hardware triggering mechanism to achieve synchronization. The synchronization control unit sends a high and low electrical pulse signal, which is simultaneously transmitted through the circuit lines to the pose sensor and the two spectrometers, commanding them to begin recording data at the same instant. This hardware-level instruction is extremely fast, with errors controllable within microseconds, ensuring that all data are strictly aligned on the timeline, which is the foundation for subsequent data fusion processing.

[0037] Based on the embodiments provided in this application, the pose sensing unit is used to collect attitude data of the UAV platform. A synchronization control unit is also included, responsible for generating and sending first control commands to the first micro-motion mechanism, and simultaneously triggering data acquisition and spectral measurement by the pose sensing unit, the water radiance probe, and the downlink irradiance probe. In water surface spectral observation scenarios, this ensures that the pose data and spectral data accurately correspond while the first micro-motion mechanism drives the water radiance probe. For example, if minor changes in the UAV's flight attitude are not recorded synchronously with spectral acquisition, it will be difficult to accurately trace the observation angle corresponding to the spectral data. The synchronization control unit eliminates this time difference, providing the data processing unit with timely and reliable basic data to accurately determine the center of the water radiance probe's field of view, further ensuring the accuracy of subsequent analysis of the water surface optical characteristics.

[0038] Furthermore, the downlink irradiance probe is connected to a second micro-motion mechanism;

[0039] The synchronization control unit is also used to generate a second control command and send the second control command to the second micro-motion mechanism to drive the receiving field of view of the downlink irradiance probe to perform a second periodic movement along the same preset trajectory as the first micro-motion mechanism;

[0040] The synchronization control unit is configured to synchronously trigger the first micro-motion mechanism and the second micro-motion mechanism;

[0041] The preset trajectory includes one of the following: a circular trajectory, a Lissajous trajectory, or a linear scan trajectory.

[0042] It's important to note that the scanning paths of the two probes (one scanning the water surface and the other the sky) are exactly the same, essentially drawing circles of the same size. However, their physical locations are separate—one above the water and the other in the sky—preventing them from colliding. To ensure perfect synchronization between the two mechanisms, they are directed by the same central control unit (synchronization control unit). This unit uses the same internal clock source to generate two sets of control commands, ensuring that these commands remain highly synchronized from the moment of initiation and throughout the entire movement, achieving simultaneous start and stop, and uniform action.

[0043] Based on the embodiments provided in this application, a synchronized and rhythmic motion observation mode is achieved between the downlink irradiance probe and the water body irradiance probe, overcoming the limitations of traditional downlink irradiance probes, which are often fixed and have a single observation angle. In actual water body observations, the spectral irradiance incident on the water surface varies spatially due to factors such as changes in solar altitude angle and cloud movement. Fixed downlink irradiance probes struggle to capture these differences comprehensively. In this embodiment, the downlink irradiance probe moves along a specific trajectory with the second micro-motion mechanism, enabling it to synchronously cover a wider observation area with the water body irradiance probe. Furthermore, the identical trajectory design ensures that the observation angles of both probes correspond at the same time. For example, when both move along a circular trajectory, they can simultaneously acquire the water surface spectral radiation signal and incident irradiance at different azimuth angles, avoiding the mismatch between irradiance and irradiance observation areas caused by different observation trajectories. Meanwhile, the selection of multiple preset trajectories can adapt to different observation needs. For example, the linear scanning trajectory is suitable for observation of narrow waters, while the Lissajous trajectory can achieve high-density grid coverage, which improves the system's adaptability to complex water scenes and reflects the innovative optimization of observation dimensions and scene adaptability.

[0044] Furthermore, the data processing unit is also used to: determine the orientation of the second field of view center of the downlink irradiance probe at each sampling moment based on the pose data of the synchronously acquired UAV platform and the second control command; and use the second control command as the second reference input signal to perform signal analysis on the synchronously acquired spectral irradiance incident on the water surface, and separate the irradiance signal component modulated by the second periodic motion.

[0045] Based on the first field of view center direction and spectral signal components at each sampling time, the spatial distribution information of the surface optical properties of the water body in the measurement area is reconstructed, including: based on the first field of view center direction and the second field of view center direction, spectral signal components and illuminance signal components at each sampling time, the spatial distribution information of remote sensing reflectance in the measurement area is reconstructed.

[0046] It's important to explain that the two probes move independently, and their data fusion relies on accurate timestamps and coordinate transformations. The spatiotemporal registration process includes: a) each data point carries a time stamp accurate to the microsecond level, generated by the synchronization system; b) the data processing unit uses the time stamp to establish a pairing relationship between the water surface irradiance signal and the sky irradiance signal collected at the same time; c) using pose data, through a series of coordinate transformation calculations, the spatial orientations corresponding to the two paired signals (one the geographic coordinates of the water surface, the other the angle of the sky) are unified to the same geographic reference frame. Finally, the system calculates the remote sensing reflectance of each point and assigns it accurate spatial location information.

[0047] It should be noted that during data processing, before generating the final regular distribution map, the system already possesses all the necessary spatial distribution information, including the geographic coordinates, spectral signal components, and illuminance signal components of each sampling point. This information itself constitutes a spatial distribution form of remote sensing reflectance.

[0048] Based on the embodiments provided in this application, the motion observation data of downlink irradiance and the motion observation data of water body radiance are deeply fused, rather than simply using fixed irradiance data for calculation. Specifically, remote sensing reflectance is a key parameter of the surface optical characteristics of water bodies, and its calculation requires accurate water body radiance and corresponding downlink irradiance data. In traditional methods, if the downlink irradiance data and water body radiance data are not accurately matched in terms of observation angle and time, it will lead to a large error in the calculation of remote sensing reflectance. In this embodiment, the data processing unit first determines the orientation of the field of view center of the two probes respectively, ensuring that each set of spectral signal components has a corresponding illuminance signal component and clear observation position information, and then calculates the remote sensing reflectance based on the two, fundamentally solving the problem of mismatch between radiance and irradiance data. For example, when the two probes move synchronously along the Lissajous trajectory, a large number of spectral and illuminance data pairs at the same position and angle can be acquired. The reconstructed spatial distribution information of remote sensing reflectance can more realistically reflect the differences in optical characteristics of different areas of the water body, providing a more accurate basis for water quality parameter inversion.

[0049] Furthermore, based on the pose data of the synchronously acquired UAV platform and the first control command, the orientation of the first field of view center of the water radiance probe at each sampling moment is determined; using the first control command as the first reference input signal, signal analysis is performed on the synchronously acquired water surface spectral radiation signal to separate the spectral signal component modulated by the first periodic motion, including:

[0050] Based on the platform's spatial position, height, and attitude angle in the pose data, and combined with the instantaneous motion offset corresponding to the first control command, the geographic coordinates pointing to the center of the first field of view at each sampling moment are calculated through coordinate transformation.

[0051] An adaptive noise cancellation algorithm is adopted, with the first control command signal as the first reference input signal and the water surface spectral radiation signal as the main input. The algorithm suppresses noise components that are not related to the first reference input signal from the main input through filtering calculation, and extracts the spectral signal components that change synchronously with the first periodic motion.

[0052] In one embodiment, the system synchronously records the UAV's spatial position (provided by the navigation satellite system module), relative altitude (provided by the barometer or laser rangefinder), and attitude angles (i.e., pitch, roll, and yaw angles) at each sampling moment.

[0053] At the same time, the system also records the precise control commands (e.g., voltage signals) of the first micro-motion mechanism at each moment, which correspond to the instantaneous displacement offset generated by the fiber optic end of the drive probe.

[0054] The data processing unit performs a coordinate transformation calculation involving multi-source data fusion. First, based on the UAV's attitude angles, it calculates the transformation relationship between the UAV's body coordinate system and the geodetic coordinate system. Then, it superimposes the instantaneous offset of the micro-motion mechanism in the body coordinate system onto the UAV's position and orientation. Finally, through a series of coordinate rotation and translation transformations, it calculates the precise latitude and longitude coordinates of the center point of the small patch of water surface "seen" by the water radiance probe at each sampling instant. In this way, each spectral reading has a clear geographical location label.

[0055] The raw water surface spectral radiation signal measured by the water radiance probe is a mixed signal, containing both effective signal variations caused by the periodic scanning of the micro-motion mechanism and random noise caused by water surface ripples, rapid cloud drift, and instrument noise. The system employs an adaptive noise cancellation algorithm to process this signal. This algorithm has two inputs: a "main input" is the mixed spectral signal measured by the probe; the other "reference input" is the first control command signal itself (this command signal precisely drives the periodic movement of the micro-motion mechanism).

[0056] The adaptive noise cancellation algorithm works by adjusting the parameters of an internal digital filter so that it predicts the components of the main input related to periodic motion based on a reference input. This predicted noise is then subtracted from the main input, leaving the purified effective spectral signal component primarily modulated by the scanning motion. An analogy is that at a noisy cocktail party, using the characteristics of the speaker's voice as a reference, a smart headset filters out background noise, allowing you to hear only the person you want to hear.

[0057] Regarding the parameters of the adaptive noise cancellation algorithm, such as the number of filter taps, the number of taps determines the algorithm's "resolution precision" of the signal. Considering the computing power of the UAV platform and the characteristics of the water spectral signal, 16 to 32 taps are selected. If the observed water body is clean (with less noise), 16 taps are sufficient to meet the requirements, effectively separating noise without consuming too much computing power. If the water body is complexly polluted (e.g., containing a large amount of suspended solids, algae, and many noise components), then 32 taps are selected to more meticulously identify and filter noise in different frequency bands, avoiding the accidental deletion of effective spectral signals.

[0058] For example, the step size factor controls the speed at which the algorithm "learns and adapts to signal changes," employing a variable step size design that is "fast at first and then slow." Initially, the step size is set to 0.01, allowing the algorithm to quickly capture the overall pattern of signal changes. After the algorithm has run for a period of time (e.g., 100 sampling periods) and the signal separation effect stabilizes, the step size automatically decreases to 0.001. This prevents the algorithm from fluctuating repeatedly around the optimal result due to an excessively large step size, ensuring a more stable and effective separated signal.

[0059] In some embodiments, before the signal enters the algorithm, normalization preprocessing is required. The specific steps are as follows: For the water surface spectral radiation signal (main input) and the first control command signal (reference input) collected over a period of time (e.g., one motion cycle), find the maximum and minimum values ​​respectively; adjust each signal value to the range of 0 to 1, using the rule: subtract the minimum value of the signal from the original signal value, and then divide by the difference between the maximum and minimum values; if there are obvious abnormal extreme values ​​in the signal (e.g., sudden signal spikes, possibly caused by electromagnetic interference), remove these extreme values ​​first (the criterion is: values ​​exceeding three times the average value of the signal are considered extreme values), and then perform normalization. This processing can eliminate unit differences between different signals (e.g., the unit of spectral signal is light intensity, and the unit of control command is voltage), preventing one type of signal from "masking" another type of signal due to its large value range, ensuring fair processing of the two types of input signals by the algorithm, and improving the stability of the algorithm.

[0060] Based on the embodiments provided in this application, the ingenuity lies in combining macroscopic pose data with microscopic motion offsets to achieve accurate positioning. Simultaneously, algorithms are used to specifically eliminate noise. In UAV hovering observation scenarios, the spatial position, altitude, and attitude angles (such as roll and pitch angles) of the UAV platform directly affect the probe's observation angle. Pose data alone cannot accurately reflect the microscopic displacement generated by the probe driven by the first micro-motion mechanism. However, by combining the instantaneous motion offset of the first control command, the macroscopic attitude of the platform and the microscopic motion of the probe can be fused through coordinate transformation, accurately calculating the geographic coordinates pointing to the center of the field of view, ensuring that every spectral data point can be traced back to a specific observation location. Furthermore, water surface spectral radiation signals are susceptible to noise interference from atmospheric scattering and the UAV's own shadow. The adaptive noise cancellation algorithm, using the first control command as a reference, can accurately identify and suppress noise unrelated to the probe's periodic motion, retaining only the effective spectral signal components synchronized with the motion. This avoids noise interference with subsequent data reconstruction and significantly improves the purity of the spectral data.

[0061] Furthermore, based on the pose data of the synchronously acquired UAV platform and the second control command, the orientation of the second field of view center of the downlink irradiance probe at each sampling moment is determined; using the second control command as the second reference input signal, signal analysis is performed on the synchronously acquired spectral irradiance incident on the water surface to separate the irradiance signal component modulated by the second periodic motion, including:

[0062] Based on the platform's spatial position, height, and attitude angle in the pose data, and combined with the instantaneous motion offset corresponding to the second control command, the zenith angle and azimuth angle pointing to the center of the second field of view at each sampling moment are calculated.

[0063] An adaptive noise cancellation algorithm is adopted, using the second control command signal as the second reference input signal and the spectral irradiance incident on the water surface as the main input. Through filtering calculation, noise components unrelated to the second reference input signal are suppressed from the main input, and the irradiance signal component that changes synchronously with the second periodic motion is extracted.

[0064] It should be noted that the process of calculating the zenith angle and azimuth angle pointed to by the center of the second field of view at each sampling moment is similar to the approach of calculating geographic coordinates, but the purpose is different. The downlink irradiance probe points to the sky and measures downlink illumination from the entire hemisphere, so it is not necessary to calculate the ground coordinates of the illumination point, but rather to know the specific direction the probe is pointing towards the sky at each instant.

[0065] Similarly, the system synchronously records the attitude angle of the UAV platform and the instantaneous control commands (i.e., drive offsets) of the second micro-motion mechanism.

[0066] First, the probe's default initial pointing (usually vertically upwards, i.e., zenith angle of 0°) is used as a reference. Then, based on the UAV's current pitch and roll angles, the change in probe pointing caused by platform tilt is calculated. Finally, the drive offset of the micro-motion mechanism in the horizontal and vertical directions is superimposed. Through calculation, the specific angle pointed to by the central axis of the downlink irradiance probe's receiving field of view at each sampling moment is obtained, namely the angle relative to the zenith (zenith angle) and the azimuth relative to true north. This ensures that the system knows under what geometric observation conditions each irradiance reading was obtained.

[0067] The raw spectral irradiance signal measured by the downlink irradiance probe is also subject to interference, such as signal fluctuations caused by slight shaking of the drone or changes in cloud cover. Its signal analysis process also employs an adaptive noise cancellation algorithm, which will not be elaborated upon here. The only difference lies in the input signal: the main input here is the spectral irradiance incident on the water surface, while the reference input is replaced by the second control command signal.

[0068] The adaptive noise cancellation algorithm identifies the effective illuminance signal within the illuminance signal that changes strictly in sync with the periodic scanning motion of the probe, using a reference input (drive command). It then suppresses interference components unrelated to the scanning motion to the greatest extent possible, thereby extracting high-quality illuminance signal components. This ensures the accuracy of downlink irradiance data in subsequent calculations of remotely sensed reflectance.

[0069] The selection of parameters for the adaptive noise cancellation algorithm can refer to the process of "extracting the spectral signal components that change synchronously with the first periodic motion", which will not be elaborated upon in this application.

[0070] It should be explained that in this system, the first periodic motion (for the water radiance probe) and the second periodic motion (for the downlink radiance probe) are the core of realizing "motion-modulated observation," and their design logic and actual operating characteristics are as follows:

[0071] The essential purpose of the motion is not to move the probe randomly, but to make the probe's receiving field of view move periodically according to a fixed pattern, so as to "mark" the target spectral signal. That is, to make the effective spectral signal change synchronously with the motion pattern, while external noise (such as atmospheric stray light, drone motor interference) usually does not have a fixed periodicity, thus providing "identifiable features" for subsequent separation of effective signal and noise.

[0072] Taking the first periodic motion as an example, the synchronous control unit generates a series of first control commands, which include parameters such as "motion direction, speed, and trajectory type". After receiving the command, the first micro-motion mechanism (such as a piezoelectric ceramic actuator) drives the incident fiber end of the water radiance probe (or the entire probe) to move cyclically along a preset trajectory. For example, when a circular trajectory is selected, the fiber end will continuously perform circular motion around a center point at a fixed speed (such as rotating around the center twice per second), and each complete circle is one motion cycle. If a linear scanning trajectory is selected, the fiber end will reciprocate within a set straight line range (such as moving 2 centimeters to the left and right) at a fixed frequency (such as going back and forth 3 times per second), and each completion of "going + returning" is one cycle.

[0073] The motion cycle is strictly matched with the spectral sampling frequency. For example, if the motion cycle is 0.5 seconds (2 cycles per second), the spectral sampling frequency is set to 20 times per second, ensuring that 10 sampling points are collected within one motion cycle. Each sampling point corresponds to a different position of the probe on the trajectory (such as 10 different angles on a circular trajectory). This collaborative design allows each spectral data point to correspond to the probe's motion position, providing a complete data chain for subsequent determination of the field of view center direction and separation of the modulation signal.

[0074] The motion logic of the second periodic motion is basically the same as that of the first periodic motion, including the same trajectory type, period length, and motion speed. Both micro-motion mechanisms are triggered simultaneously by the synchronous control unit to ensure that the water radiance probe and the downward irradiance probe are at their corresponding positions on the trajectory at the same time. For example, when both move along the Lissajous trajectory, at the sampling time, the former is located at the "upper right peak point" of its trajectory, and the latter also synchronously reaches the "upper right peak point" of its own trajectory, avoiding the problem of "mismatch between radiance and irradiance observation angles" caused by asynchronous motion.

[0075] Based on the embodiments provided in this application, for the observation target (spectral irradiance incident on the water surface) of the downlink irradiance probe, the field of view center direction is quantized into zenith angle and azimuth angle. Simultaneously, an adaptive noise cancellation algorithm is continued to ensure data quality, forming a "symmetrical and compatible" logic with the data processing of the water body irradiance probe. For spectral irradiance incident on the water surface, zenith angle and azimuth angle are key parameters describing its incident direction, directly affecting the magnitude and distribution of irradiance. Compared to geographic coordinates, zenith angle and azimuth angle better reflect the physical characteristics of downlink irradiance, more accurately reflecting the direction information of incident radiation, and providing more suitable direction parameters for subsequent remote sensing reflectance calculations. Meanwhile, the application of the adaptive noise cancellation algorithm can eliminate noise interference from atmospheric turbulence and surrounding environmental reflections on downlink irradiance measurement. For example, if the irradiance fluctuation caused by the brief obstruction of clouds is not synchronized with the movement of the second micro-motion mechanism, it will be identified as irrelevant noise and suppressed by the algorithm, ensuring that the extracted irradiance signal component is only related to the periodic movement of the probe. This forms a dual-probe, same-algorithm, and adapted mode with the water irradiance data processing, further ensuring the consistency and reliability of the two types of core data.

[0076] Furthermore, based on the direction of the first field of view center and the direction of the second field of view center at each sampling time, as well as the spectral signal components and illuminance signal components, the spatial distribution information of remote sensing reflectance within the measurement area is reconstructed, including:

[0077] Based on the set of geographic coordinates of the center of the first field of view and the center of the second field of view at all sampling times, a spatial grid of the measurement area is constructed.

[0078] Traverse each pixel in the spatial grid, execute an adaptive weighted interpolation algorithm, and generate a regularized spatial distribution map of remote sensing reflectance.

[0079] It should be noted that a regularized spatial distribution map is a specific, advanced, and deeply processed representation of spatial distribution information. It is a regular grid of data, easily interpreted by humans and analyzed by computers, transformed from irregularly sampled spatial distribution information through algorithms. For example, Figure 2 As shown, for each cell to be calculated in the spatial grid, an adaptive weighted interpolation algorithm is performed, including:

[0080] S201, Locate several sampling points adjacent to the pixel to be determined;

[0081] S202, Based on the attitude angle data corresponding to each sampling time in the pose data, calculate the rate of change of attitude angle at each sampling time as a stability weight;

[0082] S203, calculate the comprehensive weight of each sampling point; where the comprehensive weight is negatively correlated with the spatial distance from the sampling point to the pixel to be determined, and positively correlated with the stability weight;

[0083] S204. Using comprehensive weights, the spectral signal components and illuminance signal components of neighboring sampling points are weighted and averaged to obtain the estimated spectral radiance value and estimated downlink illuminance value at the pixel to be determined.

[0084] S205, calculate the ratio of the estimated spectral radiance value to the estimated downlink irradiance value, and use it as the remote sensing reflectance value of the pixel to be determined.

[0085] It's important to note that the adaptive weighted interpolation algorithm directly addresses the problem of irregular data point distribution caused by dynamic scanning. Instead of simply performing ratio calculations on the original data points, it generates a clear, well-organized spatial distribution map that facilitates subsequent analysis and application by reconstructing a regular grid. This in itself represents a significant technological advancement.

[0086] Traditional inverse distance weighted (IDW) interpolation only considers spatial distance. The adaptive weighted interpolation algorithm proposed in this embodiment introduces a second key weighting factor—an "attitude stability index" (such as a function of the attitude angular velocity and acceleration at that moment). This means that a data point collected at a slightly greater distance but under very stable conditions by the UAV may have a higher weight than a data point collected at a closer distance but under severe UAV shaking. This cleverly integrates the quality information inherent in the data into the spectral reconstruction algorithm, improving the accuracy and reliability of the final product.

[0087] In one specific implementation, step one: system initialization and synchronous scanning start: the synchronous control unit generates a first control command and a second control command with specific waveforms and frequencies, and sends them synchronously to the first micro-motion mechanism and the second micro-motion mechanism; at the same time, the synchronous control unit generates a hardware trigger pulse to synchronously start the data acquisition process of the pose sensing unit, the water radiance probe and the downlink irradiance probe.

[0088] Specifically, the UAV platform flies to the target measurement area and enters a hovering state. The signal generator built into the synchronization control unit is configured to generate a first control command C1(t) and a second control command C2(t). In this embodiment, C1(t) and C2(t) are two sets of periodic voltage signals whose amplitude, frequency, and phase can be independently set but remain synchronized. Their functional expressions are defined by the following formula:

[0089]

[0090] Where t is a time variable, representing the elapsed time since the start of the motion, in seconds (s); x1(t) and y1(t) represent the offset of the fiber optic end of the water radiance probe driven by the first micro-motion mechanism in its local plane coordinate system at time t, in micrometers (μm); Ax A y Let A represent the driving amplitude, and let Y represent the maximum displacement in the X and Y axes of the local coordinate system, respectively. In this example, let A be the driving amplitude. x =200μm,A y =150μm; f x ,f y The driving frequency is represented by f, which indicates the motion frequency in the X and Y axes, respectively, in Hertz (Hz). To achieve a non-repetitive scanning trajectory, f is taken as f. x =5Hz,f y =7Hz, and the ratio of the two is an irreducible fraction; This represents the initial phase angle, expressed in radians (rad). In this example, we take...

[0091] The synchronous control unit sends the first control command C1(t) and the second control command C2(t) to the drivers of the first micro-motion mechanism and the second micro-motion mechanism, respectively, through its digital output interface.

[0092] Simultaneously, the synchronization control unit generates a high-level hardware trigger pulse (TTL signal), which is sent to the pose sensing unit, the spectrometer built into the water radiance probe, and the spectrometer built into the downlink irradiance probe, instructing all data acquisition units to operate at the same sampling rate f. s (in this example, f) s (500Hz) Start synchronous data acquisition.

[0093] Step 2: Field of view motion and synchronous data acquisition: The first and second micro-motion mechanisms drive the receiving fields of view of the water radiance probe and the downlink irradiance probe respectively, and perform periodic motion according to the preset Lissajous trajectory; the pose sensing unit and the two spectrometers synchronously record the pose data and raw spectral data of the platform.

[0094] Specifically, the first micro-motion mechanism drives the incident fiber end of the water radiance probe to move in a plane according to the received C1(t), so that the trajectory of the projection point L1(t) of its instantaneous field of view FOV1 on the water surface is jointly determined by its displacement components x1(t) and y1(t) in two orthogonal directions; the displacement components x1(t) and y1(t) are defined by the formula in step one, and the trajectory is a Lissajous trajectory.

[0095] The second micro-motion mechanism drives the incident fiber end of the downlink irradiance probe to move in complete synchronization according to the received C2(t), thereby adjusting its field of view (FOV). sky The motion pointing to L2(t) on the sky hemisphere is controlled by displacement components that are exactly the same as the mathematical patterns of x1(t) and y1(t), thus ensuring the spatiotemporal consistency of the scanning motion of the two.

[0096] During this process, the pose sensing unit operates at a frequency f s Continuously collect and record the platform's three-dimensional position (P) t ), height (H) t ), roll angle t ), pitch angle t ) and yaw angle t The spectrometer built into the water radiance probe synchronously records the water surface spectral radiation signal S_u(t,λ), and the spectrometer built into the downlink irradiance probe synchronously records the downlink spectral irradiance signal S_d(t,λ). λ is the wavelength.

[0097] Step 3: Field of view center pointing calculation and signal analysis: For each sampling moment, the data processing unit calculates the precise geographic pointing of the field of view center based on the synchronously acquired pose data and control commands; and uses an adaptive noise cancellation algorithm to separate the effective signal component modulated by the controlled periodic motion from the original spectral signal.

[0098] Specifically, for each sampling time t_k, the data processing unit performs the following operations: Field of view pointing calculation: based on the platform pose (P) at time t_k... t_k H t_k Roll t_k Pitch t_k ,Yaw t_k The instantaneous offset (x1(t_k), y1(t_k)) calculated from C1(t_k) is used to calculate the geographic coordinates G1(t_k) on the water surface through a coordinate system transformation chain (probe local coordinate system -> UAV body coordinate system -> UAV NED coordinate system -> geodetic coordinate system). t_k Lon t_k Similarly, calculate the FOV. sky Zenith angle θ_sky(t_k) and azimuth angle φ_sky(t_k).

[0099] Signal Analysis: The Least Mean Square (LMS) adaptive filtering algorithm is employed. The waveform data of C1(t_k) is used as the reference input, and S_u(t_k,λ) is used as the main input. The filter output is the separated spectral signal component L_u(t_k,λ), after removing random noise unrelated to probe motion. Using the exact same algorithm, with C2(t_k) as the reference input, S_d(t_k,λ) is processed to obtain the pure illuminance signal component E_d(t_k,λ).

[0100] Step 4: Remote sensing reflectance distribution reconstruction and spatial interpolation: Based on the geographic coordinates of the field of view center of all sampling points and the resolved signal components, the data processing unit uses an adaptive weighted interpolation algorithm that integrates spatial proximity and data quality reliability to reconstruct a two-dimensional distribution map of remote sensing reflectance on a regular spatial grid.

[0101] Specifically, a regular two-dimensional spatial grid is defined that covers the geographic coordinates G1(t_k) of all sampling points. The grid resolution can be set according to requirements (e.g., 0.1m x 0.1m).

[0102] For each pixel j to be calculated in the grid, with center coordinates G_j, the interpolation algorithm defined by the following formula is executed:

[0103] Find neighboring points: Within a circular area centered at G_j with a radius of R (e.g., R = 0.5m), find all neighboring sampling points k.

[0104] Calculate the total weight: For each neighboring sampling point k, the total weight W_total is used for calculation. k Determined by the following formula:

[0105] W_total k =W_d k ×W_s k

[0106] Among them, W_d k The distance weights are calculated using the Euclidean distance d from sampling point k to the pixel j to be determined. kj Negative correlation. This example uses inverse squared distance weights: W_d k =1 / (d kj ) 2 ;d kj The unit is meters (m); W_s k This is the attitude stability weight, positively correlated with the platform attitude stability at sampling time t_k. In this example, it is defined as the reciprocal of the rate of change of the attitude angle at that time: W_s k =α / (ω) k +∈); ω k Let be the rate of change of attitude angle at time t_k, in degrees per second (° / s). Its calculation method is as follows: It is the sum of the absolute values ​​of the instantaneous rates of change of both roll and pitch dimensions.

[0107] in, The roll and pitch angles of the UAV platform were measured at the previous sampling time t_k-1. These two parameters are related to Roll. t_k Pitch t_kTogether, they are used to calculate the change in attitude angle between two adjacent frames; Δt is the sampling interval for data acquisition, in seconds (s). Its value is the sampling frequency f. s The reciprocal of f, i.e., Δt = 1 / f s In this embodiment, f s =500Hz, therefore Δt = 0.002s. This parameter ensures the accuracy of the rate of change calculation in the time dimension.

[0108] ω k The larger the value, the more violent the drone's shaking, the worse the data quality at that point, and the lower the weight should be. ∈ is a very small constant (e.g., ε = 0.01° / s), its unit is the same as ω. k Keep it consistent; α is the normalization coefficient, with units of (° / s), for example, α can be taken as 1° / s.

[0109] Calculate the pixel remote sensing reflectance value: using the calculated comprehensive weight W_total k The estimated spectral radiance value at pixel j is obtained by weighting and averaging L_u(t_k,λ) and E_d(t_k,λ) of the neighboring sampling points respectively.<L_u(λ)> j and estimated downlink irradiance value<E_d(λ)> j.

[0110] According to the definition of remote sensing reflectance, the remote sensing reflectance value of pixel j is calculated as: R_rs(λ)j=<L_u(λ)> j / <E_d(λ)> j.

[0111] By traversing all grid cells and performing the above calculations, a regularized, high-quality spatial distribution map of remote sensing reflectance is finally generated. This map effectively suppresses noise caused by platform jitter and clearly reflects the spatial heterogeneity of water surface optical properties within the measurement area.

[0112] Based on the embodiments provided in this application, the traditional reconstruction methods of "uniform interpolation" or "simple averaging" are abandoned. Instead, a dual weighting logic of stability weight and spatial distance weight is introduced, making the interpolation results more consistent with the reliability and spatial distribution characteristics of actual observation data. In water body observation, the reliability of observation data from different sampling points varies. For example, the reliability of data collected from sampling points when the attitude of the UAV fluctuates greatly is lower than that of data collected when the attitude is stable. In this embodiment, the stability weight calculated based on the attitude angle change rate can determine the observation stability of each sampling point. The smaller the attitude angle change rate (the higher the stability), the greater the weight, ensuring that reliable data plays a greater role in interpolation. At the same time, the comprehensive weight is negatively correlated with spatial distance, avoiding excessive interference of distant sampling points to target pixels. This adaptive weighted interpolation algorithm can effectively handle the problems of uneven distribution of sampling points and low reliability of some sampling point data. For example, in sparse spectral data areas, pixel values ​​can be accurately estimated through nearby high-reliability sampling points. The generated regularized remote sensing reflectance spatial distribution map can more clearly and accurately present the spatial differences in the optical properties of water bodies, providing high-quality visualization data for subsequent identification of water quality anomalies.

[0113] Furthermore, the data processing unit is also configured to: calculate the spectral heterogeneity index within the measurement area based on the spatial distribution information of remotely sensed reflectance; and calculate the illuminance stability index within the measurement area based on the time series data of illuminance signal components.

[0114] It's important to note that the spectral heterogeneity index measures spatial properties. It addresses the question: "How significant are the differences in optical properties across different locations within the area I scanned?" For example, is it a homogeneous body of water or is there algal blooms, pollution plumes, etc.? This index can only be calculated after the spatial distribution information has been reconstructed, as it is essentially a mathematical representation of this spatial distribution map (such as variance, gradient, information entropy, and other statistics). It focuses on the "non-uniformity of spatial distribution."

[0115] The illuminance stability index measures a time-related attribute. It addresses the question: "Were the solar illumination conditions stable or unstable during the last few seconds of the scan?" For example, were there any clouds causing drastic fluctuations in illumination? This index needs to be extracted from the time series of the illuminance signal components because it analyzes the curve of light intensity changing over time (e.g., variance, range of fluctuation). It focuses on the "volatility of the time series."

[0116] This is analogous to having two quality inspectors on a production line. The first inspector (judging illuminance stability) checks the stability of the power supply voltage (external environmental conditions) during parts processing. If it's unstable, the batch of parts is marked as potentially problematic. The second inspector (judging spectral heterogeneity) checks the consistency of the finished product's appearance (internal spatial characteristics) after final assembly. Significant differences indicate a need for further investigation.

[0117] Based on this embodiment, instead of making decisions blindly, the fundamental premise of data acquisition (whether the lighting is stable) is first checked. Only under the premise of stable lighting can decisions (hovering or cruising) made based on spatial heterogeneity be reliable and scientific. If the lighting is unstable, then any spatial distribution information and heterogeneity indicators calculated under such circumstances will be distorted, and any decisions made based on this will be wrong.

[0118] Among them, the spectral heterogeneity index is used to measure the uniformity of the optical properties of water bodies within the observation area. Based on the reconstructed spatial distribution map of remote sensing reflectance (each point represents the remote sensing reflectance of a small area), there are two calculation methods:

[0119] Standard deviation method: First, calculate the average remote sensing reflectance of all effective areas; calculate the difference between the reflectance value of each area and the average value, and then calculate the overall dispersion of these differences. The greater the dispersion, the more obvious the difference in reflectance between different areas, and the more uneven the optical properties of the water body (for example, some areas with high reflectance may be algal bloom areas; some areas with low reflectance may be clean water areas).

[0120] Average gradient method: For each region (except the edge region), calculate the absolute value of the reflectance difference between it and the adjacent regions to the right and below. Add these two differences to obtain the "local gradient value" of the region. Calculate the average value of the local gradient values ​​of all regions. The larger the average value, the more drastic the reflectance change between adjacent regions and the more prominent the spatial difference in the optical properties of the water body (for example, the gradient value will be significantly higher in the edge region of a pollution plume).

[0121] The illuminance stability index is used to determine the stability of solar illumination during the observation period. Based on the time series data of the downlink illuminance signal component (a series of illuminance values ​​arranged in chronological order), it is calculated in two ways:

[0122] Variance method: Select irradiance values ​​from 100 consecutive sampling points (corresponding to a duration of 10 seconds, which can reflect recent changes in light intensity without causing lag due to excessive time) and calculate their average value; calculate the difference between each irradiance value and the average value, and then statistically analyze the overall fluctuation of these differences—the smaller the fluctuation, the more stable the light intensity, and vice versa (for example, when clouds pass by and cause the light intensity to fluctuate, the fluctuation will be larger).

[0123] The coefficient of variation method: First, calculate the average value and overall dispersion (i.e., standard deviation) of the irradiance value over a period of time; divide the dispersion by the average value, and then multiply by 100% to get a percentage value. The larger this value is, the more obvious the relative fluctuation of the light is, and it is not affected by the absolute value of the light (for example, when there is strong light at noon and weak light at dusk, as long as the fluctuation ratio is the same, the coefficient of variation will be the same).

[0124] Among them, the stability threshold is determined by analyzing 500 sets of observation data under different weather conditions and at different times. It was found that when the relative fluctuation percentage (coefficient of variation) of illumination is between 5% and 10%, the data reliability meets the requirements. If the fluctuation percentage exceeds 10%, it indicates that the illumination changes too drastically, and the collected irradiance data will have a large error and cannot be used for calculation. If it is below 5%, the illumination is very stable and the data reliability is high. In some embodiments, users can input a value from 0 to 20% through the software interface to customize the stability threshold. For example, in cloudy weather, the threshold can be relaxed to 12% to avoid frequent invalidation of data.

[0125] Heterogeneity threshold: Statistical analysis of 30 different water bodies (lakes, rivers, coastal areas, etc.) revealed that the reflectance dispersion (standard deviation) of clean water bodies is between 0.02 and 0.05, while that of polluted water bodies can be relaxed to between 0.05 and 0.08. If the dispersion exceeds 0.05, it indicates that there are significant optical differences in the water body, requiring careful observation while hovering. If it is below 0.02, the water body is very uniform and can be cruised quickly. In some embodiments, a "water body type" selection menu (clean, lightly polluted, heavily polluted) is provided. After selection, the system will automatically load the corresponding suggested threshold, which users can adjust slightly.

[0126] The UAV-based suspended water body apparent spectrum multi-probe synchronous observation system also includes a decision control unit. The decision control unit is used to receive spectral heterogeneity index and illuminance stability index, and generate flight control commands based on a preset strategy rule set. The flight control commands are used to control the flight status of the UAV platform.

[0127] The strategy rule set includes:

[0128] If the illuminance stability index exceeds the stability threshold and the spectral heterogeneity index exceeds the heterogeneity threshold, a hovering command is generated to control the drone platform to hover.

[0129] If the illuminance stability index exceeds the stability threshold and the spectral heterogeneity index does not exceed the heterogeneity threshold, a cruise command is generated to control the UAV platform to cruise along the predetermined route.

[0130] If the illumination stability index does not exceed the stability threshold, a maintenance command is generated to mark the current data as invalid and maintain the current flight state of the UAV platform.

[0131] The methods for marking current data as invalid include: adding an invalid flag to the spectral and illuminance data collected synchronously within the current time period, and setting the quality control status of all data within that time period to invalid.

[0132] Specifically, marking the current data as invalid is not a simple Boolean check, but a structured operation integrated into data stream processing. Its specific technical means typically include one or more of the following steps:

[0133] Data frame level flag (invalid flag):

[0134] The raw data collected by the system (such as spectral radiation signals, irradiance signals, and pose data) is typically organized into data frames or data packets arranged in chronological order. Each data packet contains a header with information such as a timestamp and device ID. When the decision control unit determines that the current irradiance is unstable, it sends a signal to the data processing unit. The data processing unit then writes a specific "invalid flag" (e.g., setting a byte from 0x00 to 0xFF) into the header of all data packets generated within that determination period. This is equivalent to tagging this batch of data with an "invalid" electronic label.

[0135] Metadata records and associations (logs and reason codes):

[0136] The system maintains a system log or quality control metadata file, which is stored separately from or embedded within the original data file. When a "mark as invalid" operation is triggered, the system automatically generates a log record. This record includes at least: Invalid time range: precise start and end timestamps (UTC time); Invalid reason code: a preset code, such as CODE_102: Illuminance stability index is below the threshold. This clearly records the specific reason for the data invalidity; Index value at the time of triggering: recording the illuminance stability index and the set threshold at that time for post-event traceability and analysis. This log record is associated with the original data file via a timestamp. Anyone processing data later can quickly locate and filter out all invalid data segments by querying this log.

[0137] Data storage isolation (optional but advanced method):

[0138] Data is managed within a file system or database. In addition to storing the raw data file (RawData.bin), the system can automatically generate a corresponding quality control file (QC_Index.json). This file is a list that clearly indicates which time periods of data are valid and which are invalid. In subsequent processing, the software first reads this QC file, loading only the data from the valid time periods for calculation, while skipping invalid data, thus achieving automated data cleaning.

[0139] Based on the embodiments provided in this application, the UAV is upgraded from flying along a preset route to dynamically adjusting its flight status based on real-time observation data, breaking the limitation of traditional observation systems where flight and observation are disconnected. In actual water body observation, observation conditions change in real time. For example, a sudden strong gust of wind can cause a decrease in illumination stability, resulting in low reliability of data collected under such circumstances. If data cannot be processed in time, it will cause data redundancy. When spectral heterogeneity exceeds a threshold, it indicates that the optical characteristics of the water body in that area vary greatly, requiring hovering for more detailed observation. The decision control unit can generate targeted control commands by analyzing two indicators calculated in real time: hovering commands enable the system to acquire denser data in key areas, cruise commands ensure efficient coverage of uniform water areas, and invalid data marking commands prevent low-quality data from occupying storage resources. This dynamic adjustment mechanism improves data acquisition efficiency and ensures data quality, reflecting a functional upgrade from "passive execution" to "active decision-making" in the observation system.

[0140] Furthermore, the UAV-based suspended water body apparent spectral multi-probe synchronous observation system also includes:

[0141] The auxiliary UAV platform equipped with a multispectral camera has a data processing unit configured to process the image data transmitted back by the auxiliary UAV platform to identify spectral anomaly areas and generate navigation commands to guide the UAV platform to fly to the area for measurement.

[0142] In some embodiments, the specific steps for identifying spectral anomaly regions are as follows:

[0143] Image preprocessing: After the auxiliary UAV transmits multispectral images, the grayscale values ​​captured by the camera are first converted into actual light intensity values ​​(eliminating the differences in the camera's parameters), and then the effects of atmospheric scattering and absorption on the spectrum are eliminated (for example, the atmosphere weakens the blue light band signal, which needs to be corrected back) to obtain an image that can truly reflect the reflection of the water surface; the image is processed using a 3×3 window to remove sporadic noise points (such as small white or black dots in the image) to avoid misjudgment caused by these noises.

[0144] Vegetation index calculation: For each pixel in the image, an index (i.e., normalized vegetation index) is calculated using reflectance data in the red and near-infrared bands to reflect the amount of algae in the water body. The more algae, the higher the index value, and the index value of clean water bodies is usually lower.

[0145] Threshold setting and anomaly labeling: Empirical threshold: Through 20 algal bloom observation experiments, it was found that when the above index value exceeds 0.3, there is a high probability that algal bloom exists in the area (belonging to the spectral anomaly area); Region extraction: All adjacent pixels with an index value exceeding 0.3 are classified into an anomaly area, and the center position (latitude and longitude) of each area is calculated; Artifact removal: If the area of ​​an anomaly area is too small (for example, less than 10 square meters, about 20 pixels in size), it is judged as a false anomaly caused by noise and is not included in the target area.

[0146] Navigation command generation: The data processing unit converts the center position of the abnormal area into navigation coordinates that the UAV can recognize, and generates commands containing "target position, flight altitude (consistent with the auxiliary UAV, such as 50 meters), and flight speed (5 meters / second to ensure stability)", which are sent to the main UAV via wireless signal to guide it to the abnormal area for precise observation.

[0147] Based on the embodiments provided in this application, a "master-slave collaborative" observation mode is adopted, allowing the auxiliary UAV to undertake the task of "large-scale screening" while the master UAV focuses on "precise detailed investigation," thus changing the contradictory situation of "insufficient accuracy in large-scale rapid observation and limited coverage in precise observation" inherent in traditional single UAVs. In large-area water observation, relying solely on the master UAV platform for comprehensive observation would be inefficient due to its slow observation speed; while rapid cruise observation alone would struggle to detect local spectral anomalies (such as small-scale pollutant leak areas). The auxiliary UAV's multispectral camera can quickly acquire image data of a large area of ​​water, identify spectral anomalies through its data processing unit (such as a region with significantly different multispectral reflectance compared to its surroundings), and then guide the master UAV platform precisely to that region, utilizing the master system's multi-probe synchronous observation capability to acquire high-precision apparent spectral data of the water body. This division of labor achieves an organic combination of large-scale rapid screening and small-area precise measurement, improving the observation efficiency of large-area water while ensuring accurate capture of anomalies, demonstrating an innovative design that balances the spatial coverage and observation accuracy of the observation system.

[0148] Furthermore, both the first and second micro-motion mechanisms are piezoelectric ceramic actuators. The first micro-motion mechanism is configured to drive the incident fiber end of the water irradiance probe to move, and the second micro-motion mechanism is configured to drive the incident fiber end of the downlink irradiance probe to move synchronously, so that the receiving fields of view of the two probes can be scanned synchronously along the same preset trajectory.

[0149] In some embodiments, depending on the probe's field-of-view scanning requirements (the preset trajectory is a circular or straight trajectory with a diameter of 5 to 10 cm, and a motion frequency of 1 to 5 times / second), the piezoelectric ceramic actuator must meet the following parameters:

[0150] Displacement range: It can achieve a displacement of ±200 to ±500 micrometers. For example, to complete a circular trajectory with a diameter of 10 centimeters, the actuator needs to move in both the X and Y directions. A displacement of ±350 micrometers in a single direction is sufficient to cover the trajectory range. Response frequency: The resonant frequency is between 1000 and 5000 Hz, which is much higher than the maximum motion frequency of the probe (5 times / second), ensuring that the actuator can respond quickly to control commands without motion lag. Thrust parameters: The thrust needs to reach 5 to 20 Newtons because the fiber optic end (including the fixing fixture) weighs about 5 to 10 grams. Sufficient thrust can ensure that the actuator can still drive the fiber optic end smoothly without jamming when the UAV vibrates slightly. Control accuracy: The displacement accuracy needs to be within 0.1 micrometers to ensure that the error of the probe's field of view center pointing is less than 0.1 degrees, meeting the requirements of spectral observation for angular accuracy.

[0151] The actuator's motion platform (made of metal with a flat surface) has tiny threaded holes for securing the micro-clamp. The micro-clamp is made of titanium alloy (lightweight and high-strength) and has internal holes matching the fiber diameter (typically 0.2 to 0.5 mm). The inner walls of these holes are lined with silicone pads (moderate hardness, 50 Shore A) to secure the fiber without damaging it. The clamp consists of two parts, upper and lower, secured with two small screws. The tightening force is controlled between 0.5 and 1 N·m to ensure the fiber does not move back and forth (movement error less than 5 micrometers). After connection, the motion trajectory of the fiber end is observed using a 100x microscope. The initial X and Y axis positions of the actuator are fine-tuned to align the trajectory center with the optical center of the probe. The actuator is fixed to the drone's shock-absorbing module (made with silicone pads, reducing vibration by more than 80%) via an aluminum alloy bracket, preventing drone flight vibrations from affecting the actuator's motion accuracy.

[0152] Based on the embodiments provided in this application, a piezoelectric ceramic actuator is selected as the micro-motion mechanism, focusing on the critical component at the end of the incident optical fiber rather than driving the entire probe, thus balancing motion accuracy and miniaturization requirements. Piezoelectric ceramic actuators are characterized by fast response speed, high displacement accuracy, and small size, making them highly suitable for UAV platforms with limited space and high motion accuracy requirements. Traditional micro-motion mechanisms driven by motors tend to be bulky and occupy significant space in the UAV's payload, and their motion accuracy is insufficient to meet the minute displacement requirements of spectral observation. In this embodiment, the piezoelectric ceramic actuator directly drives the end of the incident optical fiber, requiring only a small displacement to change the probe's receiving field of view. This achieves precise periodic motion without excessively occupying payload space. Simultaneously, the two actuators synchronously drive the fiber end displacement, ensuring that the receiving fields of view of the two probes are scanned synchronously along the same trajectory, avoiding motion asynchrony caused by differences in mechanisms. This provides hardware-level accuracy assurance for "radiance and irradiance data matching" in subsequent data processing, demonstrating optimized innovation in the selection of the micro-motion mechanism and the driving method.

[0153] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A multi-probe synchronous observation system for apparent spectra of water bodies based on unmanned aerial vehicles (UAVs), characterized in that, include: A water radiance probe, mounted on a drone platform, is used to measure the spectral radiation signal of the water surface; A downlink irradiance probe, mounted on the UAV platform, is used to measure the spectral irradiance incident on the water surface; A first micro-motion mechanism is connected to the water radiance probe and is used to drive the receiving field of view of the water radiance probe to perform a first periodic movement along a preset trajectory based on the received first control command. The data processing unit is used to determine the orientation of the first field of view center of the water radiance probe at each sampling moment based on the pose data of the UAV platform collected synchronously and the first control command. Using the first control command as the first reference input signal, the synchronously acquired water surface spectral radiation signal is analyzed to separate the spectral signal component modulated by the first periodic motion; based on the direction of the first field of view center at each sampling time and the spectral signal component, the spatial distribution information of the surface optical properties of the water body within the measurement area is reconstructed. The UAV-based suspended water body apparent spectral multi-probe synchronous observation system also includes: The pose sensing unit is used to collect the pose data of the UAV platform; A synchronous control unit is used to generate the first control command and send the first control command to the first micro-motion mechanism, and to synchronously trigger the data acquisition of the pose sensing unit and the spectral measurement of the water radiance probe and the downlink irradiance probe. The downlink irradiance probe is connected to a second micro-motion mechanism; The synchronization control unit is also used to generate a second control command and send the second control command to the second micro-motion mechanism to drive the receiving field of view of the downlink irradiance probe to perform a second periodic movement along the same preset trajectory as the first micro-motion mechanism; The synchronization control unit is configured to synchronously trigger the first micro-motion mechanism and the second micro-motion mechanism; The preset trajectory includes one of a circular trajectory, a Lissajous trajectory, or a linear scan trajectory; The data processing unit is further configured to: determine the orientation of the second field of view center of the downlink irradiance probe at each sampling moment based on the pose data of the synchronously acquired UAV platform and the second control command; and use the second control command as a second reference input signal to perform signal analysis on the synchronously acquired spectral irradiance incident on the water surface, and separate the irradiance signal component modulated by the second periodic motion. The method of reconstructing the spatial distribution information of the surface optical properties of water bodies within the measurement area based on the first field of view center direction and the spectral signal component at each sampling time includes: reconstructing the spatial distribution information of remote sensing reflectance within the measurement area based on the first field of view center direction and the second field of view center direction at each sampling time, the spectral signal component and the illuminance signal component. The spatial distribution information of remote sensing reflectance within the measurement area is reconstructed based on the direction of the first field of view center and the direction of the second field of view center at each sampling time, the spectral signal components, and the illuminance signal components, including: Based on the set of geographic coordinates of the center of the first field of view and the center of the second field of view at all sampling times, a spatial grid of the measurement area is constructed. Traverse each pixel in the spatial grid, execute an adaptive weighted interpolation algorithm, and generate a regularized spatial distribution map of remote sensing reflectance. For each pixel to be calculated in the spatial grid, an adaptive weighted interpolation algorithm is performed, including: Find several sampling points adjacent to the pixel to be identified; Based on the attitude angle data corresponding to each sampling time in the pose data, the attitude angle change rate at each sampling time is calculated as a stability weight. Calculate the comprehensive weight of each sampling point; wherein, the comprehensive weight is negatively correlated with the spatial distance from the sampling point to the pixel to be determined, and positively correlated with the stability weight; The spectral signal components and illuminance signal components of neighboring sampling points are weighted and averaged using the comprehensive weights respectively to obtain the estimated spectral radiance value and estimated downlink illuminance value at the pixel to be determined. The ratio of the estimated spectral radiance value to the estimated downlink irradiance value is calculated and used as the remote sensing reflectance value of the pixel to be determined.

2. The UAV-based suspended water body apparent spectrum multi-probe synchronous observation system according to claim 1, characterized in that, The pose data of the UAV platform collected synchronously and the first control command are used to determine the orientation of the first field of view center of the water radiance probe at each sampling time. Using the first control command as the first reference input signal, signal analysis is performed on the synchronously acquired water surface spectral radiation signal to separate the spectral signal component modulated by the first periodic motion, including: Based on the platform's spatial position, height, and attitude angle in the pose data, and combined with the instantaneous motion offset corresponding to the first control command, the geographic coordinates pointing to the center of the first field of view at each sampling time are calculated through coordinate transformation. An adaptive noise cancellation algorithm is adopted, with the first control command signal as the first reference input signal and the water surface spectral radiation signal as the main input. The algorithm suppresses noise components that are not related to the first reference input signal from the main input through filtering calculation, and extracts the spectral signal components that change synchronously with the first periodic motion.

3. The UAV-based suspended water body apparent spectrum multi-probe synchronous observation system according to claim 1, characterized in that, The pose data of the UAV platform collected synchronously and the second control command are used to determine the orientation of the second field of view center of the downlink irradiance probe at each sampling time; Using the second control command as the second reference input signal, signal analysis is performed on the synchronously acquired spectral irradiance incident on the water surface to separate the irradiance signal component modulated by the second periodic motion, including: Based on the platform's spatial position, height, and attitude angle in the pose data, and combined with the instantaneous motion offset corresponding to the second control command, the zenith angle and azimuth angle pointing to the center of the second field of view at each sampling moment are calculated. An adaptive noise cancellation algorithm is adopted, using the second control command signal as the second reference input signal and the spectral irradiance incident on the water surface as the main input. The algorithm suppresses noise components that are not related to the second reference input signal from the main input through filtering calculation, and extracts the irradiance signal component that changes synchronously with the second periodic motion.

4. The UAV-based suspended water body apparent spectrum multi-probe synchronous observation system according to claim 1, characterized in that, The data processing unit is further configured to: calculate the spectral heterogeneity index within the measurement area based on the spatial distribution information of the remotely sensed reflectance; and calculate the illuminance stability index within the measurement area based on the time series data of the illuminance signal components. The UAV-based suspended water body apparent spectrum multi-probe synchronous observation system also includes a decision control unit. The decision control unit is used to receive the spectral heterogeneity index and the illuminance stability index, and generate flight control commands based on a preset strategy rule set. The flight control commands are used to control the flight status of the UAV platform. The policy rule set includes: If the illuminance stability index exceeds the stability threshold and the spectral heterogeneity index exceeds the heterogeneity threshold, a hovering command is generated to control the drone platform to hover. If the illuminance stability index exceeds the stability threshold and the spectral heterogeneity index does not exceed the heterogeneity threshold, a cruise command is generated to control the UAV platform to cruise along a predetermined route. If the illumination stability index does not exceed the stability threshold, a maintenance command is generated to mark the current data as invalid and maintain the current flight state of the UAV platform.

5. The UAV-based suspended water body apparent spectrum multi-probe synchronous observation system according to claim 1, characterized in that, The UAV-based suspended water body apparent spectrum multi-probe synchronous observation system also includes: The auxiliary unmanned aerial vehicle (UAV) platform equipped with a multispectral camera has a data processing unit configured to process the image data transmitted back by the auxiliary UAV platform to identify spectral anomaly regions and generate navigation commands to guide the UAV platform to fly to the region for measurement.

6. The UAV-based suspended water body apparent spectrum multi-probe synchronous observation system according to claim 1, characterized in that, Both the first micro-motion mechanism and the second micro-motion mechanism are piezoelectric ceramic actuators. The first micro-motion mechanism is configured to drive the incident fiber end of the water irradiance probe to move, and the second micro-motion mechanism is configured to drive the incident fiber end of the downlink irradiance probe to move synchronously, so that the receiving fields of view of the two probes can be scanned synchronously along the same preset trajectory.

Images