Multi-source heterogeneous data synchronous acquisition method used in scene that pilot operates helicopter

Abstract

The invention discloses a multi-source heterogeneous data synchronous acquisition method used in a scene that a pilot operates a helicopter, and the method comprises the following steps: constructing a multi-source sensor network, and carrying out the synchronous acquisition of data; a multi-channel data fusion algorithm based on timestamp alignment is adopted to process data, and sensor noise is eliminated through Kalman filtering; a two-stage transmission ratio calibration model based on combination of ground calibration and flight data is established, and dynamic mapping of the displacement / angle of the joystick and the pitch angle of the rotor wing and the tail rotor is achieved; developing an anti-vibration optical character recognition system, and extracting fuel quantity, engine temperature, rotor speed and rotor torque parameters from an instrument image; and generating a standardized data packet containing a quality identifier so as to support real-time monitoring of a flight state and offline analysis of flight data. According to the method, high-precision, total-factor and real-time data acquisition and visual monitoring in the helicopter operation process can be realized, and the method is suitable for test and verification scenes of helicopters of different models.

Description

Technical Field

[0001] This invention belongs to the field of helicopter control data acquisition, and specifically relates to a method for multi-source data acquisition tasks, which is applicable to integrated, intelligent and unified detection systems. Background Technology

[0002] Early helicopter simulators mostly used purely mechanical joysticks, which presented challenges such as difficult data acquisition and complex replacement. For example, mechanical joysticks simulated actions using springs and levers, but struggled to acquire accurate control attitude data in real time, and were also costly to maintain. Simultaneously, helicopter rotor systems (main rotor and tail rotor) faced harsh measurement environments during high-speed rotation, including centrifugal force, vibration, and temperature variations. Traditional methods required mounting sensors on rotating components, but limitations in power supply and data transmission technologies often led to signal interference or data loss. Some systems transmitted data via satellite or dedicated base stations (such as high-throughput satellite communication), which, while offering strong real-time performance, was prohibitively expensive and unsuitable for small and medium-sized flight schools or general aviation applications. Regarding data storage, traditional SD cards required manual insertion and removal for data export, resulting in cumbersome operations and poor timeliness. For instance, some systems required pilots to manually change storage media, easily leading to data loss or corruption. Furthermore, for data acquisition in helicopter scenarios, flight status data (such as attitude and vibration) needed to be collected simultaneously; existing systems, due to limited sensor types or incompatible communication protocols, struggled to achieve efficient integration of multimodal data.

[0003] With the development of intelligent integrated mechanical structures, multi-sensor fusion and edge computing are receiving increasing attention. For example, by integrating speed sensors, displacement sensors, and laser sensors, and combining them with edge computing modules (such as STM32 microcontrollers), local data processing and intelligent alarms (such as over-temperature and oil pressure anomalies) can be achieved. However, current research has not yet yielded a method for synchronous acquisition of multi-source heterogeneous data that strikes a balance between integration and completeness. Summary of the Invention

[0004] To address the issues of poor real-time data acquisition, incomplete parameters, and lack of intelligent processing in helicopter scenarios, this invention proposes a method for synchronous acquisition of helicopter control and flight data based on multi-source heterogeneous data. This method is applicable to different helicopter models and enables synchronous data acquisition and visual monitoring in both ground and flight testing scenarios.

[0005] To achieve the above objectives, the invention adopts the following technical solution:

[0006] A method for synchronous acquisition of multi-source heterogeneous data in helicopter pilot operation scenarios includes the following steps:

[0007] Step 1: Construct a multi-source sensor network to synchronously collect pilot control input data, helicopter attitude response data, and instrument image data; the multi-source sensor network includes at least a displacement sensing module for measuring the travel of the joystick and booster, an attitude measurement module for measuring the helicopter's position and attitude, and an image acquisition module for acquiring instrument panel video.

[0008] Step 2: A multi-channel data fusion algorithm based on timestamp alignment is used to process the multi-source data collected in Step 1, and sensor noise is eliminated by Kalman filtering.

[0009] Step 3: Establish a two-stage transmission ratio calibration model based on the combination of ground calibration and flight data to achieve dynamic mapping between joystick displacement / angle and rotor and tail rotor pitch angle;

[0010] Step 4: Develop a vibration-resistant optical character recognition (OCR) system that incorporates image stabilization processing and deep learning recognition to extract parameters such as fuel quantity, engine temperature, rotor speed, and rotor torque from instrument images;

[0011] Step 5: Generate a standardized data package containing quality identifiers to support real-time monitoring of flight status and offline analysis of flight data.

[0012] Furthermore, in step 1, the multi-source sensor network adopts a distributed architecture with an industrial control computer as the data processing core; each sensor module communicates with the industrial control computer through a unified RS422 or RS485 industrial standard protocol, and supports hot-swapping and online configuration.

[0013] Furthermore, in step 1, the displacement sensing module includes a laser displacement sensor and a non-contact angular displacement sensor; the attitude measurement module includes a navigation system composed of a fiber optic strapdown inertial navigation system and a satellite navigation box, an atmospheric data sensor, and a radio altimeter; and the image acquisition module is a high-speed camera.

[0014] Furthermore, in step 2, the multi-channel data fusion algorithm specifically includes the following processing sub-steps:

[0015] Verify the data from each port;

[0016] A uniform, high-precision timestamp is added to the verified data to achieve alignment. Timestamp alignment is achieved by combining hardware trigger signals with software synchronization algorithms to ensure the timing consistency between control inputs and helicopter response data.

[0017] Data compensation is performed on data channels that experience delays or anomalies.

[0018] Furthermore, in step 2, sensor noise is eliminated by Kalman filtering, specifically by using the Kalman filtering algorithm to fuse data from the inertial navigation system and the satellite navigation system to improve the measurement accuracy of the helicopter's attitude parameters.

[0019] Furthermore, in step 3, the transmission ratio calibration model uses the recursive least squares method to update parameters, and the update frequency is adaptively adjusted according to the flight state; the transmission ratio calibration includes the calculation of collective pitch transmission ratio, longitudinal transmission ratio, lateral transmission ratio and pedal transmission ratio.

[0020] Furthermore, in step 4, the vibration-resistant optical character recognition (OCR) system sequentially performs the following four processing stages: image preprocessing, reference point tracking, character region segmentation, and character recognition based on a deep learning model.

[0021] Furthermore, in step 4, during the reference point tracking stage, for the first instrument panel displaying fuel quantity, engine temperature, and rotor torque parameters, the following steps are performed: scanning the edges of the horizontal and vertical reference lines in the image to form a point set, fitting the point set with a straight line, and using the intersection of the two fitted straight lines as the positioning reference point; for the second instrument panel displaying rotor speed parameters, the following steps are performed: edge detection is performed on the image to extract the circular instrument outline, finding the largest bounding rectangle of the outline, and calculating the centroid of the rectangle as the positioning reference point.

[0022] Furthermore, in step 5, the standardized data packet adopts a hierarchical storage structure, including a raw data layer, a processed data layer, and an application data layer; real-time monitoring is implemented through a front-end interface based on PyQt5, which has positioning and navigation functions and can calculate and display the return direction and distance based on real-time latitude and longitude.

[0023] Furthermore, the method is applicable to both ground testing and flight testing scenarios; ground testing includes sensor calibration and transmission ratio modeling through stepped control inputs, and measuring the dynamic characteristics of the servo motor through dipole control inputs; the flight testing is used for multi-dimensional data acquisition and real-time monitoring in a real flight environment.

[0024] Beneficial effects:

[0025] Compared with the prior art, the present invention has the following significant advantages:

[0026] (1) Excellent synchronization: Through a unified hardware triggering and software timestamp alignment mechanism, high-precision synchronous acquisition of multi-source heterogeneous data is achieved, solving the problem of chaotic timing in traditional systems;

[0027] (2) High safety: The non-contact measurement scheme avoids the installation of sensors in high-risk parts such as rotors, which significantly improves the reliability and safety of the system;

[0028] (3) Complete data dimensions: Combining traditional sensors and machine vision technology, it achieves full coverage of manipulation input, posture response, environmental parameters and instrument parameters;

[0029] (4) High level of intelligence: It integrates intelligent algorithms such as Kalman filtering, recursive least squares, and deep learning to realize adaptive data processing and anomaly identification;

[0030] (5) High versatility: It adopts modular design and standardized interface, which can be adapted to different helicopter models through configuration, greatly improving the scalability and deployment efficiency of the system;

[0031] (6) Strong real-time monitoring capability: It provides a visual monitoring interface with positioning and navigation functions, supports real-time display and data analysis of flight status, and provides comprehensive technical support for flight tests.

[0032] Beneficial effects: Attached Figure Description

[0033] Figure 1 This is a diagram of the overall ground testing plan;

[0034] Figure 2 This is a diagram of the overall flight test plan. Detailed Implementation

[0035] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings. This section describes the preferred mode of implementing the present invention, but it should not be construed as a limitation on the scope of protection of the claims.

[0036] This invention provides a method for synchronous acquisition of multi-source heterogeneous data in helicopter pilot scenarios. Its specific implementation method ensures the synchronization, accuracy, and real-time nature of data acquisition through a systematic process design. The implementation steps are described in detail below, in stages.

[0037] To address the issues of poor real-time data acquisition, incomplete parameters, and lack of intelligent processing in helicopter scenarios, this invention proposes a method for synchronous acquisition of helicopter control and flight data based on multi-source heterogeneous data. This method is applicable to different helicopter models and enables synchronous data acquisition and visual monitoring in both ground and flight testing scenarios.

[0038] The core of this method lies in constructing a modular data acquisition system and employing standardized processes to process data. Addressing the five major problems inherent in traditional data acquisition methods, this invention provides corresponding solutions:

[0039] 1. Addressing the issue of data timing asynchrony: An industrial PC is used as the unified sampling master controller. All RS422 / RS485 / USB sensors are uniformly triggered for sampling by the industrial PC. A high-precision timer and a unified clock are used in the built-in Python acquisition program to align timestamps, and the sampling thread executes at 100Hz. The software architecture adopts a multi-threaded / time-driven model. After acquisition, each sensor data packet is appended with a high-precision timestamp by the acquisition thread, and the main program thread writes data synchronously at the same frequency according to a time window (10ms), recording the status word of each channel for data validity judgment.

[0040] 2. Addressing the risks associated with contact-based measurements: Replace sensors directly mounted on the rotor or high-risk areas with non-contact measurement methods (laser displacement, non-contact angular displacement, high-definition cameras, etc.), and establish a transmission model through ground calibration to calculate parameters such as the pitch angle. During the ground static calibration phase, the mapping relationship between known inputs (control stick / booster displacement) and rotor hub tilt (tilt meter) is established, and the transmission matrix is ​​fitted using least squares or SVD. During flight, only the input end displacement is collected, and the pitch angle is calculated through the model, avoiding wiring on the rotor.

[0041] 3. To address the limited acquisition of instrument and engine parameters: A camera mounted on the instrument panel, combined with OCR / visual recognition algorithms, is used to extract key parameters (speed, torque, fuel consumption, temperature, etc.) in real time. A lightweight inference model (such as a combination of YOLO / Tesseract or lightweight CNN regression) is deployed at the edge of the industrial control computer, combined with benchmark tracking (using dial feature lines or circular circumscribed rectangles) to improve anti-jitter capability. Kalman filtering is used to smooth the output when necessary.

[0042] 4. Addressing the issue of insufficient real-time monitoring and alarm capabilities: A real-time channel is established between the front-end PC and the industrial control computer. The industrial control computer implements an anomaly detection module (threshold detection, status word check, missing frame detection, etc.) at the acquisition layer, reporting anomaly information along with a timestamp to the front end and triggering audible and visual alarms or log alerts. Anomaly determination supports configurable strategies (static threshold, moving average, robust determination with noise), and provides playback and comparison tools on the interface for easy engineering backtracking analysis.

[0043] 5. Regarding system integration and scalability: A modular hardware access specification is defined (standard RS422 / RS485 / USB interfaces and a unified frame protocol). The industrial control computer provides a driver abstraction layer and configuration files, allowing adaptation to different helicopter models through configuration. Sensor description files are provided, including signal channels, calibration coefficients, and status word definitions. The industrial control computer automatically loads and parses the rules after reading these files, reducing the secondary development costs for each deployment.

[0044] The method of this invention uses an industrial control computer as the core processing unit and is compatible with multiple types of general-purpose sensors, including:

[0045] Displacement sensor: used to measure the travel of the joystick and booster;

[0046] Angle sensor: used to measure the rotation angle of the joystick;

[0047] Inertial navigation system: used to measure pose parameters;

[0048] Atmospheric data sensor: used to measure airspeed and air pressure parameters;

[0049] Radio altimeter: used for low-altitude altitude measurement;

[0050] Image acquisition equipment: used for instrument monitoring.

[0051] The data interface supports multiple industry standard protocols (such as RS422 / RS485) and high-speed interfaces (such as USB), ensuring seamless aggregation of multi-source data. The processing unit is equipped with a large-capacity storage module and a wide-voltage power supply adapter module to adapt to airborne environments. The power supply system adopts a general-purpose battery solution and is equipped with overcurrent protection devices to ensure safety and adaptability.

[0052] Example 1: Ground Test Implementation Steps

[0053] like Figure 1 As shown, the ground testing phase mainly involves sensor calibration, transmission relationship modeling, and system verification, laying the foundation for flight testing.

[0054] Step 1: Confirm the data collection target and system deployment

[0055] Based on the data required for ground testing, deploy the corresponding sensors on the helicopter:

[0056] Laser displacement sensors were used to measure the displacement of the three boosters of a helicopter.

[0057] Use a non-contact angular displacement sensor to measure the longitudinal angle of the pitch lever, the collective pitch lever angle, and the foot pedal angle.

[0058] Use an inclinometer to measure the tilt angle of the rotor hub and the angle of the tail rotor blades.

[0059] Use vernier calipers to measure the stroke of the pedal drive lever;

[0060] The specific locations where the sensors are deployed include:

[0061] The displacement sensor is installed in the linkage mechanism between the joystick and the booster;

[0062] The angle sensor is fixed to the joystick's rotation axis;

[0063] The inclinometer is positioned near the rotor hub;

[0064] All sensors are connected to the data processing unit via standard industrial interfaces (such as RS485 / RS422). Except for the inclinometer which measures the pitch angle (collective pitch angle) of each blade and the vernier caliper which measures the travel of the pedal drive rod, all other sensors output RS485 standard signals, and the data is collected on the industrial control computer via serial communication.

[0065] Step 2: System Initialization and Calibration

[0066] Start the data processing unit and load the sensor drivers and configuration parameters. Perform a system self-test to verify the communication status of each sensor and the validity of the data. Set the sampling frequency to 100Hz to ensure high-frequency data acquisition requirements.

[0067] Environmental calibration is performed under static conditions, and sensor baseline values ​​are recorded to eliminate environmental interference. Random errors are reduced by averaging multiple measurements.

[0068] Step 3: Data Acquisition and Calibration Process

[0069] A step-by-step testing method using progressive manipulation input is employed.

[0070] First, fix the non-test channels (e.g., lock the horizontal and vertical joysticks);

[0071] Perform step-by-step operation on the test channel (such as the collective pitch lever) from minimum to maximum travel;

[0072] Each travel point is held stably for 5-10 seconds, and displacement, angle and tilt data are collected simultaneously.

[0073] The data collected covers longitudinal, lateral, collective pitch, and foot-operated displacement and angle data, as well as parameters such as blade pitch angle. A custom algorithm is used to analyze the sensor data and calculate the transmission ratio and dynamic characteristics.

[0074] Step 4: Transmission ratio calculation and modeling

[0075] Based on the collected data, a mapping relationship between the control input and the blade pitch angle is established. A linear model is fitted using the least squares method to calculate the transmission ratio. Specifically, this includes the calculation of the collective pitch ratio, longitudinal transmission ratio, lateral transmission ratio, and pedal transmission ratio.

[0076] The joystick position-pitch angle relationship test was performed by deriving the joystick (collective pitch stick, control stick) transfer function using the following method:

[0077] Collective Distance Midpoint Measurement: First, fix the collective distance, longitudinal, and lateral operating lever positions with pins. This position is recorded as the midpoint of the three operating values, yielding the displacement L of the three hydraulic boosters A, B, and C. A0 L B0 L C0The blade pitch angles in four directions are obtained, and the algebraic average of the four pitch angles is calculated, which is the total pitch value at the blade root.

[0078] Collective pitch ratio measurement: With the longitudinal and lateral control levers retained (i.e., no longitudinal or lateral periodic pitch changes), remove the collective pitch center pin. Move the collective pitch lever from its lowest to its highest position. According to the collective pitch transmission structure, the collective pitch lever displacement controls the rotor collective pitch change. All boosters have the same displacement. Take any booster (for example, the left front booster A) and obtain the lowest and highest input displacements L of the booster. Acmin and L Acmax The pitch angle ψ of the corresponding blade A Acmin , ψ Acmax The total pitch ratio k can be obtained. col :

[0079]

[0080] Longitudinal transmission ratio measurement: With the collective pitch and lateral control lever pins retained (i.e., no collective pitch and lateral periodic pitch change), remove the longitudinal center pin. Move the control lever from the foremost position to the rearmost position. According to the longitudinal pitch change transmission structure, the longitudinal pitch change is transmitted by the left front booster A. Take the minimum and maximum value L corresponding to the displacement of the left front booster A and the foremost and rearmost positions of the control lever. Almin and L Almax The pitch angle ψ of the corresponding blade A Almin , ψ Almax The longitudinal transmission ratio k can be obtained. lon :

[0081]

[0082] Lateral transmission ratio measurement: With the collective pitch and longitudinal control lever pins retained (i.e., no collective pitch and longitudinal periodic pitch change), remove the lateral center pin. Move the control lever from the leftmost position to the rightmost position. According to the lateral pitch change transmission structure, the longitudinal pitch change is transmitted by the left rear booster B and the right front booster C. The displacements of the left rear booster B and the right front booster C are equal in magnitude and opposite in direction. Take the minimum and maximum value L corresponding to the leftmost and rightmost positions of the control lever. Blmin and L Blmax The pitch angle ψ of the corresponding blade B Blmin , ψ Blmax The longitudinal transmission ratio k can be obtained. lat :

[0083]

[0084] Arbitrary combination manipulation to calculate blade collective pitch and periodic pitch: Remove all pins and perform arbitrary combination manipulation to obtain the displacement of the three boosters as L. AL B L C The changes in the stroke of the three boosters are expressed as follows:

[0085]

[0086] collective pitch, longitudinal and lateral control stroke of the booster , , :

[0087]

[0088] Changes in total pitch, longitudinal periodic pitch, and lateral periodic pitch of the blades , , :

[0089]

[0090] Pedal gear ratio measurement: Move the left pedal from the foremost position to the rearmost position, and obtain the angle θ between the foremost and rearmost positions of the pedal. min θ max The minimum and maximum values ​​of the booster stroke L correspond to the following: pmin L pmax And the corresponding tail rotor deflection angle , The pedal-assist ratio k can be obtained. ph and pedal-tail propeller ratio k pt for:

[0091]

[0092] Step 5: Dynamic Characteristic Testing

[0093] Under normal operating conditions of the hydraulic system, a rapid control input (such as a dipole signal) is applied, and the booster response curve is collected. The dynamic characteristics of the servo motor are approximated as a first-order system model, and its transfer function is expressed as:

[0094]

[0095] The time constant T is obtained by fitting experimental data to evaluate the dynamic performance of the system. The test is repeated 3-5 times to ensure the reliability of the results.

[0096] Step 6: Data Validation and Storage

[0097] The data processing unit performs range checks and abrupt change detection on the collected data, automatically marking abnormal data and triggering re-collection. For example, when displacement data exceeds a preset range, the system records an anomaly log.

[0098] The verified data is stored in a timestamp-aligned text format, with each line containing synchronization data from all sensor channels. An automatic backup mechanism is configured in the storage path to prevent data loss.

[0099] In ground testing, data processing for the laser displacement sensor and angular displacement sensor was implemented on an industrial control computer using Python code. The industrial control computer synchronously reads the buffers of all serial ports and verifies the data according to the corresponding sensor's verification method. Data that passes verification is then decoded. A QTimer is used to control the storage thread to store data at a fixed frequency of 100Hz. Each line of the storage file represents data from a single storage session, with spaces used as separators to facilitate importing into Excel for subsequent data analysis and modeling.

[0100] In addition to storing all the collected objects, the parameters and status parameters of the collected objects stored at the same frequency are also packaged into a data list and sent to the laptop at the same frequency via the 422 communication protocol for display on the front-end interface of the laptop.

[0101] Example 2: Flight Test Implementation Steps

[0102] like Figure 2 As shown, the flight test phase verifies the system performance in a real flight environment, covering multi-dimensional data collection.

[0103] Step 1: Pre-flight preparation

[0104] A comprehensive check was performed to ensure the sensors were securely installed, the power supply status was correct, and the communication links were functioning properly. The data processing unit ran diagnostic programs to confirm that all modules were working correctly.

[0105] Configure data acquisition parameters according to the flight subject (e.g., steady-state flight, maneuvering flight). For example, set the inertial navigation system output frequency to 100Hz and the atmospheric data sensor sampling interval to 10ms.

[0106] Activate redundant sensors and backup power supplies to ensure stable operation of the system under harsh environments such as vibration and temperature changes.

[0107] Step 2: Confirm the data collection target for the flight test

[0108] The data collected during flight testing mainly includes:

[0109] Control inputs: lateral cyclic pitch lever input, longitudinal cyclic pitch lever input, total pitch lever input, and pedal control input;

[0110] Attitude and position changes: latitude and longitude, GPS altitude, forward velocity, lateral velocity, vertical velocity, roll angle, pitch angle, yaw angle, angular rate, acceleration, etc.

[0111] Instrument parameters: fuel quantity, T4 temperature, rotor speed, rotor torque;

[0112] Step 3: Data Collection Solution

[0113] Based on the data collection targets of the flight test, a multi-source sensor data collection scheme for helicopters was designed:

[0114] Laser displacement sensors are used to measure the displacement of the three boosters of the helicopter in order to calculate the lever position.

[0115] The angular displacement of the pedals is measured using a non-contact angular displacement sensor;

[0116] A combined navigation system consisting of fiber optic strapdown inertial navigation and satellite navigation box is used to measure attitude parameters;

[0117] Atmospheric parameters are measured using atmospheric data sensors;

[0118] Use a radio altimeter to measure radio altitude;

[0119] High-definition cameras are used to monitor helicopter instrument parameters;

[0120] Except for the high-definition camera which uses a USB interface, all other sensors output 422 standard signals and collect the data on the industrial control computer via serial communication.

[0121] Step 4: Real-time data acquisition process

[0122] Multi-source data synchronization is achieved by combining hardware triggering and software timing. All sensors receive a unified clock signal, and the data processing unit uses a high-precision timer (error less than 1ms) to achieve data alignment.

[0123] The collected content includes:

[0124] Input data: Real-time acquisition of displacement and angle values ​​of longitudinal, lateral, total distance and foot pedal operation, and back-calculation of actual operation amount through ground calibration model;

[0125] Attitude and environmental data: The inertial navigation system outputs position, velocity, acceleration, and angular velocity; atmospheric data sensors provide parameters such as altitude, airspeed, and angle of attack; and a radio altimeter measures low-altitude altitude.

[0126] Instrument data: Video from the instrument panel is captured by an image acquisition device, and parameters such as fuel quantity, engine temperature, and rotor speed are extracted using an optical character recognition algorithm;

[0127] Anomaly Handling: Real-time monitoring of data quality; if a sensor fails or data is abnormal, automatically switch to a backup channel or use historical data interpolation.

[0128] Step 5: Data Processing

[0129] Multi-source data processing during flight testing was implemented using Python code on an industrial control computer. The industrial control computer synchronously reads the buffers of all serial ports and performs data verification according to the verification methods of the corresponding sensors. The data that passes verification is then decoded.

[0130] Meanwhile, QTimer is used to control the storage thread to store data at a fixed frequency of 100Hz. Each line of the storage file represents the data stored in a single time, and the data is separated by spaces, which facilitates importing into Excel for subsequent data analysis and modeling.

[0131] In addition to storing all data collected, the system also stores sensor status words, GPS positioning time, GPS satellite count, and other status parameters at the same frequency to determine data validity. The stored data object parameters and status parameters are packaged into a data list and transmitted to the laptop via the 422 communication protocol for display on the laptop's front-end interface.

[0132] Step 6: Data Fusion and Pose Parameter Acquisition

[0133] Integrating inertial navigation system and satellite navigation data, the system fuses position, velocity, acceleration, and angular velocity information using Kalman filtering to output high-precision attitude parameters. The data processing unit performs time alignment, format unification, and redundancy removal on multi-source data, and employs Kalman filtering technology to fuse inertial and satellite navigation data, thereby improving attitude measurement accuracy.

[0134] Step 7: Machine Vision Applications and Vibration-Resistant OCR Processing

[0135] Optical character recognition technology is used to automatically extract instrument parameters (such as rotor speed and fuel quantity). To address image jitter, a reference point localization algorithm is designed—a stable reference point is extracted through image features (such as the dashboard outline or reference lines), and then the data region is segmented based on the offset to improve recognition robustness.

[0136] The vibration-resistant OCR system executes the following four processing stages sequentially:

[0137] Image preprocessing: Preprocessing the acquired instrument images;

[0138] Benchmark point tracking: Design a benchmark point localization algorithm;

[0139] For the first dashboard, a straight line is fitted by scanning the edge point set of the horizontal and vertical white lines and using the intersection point as the reference point;

[0140] For the second instrument panel, the circular outline is extracted by edge detection, and the centroid of its circumscribed rectangle is calculated as the reference point.

[0141] Character region segmentation: Data regions are segmented based on offsets;

[0142] Deep learning recognition: Automatically extract instrument parameters using optical character recognition technology.

[0143] Step 8: Real-time monitoring and front-end display

[0144] The processing unit packages the data and sends it to the display terminal. The monitoring interface dynamically displays parameter curves and status indicators (such as navigation mode and data validity), supporting real-time adjustment of test subjects.

[0145] The front-end interface display functionality is implemented on a laptop using a Python program developed based on PyQt5. The front-end layout is designed according to the requirements of flight test subjects.

[0146] The interface contains six display areas for sensor data, showing the current value and data curve (the latest 1000 sets of data) for each sensor.

[0147] The latest 10,000 sets of data are retained in the buffer data of the display interface, and the display range can be selected by dragging the scroll bar.

[0148] The interface contains five tabs: longitudinal test, lateral test, vertical test, heading test, and global parameters.

[0149] The four interfaces—longitudinal test, lateral test, vertical test, and heading test—are used to display important parameter information in real time in the form of dynamic waveforms during the frequency sweep maneuver input response test of the corresponding channel.

[0150] The status of the inertial navigation / satellite navigation system, atmospheric engine, and radio altimeter is monitored to ensure the validity of the test data.

[0151] The global parameter interface is used to display all the parameters of the collected objects in real time. The positioning and navigation function is set on the right side of the interface. The location and distance of the takeoff point are calculated in real time based on the latitude and longitude information to guide the helicopter back.

[0152] A save button is located at the bottom of the interface for customizing the storage of test data. Pressing the save button before and after the start and end of a flight test generates data for that specific test with a timestamp. Abnormal data filtering is implemented at the code level. Based on the sensor status words and the normal range of key parameters, abnormal data is identified and replaced with the previous set of data.

[0153] Step 9: Data Storage and Export

[0154] Data is stored in segments, with each flight course generating an independent file. Real-time export to standard formats (such as TXT and CSV) is supported for easy subsequent analysis. Standardized data packets with quality identifiers are generated, employing a layered storage structure comprising a raw data layer, a processed data layer, and an application data layer.

[0155] The microcontroller incorporates an adaptive filtering algorithm to dynamically adjust the acquisition strategy (such as sampling rate or verification threshold) based on data quality. The power supply system uses a general-purpose battery solution and is equipped with overcurrent protection devices to ensure safety and adaptability.

[0156] To ensure the long-term stable operation of the method of this invention in complex airborne environments, the following optimizations and safety measures have been implemented at the system hardware and software levels:

[0157] Adaptive optimization mechanism: An adaptive filtering algorithm and data quality assessment module are embedded in an industrial control computer or a dedicated microcontroller. This module analyzes the signal-to-noise ratio, packet loss rate, and status word information of the data stream in real time, and can dynamically adjust the acquisition strategy accordingly. For example, it can automatically reduce the sampling rate of non-critical parameters or adjust the data verification threshold when the signal quality is poor, thereby optimizing system resource allocation while ensuring the quality of core data.

[0158] Robust Power Supply and Electrical Safety: The entire system utilizes a standard airborne battery solution with a wide voltage input range to accommodate fluctuations in helicopter power systems. To ensure absolute safety, DC circuit breakers and fuses are connected in series in the power supply circuit, providing overcurrent and short-circuit protection. This design effectively prevents safety hazards caused by circuit faults and ensures the system's adaptability and reliability in harsh airborne environments such as vibration and temperature changes.

[0159] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for synchronous acquisition of multi-source heterogeneous data in a helicopter pilot operation scenario, characterized in that, Includes the following steps: Step 1: Construct a multi-source sensor network to synchronously collect pilot control input data, helicopter attitude response data, and instrument image data; the multi-source sensor network includes at least a displacement sensing module for measuring the travel of the joystick and booster, an attitude measurement module for measuring the helicopter's position and attitude, and an image acquisition module for acquiring instrument panel video. Step 2: A multi-channel data fusion algorithm based on timestamp alignment is used to process the multi-source data collected in Step 1, and sensor noise is eliminated by Kalman filtering. Step 3: Establish a two-stage transmission ratio calibration model based on the combination of ground calibration and flight data to achieve dynamic mapping between control stick displacement / angle and rotor and tail rotor pitch angle; Step 4: Develop a vibration-resistant optical character recognition system that incorporates image stabilization processing and deep learning recognition to extract parameters such as fuel quantity, engine temperature, rotor speed, and rotor torque from instrument images; Step 5: Generate a standardized data package containing quality identifiers to support real-time monitoring of flight status and offline analysis of flight data.

2. The method according to claim 1, characterized in that, In step 1, the multi-source sensor network adopts a distributed architecture with an industrial control computer as the data processing core. Each sensor module communicates with the industrial control computer through a unified RS422 or RS485 industrial standard protocol and supports hot-swapping and online configuration.

3. The method according to claim 1 or 2, characterized in that, In step 1, the displacement sensing module includes a laser displacement sensor and a non-contact angular displacement sensor; the attitude measurement module includes a navigation system composed of a fiber optic strapdown inertial navigation system and a satellite navigation box, an atmospheric data sensor, and a radio altimeter; and the image acquisition module is a high-speed camera.

4. The method according to claim 1, characterized in that, Step 2 of the multi-channel data fusion algorithm specifically includes the following processing sub-steps: Verify the data from each port; A uniform, high-precision timestamp is added to the verified data to achieve alignment. Timestamp alignment is achieved by combining hardware trigger signals with software synchronization algorithms to ensure the timing consistency between control inputs and helicopter response data. Data compensation is performed on data channels that experience delays or anomalies.

5. The method according to claim 1, characterized in that, In step 2, sensor noise is eliminated by Kalman filtering. Specifically, this involves using the Kalman filtering algorithm to fuse data from the inertial navigation system and the satellite navigation system to improve the measurement accuracy of the helicopter's attitude parameters.

6. The method according to claim 1, characterized in that, In step 3, the transmission ratio calibration model uses the recursive least squares method to update parameters, and the update frequency is adaptively adjusted according to the flight state; the transmission ratio calibration includes the calculation of collective pitch transmission ratio, longitudinal transmission ratio, lateral transmission ratio and pedal transmission ratio.

7. The method according to claim 1, characterized in that, In step 4, the vibration-resistant optical character recognition system sequentially performs the following four processing stages: image preprocessing, reference point tracking, character region segmentation, and character recognition based on a deep learning model.

8. The method according to claim 7, characterized in that, In step 4, during the reference point tracking stage, for the first instrument panel displaying fuel quantity, engine temperature, and rotor torque parameters, the following steps are performed: scanning the edges of the horizontal and vertical reference lines in the image to form a point set, fitting the point set with a straight line, and using the intersection of the two fitted straight lines as the positioning reference point; for the second instrument panel displaying rotor speed parameters, the following steps are performed: edge detection is performed on the image to extract the circular instrument outline, finding the largest bounding rectangle of the outline, and calculating the centroid of the rectangle as the positioning reference point.

9. The method according to claim 1, characterized in that, In step 5, the standardized data packet adopts a hierarchical storage structure, including a raw data layer, a processed data layer, and an application data layer; real-time monitoring is implemented through a front-end interface based on PyQt5, which has positioning and navigation functions and can calculate and display the return direction and distance based on real-time latitude and longitude.

10. The method according to claim 1, characterized in that, The method is applicable to both ground testing and flight testing scenarios. Ground testing includes sensor calibration and transmission ratio modeling through stepped control inputs, as well as measuring the dynamic characteristics of the servo motor through dipole control inputs. Flight testing is used for multi-dimensional data acquisition and real-time monitoring in a real flight environment.

Images