A distributed large field of view spectral displacement measurement method and system

Abstract

The embodiment of the application provides a kind of distributed large field of view spectral displacement measurement method, comprising: target layout is carried out to structure;Laser ranging value is corrected and compensated;Initial reference origin data is acquired and calibration instruction is sent to host computer by reference station;Laser distance data is rejected to outlier and the pixel coordinates of reference target image data are distorted correction, spot center subpixel extraction, pixel coordinates are solved three-dimensionally, determine the initial space three-dimensional coordinates of target;Data is corrected and compensated by six degrees of freedom attitude compensation sensor;The coordinate data and laser distance data after correction and compensation are Kalman filtering fusion, and the optimal reference station coordinate estimation value is output;It is compared with the initial space three-dimensional coordinates after correction and compensation, and the target coordinate variation is output to determine the displacement of structure. Improve the precision, degree of automation and anti-interference ability of long-distance deformation measurement of large structure under complex working conditions.

Description

Technical Field

[0001] This invention relates to the field of building structural health monitoring technology, and in particular to a distributed large field-of-view spectral displacement measurement method and system. Background Technology

[0002] Currently, buildings, tunnels, and large structures all require long-distance measurement of multi-point dynamic displacement and deformation. However, current long-distance displacement measurement mainly relies on total stations, RTK, and machine vision technologies. Total stations require manual operation, making them difficult to adapt to the needs of automated and real-time monitoring, and their long-distance aiming accuracy is easily affected by environmental interference. RTK measurement accuracy is only at the centimeter level, the equipment cost is high, and it is significantly affected by signal blockage and electromagnetic interference. Traditional machine vision measurement mostly uses single-point monitoring, which does not fully consider the three-dimensional attitude distortion of the measuring platform, optical distortion at the edge of the field of view, and environmental stray light interference. The depth of field is limited, making it difficult to achieve high-precision long-distance measurement at the hundred-meter level. The targets are mostly passive reflective, with weak anti-interference capabilities, insufficient spot center extraction accuracy, and a lack of distributed benchmark calibration mechanism. The consistency of multi-point measurements is poor, which cannot meet the needs of real-time, full-domain synchronous deformation monitoring of large structures. Summary of the Invention

[0003] In view of this, to address the problems in existing technologies for deformation monitoring of large structures under complex working conditions, such as spot extraction errors, unstable focal length calibration, low automation, low accuracy at long distances, and weak anti-interference capabilities, making it difficult to achieve full-area, real-time, and accurate monitoring, this invention provides a distributed large-field-of-view spectral displacement measurement method and system.

[0004] Firstly, a distributed large field-of-view spectral shift measurement method is provided, including: S1. Target placement is carried out on the structure. The phase difference between the laser and the two reference stations is determined by a three-axis laser interferometer. Adjustments are made based on the phase difference. After confirming that there is no relative displacement between the target and the structure, the placement record is completed. S2. The environmental sensors read the temperature, humidity and air pressure data in real time, and use the air refractive index to correct and compensate the laser ranging value of the three-axis laser interferometer. S3. Acquire initial reference origin data and use it as a reference for subsequent dynamic measurements. Send calibration instructions to the host through the reference station. The initial reference origin data consists of the corrected laser ranging value and the reference target image data acquired by the dual-spectrum imaging camera. The reference target image data is acquired synchronously by multiple cameras controlled by the control gimbal. S4. After removing outliers from the laser distance data and correcting the distortion of the pixel coordinates of the benchmark target image data, and extracting the sub-pixel center of the spot, the pixel coordinates are solved in three dimensions by combining the three-point intersection method with the least squares method to determine the initial three-dimensional spatial coordinates of the target. S5. The initial three-dimensional spatial coordinates are corrected and compensated for in terms of three-dimensional translation and three-dimensional attitude angle using a six-degree-of-freedom attitude compensation sensor, and a global target coordinate database is established and stored on the host computer. S6. The initial three-dimensional spatial coordinate data after correction and compensation is used as the state vector, and the laser distance data is used as the observation vector. Kalman filtering is performed to fuse them together to output the optimal reference station coordinate estimate. S7. Compare the estimated optimal base station coordinates with the corrected and compensated initial three-dimensional spatial coordinates, and output the target coordinate change to determine the displacement of the structure.

[0005] Secondly, a distributed large field-of-view spectral shift measurement system is provided, applied to a distributed reference monitoring station, including: Target array, imaging module, benchmark monitoring module, and main unit; The target array is used to deploy targets on the structure, determine the laser ranging value of the laser traveling back and forth between the two reference stations, and complete the deployment record after there is no relative displacement between the target and the structure. The imaging module is used to control multiple dual-spectrum imaging cameras to simultaneously acquire reference target image data via a control gimbal. The benchmark monitoring module is used to determine the phase difference of the laser traveling back and forth between two benchmark stations using a three-axis laser interferometer, adjust the laser ranging value based on the phase difference, and transmit the laser ranging value and the benchmark target image data to the host computer via an industrial control board. The host is used to perform distortion correction, spot center sub-pixel extraction, attitude distortion compensation, environmental parameter compensation, and three-dimensional calculation to determine the initial three-dimensional coordinate data of the correction compensation for the reference target image data. It also performs outlier removal processing on the laser range value and performs Kalman filtering fusion on the processed laser distance data and the corrected initial three-dimensional coordinates to determine the optimal reference station coordinate estimate. The host computer determines the displacement of the structure by comparing the estimated coordinates of the optimal reference station with the initial three-dimensional spatial coordinate data after correction and compensation, and outputting the change in the target coordinates.

[0006] In one possible embodiment, the target array, with its target axis perpendicular to the surface of the structure, is divided into upper and lower cavities. The upper cavity is a light-emitting cavity, and the lower cavity is a power supply cavity for controlling the spatial distribution of light field intensity. The cavity is separated by an insulating epoxy partition of a preset thickness, with a wire hole of a preset diameter reserved in the center of the partition. The cavity is filled with thermally conductive silicone grease.

[0007] In one possible embodiment, the target array includes: a plurality of reference targets fixed to a reference station mounting plate by stainless steel bolts; The reference target is made of alumina ceramic.

[0008] In one possible embodiment, the triaxial laser interferometer includes: a laser emitting head fixed at the center of the reference station, a triaxial mirror assembly, and a signal processing unit; The three-axis reflectors in the three-axis reflector group are respectively installed on multiple adjacent reference stations to form X, Y, and Z three-axis measurement optical paths. During measurement, the laser emitter head emits polarized laser light, which is reflected back by the reflectors. The signal processing unit is used to calculate the distance change between reference stations by measuring the round-trip phase difference of the laser.

[0009] In one possible embodiment, the imaging module includes: a feature point matching unit and a grayscale fusion unit; The feature point matching unit is used to construct Gaussian pyramids for the multi-view images captured by the dual-spectrum imaging camera, generate scale space images, detect pixel gray-level extreme points in adjacent multi-layer images of the scale space image, remove points with contrast less than the preset contrast and points with edge response greater than the edge threshold, fit the extreme points through Taylor expansion, correct the feature point coordinates to the sub-pixel level, perform region division, and normalize the feature descriptors of each region. The grayscale fusion unit is used to filter image feature descriptors by calculating the Euclidean distance of the image feature descriptors. It is used to select a grayscale window to be matched within a preset pixel range, centered on the center feature point of the filtered image, to calculate its grayscale correlation coefficient. Then, it performs quadratic polynomial interpolation on the grayscale correlation coefficient through subpixel interpolation to solve the subpixel-level optimal matching coordinates. Based on the optimal matching coordinates, it solves the homography matrix between the image to be matched and performs matrix transformation to correct the entire image.

[0010] In one possible embodiment, the host computer includes: an environmental parameter compensation unit, a data processing unit, a parallel image processing unit, and a six-degree-of-freedom distortion compensation unit; The environmental parameter compensation unit is used to perform temperature and humidity compensation, air pressure compensation, and wind speed compensation by reading temperature, humidity, and air pressure data in real time through environmental sensors, and to correct and compensate the laser ranging value of the three-axis laser interferometer by incorporating the air refractive index. The parallel image processing unit is used to perform multi-channel separation, adaptive wavelet threshold denoising, and image enhancement on the reference target image of each dual-spectrum imaging camera according to the target emission band; it is also used to locate the connected component of the spot region of the reference target image, extract it with two-dimensional Gaussian fitting, and perform weighted fusion of the center of the multispectral spot region, real-time focal length calibration, and calculation of the horizontal and vertical displacement of the target to determine the pixel coordinates of the reference target image. The data processing unit is used to perform outlier removal processing on the laser ranging data and distortion correction processing on the reference target image data; it is also used to perform three-dimensional calculation on the pixel coordinates by combining the three-point intersection method with the least squares method to determine the initial three-dimensional spatial coordinates of the target. The six-degree-of-freedom distortion compensation unit is used to collect and correct the three-dimensional translation and three-dimensional attitude angle data in the initial three-dimensional spatial coordinate data through a six-degree-of-freedom sensor; it is also used to synchronously upload the initial three-dimensional spatial coordinate data to the host with a timestamp. The data processing unit is used to perform Kalman filtering fusion on a pre-constructed sixth-order Kalman filter model, using the corrected and compensated initial three-dimensional spatial coordinate data as the state vector and the laser distance data as the observation vector, to output the optimal reference station coordinate estimate; it is also used to compare the optimal reference station coordinate estimate with the corrected and compensated initial three-dimensional spatial coordinates to output the target coordinate change and determine the displacement of the structure.

[0011] In one possible embodiment, the data processing unit further establishes and stores an initial benchmark database based on the laser ranging data that has been filtered out and the benchmark target image data that has been distorted.

[0012] In one possible embodiment, the six-degree-of-freedom distortion compensation unit includes: a six-axis inertial sensor, a tilt sensor, and a three-dimensional accelerometer. The six-axis inertial sensor is used to collect angular velocity and acceleration data in real time, and to solve for attitude angles and integral time constants through integration calculations. The tilt sensor is used to output high-precision tilt angle data, which is fused with inertial sensor data to correct attitude angle drift. The three-dimensional accelerometer is used to collect acceleration data, and combined with the gimbal position information, the translation amount is calculated by integration, sampling frequency, and translation accuracy.

[0013] In one possible embodiment, the data processing unit includes: The Kalman filter fusion subunit calculates the process noise covariance and observation noise covariance in the sixth-order Kalman filter model based on the state vector and observation vector, and outputs the optimal reference station coordinate estimate based on iterative calculation of the process noise covariance and observation noise covariance.

[0014] This invention provides a distributed large field-of-view spectral displacement measurement method, comprising: setting up a target on the structure; correcting and compensating the laser ranging value; acquiring initial reference origin data and sending calibration commands to the host through the reference station; removing outliers from the laser distance data and correcting the distortion of pixel coordinates in the reference target image data, extracting sub-pixels at the center of the light spot, and then performing three-dimensional calculation on the pixel coordinates to determine the initial three-dimensional spatial coordinates of the target; correcting and compensating the data using a six-degree-of-freedom attitude compensation sensor; fusing the corrected and compensated coordinate data and laser distance data using Kalman filtering to output the optimal reference station coordinate estimate; and comparing this estimate with the corrected and compensated initial three-dimensional spatial coordinates to output the change in target coordinates and determine the displacement of the structure. This method improves the accuracy, automation, and anti-interference capability of long-distance deformation measurement of large structures under complex working conditions. Attached Figure Description

[0015] The accompanying drawings, which form part of this specification, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings: Figure 1 A flowchart illustrating an embodiment of a distributed large field-of-view spectral displacement measurement method provided by this invention; Figure 2 This is a block diagram of an embodiment of a distributed large field-of-view spectral displacement measurement system provided by the present invention. Detailed Implementation

[0016] The present invention will now be described in detail with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described herein can be combined with each other.

[0017] The following detailed description is exemplary and intended to provide further detailed explanation of the invention. Unless otherwise specified, all technical terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in this invention is for describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention.

[0018] See Figure 1 A flowchart illustrating an embodiment of a distributed large field-of-view spectral shift measurement method provided by this invention is shown below. Figure 1 As shown, the following steps may be included: S1. Target placement is carried out on the structure. The phase difference between the laser and the two reference stations is determined by a three-axis laser interferometer. Adjustments are made based on the phase difference. After confirming that there is no relative displacement between the target and the structure, the placement record is completed.

[0019] S2. The environmental sensors read the temperature, humidity and air pressure data in real time, and use the air refractive index to correct and compensate the laser ranging value of the three-axis laser interferometer.

[0020] S3. Acquire the initial reference origin data and use it as the reference for subsequent dynamic measurements. Send calibration commands to the host through the reference station. The initial reference origin data consists of the corrected laser ranging value and the reference target image data acquired by the dual-spectrum imaging camera.

[0021] Here, the benchmark target image data is collected synchronously by multiple cameras controlled by a control gimbal.

[0022] S4. After removing outliers from the laser distance data and correcting the distortion of pixel coordinates in the benchmark target image data, and extracting sub-pixels at the center of the light spot, the pixel coordinates are calculated in three dimensions using the three-point intersection method combined with the least squares method to determine the initial three-dimensional spatial coordinates of the target.

[0023] S5. The initial three-dimensional spatial coordinates are corrected and compensated for in terms of three-dimensional translation and three-dimensional attitude angle by a six-degree-of-freedom attitude compensation sensor, and a global target coordinate database is established and stored on the host computer.

[0024] S6. The initial three-dimensional spatial coordinate data after correction and compensation is used as the state vector, and the laser distance data is used as the observation vector. Kalman filtering is performed to fuse them together to output the optimal reference station coordinate estimate.

[0025] S7. Compare the estimated coordinates of the optimal reference station with the initial three-dimensional spatial coordinates after correction and compensation, and output the change in target coordinates to determine the displacement of the structure.

[0026] In one embodiment, the above-described distributed large field-of-view spectral shift measurement method can be implemented by a distributed large field-of-view spectral shift measurement system that executes the method. This invention is primarily described using an application to a distributed reference monitoring station as an example. See also... Figure 2 A block diagram illustrating an embodiment of a distributed large field-of-view spectral displacement measurement system provided by this invention is shown below. Figure 2 The system shown may include: a target array, an imaging module, a benchmark monitoring module, and a host computer.

[0027] The target array is used to deploy targets on the structure, determine the laser ranging value of the laser traveling back and forth between two reference stations, and complete the deployment record after there is no relative displacement between the target and the structure.

[0028] The imaging module is used to control multiple dual-spectrum imaging cameras to simultaneously acquire reference target image data via a control gimbal.

[0029] The reference monitoring module is used to determine the phase difference between the laser and the two reference stations by using a three-axis laser interferometer, adjust the laser ranging value based on the phase difference, and transmit the laser ranging value and the reference target image data to the host through the industrial control board.

[0030] The host computer is used to perform distortion correction, spot center sub-pixel extraction, attitude distortion compensation, environmental parameter compensation, and three-dimensional calculation to determine the initial three-dimensional coordinate data of the correction and compensation of the reference target image data. It also performs outlier removal processing on the laser range value, and performs Kalman filtering fusion on the processed laser distance data and the corrected initial three-dimensional coordinates to determine the optimal reference station coordinate estimate.

[0031] The host computer determines the displacement of the structure by comparing the estimated coordinates of the optimal reference station with the initial three-dimensional spatial coordinate data after correction and compensation, and outputting the change in the target coordinates.

[0032] (i) The target array may include: multiple reference targets, fixed to the base station mounting plate by stainless steel bolts. The reference targets may be made of alumina ceramic. For example, the reference targets may be made of 99% alumina ceramic, with a diameter of 60 mm, a thickness of 8 mm, a flatness of ≤0.02 mm, and a surface that has been ground and polished (roughness Ra≤0.1 μm). A crosshair calibration line is etched at the center of the target surface, with a line width of 0.18 mm, a length of 40 mm, an etching depth of 0.2 mm, and a line edge straightness of ≤0.01 mm; the intersection of the crosshairs is the reference origin, with a coordinate accuracy of ≤0.02 mm. An M8 threaded hole is provided on the back of the target, which can be fixed to the base station mounting plate by stainless steel bolts. After installation, the target plane is perpendicular to the optical axis of the imaging camera, with a perpendicularity error of ≤0.03°.

[0033] (ii) A three-axis laser interferometer may include: a laser emitting head fixed at the center of the reference station, a three-axis mirror group, and a signal processing unit.

[0034] The three-axis reflectors in the three-axis reflector group are installed on multiple adjacent reference stations to form X, Y, and Z three-axis measurement optical paths. During measurement, the laser emitter head emits polarized laser light, which is reflected back by the reflectors.

[0035] The signal processing unit is used to calculate the distance change between reference stations by measuring the round-trip phase difference of the laser.

[0036] Here, a SP-300 triaxial laser interferometer can be used, with a range of 0-200m, resolution of 0.02μm, measurement accuracy of ±0.5ppm, laser wavelength of 632.8nm (helium-neon laser), and output power of 0.8mW. The equipment includes a laser emitter, a triaxial mirror assembly, and a signal processing unit. The emitter is fixed at the center of the base station, and the triaxial mirrors are installed at three adjacent base stations, forming X, Y, and Z axis measurement optical paths. During measurement, the laser emitter emits polarized laser light, which is reflected back by the mirrors. The signal processing unit calculates the distance change between the base stations by measuring the round-trip phase difference of the laser light, using the formula:

[0037] Where λ = 632.8 nm is the laser wavelength. The phase difference is ≤0.5ms for a single measurement, and the data output frequency is 1kHz.

[0038] (iii) Imaging module, which may include: feature point matching unit and grayscale fusion unit.

[0039] The feature point matching unit is used to construct Gaussian pyramids for multi-view images captured by dual-spectrum imaging cameras, generate scale-space images, detect pixel gray-level extreme points in adjacent multi-layer images of the scale-space image, remove points with contrast less than the preset contrast and points with edge response greater than the edge threshold, fit extreme points through Taylor expansion, correct feature point coordinates to the sub-pixel level, perform region division, and normalize the feature descriptors of each region.

[0040] Here, a multi-view collaborative imaging module is used, specifically employing a square frame structure with dimensions of 120mm × 120mm × 20mm, made of 7075 aluminum alloy with a wall thickness of 3mm and a flatness of ≤0.02mm. Four imaging cameras (model IM-400) are arranged at the vertices of the square, with a center-to-center distance of 70mm between adjacent cameras. The camera lens end faces are flush with the frame plane, and the parallelism error of the camera optical axis after installation is ≤0.01°. The cameras are fixed to the frame with M4 bolts, and a 0.5mm thick PTFE gasket is added to the bottom to prevent deformation caused by metal-to-metal contact. A 10mm diameter cable pass-through hole is pre-drilled in the center of the frame for power and data cables, ensuring secure cable fixation without tensile stress.

[0041] The imaging module can be mounted on the PT-200 two-dimensional rotating gimbal. This gimbal consists of a horizontal rotation mechanism, a vertical rotation mechanism, a servo motor, a photoelectric encoder, and a control driver. The horizontal rotation angle is ±180°, the vertical rotation angle is ±90°, the rotation speed is adjustable from 0.1° / s to 10° / s, the positioning accuracy is 0.005°, and the repeatability is 0.002°. The servo motor has a horizontal torque of 2N. m, vertical motor torque 3N m, response time ≤20ms; photoelectric encoder: 23-bit resolution (0.000138° / pulse), real-time feedback of rotation angle; the driver adopts pulse control mode, pulse frequency 1MHz, high control precision. The gimbal base is fixed to a rigid bracket with M6 bolts, the bracket height is 1.5m, and the base levelness error is ≤0.01° to ensure the gimbal rotates without shaking. Scanning imaging process: the gimbal rotates horizontally for scanning, pausing for 100ms every 5° rotation, 4 cameras simultaneously acquire images, a single horizontal scan covers 360°, and the time is ≤1.8s; vertical fine-tuning is performed as needed to ensure the target image is located in the center of the image area.

[0042] The camera used is the MS-500 dual-spectrum imaging camera, equipped with a 5-megapixel CMOS sensor (IMX264) with a pixel size of 3.45μm × 3.45μm, a resolution of 2560 × 1920, a frame rate of 45fps, and global exposure mode. The lens is a zoom lens with a continuously adjustable focal length of 16-24mm, an aperture of F2.0-F16, and a field of view of 28°-42° horizontally and 21°-31° vertically. The camera incorporates a beam splitter to separate the incident light into a visible light channel (400-700nm) and an infrared light channel (750-900nm). Images from both channels are acquired synchronously, and the exposure time is controlled synchronously with an error ≤1μs. The camera is mounted on top of the base station, with the lens center at a height of 250mm and the optical axis horizontal. Calibration using a level instrument ensures a levelness error ≤0.01°.

[0043] In one embodiment, the process of feature point matching is illustrated by example: Scale space construction: Gaussian pyramids are constructed for the four images respectively, with a Gaussian kernel σ=1.6, 6 pyramid layers, and 3 layers in each group, to generate scale space images.

[0044] Extreme point detection: In three adjacent layers of images in scale space, detect pixel gray-level extreme points (8 neighborhoods for comparison + 9 neighborhoods at the upper and lower scales, for a total of 26 points), and remove low contrast points with contrast <0.04 and points with strong edge response (edge ​​threshold 10).

[0045] Precise feature point localization: Extreme points are fitted using Taylor expansion, feature point coordinates are corrected to the sub-pixel level, and points with errors > 0.03 pixels are removed.

[0046] Direction assignment: Calculate the gradient direction histogram of the neighborhood (16×16 pixels) of the feature point. The main direction is the direction of the peak value of the histogram, and the auxiliary direction is the direction of the peak value above 80%, to ensure the rotation invariance of the feature point.

[0047] Feature descriptor generation: The neighborhood of a feature point is divided into 4×4 sub-regions, the 8-direction gradient histogram of each sub-region is calculated, a 128-dimensional feature descriptor is generated, and the descriptor is output after normalization. 900-1100 feature points are extracted from each image.

[0048] The grayscale fusion unit is used to filter image feature descriptors by calculating the Euclidean distance of the image feature descriptors. It is used to select a grayscale window to be matched within a preset pixel range, centered on the center feature point of the filtered image, to calculate its grayscale correlation coefficient. Then, through subpixel interpolation, it performs quadratic polynomial interpolation on the grayscale correlation coefficient to solve the subpixel-level optimal matching coordinates. Based on the optimal matching coordinates, it solves the homography matrix between the image and the image to be matched and performs matrix transformation to correct the entire image.

[0049] In one embodiment, an example of continuing the above feature point matching in the grayscale fusion process is illustrated: Coarse matching: Using image 1 as the reference, match feature points with images 2, 3, and 4 respectively. Calculate the Euclidean distance between the feature descriptors of the reference image and the images to be matched, using the formula:

[0050] in, , These are the feature descriptors for the reference image and the image to be matched, respectively.

[0051] Mismatches are eliminated using a ratio method: the nearest neighbor distance d1 and the second nearest neighbor distance d2 are taken. If d1 / d2 ≤ 0.75, the matching pair is retained; otherwise, it is eliminated. After coarse matching, the number of matching points for each pair of cameras is ≥ 600, and the matching error is ≤ 1.8 pixels.

[0052] Fine registration: Determine the matching window. Centered on the coarse matching feature points, select a 15×15 pixel grayscale window to ensure that the window contains the complete light spot area without large-area background interference.

[0053] Gray-level correlation calculation: Calculate the gray-level correlation coefficient between the reference window and the window to be matched. Formula:

[0054] Where I1 and I2 are the pixel gray levels of the two windows, , The average gray level of the window. For each window of pixels, the peak value of the correlation coefficient corresponds to the optimal matching position.

[0055] Subpixel interpolation: Perform quadratic polynomial interpolation on the neighborhood (3×3) of the correlation coefficient peak to solve for the subpixel-level optimal matching coordinates with an interpolation accuracy of 0.01 pixels.

[0056] Coordinate transformation: Based on the precise matching point pairs, the homography matrix H between the image to be matched and the reference image is solved. The entire image is corrected through matrix transformation. Finally, the registration accuracy of the 4-eye image is ≤0.03 pixels, ensuring that the position of the light spots in the multi-eye image is highly consistent.

[0057] (iv) The host computer includes: an environmental parameter compensation unit, a data processing unit, a parallel image processing unit, and a six-degree-of-freedom distortion compensation unit.

[0058] The environmental parameter compensation unit is used to perform temperature and humidity compensation, air pressure compensation, and wind speed compensation by reading temperature, humidity, and air pressure data in real time through environmental sensors, and to correct and compensate the laser ranging value of the three-axis laser interferometer by incorporating the air refractive index.

[0059] In one embodiment, the environmental sensor can be an integrated temperature, humidity, and barometric pressure sensor, model SHT30+MS5803. Its temperature measurement range is -40℃ to 85℃, with an accuracy of ±0.1℃ and a resolution of 0.01℃; humidity measurement range is 0-100%RH, with an accuracy of ±1.2%RH and a resolution of 0.01%RH; and barometric pressure measurement range is 30-120kPa, with an accuracy of ±0.05kPa and a resolution of 0.01kPa. The sensor is installed inside the base station housing at a ventilated location, collecting environmental parameters in real time at a sampling frequency of 10Hz. The data is transmitted to the industrial control board via an I2C bus for subsequent environmental compensation calculations.

[0060] Furthermore, an example is given to illustrate the process of adaptive compensation for environmental parameters: 1. Construction of a multi-factor environmental compensation model Real-time acquisition of environmental parameters such as temperature (T), relative humidity (H), atmospheric pressure (P), and wind speed (V) is used to construct a four-factor compensation model for temperature, humidity, air pressure, and wind speed, thereby eliminating the influence of the environment on light propagation, camera parameters, and target size.

[0061] 2. Temperature compensation Camera focal length temperature compensation: Temperature changes cause the lens to expand and contract, and the formula for focal length change is:

[0062] in, =20mm (standard focal length at 25℃). =6.5×10^-6 / ℃ (focal length temperature coefficient). =25℃, T is the real-time temperature (℃).

[0063] Target size temperature compensation: thermal expansion coefficient of aluminum alloy target =2.8×10^-6 / ℃, target spacing compensation:

[0064] in, The standard spacing is 25°C.

[0065] Air refractive index temperature compensation: The compensation coefficient for the change of air refractive index with temperature. =9.2×10^-7 / ℃, refractive index: n0=1.000292 (standard refractive index at 25℃).

[0066] 3. Humidity compensation Humidity affects the refractive index of air; the compensation formula is as follows:

[0067] H represents the real-time relative humidity (%RH), and the refractive index error after compensation is ≤1.2×10-8.

[0068] 4. Air pressure compensation Air pressure affects the path length of light over long distances; compensation formula:

[0069] in, The distance is at standard atmospheric pressure (P0 = 101.325 kPa), where P is the real-time atmospheric pressure (kPa).

[0070] 5. Wind speed compensation Wind speed causes air turbulence, resulting in light spot flickering. Compensation formula: ΔuV= 0.02Vt, ΔvV= 0.015Vt Where V is the wind speed (m / s), t is the exposure time (s), and the pixel offset caused by jitter is deducted.

[0071] The parallel image processing unit is used to perform multi-channel separation, adaptive wavelet threshold denoising, and image enhancement on the reference target image of each dual-spectrum imaging camera according to the target emission band; it is also used to locate the connected components of the spot region of the reference target image, extract it with two-dimensional Gaussian fitting, and perform weighted fusion of the center of the multispectral spot region, real-time focal length calibration, and calculation of the horizontal and vertical displacement of the target to determine the pixel coordinates of the reference target image.

[0072] In one embodiment, the parallel image processing process is illustrated as follows: a multi-channel image parallel preprocessing process can be used first. Specifically, 4 eyes and 6 channels of images can be acquired, with 6 spectra per camera, for a total of 24 channels of image data. The data is transmitted to the parallel computing host via gigabit Ethernet. The host uses multi-threaded parallel processing (24 threads running synchronously). The preprocessing time for a single image is ≤0.8ms, and the total estimated time is ≤1.2ms.

[0073] 1. Multi-channel separation Based on the target's emission wavelength, the composite image from each camera is separated into six single-channel images (multicolor light): blue (455nm), green (525nm), yellow (590nm), red (660nm), near-infrared (790nm), and far-infrared (860nm). The separation algorithm is based on a spectral filtering matrix, the matrix parameters of which are determined through calibration. The inter-channel interference suppression ratio is ≥42dB, effectively avoiding crosstalk between wavelengths.

[0074] 2. Adaptive wavelet threshold denoising Wavelet basis selection: The db4 wavelet basis is selected, which has compact support, orthogonality, and smoothness, making it suitable for noise reduction of light spot images.

[0075] Wavelet decomposition: A single-channel image is decomposed into 5 layers of wavelets to obtain 1 layer of low-frequency coefficients (approximate components) and 5 layers of high-frequency coefficients (detail components). The image size is halved at each level after decomposition, while preserving image features.

[0076] Threshold determination: A hierarchical adaptive thresholding method is adopted, with a threshold of 0.10 for high-frequency coefficients (layers 1-3), a threshold of 0.03 for high-frequency coefficients (layers 4-5), and no noise reduction for low-frequency coefficients; a soft thresholding function is used.

[0077] Where w is the wavelet coefficient and T is the threshold.

[0078] Wavelet reconstruction: The inverse wavelet transform is performed on the denoised wavelet coefficients to reconstruct the denoised image, which improves the image signal-to-noise ratio by 28dB and effectively filters out ambient stray light and electromagnetic interference noise.

[0079] 3. CLAHE Image Enhancement Image segmentation: The denoised image is divided into 8×8 pixel non-overlapping sub-blocks to ensure uniform grayscale distribution within the sub-blocks.

[0080] Contrast Limit: Set the contrast limit factor to 4.0 to suppress excessive local contrast enhancement and avoid noise amplification.

[0081] Histogram equalization: Histogram equalization is performed on each sub-block, the gray-level histogram of the sub-block is calculated, the gray-level values ​​are mapped, the contrast of the spot area is enhanced, the edge of the spot is clear and the internal gray level is uniform, and the contrast between the spot and the background is improved by 35%.

[0082] 4. Otsu Adaptive Threshold Segmentation Gray-level histogram calculation: Statistically analyze the gray-level histogram (0-255 levels) of the enhanced image and calculate the number of pixels at each gray level.

[0083] Calculate the maximum inter-class variance: Iterate through all possible thresholds t (0-255) and calculate the inter-class variance σ²(t) between the background and the spot region.

[0084] in, , For background and light spot pixel ratio, , The average gray level of the background and light spots.

[0085] Determining the optimal threshold: The threshold t0 corresponding to the maximum inter-class variance is selected as the optimal segmentation threshold. Threshold segmentation generates a binary image (spot region 255, background 0), with a segmentation accuracy of ≥99.85%. The spot region is complete and without missing parts, and the background has no residual noise.

[0086] The subsequent improved centroid-Gaussian fitting fusion spot center extraction may include: 1. Spot connectivity location An 8-connected component analysis is performed on the binarized image. All connected components are labeled, and the area (number of pixels) of each connected component is calculated. Noisy connected components with an area <50 pixels are removed, and the effective connected components of the light spot are retained. The light spot contour is extracted using a contour detection algorithm, and the minimum bounding rectangle of the contour is calculated to determine the effective area of ​​the light spot (5 pixels inside the rectangle). The boundary error of the area is ≤0.8 pixels to ensure that the complete light spot is included.

[0087] 2. Coarse extraction of the center of gravity Within the effective area of ​​the light spot, calculate the pixel centroid coordinates (u_c, v_c) using the following formula:

[0088]

[0089] in, , Here, represents the column and row coordinates of the pixels within the region, Imn represents the pixel grayscale value (0-255), and M and N represent the number of pixels in the region's rows and columns. After coarse extraction, the centroid coordinate accuracy is ≤0.18 pixels, which is used as the initial value for Gaussian fitting.

[0090] 3. Two-dimensional Gaussian fitting and fine extraction Fitting region selection: based on centroid coordinates ( , Centered on the spot, a 3×3 pixel neighborhood is selected, which includes the high grayscale area at the center of the spot to ensure fitting accuracy.

[0091] Two-dimensional Gaussian model: The gray-level distribution of the light spot conforms to a two-dimensional Gaussian function.

[0092] Where A is the peak gray level and B is the background gray level. , For the precise coordinates of the light spot center, , The standard deviations of the Gaussian distributions in the x and y directions are given, and there are a total of 5 parameters to be fitted.

[0093] Least squares fitting: Constructing the residual sum of squares function:

[0094] The LM iterative algorithm is used to solve for the parameters. The initial values ​​of the iteration are: A=220, B=10. = , = , = =1.5, Iteration convergence condition: Residual change < The number of iterations should be ≤50.

[0095] Center coordinate output: After iterative convergence, output the optimal parameters. , This refers to the precise coordinates of the light spot center, with an extraction accuracy of ≤0.05 pixels, sub-pixel level accuracy, far exceeding traditional centroid algorithms.

[0096] Then, multispectral data fusion and target displacement calculation are performed, which may include: 1. Multispectral spot center weighted fusion ①Signal-to-noise ratio dynamic weight allocation Calculate the signal-to-noise ratio (SNR) of the spot image with 6 spectral channels. The SNR formula is:

[0097] Signal-to-noise ratio (SNR) ranges for each channel: Blue light 25-30dB, Green light 32-38dB, Yellow light 28-33dB, Red light 35-40dB, Near-infrared 40-45dB, Far-infrared 38-43dB. Weights are assigned based on normalized SNR. k=1,2,...,6.

[0098] Example of setting the weight range: Blue light 0.1, Green light 0.21, Yellow light 0.11, Red light 0.16, Near infrared 0.23, Far infrared 0.19 (dynamic fine-tuning, total = 1).

[0099] ② Calculation of fused coordinates The center coordinates of the 6-channel spot are (u1, v1) ~ (u6, v6), and the fused coordinates are (uf, vf):

[0100]

[0101] After fusion, the coordinate stability is improved by 45%, effectively eliminating coordinate drift caused by single-band environmental interference (such as strong light and shadow).

[0102] 2. Real-time focus calibration Four reference targets with known spacing (precise to 0.01 mm) were selected. After imaging, the center coordinates of the light spot were extracted, and the focal length f was calibrated based on the pinhole imaging principle.

[0103] Where d is the actual distance between the targets, L is the distance from the camera to the target, and Δu is the pixel spacing at the center of the light spot. The optimal focal length is solved by fitting four sets of target data, with a calibration accuracy of ≤0.02mm. The focal length parameters are updated in real time to eliminate focal length errors caused by zoom or temperature.

[0104] 3. Calculation of horizontal and vertical displacement of the target Based on the change in the center pixel of the fused light spot, the calibrated focal length, the attitude compensation data, and the reference coordinates, the horizontal displacement of the target is calculated. Vertical displacement ,formula:

[0105]

[0106] in, , : Change in the number of pixels (in pixels) at the center of the fused light spot; L: Initial distance between the base station and the target (mm, accuracy 0.1mm). p: Camera pixel size (3.45μm=3.45×10^-3mm); f: Camera focal length after calibration (mm, accuracy 0.02mm); , , Initial three-dimensional coordinates of the target (mm, accuracy 0.07mm); , , Initial 3D coordinates of the imaging module (mm, accuracy 0.01mm); , , : Change in the attitude angle of the imaging module (°, accuracy 0.002°); , : Horizontal and vertical displacement changes of the imaging module (mm, accuracy 0.01mm).

[0107] Here, displacement calculation is output in real time, with a sampling frequency of 50Hz and a single calculation time of ≤1.0ms. It can capture the dynamic deformation of structures (frequency ≤20Hz) and the measurement accuracy is ≤0.1mm / 100m, meeting the requirements for high-precision deformation monitoring.

[0108] The data processing unit is used to perform outlier removal processing on the laser ranging data and distortion correction processing on the reference target image data; it is also used to perform three-dimensional calculation of the pixel coordinates by combining the three-point intersection method with the least squares method to determine the initial three-dimensional spatial coordinates of the target.

[0109] In one embodiment, a two-order optical distortion correction algorithm is used to perform outlier removal on laser ranging data and distortion correction on reference target image data. Specifically, this may include: 1. First-order: Zhang's calibration method for lens inherent distortion correction Calibration preparation: A high-precision checkerboard calibration plate is used, with dimensions of 300mm×240mm, square size of 10mm×10mm, number of rows and columns of 12×9, checkerboard flatness ≤0.01mm, and corner point coordinate accuracy ≤0.005mm.

[0110] Image acquisition: Place the calibration board within 1-5m in front of the camera, adjust the angle and distance, and acquire 22 sets of calibration images in different poses (including frontal view, off-angle view, and near and far views). The image resolution is 2560×1920, the exposure time is 1 / 5000s, and the corner points are clear and without blur.

[0111] Corner extraction: A sub-pixel-level corner detection algorithm is used to extract the pixel coordinates of the checkerboard corners in each image with an extraction accuracy of ≤0.01 pixels. Invalid images with a corner extraction error >0.03 pixels are removed.

[0112] Intrinsic parameter distortion coefficient solution: Construct the camera imaging model, intrinsic parameter matrix K:

[0113] in, , Let x and y be the focal lengths, and p1 and p2 be the pixel coordinates of the image center. The distortion model includes radial distortion (k1, k2, k3) and tangential distortion (p1, p2).

[0114] in, The normalized image coordinate radius is denoted as . The parameters are solved using maximum likelihood estimation, converging after 100 iterations. The final parameters are: =5792.34μm, =5791.97μm, =1280.56 pixels, =960.32.

[0115] Distortion correction: For the acquired target image, the distortion offset is calculated by substituting it into the distortion model, and the coordinates are corrected pixel by pixel. After correction, the inherent distortion error of the lens is ≤0.04 pixels.

[0116] 2. Second-order: Field-edge distortion mapping correction Polar coordinate transformation: based on the image center ( , Using the origin as the coordinate system, convert the image pixel coordinates (u, v) to polar coordinates (ρ, θ):

[0117]

[0118] Distortion fitting: Based on calibration data, fit the distortion polynomial at the edge of the field of view:

[0119] Fit coefficients: =0, =1.0032, =-0.00021, =0.000008, fitting residual ≤0.02 pixels.

[0120] Inverse coordinate transformation: transforming the corrected polar coordinates ( Convert θ back to pixel coordinates ( , The distortion correction at the edge of the field of view is completed, and the distortion error at the edge of the field of view (area > 800 pixels from the center) after correction is ≤ 0.05 pixels.

[0121] Furthermore, by using the three-point intersection method combined with the least squares method to perform three-dimensional calculations on the pixel coordinates to determine the initial three-dimensional spatial coordinates of the target, the following steps can be taken: In one embodiment, the explanation is based on the static calibration after the initial power-on of the base station: The host computer constructs a global reference coordinate system using a spatial distance-image coordinate fusion algorithm: with reference station 1 as the origin (0,0,0), the line connecting reference stations 1 and 2 as the X-axis, the line connecting reference stations 1 and 4 projected onto the horizontal plane as the Y-axis, and the Z-axis perpendicular to the horizontal plane and upwards. After target reception, initial coordinate calibration is completed using the three-point intersection method, with an accuracy ≤0.07mm. An approximate three-dimensional coordinate system is calibrated, and then the initial three-dimensional coordinates of each reference station are iteratively solved using the least squares method, with a coordinate residual ≤0.04mm. An initial reference database can be established and stored, and further updated after subsequent dynamic calibration, i.e., six-degree-of-freedom attitude calibration. Alternatively, the database can be re-established after subsequent six-degree-of-freedom attitude calibration.

[0122] The six-degree-of-freedom distortion compensation unit is used to collect and correct the three-dimensional translation and three-dimensional attitude angle data in the initial three-dimensional spatial coordinate data through a six-degree-of-freedom sensor; it is also used to synchronously upload the initial three-dimensional spatial coordinate data to the host with a timestamp.

[0123] In one embodiment, the six-DOF distortion compensation unit can collect its own pitch, yaw, and roll angle changes. In addition, the attitude sensing unit hardware can be selected and installed, and may include: The six-degree-of-freedom distortion compensation unit may include a six-axis MEMS inertial sensor, a high-precision tilt sensor, and a three-dimensional accelerometer. It is mounted on a rigid base (5mm thick, 7075 aluminum alloy) at the bottom of the imaging module frame and fixed with M3 bolts. Thermal grease is applied to the contact surface to ensure a rigid connection between the sensor and the module without relative deformation.

[0124] Six-axis MEMS inertial sensor (model MPU-6050): integrates a three-axis gyroscope and a three-axis accelerometer. The gyroscope has a measurement range of ±500° / s, a resolution of 0.001° / s, and a noise density of 0.004° / s / √Hz; the accelerometer has a measurement range of ±16g, a resolution of 0.0001g, and a noise density of 0.0008g / √Hz; the sampling frequency is 150Hz, and it has an I2C interface output.

[0125] High-precision tilt sensor (model SCA100T): Measurement range ±15°, resolution 0.001°, zero drift ≤0.005° / year, nonlinearity ≤0.01%FS, sampling frequency 100Hz, analog output (0-5V).

[0126] 3D accelerometer (model ADXL355): Measurement range ±10g, resolution 0.0005g, bandwidth 1kHz, low noise, SPI interface output.

[0127] Specifically, a six-axis inertial sensor is used to collect angular velocity and acceleration data in real time, and to solve for attitude angles and integral time constants through integration calculations.

[0128] Tilt sensors are used to output high-precision tilt angle data, which is fused with inertial sensor data to correct attitude angle drift.

[0129] A three-dimensional accelerometer is used to collect acceleration data. Combined with gimbal position information, the translation amount is calculated by integration. The sampling frequency and translation accuracy are also considered.

[0130] In one embodiment, the six degrees of freedom parameters acquired by the imaging module can be defined as follows: 3D translation: ΔXs (horizontal X direction), ΔYs (horizontal Y direction), ΔZs (vertical Z direction), unit mm; Three-dimensional attitude angles: Δαs (pitch angle, about the X-axis), Δβs (yaw angle, about the Y-axis), Δγs (roll angle, about the Z-axis), in degrees.

[0131] The data acquisition process can be as follows: The sensor is powered on and warmed up for 10 minutes to enter a stable working state; the six-axis inertial sensor acquires angular velocity and acceleration data in real time, and solves for attitude angles Δαs, Δβs, and Δγs through integration calculations, with an integration time constant of 100ms and an attitude angle accuracy ≤0.002° after error compensation; the tilt sensor directly outputs high-precision tilt angle data, which is fused with the inertial sensor data to correct attitude angle drift; the three-dimensional accelerometer acquires acceleration data, and combines it with gimbal position information to solve for translational amounts ΔXs, ΔYs, and ΔZs through integration, with a sampling frequency of 150Hz and a translational accuracy ≤0.01mm. All parameters are timestamped (accuracy 1μs) and synchronously uploaded to the parallel computing host, with a transmission delay ≤40μs.

[0132] Furthermore, the construction and error compensation of the six-degree-of-freedom distortion compensation model may include: 1. Definition of coordinate system Reference coordinate system O-XYZ: The origin O is the center of reference station No. 1, the X-axis is horizontal to the right, the Y-axis is horizontal forward, and the Z-axis is vertical upward. It is a right-hand rectangular coordinate system. Imaging module coordinate system Os-XsYsZs: The origin Os is the intersection of the optical centers of the four cameras. In the initial state, it coincides with the origin of the reference coordinate system and the coordinate axes are parallel. Target coordinate system Oi-XiYiZi: The origin Oi is the center of the target, and the initial state is consistent with the coordinate system of the reference coordinate system.

[0133] 2. Spatial coordinate transformation model When the imaging module undergoes six-degree-of-freedom distortion, the transformation relationship between the module coordinate system and the reference coordinate system is as follows:

[0134] Where R is a 3×3 rotation matrix, calculated from attitude angles Δαs, Δβs, and Δγs:

[0135]

[0136]

[0137]

[0138] T is a 3×1 translation matrix: T=[ΔXs,ΔYs,ΔZs]T.

[0139] The projection relationship between the target world coordinates (Xi, Yi, Zi) and the image pixel coordinates (u, v):

[0140] K is the camera intrinsic parameter matrix, and Zs is the distance from the optical center of the module to the target.

[0141] 3. Calculation of pose distortion pixel offset Based on the coordinate transformation model, the pixel offsets Δus and Δvs of the light spot center caused by pose distortion are derived:

[0142]

[0143] Where fx and fy are the focal lengths of the camera in the x and y directions, (Xi, Yi, Zi) are the initial coordinates of the target, and (Xs, Ys, Zs) are the initial coordinates of the module.

[0144] 4. Real-time error compensation The original pixel changes Δuraw and Δvraw at the center of the target image spot are collected. After subtracting the pose distortion offset, the compensated pixel changes are obtained. △ucorr=△uraw-△us △vcorr=△vraw-△vs The compensation calculation is synchronized in real time, with a single calculation time of ≤20μs and a compensation accuracy of ≤0.04 pixels, completely eliminating the influence of imaging module attitude distortion on spot coordinate measurement.

[0145] The data processing unit uses the corrected and compensated initial three-dimensional spatial coordinate data as the state vector and the laser distance data as the observation vector, and performs Kalman filtering fusion on a pre-constructed sixth-order Kalman filter model to output the optimal reference station coordinate estimate. It also compares the optimal reference station coordinate estimate with the corrected and compensated initial three-dimensional spatial coordinates to output the target coordinate change and determine the structure's displacement. Furthermore, it establishes and stores an initial reference database based on the filtered laser range data and the distortion-compensated reference target image data.

[0146] In one embodiment, the state vector is the initial three-dimensional spatial coordinate data after correction and compensation, which may include: the three-dimensional coordinates of each reference station and the three-dimensional attitude angle. The observation vector is the laser distance data, which may include: laser distance, image coordinates, attitude angle data, and filtering parameters. A sixth-order Kalman filter model is constructed based on the state vector and the observation vector. The process noise covariance Q and the observation noise covariance R are calculated in the model. The filtering iteration period is 250ms, and the optimal reference station coordinate estimate is output.

[0147] Furthermore, the output, which integrates multi-camera parallax, laser absolute distance, and IMU six-free attitude compensation, compares the estimated coordinates of the optimal reference station with the corrected and compensated initial three-dimensional spatial coordinates to determine the displacement of the structure by outputting the change in target coordinates. This displacement can be in the form of three-dimensional coordinates: ( (), whose unit can be: millimeter or micrometer.

[0148] The establishment of the database has already been explained in the aforementioned parallel image processing section, and will not be repeated here.

[0149] This invention provides a distributed large field-of-view spectral displacement measurement method, comprising: setting up a target on the structure; correcting and compensating the laser ranging value; acquiring initial reference origin data and sending calibration commands to the host through the reference station; removing outliers from the laser distance data and correcting the distortion of pixel coordinates in the reference target image data, extracting sub-pixels at the center of the light spot, and then performing three-dimensional calculation on the pixel coordinates to determine the initial three-dimensional spatial coordinates of the target; correcting and compensating the data using a six-degree-of-freedom attitude compensation sensor; fusing the corrected and compensated coordinate data and laser distance data using Kalman filtering to output the optimal reference station coordinate estimate; and comparing this estimate with the corrected and compensated initial three-dimensional spatial coordinates to output the change in target coordinates and determine the displacement of the structure. This method improves the accuracy, automation, and anti-interference capability of long-distance deformation measurement of large structures under complex working conditions.

[0150] As is known from common technical knowledge, this invention can be implemented through other embodiments that do not depart from its spirit or essential characteristics. Therefore, the disclosed embodiments described above are merely illustrative in all respects and are not the only ones. All modifications within the scope of this invention or its equivalents are included in this invention.

[0151] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0152] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0153] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0154] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0155] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A distributed large field-of-view spectral shift measurement method, characterized in that, include: S1. Target placement is carried out on the structure. The phase difference between the laser and the two reference stations is determined by a three-axis laser interferometer. Adjustments are made based on the phase difference. After confirming that there is no relative displacement between the target and the structure, the placement record is completed. S2. The environmental sensors read the temperature, humidity and air pressure data in real time, and use the air refractive index to correct and compensate the laser ranging value of the three-axis laser interferometer. S3. Acquire initial reference origin data and use it as a reference for subsequent dynamic measurements. Send calibration instructions to the host through the reference station. The initial reference origin data consists of the corrected laser ranging value and the reference target image data acquired by the dual-spectrum imaging camera. The reference target image data is acquired synchronously by multiple cameras controlled by the control gimbal. S4. After removing outliers from the laser distance data and correcting the distortion of the pixel coordinates of the benchmark target image data, and extracting the sub-pixel center of the spot, the pixel coordinates are solved in three dimensions by combining the three-point intersection method with the least squares method to determine the initial three-dimensional spatial coordinates of the target. S5. The initial three-dimensional spatial coordinates are corrected and compensated for in terms of three-dimensional translation and three-dimensional attitude angle using a six-degree-of-freedom attitude compensation sensor, and a global target coordinate database is established and stored on the host computer. S6. The initial three-dimensional spatial coordinate data after correction and compensation is used as the state vector, and the laser distance data is used as the observation vector. Kalman filtering is performed to fuse them together to output the optimal reference station coordinate estimate. S7. Compare the estimated optimal base station coordinates with the corrected and compensated initial three-dimensional spatial coordinates, and output the target coordinate change to determine the displacement of the structure.

2. A distributed large field-of-view spectral displacement measurement system, applied to a distributed reference monitoring station, characterized in that, include: Target array, imaging module, benchmark monitoring module, and main unit; The target array is used to deploy targets on the structure, determine the laser ranging value of the laser traveling back and forth between the two reference stations, and complete the deployment record after there is no relative displacement between the target and the structure. The imaging module is used to control multiple dual-spectrum imaging cameras to simultaneously acquire reference target image data via a control gimbal. The benchmark monitoring module is used to determine the phase difference of the laser traveling back and forth between two benchmark stations using a three-axis laser interferometer, adjust the laser ranging value based on the phase difference, and transmit the laser ranging value and the benchmark target image data to the host computer via an industrial control board. The host is used to perform distortion correction, spot center sub-pixel extraction, attitude distortion compensation, environmental parameter compensation, and three-dimensional calculation to determine the initial three-dimensional coordinate data of the correction compensation for the reference target image data. It also performs outlier removal processing on the laser range value and performs Kalman filtering fusion on the processed laser distance data and the corrected initial three-dimensional coordinates to determine the optimal reference station coordinate estimate. The host computer determines the displacement of the structure by comparing the target coordinate changes with the estimated coordinates of the optimal reference station and the initial three-dimensional spatial coordinate data after correction and compensation.

3. The system according to claim 2, characterized in that, The target array has its target axis perpendicular to the surface of the structure and is divided into upper and lower cavities. The upper cavity is a light-emitting cavity, and the lower cavity is a power supply cavity for controlling the spatial distribution of light field intensity. The cavity is separated by an insulating epoxy partition of a preset thickness. A wire hole of a preset diameter is reserved in the center of the partition, and the cavity is filled with thermally conductive silicone grease.

4. The system according to claim 2, characterized in that, The target array includes: multiple reference targets, which are fixed to the reference station mounting plate by stainless steel bolts; The reference target is made of alumina ceramic.

5. The system according to claim 2, characterized in that, The triaxial laser interferometer includes: a laser emitting head fixed at the center of the reference station, a triaxial mirror group, and a signal processing unit; The three-axis reflectors in the three-axis reflector group are respectively installed on multiple adjacent reference stations to form X, Y, and Z three-axis measurement optical paths. During measurement, the laser emitter head emits polarized laser light, which is reflected back by the reflectors. The signal processing unit is used to calculate the distance change between reference stations by measuring the round-trip phase difference of the laser.

6. The system according to claim 2, characterized in that, The imaging module includes: a feature point matching unit and a grayscale fusion unit; The feature point matching unit is used to construct Gaussian pyramids for the multi-view images captured by the dual-spectrum imaging camera, generate scale space images, detect pixel gray-level extreme points in adjacent multi-layer images of the scale space image, remove points with contrast less than the preset contrast and points with edge response greater than the edge threshold, fit the extreme points through Taylor expansion, correct the feature point coordinates to the sub-pixel level, perform region division, and normalize the feature descriptors of each region. The grayscale fusion unit is used to filter image feature descriptors by calculating the Euclidean distance of the image feature descriptors. It is used to select a grayscale window to be matched within a preset pixel range, centered on the center feature point of the filtered image, to calculate its grayscale correlation coefficient. Then, it performs quadratic polynomial interpolation on the grayscale correlation coefficient through subpixel interpolation to solve the subpixel-level optimal matching coordinates. Based on the optimal matching coordinates, it solves the homography matrix between the image to be matched and performs matrix transformation to correct the entire image.

7. The system according to claim 2, characterized in that, The host computer includes: an environmental parameter compensation unit, a data processing unit, a parallel image processing unit, and a six-degree-of-freedom distortion compensation unit; The environmental parameter compensation unit is used to perform temperature and humidity compensation, air pressure compensation, and wind speed compensation by reading temperature, humidity, and air pressure data in real time through environmental sensors, and to correct and compensate the laser ranging value of the three-axis laser interferometer by incorporating the air refractive index. The parallel image processing unit is used to perform multi-channel separation, adaptive wavelet threshold denoising, and image enhancement on the reference target image of each dual-spectrum imaging camera according to the target emission band; it is also used to locate the connected component of the spot region of the reference target image, extract it with two-dimensional Gaussian fitting, and perform weighted fusion of the center of the multispectral spot region, real-time focal length calibration, and calculation of the horizontal and vertical displacement of the target to determine the pixel coordinates of the reference target image. The data processing unit is used to perform outlier removal processing on the laser ranging data and distortion correction processing on the reference target image data; it is also used to perform three-dimensional calculation on the pixel coordinates by combining the three-point intersection method with the least squares method to determine the initial three-dimensional spatial coordinates of the target. The six-degree-of-freedom distortion compensation unit is used to collect and correct the three-dimensional translation and three-dimensional attitude angle data in the initial three-dimensional spatial coordinate data through a six-degree-of-freedom sensor; it is also used to synchronously upload the initial three-dimensional spatial coordinate data to the host with a timestamp. The data processing unit is used to perform Kalman filtering fusion on a pre-constructed sixth-order Kalman filter model, using the corrected and compensated initial three-dimensional spatial coordinate data as the state vector and the laser distance data as the observation vector, to output the optimal reference station coordinate estimate; it is also used to compare the optimal reference station coordinate estimate with the corrected and compensated initial three-dimensional spatial coordinates to output the target coordinate change and determine the displacement of the structure.

8. The system according to claim 7, characterized in that, The data processing unit also establishes and stores an initial benchmark database based on the laser ranging data that has been filtered and the benchmark target image data that has been distorted.

9. The system according to claim 7, characterized in that, The six-degree-of-freedom distortion compensation unit includes: a six-axis inertial sensor, a tilt sensor, and a three-dimensional accelerometer. The six-axis inertial sensor is used to collect angular velocity and acceleration data in real time, and to solve for attitude angles and integral time constants through integral calculations. The tilt sensor is used to output high-precision tilt angle data, which is fused with inertial sensor data to correct attitude angle drift. The three-dimensional accelerometer is used to collect acceleration data, and combined with the gimbal position information, the translation amount is calculated by integration, sampling frequency, and translation accuracy.

10. The system according to claim 7, characterized in that, The data processing unit includes: The Kalman filter fusion subunit calculates the process noise covariance and observation noise covariance in the sixth-order Kalman filter model based on the state vector and observation vector, and outputs the optimal reference station coordinate estimate based on iterative calculation of the process noise covariance and observation noise covariance.

Images