Single-photon unmanned aerial vehicle airborne high-precision imaging system and method

Abstract

The application provides a kind of single-photon unmanned aerial vehicle airborne high-precision imaging system and method, it is related to airborne imaging system field, including: real-time attitude sensing jitter acquisition unit, high-speed transceiver control signal acquisition and emission unit, spatial scanning two-dimensional scanning unit, real-time data processing motion compensation unit, and overall global timing control main control unit.Signal acquisition unit adopts transceiving integrated coaxial optical structure, introduces dynamic light leakage suppression mechanism.Jitter acquisition unit integrates high-frequency inertial measurement unit (IMU), and through compensation algorithm, its sampling frequency is higher than laser pulse frequency, to high time resolution real-time acquisition body in dynamic environment under three-axis attitude and drift angle data.Motion compensation unit registers IMU coordinate system, laser radar coordinate system and earth-fixed coordinate system, calculates instantaneous attitude at each laser pulse transmission time and calibrates the position information of photon event, solves point cloud motion distortion.

Description

Technical Field

[0001] This invention relates to the field of airborne imaging systems, and more specifically, to a single-photon unmanned aerial vehicle (UAV) airborne high-precision imaging system and method. Background Technology

[0002] 3D imaging technology is of great value in fields such as environmental perception, autonomous driving, surveying and mapping, and intelligent inspection. LiDAR, due to its high resolution and high-precision ranging capabilities, has become the core means of acquiring 3D point clouds. However, when LiDAR is mounted on high-dynamic platforms such as unmanned aerial vehicles (UAVs), the system performance is severely limited.

[0003] During flight, UAVs are subject to disturbances such as rotor vibration, gusts, and airflow turbulence, causing high-frequency and unpredictable changes in their attitude (pitch angle, roll angle, and yaw angle). The different attitudes of the aircraft during each laser pulse transmission and reception cause "distortion" and "deformation" of the point cloud, thus affecting the accuracy of 3D reconstruction and severely impacting the execution of high-precision mapping and navigation tasks.

[0004] In existing technologies, some solutions use high-frequency lidar or SLAM-based post-processing registration methods, but these are costly, computationally complex, or have large latency, failing to meet real-time requirements. Another type of method relies on lidar odometry or inertial navigation systems to provide frame start and end poses and performs interpolation compensation based on the assumption of uniform velocity, but these methods cannot effectively cope with the high-frequency vibrations of UAV platforms.

[0005] Furthermore, single-photon imaging also faces the problem of low signal-to-noise ratio, especially under the influence of background noise during the day and the device's dark count. Existing suppression methods, such as narrowband filtering or increasing laser power, are limited in effectiveness by the constraints of UAV payload and power consumption.

[0006] In summary, the main challenges currently faced by UAV-borne single-photon imaging radar are motion distortion, insufficient sensor synchronization accuracy, and low signal-to-noise ratio. These issues affect the system's high-precision measurement and real-time performance, and there is an urgent need to propose an effective technical solution to address these problems. Summary of the Invention

[0007] The purpose of this invention is to solve the problems of imaging motion distortion caused by severe vibration of the UAV platform and airflow turbulence in the existing technology, as well as the technical problems of insufficient signal-to-noise ratio of single-photon detection caused by light leakage and near-field backscattering inside the optical system, and to provide a single-photon UAV airborne high-precision imaging system and method.

[0008] To achieve the above objectives, the technical solution provided by this invention is as follows:

[0009] A single-photon unmanned aerial vehicle (UAV) airborne high-precision imaging system includes: a main control unit, a jitter acquisition unit, a signal acquisition and transmission unit, a two-dimensional scanning unit, and a motion compensation unit;

[0010] The main control unit is electrically connected to the jitter acquisition unit, the signal acquisition and transmission unit, the two-dimensional scanning unit and the motion compensation unit, respectively, and is used to provide picosecond-level global timing synchronization signals.

[0011] The jitter acquisition unit is communicatively connected to the motion compensation unit and is used to transmit inertial navigation unit (IMU) data to the motion compensation unit in real time.

[0012] The signal acquisition and transmission unit is optically connected to the two-dimensional scanning unit and is used to emit laser pulses through the two-dimensional scanning unit to the target area.

[0013] The signal acquisition and transmission unit is data-connected to the motion compensation unit and is used to transmit the photon event timestamp to the motion compensation unit.

[0014] The motion compensation unit is used to align the inertial navigation unit (IMU) data with the photon event timestamps, calculate the instantaneous attitude of each photon event, and perform point-by-point correction on the spatial position of the photon event to eliminate point cloud motion distortion.

[0015] Furthermore, the signal acquisition and transmission unit adopts a coaxial optical structure integrating transmission and reception, and is equipped with a dynamic light leakage suppression mechanism. The dynamic light leakage suppression mechanism includes a time-to-digital converter (TDC) to accurately measure the timestamps of each laser pulse emission and echo photon reception, and a nanosecond-level gate control of the two acousto-optic modulators (AOM) by the main control unit, so as to physically cut off the receiving optical path before the echo photon arrives in order to suppress light leakage and near-field scattered light in the system.

[0016] Furthermore, the two-dimensional scanning unit can be a MEMS micro-mirror, which functions to perform high-speed scanning of the laser pulse beam in space to cover the target area and generate point cloud data.

[0017] Furthermore, the motion compensation unit employs a single-photon event-level jitter compensation algorithm, which uses an IMU pre-integration sensor fusion algorithm instead of the traditional linear interpolation method. This algorithm pre-integrates the inertial data from the IMU to accurately calculate the system's motion and attitude increments between two IMU measurement moments. By temporally aligning the IMU data and laser data, the attitude at each photon event moment can be obtained. This allows for real-time acquisition of the UAV's three-axis attitude (pitch, roll, and yaw) data in a dynamic environment with high temporal resolution, thus solving the data sparsity problem caused by the IMU data sampling frequency being lower than the lidar pulse frequency. Subsequently, the algorithm uses the pre-integrated IMU data and the echo data of each laser pulse as factors, and optimizes them through joint pixel geometric constraints and motion compensation constraints. In this way, the algorithm can accurately calculate the instantaneous attitude at each photon transmission and reception event.

[0018] Furthermore, the timestamps of photon events at the nanosecond level are precisely calculated. The three-axis attitude information corresponding to the timestamps of emitted and received photon events is extracted.

[0019] Furthermore, utilizing a unique jitter compensation algorithm, the time-of-flight (TOF) and spatial position of the photon are precisely calibrated, accurately correcting the position information of each photon event to the Earth coordinate system. This point-by-point correction, rather than traditional linear interpolation, fundamentally eliminates point cloud distortion caused by nonlinear motion, improving point cloud accuracy from the centimeter level to the micrometer level.

[0020] A high-precision imaging method for single-photon unmanned aerial vehicles (UAVs) includes the following steps:

[0021] Step 1: Use the FPGA synchronization signal as the global clock to synchronize the pulsed laser, time-to-digital converter (TDC), two-dimensional galvanometer, and high-frequency IMU.

[0022] Step 2: The main control unit triggers the pulsed laser to emit an ultrashort pulse and starts the TDC timing, recording the timestamp of the emitted photon event;

[0023] Step 3: The main control unit closes the receiving optical path with nanosecond-level gating within the predetermined dead time and opens the receiving optical path within the expected time window of the echo photon arrival to achieve dynamic light leakage suppression.

[0024] Step 4: The echo photon enters the single-photon detector SPAD through gating, triggering the TDC to stop timing and record the timestamp of the received photon event;

[0025] Step 5: The high-frequency IMU acquires attitude data and aligns it with the photon emission / reception timestamps. Combined with the system-calibrated emission and reception optical path delays, the three-axis instantaneous attitude corresponding to each photon event is obtained.

[0026] Step 6: Employ a single-photon event-level jitter compensation algorithm to perform high-precision calibration of the flight time and spatial position of each photon event and eliminate point cloud motion distortion;

[0027] Step 7: Perform denoising and sparse point cloud reconstruction on the motion-compensated point cloud to output a stable and clear 3D scene.

[0028] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0029] 1. This invention adopts a "single-photon event-level" high-precision motion compensation scheme supported by high-frequency IMU data, which fundamentally solves the problem of point cloud distortion caused by nonlinearity and high-frequency jitter of UAV platforms that traditional methods cannot effectively handle. It improves the distortion correction accuracy from the centimeter level of traditional methods to the millimeter level, thereby obtaining more stable and clearer 3D scene data.

[0030] 2. This invention effectively blocks the impact of near-field strong light and leakage light from the ring fiber optic cable on the single-photon detector through a dynamic time gating mechanism in which TDC and AOM work together, solving the problems of detector saturation and dead time, and greatly improving the signal-to-noise ratio and detection robustness of the system.

[0031] 3. The combination of high signal-to-noise ratio and high sensitivity enables the system to effectively detect weak photon echoes from a greater distance, thereby expanding the effective detection range and application scope of single-photon radar.

[0032] 4. The application of hardware-level high-precision timing synchronization and multi-sensor fusion algorithms enables this system to maintain high-precision perception and reconstruction capabilities even in the most severe dynamic environments, providing strong technical support for the application of UAVs in high-precision 3D mapping, intelligent inspection and automatic navigation. Attached Figure Description

[0033] Figure 1 This is a schematic diagram of the overall architecture of the UAV-borne single-photon imaging radar system of the present invention;

[0034] Figure 2 This is a timing diagram of the dynamic light leakage suppression mechanism of the present invention;

[0035] Figure 3 This is a schematic diagram of the multi-sensor coordinate system registration and transformation relationship of the present invention;

[0036] Figure 4 This is a flowchart of the single-photon event-level jitter compensation algorithm of the present invention. Detailed Implementation

[0037] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other. To achieve the above objectives, this invention adopts the following technical solution.

[0038] This invention provides a single-photon unmanned aerial vehicle (UAV) airborne high-precision imaging system, such as... Figure 1 As shown, it includes a laser, an acousto-optic modulator A (AOM_A), an acousto-optic modulator B (AOM_B), an optical fiber circulator, a single-photon detector SPAD, a time-to-digital converter TDC, a main control module FPGA, an inertial navigation unit IMU, and a two-dimensional scanning mirror MEMS.

[0039] The connections and functions of each component are as follows: The laser output light is modulated by AOM_A and coupled into the input end of the fiber optic circulator. The fiber optic circulator guides the laser pulse to the two-dimensional scanning mirror MEMS for scanning and emission into space. The two-dimensional scanning mirror MEMS oscillates at high speed along the horizontal X-axis and vertical Y-axis to achieve two-dimensional beam scanning of the target area. After the laser pulse irradiates the target, its reflected echo signal returns along the original path and is coupled into the fiber optic circulator via the two-dimensional scanning mirror MEMS. The fiber optic circulator guides the echo signal to AOM_B. AOM_B gates and modulates the returned light path before transmitting the signal to a single-photon detector. The single-photon detector, for example, uses an avalanche photodiode (SPAD) operating in Geiger mode. When a single photon event is detected, it outputs a corresponding electrical signal pulse. This pulse signal is input to a time-to-digital converter (TDC), which accurately records the photon arrival timestamp or time-of-flight (TOF) data. The FPGA is electrically connected to the aforementioned components and provides real-time control and data acquisition: the FPGA sends a synchronization trigger signal to the laser to emit laser pulses, sends gating control signals to AOM_A and AOM_B to sequentially open / close the optical path, and acquires photon time-of-flight measurements from the TDC. Simultaneously, the FPGA is also connected to a high-frequency IMU, providing a picosecond-level absolute time reference synchronization clock, while the IMU outputs attitude angular velocity and acceleration data of the UAV platform at a high 2kHz sampling rate. These data are then fused by the processor to provide real-time attitude data. Through this structure, the modules collaboratively achieve synchronized acquisition of single-photon laser ranging and attitude, providing fundamental data for subsequent motion compensation and point cloud reconstruction.

[0040] like Figure 2 As shown, this invention addresses the low signal-to-noise ratio problem caused by internal light leakage and environmental noise in lidar during high-speed flight by designing a dynamic light leakage suppression mechanism with dual AOM gating. Figure 2As shown, the FPGA generates two synchronous gating signals to control AOM_A and AOM_B respectively, dividing the emission window and the receiving window in each laser emission cycle.

[0041] Specifically, at the initial stage of each laser ranging cycle, the FPGA first outputs a synchronization pulse signal to trigger the laser to emit a narrow 1.5ns laser pulse. Almost simultaneously, the FPGA sends an enable signal to AOM_A, switching AOM_A to a high-transmittance state to allow the laser pulse to pass through and enter the transmission optical path. After a laser pulse of one pulse width successfully passes through, the FPGA quickly turns AOM_A off to a low-transmittance (high-loss) state, thereby blocking any subsequent possible leakage or stray light from the laser from entering the optical path. During this transmission window, AOM_B remains off (low-transmittance state) to prevent the transmitted pulse from directly coupling to the receiver via the fiber optic circulator, causing saturation interference. Subsequently, during the reception phase of this cycle, the FPGA sends a control signal to enable AOM_B according to a preset delay window, meaning that AOM_B is only in a high-transmittance state within the expected time window of the target echo arrival, allowing the echo photons to pass through and enter the SPAD detector. Once the reception window ends, AOM_B is quickly turned off again, returning to the light-blocking state.

[0042] Through the aforementioned dual AOM gating control, the laser emission and reception channels are strictly separated in time: at the moment of laser emission, the reception channel is turned off to suppress direct light leakage; during the echo waiting period, the emission channel is turned off to avoid continuous light emission interference, thereby significantly reducing the possibility of internal crosstalk and stray light from the environment entering the detector. The FPGA precisely controls the opening and closing timing of AOM_A and AOM_B at the nanosecond level, ensuring that the reception window width covers only a very narrow time period near the arrival time of the desired signal. The delay and width of the reception window can be adjusted in real time according to the approximate distance between the UAV and the target, ensuring that the actual signal photons fall during the detector's on period while most noise photons are rejected due to asynchrony.

[0043] The TDC, in conjunction with the FPGA logic, measures the time of each photon event entering the receiving window. When a photon is detected, its corresponding flight time is recorded, while noise triggering outside the window is ignored. Thanks to the aforementioned dynamic light leakage suppression design, this system can maintain a high signal-to-noise ratio for single-photon detection even in strong sunlight or complex airborne optical reflection environments.

[0044] For example, the optical isolation of dual AOMs ensures that the detector is not affected by saturation for hundreds of nanoseconds after the emitted pulse. Combined with narrowband filtering and time filtering, the noise count rate can be reduced to the level of single photon counts, significantly improving ranging accuracy and effective detection range.

[0045] like Figure 3The image shows the pixel drift caused by jitter at the moment of photon emission and reception. When a drone hovers at high altitude, it will generate high-frequency jitter based on the drone itself, that is, the drone's coordinate system will have a three-axis angular difference with the world coordinate system. However, the photon transmission and reception system is based on the drone's coordinate system, while the target object is based on the world coordinate system, which will cause deviations in imaging pixels, intensity, and depth information.

[0046] like Figure 4 The diagram illustrates the single-photon event-level jitter compensation algorithm of this invention. The photon-level pixel correction includes: obtaining prior delay information through optical path delay calibration and synchronizing it with the FPGA's reference clock to obtain the precise timestamp of the photon emitted from the system's transmitter; registering the high-frequency inertial navigation data with the timestamp to obtain the UAV attitude information corresponding to the emitted photon timestamp, specifically the roll and pitch angles; registering the attitude information with the X-axis and Y-axis offset angle information of the MEMS galvanometer at that moment to obtain the emission angle in the world coordinate system corresponding to the emitted photon timestamp; and correcting the position of the actual pixel in the actual scanned image using this angle.

[0047] Intensity information correction includes: obtaining prior delay information through transmit and receive optical path delay calibration in advance, and synchronizing it with the FPGA's reference clock to obtain the precise timestamps of photon events at the system's transmitter and receiver; registering the high-frequency inertial navigation data with the above timestamps to obtain the UAV attitude information corresponding to the photon event timestamps, specifically the heading angles at the two photon moments; since there is a position difference (heading angle difference) between the transmit and receive moments, the number of received photons will decrease, and the greater the position difference, the fewer photons are received, i.e., the heading angle difference is used for photon quantity correction.

[0048] Micrometer-level imaging depth correction includes: obtaining prior delay information through optical path delay calibration and synchronizing it with the FPGA's reference clock to obtain the precise timestamp of the photon event at the system receiver; registering the high-frequency inertial navigation data with the timestamp to obtain the UAV attitude information corresponding to the photon event timestamp, specifically the roll and pitch angles; registering the time-of-flight (TOF) data of the photon with the timestamp to obtain the TOF of the photon event, and also obtaining the aircraft path information; since different angle information corresponds to different flight path information; the angle fusion vector can be obtained from the roll and pitch angles; by correcting the flight path information with the angle information, the true TOF data of the photon event can be obtained.

[0049] Specifically, the FPGA or its connected processing unit pre-integrates the angular velocity and acceleration data output from the high-frequency IMU to calculate the relative attitude changes of the UAV at the moments of photon emission and reception. For each laser pulse event, a timestamp provided by the TDC records the emission time and the corresponding single-photon return time. By pre-integrating and estimating the IMU data, attitude data with the same frequency as the laser pulse can be obtained. Simultaneously, the emission and reception optical path delays in the optical system are obtained through calibration. By accurately calculating the specific times of the two timestamps, the attitude (heading, roll, pitch) and position of the UAV at those two timestamps can be obtained. This allows for the determination of the amount of attitude change (e.g., [missing information]) that occurs during the nanosecond-level flight time. Figure 3 (The coordinate system of the aircraft changes slightly relative to the geographic coordinate system at the time of transmission and reception, which will cause the uncorrected point cloud to shift and distort.)

[0050] This invention treats each photon ranging as a motion compensation unit: first, the emission direction of the laser pulse in the airborne coordinate system is determined by the attitude at the time of emission, and the initial spatial ray is determined by combining the deflection angle of the scanning galvanometer. Specifically, the roll angle of the aircraft corresponds to the X-axis of the MEMS galvanometer, and the pitch angle corresponds to the Y-axis of the MEMS, and photon event pixel correction is performed.

[0051] Then, the change in the UAV coordinate system relative to the geographic reference system is determined using the attitude at the moment of reception. The relationship between the spatial point obtained from each photon ranging and the UAV pose is used as an observation factor. The received attitude transformation is converted into an angle fusion vector, and the fused angle is calibrated with depth information (e.g., depth information). Figure 3 The difference in one-dimensional depth information is shown, and micron-level imaging depth correction is performed accordingly.

[0052] Next, the pose of the UAV at each laser emission / reception moment is used as a state variable node. When the received photon is regarded as returning along the original ray but the sensor attitude has changed, the greater the difference in attitude angle, the greater the loss of intensity information. The heading angles corresponding to the two timestamps are differentially inverted to correct the intensity information of the received reflected photon, thus forming a complete motion compensation optimization problem.

[0053] The coordinates of the reflection point corresponding to each photon event are accurately estimated in the global coordinate system, greatly reducing the misalignment error of the airborne radar point cloud caused by high-speed attitude changes. Simply put, the method of this invention is equivalent to performing "spatiotemporal correction" on each laser point: based on the attitude differences at the time of transmission and reception, the true projection angle and distance of the light in the geographic coordinate system are deduced, thereby accurately restoring the position of each photon event in the geographic coordinate system and achieving point-by-point distortion correction of the point cloud. This motion compensation is particularly effective under severe UAV maneuvers, ensuring that the reconstructed 3D point cloud matches the actual terrain and target.

[0054] Implementation Process: The specific workflow of this system includes the following stages. First, system initialization and synchronization are established: the FPGA main control and all sensor modules are started, the AOM is reset to its initial off state, and the laser enters standby mode. The FPGA provides a standard time pulse to synchronize the system clock to the absolute time reference and performs unified reference alignment for the data time axes of the TDC and IMU. Then, the measurement cycle begins: the FPGA sends a trigger signal to the laser at a predetermined pulse repetition frequency of 1MHz and simultaneously generates a synchronization signal sent to the TDC as a starting reference. Each time the laser receives a trigger, it emits a laser pulse, which enters the optical fiber through the gated window of AOM_A and is emitted in a specific direction by the galvanometer. Immediately afterwards, the FPGA controls AOM_A to turn off and waits for a preset delay before opening the AOM_B receiving window. If the single-photon signal reflected from the target returns during this window period, it passes through AOM_B and is detected by the SPAD. The SPAD generates an electrical signal that is compared with the starting synchronization signal via the TDC to obtain the time of flight (TOF) of the photon. The FPGA reads the TOF value recorded by the TDC and adds a time stamp to the ranging event. Simultaneously, throughout the process, the IMU outputs inertial data at high frequency, and the FPGA receives and buffers the IMU data stream in real time. Each time a laser transmit-receive cycle is completed, the FPGA calculates the attitude change using a pre-integration algorithm on the IMU data for the corresponding time period and obtains the transmit and receive attitude information for that ranging measurement for subsequent distortion correction. The system continuously repeats the above pulse transmission and data acquisition process, while the galvanometer gradually changes its exit angle according to the set scanning pattern, ensuring the laser beam covers the predetermined scan line or field of view in the target scene. Throughout the operation, the FPGA coordinates the synchronization of each module: ensuring that laser triggering, AOM gating, IMU sampling, and TDC timing all operate strictly according to a unified time base, with typical synchronization accuracy better than 1 nanosecond, ensuring that multi-sensor data fusion errors are reduced to a negligible level.

[0055] In the data processing stage, the first step is to perform time difference conversion and preliminary point cloud reconstruction: the corresponding distance is calculated based on the photon TOF measurement value, and the spatial coordinates are determined by combining the current galvanometer deflection angle and attitude. Then, distortion correction and point cloud alignment are performed: the aforementioned factor map optimization or extrapolation algorithm is applied to smooth the trajectory and attitude of the UAV based on the spatiotemporal information of all acquired photon events, correcting the spatial position deviation of each point and aligning the point cloud in the geographic coordinate system. Next, point cloud filtering and denoising are performed: since single-photon detection inherently has a certain probability of noise triggering and multi-photon aliasing, the histogram accumulation method is used to identify effective echoes (if multiple photons are detected near the same location in a single scan and the distance is consistent, it is judged as a real target) and eliminate abnormal outliers and non-real ranging points. Finally, the filtered 3D point cloud data is reconstructed and output: depending on the application requirements, the point cloud can be converted into a digital elevation model (DEM) or a 3D mesh, or directly superimposed onto a map model in a surveying coordinate system; at the same time, pseudo-color rendering can be performed by combining the return intensity information of each pixel (obtained by multiple echo counting or photon counting statistics) to generate a more readable radar image.

[0056] Example of key hardware structure and parameter settings: In this embodiment, each core hardware component uses high-performance devices, and the parameters are optimized according to the requirements of single-photon detection. The laser is preferably a pulsed fiber laser with a center wavelength in the near-infrared 1550 nm band, a single pulse energy of microjoules, a pulse width of about 1.5 ns, and a pulse repetition frequency of over 1 MHz to provide high detection sensitivity and sufficient point cloud density; AOM_A and AOM_B use high-speed acousto-optic modulators with rise / fall time not exceeding tens of nanoseconds, insertion loss in the on state less than 3dB, and a high extinction ratio (>50dB) in the off state to ensure that the gated window can open and close quickly and effectively isolate direct light; the time-to-digital converter (TDC) has picosecond-level time resolution (e.g., a single-step resolution of about 10 picoseconds), corresponding to a distance resolution in the millimeter range.

[0057] High-frequency IMUs, for example, use a combination of a three-axis gyroscope and an accelerometer, with a sampling frequency of 2kHz, a low angle random walk coefficient, and a dynamic range that meets the maneuverability range of UAVs, to ensure accurate attitude measurement under high dynamic conditions; MEMS two-dimensional galvanometers have scanning optical angle ranges of, for example, ±15° horizontally and ±10° vertically, with bandwidths of hundreds of hertz, which can support forward sector scanning imaging of small UAV platforms.

[0058] The FPGA main control module uses high-speed devices, and its internal clock synchronization accuracy can reach the nanosecond level. It also integrates a high-speed ADC / DAC interface for sensor control and data acquisition and storage. Through the above device configuration and parameter selection, this system can still ensure the ranging accuracy (typical distance accuracy can be better than a few centimeters), spatial resolution and reliability of the single-photon lidar system in the environment of high-speed flight and severe vibration of UAV.

[0059] Data Flow and Signal Control Logic: The data and synchronization signal interaction process between modules in this system is clear and rigorous, and as mentioned earlier, it is uniformly coordinated by the FPGA. The specific signal flow is as follows: A high-precision timing generator runs inside the FPGA, generating a synchronization trigger pulse at the beginning of each cycle. This pulse is output through the laser driver interface to excite the laser to emit a pulse, and is also sent to the TDC as a reference marker for the zero point of the cycle. Subsequently, the FPGA sets the AOM_A control terminal high after a delay based on the internal timer, maintaining the on state for the agreed pulse width time, so that the laser pulse returns to low level immediately after passing through AOM_A completely, closing the transmission optical path. After the laser pulse is emitted through the circulator and galvanometer and interacts with the target, the FPGA waits for the pre-calculated time-of-flight delay, and then sets the AOM_B control terminal high to open the receiving window. If the target distance changes dynamically or is uncertain, the FPGA can adaptively adjust its value based on the airborne altitude sensor or historical measurement data to achieve "sliding window" type dynamic gating. When the SPAD detector receives photons during the AOM_B period, it generates an electrical signal, which is transmitted to the TDC timing module through a high-speed circuit. The TDC measures the time delay of the photon signal relative to the reference synchronization pulse and feeds this time data back to the FPGA via a high-speed serial interface. The FPGA registers and matches each measured time data with the current galvanometer scanning angle and IMU attitude data according to time tags (e.g., Figure 4 The data registration module shown forms the complete measurement information for this photon event. Simultaneously, the raw measurement data from the IMU is continuously uploaded via the FPGA's internal bus and stored in a cache at fixed intervals for subsequent pre-integration calculations.

[0060] Optional technical solutions or variant embodiments: The main embodiments of the present invention have been described above with reference to the accompanying drawings, but the present invention is not limited thereto. In other variations, each module device can be equivalently replaced or its functions expanded according to application requirements. For example, acousto-optic modulators A and B can be replaced by high-speed electro-optic modulators or fiber optic switching devices, as long as nanosecond-level optical path switching functions can be achieved; single-photon detectors are not limited to a single APD device, but can also use a multi-pixel SPAD array to form a surface array receiver, thereby improving spatial coverage (in which case the scanning mechanism can be appropriately simplified); the scanning galvanometer assembly can be replaced by other scanning methods, such as rotating polygon mirrors or MEMS phased arrays, to adapt to different platform size and power consumption requirements. Furthermore, in motion compensation algorithms, different implementation methods can be selected according to real-time requirements: for offline data processing, factor graph global optimization can be used to obtain the highest accuracy trajectory and point cloud correction results; for real-time applications, tightly coupled solution based on Kalman filtering can be used, and the position of each laser point can be corrected instantly using IMU increments, thereby reducing computational latency. Furthermore, the system can perform SLAM (Simultaneous Localization and Matching) solely based on IMU and laser point cloud features, which is an extended application of the technical solution of this invention. In summary, those skilled in the art can make various substitutions and improvements to the system components and method steps without departing from the original concept of this invention, and all such modifications should be considered to fall within the protection scope of this invention.

Claims

1. A high-precision imaging system for single-photon unmanned aerial vehicles (UAVs), characterized in that, include: Main control unit, jitter acquisition unit, signal acquisition and transmission unit, two-dimensional scanning unit and motion compensation unit; The main control unit is electrically connected to the jitter acquisition unit, the signal acquisition and transmission unit, the two-dimensional scanning unit and the motion compensation unit, respectively, and is used to provide picosecond-level global timing synchronization signals. The jitter acquisition unit is communicatively connected to the motion compensation unit, including an inertial navigation unit (IMU), and transmits IMU data to the motion compensation unit in real time. The signal acquisition and transmission unit is optically connected to the two-dimensional scanning unit and is used to emit laser pulses through the two-dimensional scanning unit to the target area. The signal acquisition and transmission unit is data-connected to the motion compensation unit and is used to transmit the photon event timestamp to the motion compensation unit. The motion compensation unit is used to align the inertial navigation unit (IMU) data with the photon event timestamps, calculate the instantaneous attitude at the moment each photon event occurs, and perform point-by-point correction on the spatial position of the photon event to eliminate point cloud motion distortion. The signal acquisition and transmission unit includes a pulsed laser, a single-photon detector (SPAD), an acousto-optic modulator (AOM1), an acousto-optic modulator (AOM2), and a time-to-digital converter (TDC). AOM1 is positioned in the transmission optical path, and AOM2 is positioned in the reception optical path. The TDC is electrically connected to both the SPAD and the main control unit, and is used to record the timestamps of laser emission and echo photon reception. The main control unit precisely controls the opening and closing timing of AOM1 and AOM2 in the transmission optical path to the nanosecond level, ensuring that the reception window width covers only a very narrow time period near the expected signal arrival time. The delay and width of the reception window are adjusted in real-time according to the approximate distance between the UAV and the target, ensuring that the actual signal photons fall during the detector's on period while most noise photons are rejected due to asynchrony. This ensures strict temporal separation between the transmission and reception windows, suppressing internal light leakage and near-field scattered light. The jitter acquisition unit includes an inertial navigation unit (IMU), and uses a compensation algorithm to increase its sampling frequency to a higher level than the laser pulse frequency. This allows for real-time acquisition of the UAV's pitch, roll, and yaw angles with high temporal resolution. The compensation algorithm works as follows: First, the timestamps of the IMU data are mapped to the FPGA's time base to ensure alignment between the IMU and FPGA times. Then, within each time window, the angular velocity and acceleration of the IMU are integrated to calculate the increments in attitude, velocity, and position. At each photon event or when the FPGA timestamp arrives, the calculated increments are used to update the state, increasing the IMU's frequency to the nanosecond level.

2. The system as described in claim 1, characterized in that, The MEMS micro-mirror of the two-dimensional scanning unit swings at high speed along the horizontal and vertical degrees of freedom to generate a preset scanning pattern. The mirror deflection angle and the timestamp of the laser emission / reception event are used together for spatial coordinate calculation.

3. The system as described in claim 1, characterized in that, The motion compensation unit registers the inertial navigation unit (IMU) coordinate system, the lidar coordinate system, and the Earth fixed coordinate system. It uses delay calibration to estimate the attitude increment between the two instants of transmission and reception, and fuses the IMU data and photon event data as factors. Through joint optimization of pixel geometric constraints and motion compensation constraints, it performs pixel, intensity, and depth corrections point by point for each photon event.

4. The system as described in claim 1, characterized in that, The main control unit implements picosecond-level global timing based on the parallel and configurable logic of the FPGA, synchronously distributes trigger / gating / sampling control signals to the laser, acousto-optic modulator (AOM), time-to-digital converter (TDC), and IMU, and adds a time tag with a unified time reference to the time-of-flight measurement value returned by the TDC.

5. The system as described in claim 1, characterized in that, After completing the single-photon event-level motion distortion correction, the motion compensation unit further performs point cloud filtering and alignment, and maps the point-by-point corrected 3D point cloud onto the Earth fixed coordinate system to obtain aligned 3D scene data.

6. The system as described in claim 1, characterized in that, The motion compensation unit also includes a point cloud denoising and reconstruction module based on deep learning. The module uses a convolutional neural network and integrates residual learning, batch normalization and attention mechanisms to denoise the calibrated point cloud and reconstruct the sparse point cloud into a continuous geometric surface model.

7. The system as described in claim 2, characterized in that, The acousto-optic modulator can be replaced with a high-speed electro-optic modulator or fiber optic switch capable of nanosecond-level optical path switching.

8. The system as described in claim 1, characterized in that, The two-dimensional scanning unit can be replaced by a rotating multifaceted mirror scanning mechanism or a MEMS phased array.

9. A high-precision airborne imaging method for single-photon unmanned aerial vehicles, which is the imaging method of the system described in any one of claims 1-8, characterized in that, Includes the following steps: Step 1: Use the FPGA synchronization signal as the global clock to synchronize the pulsed laser, time-to-digital converter (TDC), two-dimensional galvanometer, and high-frequency IMU. Step 2: The main control unit triggers the pulsed laser to emit an ultrashort pulse and starts the TDC timing, recording the timestamp of the emitted photon event; Step 3: The main control unit closes the receiving optical path with nanosecond-level gating within the predetermined dead time and opens the receiving optical path within the expected time window of the echo photon arrival to achieve dynamic light leakage suppression. Step 4: The echo photon enters the single-photon detector SPAD through gating, triggering the TDC to stop timing and record the timestamp of the received photon event; Step 5: The high-frequency IMU acquires attitude data and aligns it with the photon emission / reception timestamps. Combined with the system-calibrated emission and reception optical path delays, the three-axis instantaneous attitude corresponding to each photon event is obtained. Step 6: Use a single-photon event-level jitter compensation algorithm to calibrate the flight time and spatial position of each photon event and eliminate point cloud motion distortion; Step 7: Perform denoising and sparse point cloud reconstruction on the motion-compensated point cloud to output a stable and clear 3D scene.

10. The method as described in claim 9, characterized in that, The dead time in step 3 is on the order of nanoseconds. The main control unit performs time-division gating on the transmitting optical path and the receiving optical path respectively, so that the two are strictly separated in time within a single ranging cycle and adaptively matched with the echo arrival window.

11. The method as described in claim 9, characterized in that, In step 5, the IMU attitude data and photon event timestamps are aligned under a unified time reference, and the attitude of each photon event is precisely matched with its timestamp through transmit / receive optical path delay calibration.

12. The method as described in claim 9, characterized in that, In step 6, IMU pre-integration and motion compensation are used for joint optimization to constrain the attitude during transmission and reception, and pixel geometry and depth correction are performed point by point for each photon event.

13. The method as described in claim 9, characterized in that, The denoising in step 7 employs a convolutional neural network that integrates residual learning, batch normalization, and attention mechanisms. The reconstruction uses a deep learning algorithm to convert the denoised point cloud into a continuous geometric surface model.

Images