Bridge impact signal recognition device

Abstract

The application discloses a bridge impact signal recognition device and relates to the technical field of image visual displacement monitoring.The device comprises a communication module, a meteorological module, a positioning module, a processing unit, a flash memory module, a laser ranging module, an image acquisition module and a target target.The target target comprises a photovoltaic cell, a timer and a motor, and is installed on a fixed pile independent of a bridge pile.The meteorological module acquires meteorological information from outside through the communication module and transmits the information to the processing unit.The device performs gradient definition index sorting and screening according to image frame sequences, reduces displacement measurement noise power caused by turbulence, cross-verify periodic ranging of the target target through the laser ranging module, realizes automatic checking and drift correction of measurement accuracy, and realizes dynamic quantitative marking of measurement data reliability by driving the processing unit to calculate data quality weights and adding the weights to each displacement record through real-time visibility values obtained by the meteorological module.

Description

Technical Field

[0001] This invention relates to the field of image visual displacement monitoring technology, specifically a bridge impact signal recognition device. Background Technology

[0002] The essence of bridge impact is a transient impact event, and its displacement signal has three characteristics that distinguish it from conventional loads: abrupt changes in displacement amplitude, extremely short response time, and subsequent free decaying vibration. The abrupt changes in displacement amplitude include step changes of several millimeters to tens of millimeters; the extremely short response time includes displacement jumps that typically occur within 0.1 to 0.5 seconds; and the subsequent free decaying vibration includes the structure oscillating at its natural frequency after the impact is removed. Therefore, the identification device does not rely on seeing the impact process but captures the above signal characteristics through high-frequency displacement monitoring. After the identification device and target are deployed, a reliable displacement abrupt signal is captured immediately after the impact. Image visual displacement monitoring technology captures continuous images of the target area on the structural surface through an optical lens, uses image matching algorithms to track the pixel offset of the target point in the image sequence, and then converts the pixel offset into a structural displacement time history curve expressed in distance units through sub-pixel positioning technology and coordinate transformation. This technology is widely used in scenarios such as bridge deflection monitoring, wind vibration monitoring of tall structures, and tunnel deformation monitoring. Its non-contact characteristics allow for long-distance dynamic displacement measurement on structures where it is difficult to install contact sensors.

[0003] The existing technology has the following drawbacks: 1. Atmospheric turbulence causes blurred imaging and pixel jitter at long distances, and the monitoring accuracy drops sharply under long-distance conditions; 2. The nominal accuracy lacks an independent verification method, and the error caused by drift during long-term operation cannot be detected; 3. Severe weather and lens condensation cause monitoring data interruption or unknown quality; 4. Dust accumulation and dirt adhesion on the target surface cause the infrared signal intensity to continuously decrease, and the manual cleaning cycle is unpredictable. Therefore, it is necessary to design a bridge impact signal recognition device to solve the above-mentioned technical problems. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a bridge impact signal identification device, which solves the problems mentioned in the background section.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a bridge impact signal identification device, comprising a communication module, a meteorological module, a positioning module, a processing unit, a flash memory module, a laser ranging module, an image acquisition module, and a target. The target includes a photovoltaic cell, a timer, and a motor. The target is installed on a fixed pile independent of the bridge piles, and the fixed pile does not directly contact the bridge. The fixed pile is 5-15 meters away from the image acquisition module. The image acquisition module acquires images of other target points from the surface of the bridge piers. The meteorological module obtains meteorological information from the outside through the communication module and transmits it to the processing unit. The laser ranging module acquires its own distance information from the target in real time and transmits it to the processing unit. The image acquisition module acquires image information of the target and other target points and transmits it to the processing unit. The power output terminal of the photovoltaic cell inside the target is connected to the power input terminal of the timer. The photovoltaic cell converts solar energy into DC power and continuously supplies power to the timer. The trigger signal output terminal of the timer is connected to the control signal input terminal of the motor. After the timer reaches the preset timing interval, it outputs a high-level trigger signal to drive the motor to rotate. The motor shaft is mechanically coupled to the brush. When the motor rotates, it drives the brush to make a complete circular motion around the target, scraping off the dust and dirt on the surface. The target does not establish an electrical connection with other modules. The image acquisition module consists of an industrial camera, a replaceable optical lens, and a lens hood. The inner wall of the lens hood is embedded with a heating wire. The image data output terminal of the image acquisition module is connected to the image data input terminal of the processing unit via a GigE or USB 3.0 data cable. The image acquisition module continuously transmits the original image frames acquired by the industrial camera to the processing unit at a fixed frame rate. The heating wire control signal input terminal of the image acquisition module is electrically connected to the control signal output terminal of the processing unit. The processing unit transmits a PWM heating power control signal to the heating wire. The laser ranging module is a phase-type laser ranging sensor. The data output terminal of the laser ranging module is connected to the ranging data input terminal of the processing unit through a serial bus. The laser ranging module transmits the measured distance value to the processing unit. The trigger signal input terminal of the laser ranging module is connected to the ranging trigger signal output terminal of the processing unit. The processing unit sends a ranging trigger command to the laser ranging module according to a preset cycle. The laser emission window of the laser ranging module is aligned with the target surface of the target. The emitted ranging laser pulse is reflected by the target surface and captured by the receiving terminal of the laser ranging module, thus completing one distance measurement. The positioning module is a satellite positioning chip. The data output end of the positioning module is connected to the location information input end of the meteorological module through an internal bus. The positioning module transmits the latitude and longitude coordinates and altitude of the system installation location to the meteorological module. The communication module is a mobile network communication unit. The first data port of the communication module is connected to the weather request output terminal of the weather module. The weather module sends a weather query request containing latitude and longitude coordinates to the weather data server through the communication module. The second data port of the communication module is connected to the weather data receiving terminal of the weather module. The communication module forwards the real-time weather data returned from the weather data server to the weather module. The third data port of the communication module is connected to the external data transmission terminal of the processing unit. The processing unit sends the warning information and identification results to the remote server through the communication module. The communication module transmits displacement data and alarm data simultaneously, rather than choosing one of the two. The location information input terminal of the meteorological module is connected to the positioning module to obtain the system's latitude and longitude coordinates. The meteorological request output terminal of the meteorological module is connected to the first data port of the communication module. The meteorological module sends a weather query request to the meteorological data server through the communication module. The meteorological data receiving terminal of the meteorological module is connected to the second data port of the communication module. The communication module forwards the real-time weather data returned by the server to the meteorological module. The processing result output terminal of the meteorological module is connected to the meteorological data input terminal of the processing unit. The meteorological module transmits the parsed visibility value, temperature value, and relative humidity value to the processing unit. The flash memory module is a high-capacity NAND Flash storage chip. The data bus of the flash memory module is bidirectionally connected to the storage control interface of the processing unit. The processing unit writes the acquired original image frames, the extracted target infrared spot image area, the laser ranging record and the intermediate displacement calculation results into the flash memory module. The processing unit reads historical image data and calibration parameter files from the flash memory module. The processing unit is an industrial-grade embedded computer that runs image information filtering algorithms, normalized cross-correlation matching algorithms, Gaussian surface sub-pixel fitting algorithms, coordinate transformation algorithms, drift correction algorithms, adaptive multi-frame accumulation noise reduction algorithms, and heating power calculation programs. The processing unit includes an image data input terminal, a ranging data input terminal, a ranging trigger signal output terminal, a meteorological data input terminal, a control signal output terminal, an external data transmission terminal, and a storage control interface. The image data input terminal is connected to the image acquisition module, the ranging data input terminal is connected to the laser ranging module, the ranging trigger signal output terminal is connected to the laser ranging module, the meteorological data input terminal is connected to the meteorological module, the control signal output terminal is connected to the heating wire of the image acquisition module, the external data transmission terminal is connected to the third data port of the communication module, and the storage control interface is connected to the data bus of the flash memory module. The processing unit executes an image recognition program to obtain physical displacement values ​​with additional timestamps. The processing unit uses a laser ranging module to perform target drift correction. The processing unit dynamically weights the acquired physical displacement values ​​according to meteorological information. It intelligently heats the lens of the image acquisition module to reduce condensation. In low-light environments, the processing unit adopts adaptive multi-frame accumulation. The target executes a timed self-cleaning program. The processing unit uploads the physical displacement values ​​to the outside world through a communication module, realizing remote unmanned monitoring of the recognition device.

[0006] Further, the processing unit executes an image recognition program to obtain physical displacement values ​​with additional timestamps. This includes the processing unit performing initialization and calibration parameter loading. After the recognition device is powered on, the processing unit loads calibration parameters from the flash memory module and sends a start command to the image acquisition module. The processing unit performs image information filtering, calculating the Brenner gradient sharpness index for each frame. Frames with sharpness below the lower quartile of the circular queue are marked as turbulent blurred frames and discarded. Frames with sharpness not below the lower quartile are included in subsequent displacement calculations. After filtering, normalized cross-correlation template matching is performed. For the retained sharp frames, processing is performed... The processing unit performs normalized cross-correlation matching in the target area and within the target area using the initial frame target area as a template. The pixel coordinates with the largest correlation coefficient are taken as the integer pixel matching positions. After the matching is completed, Gaussian surface sub-pixel fitting is performed. The processing unit takes the natural logarithm of the correlation coefficient value in the 3×3 neighborhood of the integer matching position and then uses the least squares method to fit the Gaussian surface to obtain the sub-pixel peak coordinates. After the fitting is completed, coordinate transformation and displacement results are output. The processing unit multiplies the sub-pixel offset of each target point by the ratio of the calibration distance to the pixel focal length after angle projection correction, converts it into a physical displacement value, adds a timestamp, and writes it to the flash memory module. The processing unit performs image information filtering. After the recognition device is powered on, the processing unit reads the calibration parameter file from the flash memory module and loads the calibration distance value from the lens of the image acquisition module to the target surface, the horizontal angle, the vertical angle, the lens focal length, and the initial position of the target point in the image coordinate system into memory. The processing unit sends an acquisition start command to the image acquisition module. The image acquisition module starts to continuously acquire image frames at a fixed frame rate. Each frame of image is transferred to the memory ring buffer of the processing unit through the data interface in DMA mode. After the processing unit retrieves the current image frame from the circular buffer, it calculates the Brenner gradient sharpness index of the frame. Specifically, it iterates through all pixel rows and columns of the image, takes the square of the grayscale difference between two pixels horizontally at each pixel position, and accumulates the sum. The sum is the sharpness index value of the frame. The processing unit sorts the queue according to the sharpness index, marks the frames at the end of the queue as turbulent blurred frames and discards them directly, releases the memory buffer they occupy, marks the frames at the beginning of the queue as sharp frames, overwrites the oldest frame in the sorted queue, and then enters the subsequent displacement calculation process. After fitting is completed, coordinate transformation and displacement results are output. The processing unit converts the target point and the sub-pixel offset of each target point into physical displacement values. During the conversion, the processing unit multiplies the calibration distance value by the composite amount of the horizontal and vertical components of the sub-pixel offset in the image coordinate system after angle projection correction, and then divides it by the lens focal length value in pixels to obtain the physical displacement value of the target point at the current moment. The physical displacement value is then written into the displacement recording area of ​​the flash memory module after being appended with a timestamp.

[0007] Furthermore, the processing unit uses a laser ranging module to perform target drift correction. The processing unit triggers the laser ranging module to measure the target distance. When the deviation between the measured value and the calibration value continuously exceeds a threshold, a calibration alarm is generated. When the deviation does not exceed the threshold, the deviation is added as a zero-bias correction to all target displacement values. This includes the processing unit sending a ranging trigger command to the laser ranging module at fixed intervals. After receiving the trigger command, the laser ranging module emits a ranging laser pulse to the target surface, receives the reflected pulse, calculates the flight time, converts it into a distance measurement value, and returns the measurement value to the processing unit via a serial bus. The processing unit calculates the difference between the measured value and the calibration distance value stored in the flash memory module. The absolute value of the difference is... This is the current calibration deviation. When the calibration deviation continuously exceeds the nominal accuracy of the laser ranging module, the processing unit generates calibration alarm information, writes the alarm information and the current timestamp into the flash memory module, and simultaneously sends a calibration alarm data packet to the remote server through the communication module. When the calibration deviation does not exceed the threshold, the processing unit linearly superimposes the deviation value as a zero-bias correction amount into the physical displacement value of all target points. The measurement accuracy of the laser ranging module is determined by the time-of-flight measurement circuit and is not affected by atmospheric turbulence and lighting conditions. Its ranging results can serve as an independent verification benchmark for visual displacement measurement drift. The periodic cross-validation at intervals enables automatic drift detection and correction during long-term operation without significantly increasing power consumption.

[0008] Furthermore, the processing unit dynamically weights the acquired physical displacement values ​​based on meteorological information. After the positioning module acquires the latitude and longitude coordinates, it transmits them to the meteorological module. The meteorological module sends a weather query request to the meteorological server through the communication module and receives the returned real-time weather data. The meteorological module transmits the visibility value to the processing unit. The processing unit calculates the data quality weight according to the piecewise linear mapping function and appends it to the quality label field of each displacement data. This includes the positioning module continuously receiving satellite signals after power-on. After completing the initial positioning, it transmits the latitude and longitude coordinates and altitude to the meteorological module through the internal bus. The meteorological module encapsulates the received latitude and longitude coordinates into a weather query request data packet and sends it to the meteorological data server through the communication module. After receiving the real-time weather data returned by the server, the communication module forwards it to the meteorological module through the second data port. The meteorological module parses the data packet to extract the visibility value, temperature value, and relative humidity value. The above meteorological data acquisition process is repeated once at fixed intervals. The latitude and longitude coordinates provided by the positioning module enable the meteorological module to obtain accurate local weather data of the equipment installation location, rather than relying on the rough regional forecast of remote city meteorological stations, ensuring that the visibility data is consistent with the actual monitoring environment. The meteorological module transmits the parsed visibility, temperature, and relative humidity values ​​to the processing unit. The processing unit runs a data quality weight calculation program to obtain weight values, and appends the calculated weight values ​​to the quality tag field of each displacement data point within the corresponding time period. These weight values ​​are then written to the flash memory module along with the displacement data and transmitted to the remote server via the communication module. Decreased visibility means enhanced atmospheric scattering, which intensifies the attenuation of infrared light from the target on the transmission path, reduces the image signal-to-noise ratio, and consequently deteriorates the accuracy of displacement calculation. By appending dynamic weight values ​​to the data output, remote data users can assign different confidence levels to data from different time periods based on the weight values ​​in subsequent analysis.

[0009] Furthermore, the processing unit intelligently heats the lens of the image acquisition module to reduce condensation. When the air temperature is below the temperature threshold and the humidity is above the humidity threshold, the processing unit calculates the heating power proportional to the degree of temperature and humidity deviation and outputs it to the lens heating wire as a PWM signal. Under other conditions, the heating power is zero. This includes the processing unit running a heating power calculation program, obtaining the current air temperature and relative humidity values ​​from the meteorological module, and setting the temperature and humidity thresholds. When the air temperature is below the temperature threshold and the relative humidity is above the humidity threshold, the processing unit determines that there is a risk of condensation on the lens, calculates the heating power, quantifies the heating power into a PWM duty cycle signal, and transmits it to the lens heating wire drive circuit of the image acquisition module through the control signal output terminal. When the air temperature is not below the temperature threshold or the relative humidity is not above the humidity threshold, the heating power is set to zero. Lens condensation only occurs when both low temperature and high humidity conditions are met simultaneously. Under conditions with low condensation risk, heating is completely turned off to avoid unnecessary power consumption and thermal noise. Under risk conditions, the heating power is dynamically adjusted according to the degree of deviation, which prevents lens fogging and avoids excessive heating that could lead to thermal expansion of the lens and introduce additional calibration errors.

[0010] Furthermore, in low-light environments, the processing unit employs adaptive multi-frame accumulation. When visibility falls below a distance threshold, the processing unit averages the grayscale values ​​of several consecutive clear images and calculates the displacement. The number of accumulated frames increases as visibility decreases. Once visibility recovers, the accumulation mode is exited. This includes the processing unit setting a distance threshold. When the visibility value provided by the weather module is lower than the distance threshold, the processing unit automatically switches the sampling strategy to multi-frame accumulation mode. In multi-frame accumulation mode, the processing unit no longer outputs displacement values ​​frame by frame, but instead performs pixel-level grayscale averaging on several consecutive selected clear images. After accumulation, a denoised image is generated. Displacement calculation is performed on this denoised image. The processing unit uses the center time of the accumulation window as the timestamp of the displacement measurement result. When the visibility value recovers to above the distance threshold and remains above the time threshold, the processing unit exits the multi-frame accumulation mode and resumes frame-by-frame displacement calculation and effective output frequency. Under haze and low light conditions, the signal-to-noise ratio of a single frame image is insufficient to support reliable sub-pixel matching. By sacrificing temporal resolution, the signal-to-noise ratio of each frame image is improved. The number of accumulated frames increases as visibility decreases, so that the signal-to-noise ratio gain is adaptively matched with the severity of haze.

[0011] Furthermore, the target performs a timed self-cleaning program. The photovoltaic cells continuously supply power to the timer, and the timer outputs a trigger signal to drive the motor to drive the brush to clean the dust and dirt on the surface of the LED lamp array. This includes setting a countdown for the timer. The photovoltaic cells inside the target receive solar radiation during the day and convert it into DC power to continuously charge the energy storage element. When the countdown reaches zero, the timer outputs a high-level trigger signal for a continuous duration to the motor drive circuit. The motor rotates a full revolution, driving the brush to remove the dust and dirt from the surface of the target. After the trigger signal ends, the timer resets the countdown and starts the next countdown cycle. The target operates entirely on its own power. The combination of photovoltaic cells and energy storage elements can maintain the timer and motor's daily cleaning action even under extreme conditions of continuous rainy days, without the need for external wiring power supply. The fixed daily cleaning frequency can control the attenuation of infrared signal intensity due to dust accumulation to a negligible range.

[0012] Further, after the filtering is completed, normalized cross-correlation template matching is performed. This includes the processing unit performing normalized cross-correlation template matching on each clear frame after filtering, in the region where the target point is located and on each target point. The processing unit uses the target point selected in the initial frame as the matching template, slides the template pixel by pixel in the corresponding search window of the current frame and calculates the normalized cross-correlation coefficient. The numerator of the normalized cross-correlation coefficient is the inner product of the template and the mean-free gray value of the search sub-image, and the denominator is the product of the standard deviation of the template gray value and the standard deviation of the search sub-image gray value. The coefficient range is from negative one to positive one. The closer the value is to positive one, the higher the matching degree. After the processing unit traverses all pixel positions in the search window, it takes the pixel coordinate with the largest normalized cross-correlation coefficient as the integer pixel matching position. The normalized cross-correlation is not sensitive to the linear change of light intensity. Under the two extreme lighting conditions of daytime solar radiation intensity fluctuation and insufficient ambient light at night, the matching result is not affected by the overall brightness shift of the image.

[0013] Furthermore, after the matching is completed, Gaussian surface subpixel fitting is performed. This includes the processing unit obtaining the integer pixel matching position, and then performing Gaussian surface fitting in the neighborhood of that position. The processing unit takes the natural logarithm of the normalized cross-correlation values ​​of the pixels in the neighborhood, and uses the logarithmic values ​​to construct a quadratic overdetermined system of equations about the pixel coordinates. The least squares method is used to solve for the subpixel coordinates corresponding to the peak of the Gaussian surface. The difference between the subpixel coordinates and the integer pixel coordinates is the subpixel offset of the target point. The normalized cross-correlation function approximately follows a two-dimensional Gaussian distribution near the peak. Through surface fitting, the displacement measurement resolution can be improved from the single-pixel level to the 0.02 to 0.1 pixel level.

[0014] Furthermore, the port of the communication module establishes bidirectional communication with the port of the meteorological module, the output end of the positioning module is electrically connected to the input end of the meteorological module, the output end of the meteorological module is electrically connected to the input end of the processing unit, the output end of the processing unit is electrically connected to the input end of the communication module, the port of the flash memory module establishes bidirectional communication with the port of the processing unit, the output end of the image acquisition module is electrically connected to the input end of the processing unit, and the port of the processing unit establishes bidirectional communication with the port of the laser ranging module. In the target, the output terminal of the photovoltaic cell is electrically connected to the input terminal of the timer, the output terminal of the timer is electrically connected to the input terminal of the motor, and the light signal emitted by the laser ranging module is reflected back to the laser ranging module after being reflected by the target.

[0015] The present invention has the following beneficial effects: 1. Image information is acquired at a fixed frequency through the image acquisition module, and sorted and filtered according to the gradient sharpness index according to the image frame sequence. The sharp frames that are frozen by atmospheric turbulence are retained and the blurry frames are discarded, so as to achieve adaptive suppression of atmospheric turbulence interference. Compared with the visual displacement measurement system that does not introduce a frame selection mechanism, this scheme reduces the displacement measurement noise power caused by turbulence.

[0016] 2. Through periodic ranging cross-verification of the target using the laser ranging module, the processing unit compares the ranging value with the calibration distance value stored in the flash memory module and detects drift, thereby realizing automatic verification and drift correction of measurement accuracy. Compared with the visual displacement measurement system that relies solely on manual recalibration, this scheme automatically detects system drift error and adds the zero-bias correction amount to the displacement measurement result without human intervention.

[0017] 3. The real-time visibility value obtained by the meteorological module drives the processing unit to calculate the data quality weight and attach it to each displacement record, realizing the dynamic quantitative labeling of the reliability of the measurement data. Compared with the visual displacement measurement system without attaching environmental quality labels, this scheme provides remote data users with an independent basis for judging the reliability of each displacement data. The visibility value obtained by the meteorological module drives the processing unit to automatically switch the multi-frame accumulation and noise reduction mode. The number of accumulated frames increases adaptively as the visibility decreases, realizing the maintenance of the signal-to-noise ratio of displacement measurement under low visibility conditions. Compared with the visual displacement measurement system with a fixed sampling frequency and no accumulation processing, this scheme provides an equivalent signal-to-noise ratio gain when visibility is low.

[0018] 4. The temperature and humidity data obtained by the meteorological module drive the processing unit to calculate the PWM heating power of the lens heating wire, realizing automatic prevention of lens condensation and icing under low temperature and high humidity conditions. Compared with visual displacement monitoring equipment that does not perform active anti-condensation treatment, this solution automatically maintains the imaging clarity of the lens optical surface under condensation risk conditions. The photovoltaic cells inside the target independently power the timer and motor. The timer periodically drives the motor to drive the brush to clean the target surface, realizing timed self-cleaning of the target. Compared with the target maintenance method that relies on manual inspection and cleaning, this solution fixes the cleaning cycle from unpredictable manual intervals to once a day for automatic execution.

[0019] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a system block diagram of a bridge impact signal identification device according to the present invention. Detailed Implementation

[0022] 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 some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Please see Figure 1 This invention provides a technical solution: a bridge impact signal identification device, comprising a communication module, a meteorological module, a positioning module, a processing unit, a flash memory module, a laser ranging module, an image acquisition module, and a target. The target includes a photovoltaic cell, a timer, and a motor. The target is installed on a fixed pile independent of the bridge piles, and the fixed pile does not directly contact the bridge. The fixed pile is 5-15 meters away from the image acquisition module. The image acquisition module acquires the remaining target points from the surface of the bridge piers. The meteorological module obtains meteorological information from the outside through the communication module and transmits it to the processing unit. The laser ranging module acquires the distance information between itself and the target in real time and transmits it to the processing unit. The image acquisition module acquires the image information of the target and the remaining target points and transmits it to the processing unit. The processing unit executes an image recognition program to obtain physical displacement values ​​with additional timestamps. The processing unit uses a laser ranging module to perform target drift correction. The processing unit dynamically weights the acquired physical displacement values ​​based on meteorological information. It intelligently heats the lens of the image acquisition module to reduce condensation. In low-light environments, the processing unit adopts adaptive multi-frame accumulation. The target executes a timed self-cleaning program. The processing unit uploads the physical displacement values ​​to the outside world through the communication module, realizing remote unmanned monitoring of the recognition device.

[0024] The processing unit executes an image recognition program to obtain physical displacement values ​​with additional timestamps. This includes initialization and calibration parameter loading. After the recognition device is powered on, the processing unit loads calibration parameters from the flash memory module and sends a start command to the image acquisition module. The processing unit performs image information filtering, calculating the Brenner gradient sharpness index for each frame. Frames with sharpness below the lower quartile of the circular queue are marked as turbulent blurred frames and discarded. Frames with sharpness not below the lower quartile are included in subsequent displacement calculations. After filtering, normalized cross-correlation template matching is performed. For the retained sharp frames, the processing unit... Normalized cross-correlation matching is performed in the target area and within the target area using the initial frame target area as a template. The pixel coordinates with the largest correlation coefficient are taken as the integer pixel matching position. After the matching is completed, Gaussian surface sub-pixel fitting is performed. The processing unit takes the natural logarithm of the correlation coefficient value in the 3×3 neighborhood of the integer matching position and then uses the least squares method to fit the Gaussian surface to obtain the sub-pixel peak coordinates. After the fitting is completed, coordinate transformation and displacement results are output. The processing unit multiplies the sub-pixel offset of each target point by the ratio of the calibration distance to the pixel focal length after angle projection correction, converts it into a physical displacement value, and adds a timestamp to write it into the flash memory module. The processing unit performs image information filtering. After the recognition device is powered on, the processing unit reads the calibration parameter file from the flash memory module and loads the calibration distance value from the lens of the image acquisition module to the target surface, the horizontal angle, the vertical angle, the lens focal length, and the initial position of the target point in the image coordinate system into the memory. The processing unit sends an acquisition start command to the image acquisition module. The image acquisition module starts to continuously acquire image frames at a fixed frequency of 50Hz. Each frame of image is transferred to the memory ring buffer of the processing unit through the data interface in DMA mode. After retrieving the current image frame from the circular buffer, the processing unit calculates the Brenner gradient sharpness index for that frame. Specifically, it iterates through all pixel rows and columns of the image, taking the square of the grayscale difference between two pixels horizontally at each pixel location and summing the results. The sum is the sharpness index value for that frame. The processing unit maintains a sharpness index sorting queue with a length of 15 frames. Specifically, when the sharpness index of a new frame is lower than the fourth value (lower quartile) in the queue (arranged from smallest to largest), the frame at the end of the queue is marked as a turbulent blur frame and discarded directly. The memory buffer it occupies marks the frames at the front of the queue as clear frames when the sharpness index is not lower than the lower quartile. After overwriting the oldest frame in the sorting queue, it enters the subsequent displacement calculation process. This process realizes real-time filtering of atmospheric turbulence blurred frames. It should be noted that the image blur caused by atmospheric turbulence is intermittent in time. At a sampling rate of 50Hz, there are an average of 15 to 25 clear images of atmospheric turbulence frozen momentarily per second. By sorting and selecting by sharpness index rather than simple threshold filtering, it can adapt to the changes in turbulence intensity under different meteorological conditions. After fitting is completed, coordinate transformation and displacement results are output. The processing unit converts the target point and the sub-pixel offset of each target point into physical displacement values. During the conversion, the processing unit multiplies the calibration distance value by the composite amount of the horizontal and vertical components of the sub-pixel offset in the image coordinate system after angle projection correction, and then divides it by the lens focal length value in pixels to obtain the physical displacement value of the target point at the current moment. The physical displacement value is then written to the displacement recording area of ​​the flash memory module after being appended with a timestamp.

[0025] The processing unit uses a laser ranging module to perform target drift correction. Every 60 seconds, the processing unit triggers the laser ranging module to measure the target distance. If the deviation between the measured value and the calibrated value exceeds a threshold five times consecutively, a calibration alarm is generated. If the deviation does not exceed the threshold, it is added as a zero-bias correction to all target displacement values. This includes the processing unit sending a ranging trigger command to the laser ranging module every 60 seconds. Upon receiving the trigger command, the laser ranging module emits a ranging laser pulse to the target surface, receives the reflected pulse, calculates the flight time, converts it into a distance measurement value, and returns the measurement value to the processing unit via a serial bus. The processing unit then subtracts the measured value from the calibrated distance value stored in the flash memory module; the absolute value of the difference is the distance difference. The current calibration deviation is considered. When the calibration deviation exceeds three times the nominal accuracy of the laser ranging module for five consecutive times, the processing unit generates a calibration alarm message, writes the alarm message and the current timestamp to the flash memory module, and sends a calibration alarm data packet to the remote server through the communication module. When the calibration deviation does not exceed the threshold, the processing unit linearly adds the deviation value as a zero-bias correction to the physical displacement value of all target points. The measurement accuracy of the laser ranging module is determined by the time-of-flight measurement circuit and is not affected by atmospheric turbulence and lighting conditions. Its ranging results can be used as an independent verification benchmark for visual displacement measurement drift. Periodic cross-validation at 60-second intervals enables automatic drift detection and correction during long-term operation without significantly increasing power consumption.

[0026] The processing unit dynamically weights the acquired physical displacement values ​​based on meteorological information. The positioning module acquires latitude and longitude coordinates and transmits them to the meteorological module. The meteorological module sends a weather query request to the meteorological server via the communication module and receives the returned real-time weather data, repeating this process every 300 seconds. The meteorological module transmits visibility values ​​to the processing unit. The processing unit calculates data quality weights using a piecewise linear mapping function and appends them to the quality label field of each displacement data point. This includes the positioning module continuously receiving satellite signals after power-on. After completing the initial positioning, it transmits the latitude and longitude coordinates and altitude to the meteorological module via an internal bus. The module encapsulates the received latitude and longitude coordinates into a weather query request data packet and sends it to the meteorological data server through the communication module. After receiving the real-time weather data returned by the server, the communication module forwards it to the meteorological module through the second data port. The meteorological module parses the data packet to extract the visibility value, temperature value and relative humidity value. The above meteorological data acquisition process is repeated every fixed time interval of 300 seconds. The latitude and longitude coordinates provided by the positioning module enable the meteorological module to obtain accurate local weather data of the equipment installation location, rather than relying on the rough regional forecast of remote city meteorological stations, to ensure that the visibility data is consistent with the actual monitoring environment. The meteorological module transmits the parsed visibility, temperature, and relative humidity values ​​to the processing unit. The processing unit runs a data quality weight calculation program to obtain weight values. When the visibility value is greater than 5 kilometers, the data quality weight is set to 1.0. When the visibility value is greater than 1 kilometer but less than 5 kilometers, the data quality weight is set to 0.3 plus 0.14 multiplied by the linear mapping result of the visibility value. The weight decreases linearly from 1.0 to 0.44 as visibility decreases. When the visibility value is less than 1 kilometer, the data quality weight is set to 0. The calculated weight values ​​are appended to the quality tag field of each displacement data point within the corresponding time period. This data is written to the flash memory module along with the displacement data and transmitted to the remote server via the communication module. Decreasing visibility means enhanced atmospheric scattering, increased attenuation of infrared light from the target on the transmission path, reduced image signal-to-noise ratio, and consequently, deterioration of displacement calculation accuracy. By appending dynamic weight values ​​to the data output, remote data users can assign different confidence levels to data from different time periods based on the weight values ​​in subsequent analysis.

[0027] The processing unit intelligently heats the lens of the image acquisition module to reduce condensation. When the temperature is below the 5°C temperature threshold and the humidity is above the 85% humidity threshold, the processing unit calculates the heating power proportional to the temperature and humidity deviation and outputs it to the lens heating wire as a PWM signal. Under other conditions, the heating power is zero. This includes the processing unit running a heating power calculation program, obtaining the current temperature and relative humidity values ​​from the meteorological module, and setting temperature and humidity thresholds. When the temperature is below the 5°C temperature threshold and the relative humidity is above the 85% humidity threshold, the processing unit determines that there is a risk of condensation on the lens and calculates the heating power as the maximum rated power multiplied by the temperature deviation. The product of the normalized difference of 5 degrees Celsius and the normalized difference of humidity deviation of 85% is used to quantify the heating power into a PWM duty cycle signal. This signal is transmitted to the lens heating wire drive circuit of the image acquisition module through the control signal output terminal. When the air temperature is not lower than the temperature threshold of 5 degrees Celsius or the relative humidity is not higher than the humidity threshold of 85%, the heating power is set to zero. Lens condensation only occurs when both low temperature and high humidity conditions are met simultaneously. Under conditions with low condensation risk, heating is completely turned off to avoid unnecessary power consumption and thermal noise. Under risk conditions, the heating power is dynamically adjusted according to the degree of deviation, which prevents lens fogging and avoids excessive heating that could lead to lens thermal expansion and introduce additional calibration errors.

[0028] In low-light environments, the processing unit employs adaptive multi-frame accumulation. When visibility falls below a distance threshold of 2 kilometers, the processing unit averages the grayscale values ​​of several consecutive clear images and calculates the displacement. The number of accumulated frames increases as visibility decreases. Once visibility recovers, the accumulation mode is exited. This includes situations where the processing unit sets a distance threshold of 2 kilometers. When the visibility value provided by the weather module is lower than the distance threshold, the processing unit automatically switches the sampling strategy to multi-frame accumulation mode. In multi-frame accumulation mode, the processing unit no longer outputs displacement values ​​frame by frame. Instead, it averages the grayscale values ​​of several consecutive selected clear images at the pixel level, and the number of accumulated frames is the visibility value divided by 10. The larger of the quotient rounded up and 2 is accumulated to generate a denoised image. Displacement calculation is performed on this denoised image. The processing unit uses the center time of the accumulation window as the timestamp of the displacement measurement result. When the visibility value recovers to above the distance threshold and is maintained for more than 600 seconds, the processing unit exits the multi-frame accumulation mode and resumes frame-by-frame displacement calculation and 50Hz effective output frequency. Under haze and low light conditions, the signal-to-noise ratio of a single frame image is insufficient to support reliable sub-pixel matching. By sacrificing temporal resolution, the signal-to-noise ratio of each frame image is improved. The number of accumulated frames increases as visibility decreases, so that the signal-to-noise ratio gain is adaptively matched with the severity of haze.

[0029] The target device performs a timed self-cleaning procedure. The photovoltaic cells continuously power the timer, which outputs a 3-second trigger signal every 86,400 seconds to drive the motor and brush to clean the dust and dirt on the surface of the LED array. The timer countdown is set to 86,400 seconds. The photovoltaic cells inside the target device receive solar radiation during the day and convert it into DC power to continuously charge the energy storage element. When the countdown reaches zero, the timer outputs a high-level trigger signal for 3 seconds to the motor drive circuit. The motor rotates a full revolution, driving the brush to remove the dust and dirt from the target device surface. After the trigger signal ends, the timer resets the countdown and starts the next countdown cycle. The target device operates entirely on its own power. The combination of photovoltaic cells and energy storage elements can maintain the timer and motor's daily cleaning action even under extreme conditions of seven consecutive rainy days, without the need for external wiring power supply. The fixed daily cleaning frequency can control the attenuation of infrared signal intensity due to dust accumulation to a negligible range.

[0030] After screening, normalized cross-correlation template matching is performed. This involves the processing unit performing normalized cross-correlation template matching on each clear frame after screening, within the target area and on each target point. The processing unit uses the target points selected in the initial frame as the matching template, slides the template pixel by pixel within the corresponding search window of the current frame, and calculates the normalized cross-correlation coefficient. The numerator of the normalized cross-correlation coefficient is the inner product of the template and the mean-free grayscale value of the search sub-image, and the denominator is the product of the standard deviation of the template grayscale and the standard deviation of the search sub-image grayscale. The coefficient ranges from -1 to +1, with a value closer to +1 indicating a higher matching degree. After traversing all pixel positions in the search window, the processing unit takes the pixel coordinates with the largest normalized cross-correlation coefficient as the integer pixel matching position. Normalized cross-correlation is not sensitive to linear changes in light intensity. Under extreme lighting conditions, such as daytime solar radiation intensity fluctuations and insufficient ambient light at night, the matching result is not affected by the overall brightness shift of the image.

[0031] After matching is completed, Gaussian surface subpixel fitting is performed. This includes the processing unit obtaining the integer pixel matching position, and then performing Gaussian surface fitting in the 3×3 neighborhood of that position. The processing unit takes the natural logarithm of the normalized cross-correlation values ​​of the 9 pixels in the neighborhood, and uses the logarithmic values ​​to construct a quadratic overdetermined system of equations about the pixel coordinates. The least squares method is used to solve for the subpixel coordinates corresponding to the peak of the Gaussian surface. The difference between the subpixel coordinates and the integer pixel coordinates is the subpixel offset of the target point. The normalized cross-correlation function approximately follows a two-dimensional Gaussian distribution near the peak. Through surface fitting, the displacement measurement resolution can be improved from the single pixel level to the 0.02 to 0.1 pixel level.

[0032] Specifically, the communication module's port establishes bidirectional communication with the meteorological module's port; the positioning module's output is electrically connected to the meteorological module's input; the meteorological module's output is electrically connected to the processing unit's input; the processing unit's output is electrically connected to the communication module's input; the flash memory module's port establishes bidirectional communication with the processing unit's port; the image acquisition module's output is electrically connected to the processing unit's input; and the processing unit's port establishes bidirectional communication with the laser ranging module's port. In the target, the output of the photovoltaic cell is electrically connected to the input of the timer, the output of the timer is electrically connected to the input of the motor, and the light signal emitted by the laser ranging module is reflected back to the laser ranging module after being reflected by the target.

[0033] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A bridge impact signal identification device, comprising a communication module, a meteorological module, a positioning module, a processing unit, a flash memory module, a laser ranging module, an image acquisition module, and a target, characterized in that: The target includes a photovoltaic cell, a timer, and a motor. The target is installed on a fixed pile independent of the bridge pile. The meteorological module acquires meteorological information from the outside through the communication module and transmits it to the processing unit. The laser ranging module collects the distance information of the target in real time. The image acquisition module acquires image information of the target and other target points and transmits it to the processing unit. The processing unit executes an image recognition program to obtain physical displacement values ​​with additional timestamps. The processing unit uses a laser ranging module to perform target drift correction. The processing unit dynamically weights the acquired physical displacement values ​​according to meteorological information. It intelligently heats the lens of the image acquisition module to reduce condensation. In low-light environments, the processing unit adopts adaptive multi-frame accumulation. The target executes a timed self-cleaning program. The processing unit uploads the physical displacement values ​​to the outside world through a communication module.

2. The bridge impact signal identification device according to claim 1, characterized in that, The processing unit executes an image recognition program to obtain physical displacement values ​​with additional timestamps, including the processing unit performing initialization and calibration parameter loading, the processing unit performing image information filtering, after filtering, performing normalized cross-correlation template matching, after matching, performing Gaussian surface subpixel fitting, and after fitting, performing coordinate transformation and outputting displacement results. The processing unit reads the calibration parameter file from the flash memory module, loads the calibration distance value from the image acquisition module to the target surface, the horizontal angle, the vertical angle, the lens focal length, and the initial position of the target point in the image coordinate system into memory, and sends an acquisition start command to the image acquisition module. The image acquisition module starts to continuously acquire image frames at a fixed frame rate, and each frame of image is transmitted to the memory ring buffer of the processing unit through the data interface. After the processing unit retrieves the current image frame from the circular buffer, it calculates the gradient sharpness index of the frame and sorts the queue according to the sharpness index. The processing unit converts the target point and the sub-pixel offset of each target point into a physical displacement value. During the conversion, the processing unit multiplies the calibration distance value by the composite amount of the horizontal and vertical components of the sub-pixel offset in the image coordinate system after angle projection correction, and then divides it by the lens focal length value in pixels to obtain the physical displacement value of the target point at the current moment. The physical displacement value is then written into the displacement recording area of ​​the flash memory module after being appended with a timestamp.

3. The bridge impact signal identification device according to claim 1, characterized in that, The processing unit uses a laser ranging module to perform target drift correction. This includes the processing unit sending a ranging trigger command to the laser ranging module at fixed intervals. Upon receiving the trigger command, the laser ranging module emits a ranging laser pulse to the target surface, receives the reflected pulse, calculates the flight time, converts it into a distance measurement value, and returns the measurement value to the processing unit via a serial bus. The processing unit calculates the difference between the measurement value and the calibrated distance value stored in the flash memory module. The absolute value of the difference is the current calibration deviation. When the calibration deviation continuously exceeds the nominal accuracy of the laser ranging module, the processing unit generates a calibration alarm message, writes the alarm message and the current timestamp into the flash memory module, and simultaneously sends a calibration alarm data packet to a remote server via the communication module. When the calibration deviation does not exceed the threshold, the processing unit linearly adds the deviation value as a zero-bias correction to the physical displacement value of all target points.

4. The bridge impact signal identification device according to claim 1, characterized in that, The processing unit dynamically weights the acquired physical displacement values ​​based on meteorological information. This includes the positioning module continuously receiving satellite signals after power-on, transmitting latitude and longitude coordinates and altitude to the meteorological module via an internal bus after completing the initial positioning. The meteorological module encapsulates the received latitude and longitude coordinates into a weather query request data packet and sends it to the meteorological data server via the communication module. After receiving the real-time weather data returned by the server, the communication module forwards it to the meteorological module via a second data port. The meteorological module parses the data packet to extract visibility, temperature, and relative humidity values. The above meteorological data acquisition process is repeated at fixed intervals. The meteorological module transmits the parsed visibility, temperature, and relative humidity values ​​to the processing unit. The processing unit runs a data quality weight calculation program to obtain weight values, and appends the calculated weight values ​​to the quality tag field of each displacement data in the corresponding time period. The weight values ​​are then written to the flash memory module along with the displacement data and transmitted to the remote server via the communication module.

5. A bridge impact signal identification device according to claim 1, characterized in that, The processing unit intelligently heats the lens of the image acquisition module to reduce condensation. This includes the processing unit running a heating power calculation program, obtaining the current temperature and relative humidity values ​​from the meteorological module, setting temperature and humidity thresholds, and determining that there is a risk of condensation on the lens when the temperature is below the temperature threshold and the relative humidity is above the humidity threshold. The processing unit calculates the heating power and quantifies it into a PWM duty cycle signal, which is transmitted to the lens heating wire drive circuit of the image acquisition module through the control signal output terminal. When the temperature is not below the temperature threshold or the relative humidity is not above the humidity threshold, the heating power is set to zero.

6. A bridge impact signal identification device according to claim 1, characterized in that, In low-light environments, the processing unit adopts adaptive multi-frame accumulation. This includes setting a distance threshold. When the visibility value provided by the weather module is lower than the distance threshold, the processing unit automatically switches the sampling strategy to multi-frame accumulation mode. In multi-frame accumulation mode, the processing unit no longer outputs displacement values ​​frame by frame. Instead, it performs pixel-level grayscale averaging on several consecutive clear images after filtering, accumulates them to generate a denoised image, and performs displacement calculation on this denoised image. The processing unit uses the center time of the accumulation window as the timestamp of the displacement measurement result. When the visibility value recovers to above the distance threshold and remains above the time threshold, the processing unit exits the multi-frame accumulation mode and resumes frame-by-frame displacement calculation and effective output frequency.

7. A bridge impact signal identification device according to claim 1, characterized in that, The target performs a timed self-cleaning procedure, including setting a countdown timer. The photovoltaic cells inside the target receive solar radiation during the day and convert it into DC power to continuously charge the energy storage element. When the countdown reaches zero, the timer outputs a high-level trigger signal for a continuous duration to the motor drive circuit. The motor rotates a full revolution, driving the brush to remove dust and dirt from the surface of the target. After the trigger signal ends, the timer resets the countdown and starts the next countdown cycle.

8. A bridge impact signal identification device according to claim 2, characterized in that, After the filtering is completed, normalized cross-correlation template matching is performed. This includes the processing unit performing normalized cross-correlation template matching on each clear frame after filtering, in the region where the target point is located and on each target point. The processing unit uses the target point selected in the initial frame as the matching template, slides the template pixel by pixel in the corresponding search window of the current frame and calculates the normalized cross-correlation coefficient. The numerator of the normalized cross-correlation coefficient is the inner product of the template and the mean-free gray value of the search sub-image, and the denominator is the product of the standard deviation of the template gray value and the standard deviation of the search sub-image gray value. The coefficient range is from negative one to positive one. After the processing unit traverses all pixel positions in the search window, it takes the pixel coordinates with the largest normalized cross-correlation coefficient as the integer pixel matching position.

9. A bridge impact signal identification device according to claim 2, characterized in that, After the matching is completed, Gaussian surface subpixel fitting is performed. This includes the processing unit obtaining the integer pixel matching position, and then performing Gaussian surface fitting in the neighborhood of that position. The processing unit takes the natural logarithm of the normalized cross-correlation values ​​of the pixels in the neighborhood, and uses the logarithmic values ​​to construct a quadratic overdetermined system of equations about the pixel coordinates. The least squares method is used to solve for the subpixel coordinates corresponding to the peak of the Gaussian surface. The difference between the subpixel coordinates and the integer pixel coordinates is the subpixel offset of the target point.

10. A bridge impact signal identification device according to claim 1, characterized in that, The communication module establishes bidirectional communication with the meteorological module, the output of the positioning module is electrically connected to the input of the meteorological module, the output of the meteorological module is electrically connected to the input of the processing unit, the output of the processing unit is electrically connected to the input of the communication module, the flash memory module establishes bidirectional communication with the processing unit, the image acquisition module's output is electrically connected to the processing unit's input, and the processing unit's port establishes bidirectional communication with the laser ranging module. In the target, the output terminal of the photovoltaic cell is electrically connected to the input terminal of the timer, the output terminal of the timer is electrically connected to the input terminal of the motor, and the light signal emitted by the laser ranging module is reflected back to the laser ranging module after being reflected by the target.

Images