A matrix-type multi-point laser positioning installation method and system

Abstract

A matrix-style multi-point laser positioning and installation method and system are disclosed to address the problems of low installation accuracy, poor efficiency, and insufficient automation in air deflector installation. The method includes: calibrating the intrinsic parameters of an all-solid-state area array laser sensor; simultaneously emitting an M-row, N-column matrix laser beam into the installation area; obtaining the original distance values ​​of each point based on time-of-flight or triangulation principles; performing statistical filtering to remove outliers; generating original point cloud data in the sensor coordinate system; transforming the data to the global coordinate system via rigid body transformation; registering the global point cloud with the air deflector theoretical model and calculating the pose deviation; generating adjustment values ​​for each degree of freedom; and driving the actuator for closed-loop adjustment until the deviation is less than a threshold. The system includes modules for an all-solid-state area array laser sensor, a calibration and storage module, a ranging calculation module, point cloud generation, coordinate transformation, registration calculation, adjustment value calculation, and drive control. This invention achieves high-precision, high-efficiency, and fully automated air deflector installation and positioning.

Description

Technical Field

[0001] This invention relates to a method and system thereof, and more particularly to a matrix-type multi-point laser positioning and installation method and system thereof. Background Technology

[0002] In fields such as industrial manufacturing, aerospace, shipbuilding, and civil engineering, the installation accuracy of deflectors (such as production line side guides, aircraft pipe supports, ship ventilation deflectors, and hydraulic engineering flow-guiding structures) directly affects the operating efficiency and safety of equipment. Traditional deflector installation and positioning methods mainly rely on manual measuring tools, such as tape measures, spirit levels, plumb lines, and scribing lines.

[0003] With the development of laser technology, various laser positioning methods have emerged to improve the installation accuracy of guide plates. For example: using laser time-of-flight (ToF) sensors for real-time distance measurement for side guide plate positioning in automated production lines; using laser projection systems to directly project CAD design contours onto the workpiece surface to guide complex assembly; using laser trackers in conjunction with target balls to obtain three-dimensional spatial coordinates for high-precision calibration of large structures; and using laser line projectors or scribing instruments to establish horizontal or vertical baselines for construction positioning.

[0004] However, the aforementioned existing technologies still have significant shortcomings. First, the measurement accuracy of traditional manual measurement methods is greatly affected by the operator's experience, easily leading to cumulative errors and making it difficult to meet high-precision installation requirements (such as micrometer or millimeter level). Second, most existing laser positioning technologies are single-point measurement or line scanning modes, with limited coverage efficiency: single-point or single-line measurements require point-by-point and line-by-line measurements, which is time-consuming for large-size or multi-point guide vane installations, making rapid synchronous positioning difficult. Third, single-point or single-line measurements cannot obtain complete two-dimensional or three-dimensional topographic data of the installation area at once, resulting in incomplete spatial information and a lack of global reference for matching and installing curved or irregularly shaped guide vanes. Furthermore, some existing laser methods (such as laser trackers requiring handheld target balls and laser projection requiring manual adjustment) still require considerable manual intervention, resulting in insufficient automation and difficulty in integrating them into fully automated installation processes. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention discloses a matrix-type multi-point laser positioning and installation method, characterized by comprising the following steps: Step 1: Perform intrinsic parameter calibration on the all-solid-state area array laser sensor to obtain the mapping relationship between the emitted optical axis direction vector and the receiving pixel coordinates of each laser ranging unit; Step 2: A matrix-shaped laser beam is emitted into the target installation area at one time using an all-solid-state area array laser sensor. The matrix-shaped laser beam consists of M rows and N columns of laser ranging points, where M and N are both integers greater than 1. Step 3: Receive the reflected signals from each laser ranging point after reflection from the target, and calculate the original distance value of each laser ranging point based on the ranging principle; Step 4: Perform statistical filtering on the original distance values ​​to remove outliers and noise, and obtain the filtered distance values; Step 5: Based on the filtered distance value and the intrinsic parameters calibrated in Step 1, calculate the three-dimensional coordinates of each ranging point in the sensor coordinate system to form the original point cloud data; Step 6: Transform the original point cloud data to the global mounting coordinate system using rigid body transformation to obtain global point cloud data; Step 7: Register the global point cloud data with the theoretical model of the guide vane installation, and calculate the pose deviation between the actual shape and the theoretical shape of the installation area. The pose deviation includes rotational deviation components and translational deviation components. Step 8: Based on the aforementioned pose deviation and the current actual pose of the guide vane, generate the adjustment amount for each degree of freedom; Step 9: Convert the adjustment amount into a drive command for the installation actuator. Drive the actuator to adjust the guide plate to the target installation position and return to Step 2 to form a closed-loop control until the deviation is less than the preset threshold.

[0006] This invention also discloses a matrix-type multi-point laser positioning and installation system, characterized in that it includes: A fully solid-state area array laser sensor is used to emit a matrix-shaped laser beam into the target installation area at one time and receive the reflected signal. The matrix-shaped laser beam consists of M rows and N columns of laser ranging points, where M and N are both integers greater than 1. The calibration storage module is used to store the intrinsic parameter calibration data of the sensor, which includes the mapping relationship between the emitted optical axis direction vector and the received pixel coordinates of each laser ranging unit. The ranging calculation module is connected to the all-solid-state array laser sensor and is used to calculate the original distance value of each laser ranging point based on the ranging principle. The filtering module is used to perform statistical filtering on the original distance value, remove outlier noise, and obtain the filtered distance value. The point cloud generation module is used to calculate the three-dimensional coordinates of each ranging point in the sensor coordinate system based on the filtered distance value and the intrinsic parameters in the calibration storage module, thus forming the original point cloud data. The coordinate transformation module is used to transform the original point cloud data to the global mounting coordinate system through rigid body transformation to obtain global point cloud data; The registration calculation module is used to register global point cloud data with the theoretical model of the guide vane installation, and calculate the pose deviation between the actual shape and the theoretical shape of the installation area. The pose deviation includes rotational deviation components and translational deviation components. The adjustment calculation module is used to generate the adjustment amount for each degree of freedom based on the pose deviation and the current actual pose of the guide plate. The drive control module is used to convert the adjustment amount into drive commands for the installation actuator. The drive actuator adjusts the guide plate to the target installation position and forms a closed-loop control until the deviation is less than the preset threshold.

[0007] Beneficial effects: 1. Significantly Improved Positioning Accuracy: This invention employs matrix-type multi-point laser transmission and reception, combined with direct time-of-flight ranging or dynamic triangulation principles. It eliminates outliers through statistical filtering and introduces a temperature compensation module to correct distance errors caused by temperature drift in real time. Simultaneously, it utilizes an iterative nearest-point algorithm to achieve high-precision registration between the point cloud and the theoretical model. Finally, it achieves precise compensation for pose deviation through closed-loop proportional-integral-derivative control. This enables installation positioning accuracy to reach the micrometer to sub-millimeter level (e.g., ±0.38 mm), far exceeding traditional manual measurement methods, and the measurement results are unaffected by the operator's skill.

[0008] 2. Significantly Improved Installation Efficiency: By emitting a matrix-shaped laser beam (e.g., M rows and N columns of ranging points) simultaneously using an all-solid-state area array laser sensor, the system can acquire the 3D coordinates of a large number of discrete points within the target installation area without requiring point-by-point or line-by-line scanning. Combined with a pipelined data buffer and direct memory access channels, point cloud generation and coordinate transformation can be performed in parallel, significantly reducing the measurement and data processing time for large-size or multi-point guide vane installations. Furthermore, a global synchronous clock controller ensures that laser emission, signal reception, distance calculation, and point cloud generation are completed within the same clock cycle, further improving the system's real-time response speed.

[0009] 3. Obtain complete installation area topography information: The dense point cloud data generated in one go can completely describe the three-dimensional topography of the installation area, overcoming the shortcomings of incomplete spatial information from single-point or single-line measurements. After registering the global point cloud data with the theoretical model of the guide vane installation, the normal and tangential deviations at each position on the installation surface can be obtained, providing a global installation reference for curved and irregularly shaped guide vanes, effectively avoiding cumulative errors caused by local deviations, and ensuring high-quality matching and installation of complex structural components.

[0010] 4. Achieve Fully Automated Closed-Loop Installation: This invention forms a complete closed-loop control process, from sensor intrinsic parameter calibration, matrix laser emission, distance calculation, filtering and noise reduction, point cloud generation and coordinate transformation, to attitude deviation calculation, adjustment amount generation and actuator driving. It eliminates the need for manual handling of the target ball or manual adjustment of the projection equipment, and can be directly integrated into a six-degree-of-freedom industrial robot or electric displacement platform to achieve automatic sensing, automatic decision-making, and automatic execution of the guide vane installation, while supporting data recording and traceability throughout the installation process.

[0011] 5. Excellent environmental adaptability and robustness: By adjusting the duty cycle of the laser emission pulse through a pulse width modulation control circuit, it can adapt to target surfaces with different reflectivities; the temperature compensation module corrects the distance measurement value in real time based on the pre-stored temperature-wavelength drift curve, reducing the impact of ambient temperature changes on accuracy; the statistical filtering algorithm effectively suppresses outlier noise and multipath interference. These features enable the invention to maintain stable and reliable performance even under complex lighting, temperature fluctuations, and uneven surface materials commonly encountered in industrial environments.

[0012] 6. Covering Multiple Ranging Principles and Application Scenarios: The method and system of this invention support both direct time-of-flight ranging mode (suitable for short-distance, space-constrained robot installation scenarios) and dynamic triangulation mode (suitable for long-distance, high-precision industrial inspection scenarios). Users can flexibly select or switch modes according to actual installation distance and accuracy requirements, demonstrating strong scenario adaptability and technical scalability. Attached Figure Description

[0013] Figure 1 This is a schematic diagram of the method flow of the present invention. Detailed Implementation

[0014] Example 1 The matrix-type multi-point laser positioning and installation method of the present invention will be described in detail below with reference to specific implementation methods. This embodiment takes the automated installation process of a curved guide vane of a large ship as an example to illustrate how the method can achieve high-precision and high-efficiency closed-loop positioning and adjustment, but the present invention is not limited to this application scenario.

[0015] First, before installation, the all-solid-state area laser sensor needs to be calibrated internally. This sensor integrates an array of vertical-cavity surface-emitting lasers (VCSELs) arranged in a 64x8 matrix, comprising 512 independent laser ranging units. Each laser unit has its own optical axis direction; due to manufacturing tolerances and assembly errors, there are slight deviations between the actual and theoretically designed directions of each optical axis. Simultaneously, the mapping relationship between each pixel of the sensor's internal receiving optics system (composed of a single-photon avalanche diode (SPAD) array) and the emitted optical axis also needs to be precisely determined. This embodiment employs a calibration method based on a planar target: a checkerboard planar target of known geometric dimensions is placed in front of the sensor at multiple distances and orientations, emitting matrix laser beams and recording the received signal position of each ranging unit. The emitted optical axis direction vector of each unit is solved by minimizing the reprojection error. and receive pixel coordinate mapping function ,in Once calibration is complete, these intrinsic parameter data are stored in the system's non-volatile memory for later real-time retrieval.

[0016] After calibration, the actual installation phase begins. Step two is executed first: a matrix-shaped laser beam is emitted simultaneously from the all-solid-state area array laser sensor towards the target installation area (i.e., the contact surface of the guide plate to be installed). This laser beam consists of 64 rows and 8 columns of laser pulses, with the points in each row and column arranged in a strictly rectangular grid in space. Unlike traditional single-point scanning or linear array lasers, the matrix-type emission used in this embodiment can simultaneously acquire echo signals from 512 independent ranging points within a single emission cycle (typically lasting in microseconds), thus significantly shortening the data acquisition time. If a single-point scanning method is used instead of a matrix-type method, completing 512 points would require at least 512 emission cycles and a precise mechanical scanning mechanism, taking several seconds or even longer and introducing motion errors. Therefore, the matrix structure is one of the key technical features for achieving high efficiency and high precision.

[0017] Step 3: Receive the reflected signals from each laser ranging point after reflection from the target, and calculate the original distance value based on the ranging principle. In this embodiment, the Direct Time-of-Flight (DTOF) ranging principle is selected based on the distance requirements of the installation environment (the distance between the guide vane and the sensor is approximately 1.5 to 3 meters). The timing circuit inside the sensor records the time difference between the emission and reception of each laser pulse. Then the first Original distance values ​​of each ranging point Calculated by the following formula: ;in The speed of light in a vacuum; Let c be the time difference between the laser emission and reception at the (i,j)th ranging point. It should be noted that although the speed of light in air is slightly reduced, this embodiment compensates for this by pre-measuring the ambient temperature and humidity and introducing a correction factor. The compensation formula is the speed of light in air: The refractive index of air can be calculated using the Edlen formula. For reference, under standard atmospheric conditions (temperature 20°C, air pressure 101325 Pa, relative humidity 50%), The value is 1.00027. In practical applications, based on real-time measured temperature, humidity, and air pressure, the value is determined using the Edlen formula. The calculation is performed, where T is the temperature in Celsius, P is the air pressure (Pa), and RH is the relative humidity (%). The reason for using DTOF instead of triangulation is that within the distance range of 1.5 to 3 meters, the measurement noise of DTOF is linearly related to distance and increases slowly with distance, while the depth error of triangulation is proportional to the square of the distance, and its accuracy drops sharply at long distances. Experimental comparisons show that at 3 meters, the repeatability of DTOF can reach ±0.5 mm, while the accuracy of triangulation with the same baseline length drops to over ±2 mm. Therefore, this embodiment uses the DTOF principle to ensure optimal accuracy at this distance.

[0018] However, in actual measurements, due to ambient stray light, multipath reflections, and noise in the sensor's internal circuitry, the original distance value... The data inevitably contains outliers and noise. Therefore, step four involves statistical filtering. In this embodiment, for each ranging point, a neighborhood set is formed by taking all points within its spatial neighborhood (e.g., neighboring points in the matrix grid whose Manhattan distance to the current point is no more than 2). For a 64×8 matrix grid, the neighborhood with a Manhattan distance of no more than 2 contains at most (2×2+1)2 neighbors. 1 = 24 adjacent points (excluding the center point itself), actual number of valid points | N ij | The value varies from 8 to 24 depending on the boundary conditions. Calculate the average distance between all points in this set. and standard deviation Then, only retain the points that satisfy the following equation: ;in As preset coefficients, this embodiment uses Monte Carlo simulation and field tests to determine the coefficients. = 2.5, theoretically capable of removing over 99% of Gaussian distribution outliers. While median or mean filtering is computationally simple, it cannot effectively remove isolated, large-amplitude noise (e.g., spurious signals caused by specular reflection) because these methods mix noise into the average, causing the filtered data to still deviate from the true value. The statistical filtering in this embodiment, however, uses standard deviation to determine the degree of outlier, thus removing outliers more cleanly while preserving true structural edge information. After filtering, the result is... .

[0019] Step 5: Based on the filtered distance values ​​and the intrinsic parameters calibrated in Step 1, calculate the 3D coordinates of each ranging point in the sensor coordinate system to form the original point cloud data. For the (i,j)th ranging point, its coordinates in the sensor coordinate system are... The calculation formula is:

[0020] in and These are the zenith angle and azimuth angle of the optical axis emitted by the ranging unit, respectively, and the optical axis direction vector calibrated in step one. The distance is obtained through analysis. Note: For DTOF mode, the directly measured distance is a straight-line distance along the optical axis, therefore it needs to be decomposed into three-dimensional coordinates according to the optical axis direction. If optical axis calibration is not performed and all optical axes are assumed to be parallel, a large systematic error (up to several millimeters) will occur for points on the edge of the matrix. This embodiment eliminates this error source through point-by-point calibration, which is an important step in ensuring the overall point cloud accuracy.

[0021] Step Six: Transform the raw point cloud data to the global mounting coordinate system using rigid body transformation to obtain global point cloud data. Since sensors are typically mounted on the end effector of a robotic arm or a movable platform, there is a fixed or real-time changing rigid body transformation relationship between the sensor coordinate system and the global mounting coordinate system. In this embodiment, the sensor is fixed to the flange of a six-DOF industrial robot, and the robot's forward kinematics can provide the pose matrix of the flange relative to the base (i.e., the global coordinate system) in real time. Meanwhile, the sensor's installation relationship relative to the flange has been determined by obtaining the transformation matrix through hand-eye calibration. The point cloud coordinates in the global coordinate system are:

[0022] This coordinate transformation unifies the point cloud data into the same coordinate system as the theoretical model of the guide vane, laying the foundation for subsequent registration.

[0023] Step 7: Register the global point cloud data with the theoretical model of the deflector installation, and calculate the pose deviation between the actual and theoretical shapes of the installation area. The theoretical model is an ideal surface provided by the CAD design file, usually stored in the form of a triangular mesh or NURBS surface. This embodiment uses the Iterative Closest Point (ICP) algorithm for rigid body registration. Let the global point cloud data be... The corresponding set of closest points in the theoretical model is Where K=512 (all valid points). The ICP algorithm solves for the optimal rotation matrix R and translation vector t by minimizing the following objective function:

[0024] The solution process uses the Singular Value Decomposition (SVD) method: First, the centroids of the two point sets are calculated. Then construct the cross-covariance matrix. SVD decomposition of H Then the rotation matrix Translation vector After multiple iterations (in this embodiment, the maximum number of iterations is set to 50, and the convergence threshold is 1e-6 meters), the final pose deviation is obtained. This pose deviation includes a rotational deviation component (expressed as Euler angles about the X, Y, and Z axes of the global coordinate system). (Expression) Translational deviation components Compared to other registration algorithms (such as feature-based registration and Normal Distribution Transform (NDT)), the ICP algorithm does not require the extraction of handcrafted features and directly utilizes all point cloud data, making it more universally applicable to curved surfaces like the deflector that lack significant texture features. Although ICP is sensitive to the initial pose, in this embodiment, since the initial installation position of the deflector has been roughly located by the robotic arm (error < 5mm), ICP can converge quickly.

[0025] Step 8: Based on the pose deviation and the current actual pose of the guide vane, generate the adjustment amounts for each degree of freedom. Since the deviation is expressed in the global coordinate system, while the actuator (six-DOF industrial robot) requires motion commands in its joint space or tool coordinate system, this embodiment uses inverse kinematics to map the pose deviation in Cartesian space into joint angle increments. ,in This is the pseudo-inverse of the Jacobian matrix for the robot.

[0026] Step Nine: Convert the adjustment amount into a drive command for the installation actuator. Drive the actuator to adjust the guide vane to the target installation position and return to Step Two to form a closed-loop control until the deviation is less than a preset threshold (in this embodiment, the preset threshold is 0.1 mm and 0.05 degrees). The drive command is generated by a proportional-integral-derivative (PID) controller, and its control law is: ; in, This is the pose deviation vector (6-dimensional) at the current moment. These are 6×6 diagonal coefficient matrices. In this embodiment, the parameters are tuned using the Ziegler-Nichols method: first, let... Increase Record the critical gain until the system oscillates with constant amplitude. and oscillation period Then calculate according to the empirical formula. The reason for using PID closed-loop control instead of a simple open-loop one-step solution is that various disturbances exist during actual installation (such as robotic arm motion errors, guide vane elastic deformation, environmental vibrations, etc.), and open-loop control cannot eliminate the final deviation caused by these disturbances. Closed-loop PID, through real-time feedback correction, can suppress the final error to below the sensor resolution. If simple proportional control (P control) is used, steady-state error cannot be eliminated; if PI control is used, although steady-state error can be eliminated, the dynamic response is slow and prone to overshoot. This embodiment uses a complete PID structure, balancing fast response, zero steady-state error, and oscillation suppression.

[0027] In this embodiment, each control cycle (i.e., a complete data acquisition, registration, and driving process) takes approximately 50 milliseconds, of which laser emission and reception take 1 millisecond, statistical filtering and point cloud generation take 10 milliseconds, ICP registration takes 30 milliseconds, and inverse kinematics and PID calculation take 9 milliseconds. After 3 to 5 control cycles, the deviation can be stabilized within the threshold. The entire automatic installation process takes no more than 0.5 seconds, which is much faster than the several minutes required for traditional manual measurement and adjustment.

[0028] To further demonstrate the inventiveness of the technical solution in this embodiment, several possible alternative technical solutions and their drawbacks are compared below: Alternative Solution 1: Use a single-point laser rangefinder in conjunction with a 2D turntable for point-by-point scanning. This solution requires mechanical moving parts, and scanning 512 points requires at least 512 times the turntable response time (typically >5 seconds). Furthermore, turntable movement introduces accumulated positioning errors, and it cannot acquire the global topography in real time; it can only perform point-by-point measurements followed by offline fitting. This embodiment, however, uses a matrix-style single-shot acquisition method to obtain the entire point cloud without motion errors, improving speed by more than two orders of magnitude.

[0029] Alternative Solution 2: Employing a structured light 3D scanner. Structured light methods require projecting coded patterns and capturing multiple images, necessitating multiple stitching steps for large curved surfaces, and are sensitive to ambient lighting. In contrast, the DTOF direct ranging method in this embodiment is unaffected by surface texture and lighting changes, and a single frame can cover the entire installation area without the need for stitching.

[0030] Alternative Solution 3: Use Kalman filtering instead of statistical filtering. Kalman filtering assumes noise follows a Gaussian distribution and requires a system state equation, but it cannot effectively remove large non-Gaussian noise points generated by multipath reflections. The statistical filtering in this embodiment is directly based on neighborhood outlier detection, requiring no prior models, and is more suitable for unknown interference in industrial environments.

[0031] Alternative Solution 4: Use Fast Point Feature Histogram (FPFH) registration instead of ICP. FPFH requires calculating the feature descriptor for each point, which has a much higher computational complexity than ICP and is prone to mismatches on smooth surfaces. In this embodiment, since the initial poses are already roughly aligned, point-to-surface ICP can be used directly to achieve sub-millimeter accuracy.

[0032] Alternative Option 5: Use fuzzy PID or adaptive PID instead of classic PID. Although adaptive PID performs better in some situations, its parameter identification process is complex, and in this embodiment, the robot's motion inertia and environmental disturbances are relatively stable, while classic PID, after tuning, can already meet the accuracy requirements. Adding overly complex control algorithms will only increase computational latency and reduce closed-loop bandwidth.

[0033] In summary, this embodiment, through the organic combination of a series of technical means such as matrix laser emission, DTOF ranging, statistical filtering, ICP registration, and PID closed-loop control, solves the problems of low accuracy, poor efficiency, incomplete information, and insufficient automation in existing guide vane installation and positioning methods, achieving significant technological progress. The specific parameters given in this embodiment (such as a 64×8 matrix, α=2.5, 50 ICP iterations, and PID tuning methods) are all practically feasible preferred examples. Those skilled in the art can adaptively adjust the parameters according to specific application scenarios without departing from the principles of this invention; such adjustments still fall within the protection scope of this invention.

[0034] Example 2 This embodiment provides a matrix-type multi-point laser positioning and installation system for achieving highly reliable automated closed-loop positioning and installation in industrial environments with large temperature fluctuations and significant differences in target surface reflectivity (e.g., installation of curved guide vanes on ships in a dry dock during summer). The system structure of this embodiment corresponds to claims 6 to 10, and will be described in detail below with reference to specific implementations.

[0035] The core of the system is an all-solid-state area array laser sensor, which integrates an emitting array composed of vertical-cavity surface-emitting lasers (VCSELs) and a receiving array composed of single-photon avalanche diodes (SPADs). The emitting array is arranged in a 64x8 matrix, with a total of 512 independent laser ranging units. Unlike conventional designs, in this embodiment, each laser corresponds to an independent drive channel, and each drive channel integrates a pulse width modulation (PWM) control circuit. This PWM circuit can adjust the duty cycle of the laser emission pulse in real time according to the reflectivity of the target surface. The principle is as follows: when the sensor scans a high-reflectivity surface (such as a polished metal guide plate), the excessively strong reflected signal may cause the SPAD receiving array to saturate, resulting in distance measurement errors or even damage to the receiving units; at this time, the PWM control circuit automatically reduces the pulse duty cycle, reduces the single pulse energy, and keeps the reflected signal intensity within the linear response region of the SPAD. Conversely, when scanning a low-reflectivity surface (such as a black light-absorbing coating or a rusted surface), the PWM control circuit increases the duty cycle to enhance the emission energy, ensuring that the echo signal can be reliably detected. If a point-by-point adjustable duty cycle driving method is not adopted, and a fixed transmit power is used instead, the system can only adapt to a single reflectivity range. For high reflectivity surfaces, saturation distortion will occur, leading to ranging failure, while for low reflectivity surfaces, the signal-to-noise ratio may be insufficient. This embodiment, by integrating a PWM drive channel, enables the same sensor to adaptively handle various surfaces with reflectivity ranging from 3% to 95%, without the need for manual sensor replacement or hardware adjustment.

[0036] To ensure time synchronization and data consistency throughout the measurement process, this embodiment also integrates a synchronization trigger controller. This controller is connected to the all-solid-state area array laser sensor, the ranging calculation module, the filtering module, and the point cloud generation module. Its core function is to generate a global synchronization clock signal, enabling the four operations—laser emission, signal reception, distance calculation, and point cloud generation—to be completed under the same clock cycle. Specifically, the synchronization trigger controller first sends a trigger pulse to the transmitting array, activating 512 lasers to simultaneously emit matrix laser pulses. Simultaneously, after passing through a precise programmable delay line, the trigger pulse opens the integration window of the receiving array. After the SPAD array captures the reflected signal, the timestamp data generated by each pixel is sent to the ranging calculation module via a parallel bus. The ranging calculation module completes the time-of-flight calculation for all 512 points in the next clock cycle. Then, the filtering module and the point cloud generation module complete denoising and coordinate transformation in the following two clock cycles. The entire pipeline is driven by the same clock, with no asynchronous waiting or buffer conflicts. More importantly, the frequency of this global synchronization clock signal can adaptively adjust according to the distance range of the target installation area: when the installation distance is close (e.g., less than 1 meter), the echo signal returns quickly, and the synchronization trigger controller increases the clock frequency (e.g., to 100MHz) to improve the measurement refresh rate; when the installation distance is far (e.g., more than 4 meters), the echo signal returns slowly, and the controller automatically reduces the clock frequency (e.g., to 25MHz) to ensure that the receiving window covers the complete echo time. In contrast, in traditional systems, each module is usually driven by its own independent clock, exchanging data asynchronously through handshake signals or FIFOs. This asynchronous method introduces uncertain delays and jitter, which can lead to significant errors in microsecond-level time-of-flight measurements. The synchronization trigger mechanism in this embodiment ensures the accuracy and repeatability of time measurements, providing key hardware support for achieving sub-millimeter-level positioning accuracy.

[0037] Ambient temperature variation is another important factor affecting the accuracy of laser ranging. The emission wavelength of a semiconductor laser redshifts with increasing junction temperature, with a typical drift coefficient of approximately 0.06 nm / ℃. This wavelength change alters the air refractive index and the quantum efficiency response of the SPAD, leading to systematic deviations in time-of-flight measurements. To eliminate this effect, this embodiment embeds a temperature compensation module within the sensor. This module consists of a thermistor (or integrated temperature sensor) attached to the laser diode chip and a pre-stored temperature-wavelength drift curve lookup table. The temperature sensor monitors the junction temperature of the laser diode in real time, with a sampling frequency synchronized with the laser emission frequency (i.e., the temperature is read once per measurement cycle). The compensation module reads the corresponding thermal expansion compensation coefficient β (typically 6.5 × 10⁻⁶) from the lookup table based on the difference ΔT between the current temperature and a reference temperature (e.g., 25℃). -5 / ℃), and then the original distance value output by the distance calculation module. Perform the following corrections:

[0038] The physical basis of this formula is that increased temperature leads to increased wavelength, decreased air refractive index (with minimal change), and a slight increase in SPAD response time. The combined effect is a systematic stretching of the measured distance; therefore, linear multiplicative compensation is sufficiently accurate. Experiments show that within a temperature range of 0℃ to 50℃, the uncompensated system ranging drift can reach ±1.2mm, while after linear multiplicative compensation in this embodiment, the residual error is less than ±0.1mm. In contrast, some existing technologies use post-processing software correction or ignore temperature effects, either requiring additional reference targets for real-time calibration (increasing system complexity and cost) or failing to guarantee accuracy in environments with temperature variations. The hardware-level real-time temperature compensation in this embodiment is simple, effective, and does not consume the main control chip's computing resources.

[0039] In terms of the data processing pipeline, this embodiment specifically designs a cascaded data buffer between the point cloud generation module and the coordinate transformation module. This buffer is implemented using dual-port random access memory (DPRAM) with a depth of 512 point cloud data units (each unit contains x, y, z coordinates and a corresponding confidence flag). After the point cloud generation module calculates the sensor coordinates of each ranging point, it writes them to a free address in the DPRAM. Simultaneously, the coordinate transformation module independently reads the written data from the DPRAM via a direct memory access (DMA) channel and performs a rigid body transformation to convert it to global coordinates. These two operations can be performed completely in parallel: the point cloud generation module continuously writes new data to the tail of the buffer, while the coordinate transformation module reads and processes data from the head of the buffer. The read and write pointers are managed by hardware arbitration logic to avoid conflicts. Without this dual-port buffer + DMA architecture, it would be necessary to wait for all 512 points of the point cloud to be generated before the CPU performs coordinate transformation point by point. This serial processing method would reduce data throughput by approximately 30% to 50% (depending on the CPU load). The parallel architecture in this embodiment makes the total time for point cloud generation and coordinate transformation basically equal to the longer of the two (about 10 milliseconds), rather than the sum of the two (about 18 milliseconds), thereby reducing the overall system control cycle from about 60 milliseconds to about 50 milliseconds and improving the bandwidth of closed-loop control.

[0040] The complete workflow of the above modules working together is as follows: 1. After the system is powered on, the calibration storage module loads the pre-calibrated intrinsic parameter data (optical axis direction vector of each laser ranging unit, receiving pixel mapping relationship, and temperature compensation coefficient β).

[0041] 2. The synchronous trigger controller generates a start pulse, with the frequency adaptively set to 40MHz based on the current preset installation distance (e.g., 2.5 meters).

[0042] 3. After receiving the start-up pulse, the all-solid-state area array laser sensor uses the PWM control circuit to dynamically adjust the pulse duty cycle of each laser based on the feedback of the reflection intensity of each channel obtained in the previous measurement (the default duty cycle of 50% is used during the first start-up), and then all VCSELs emit matrix laser pulses simultaneously.

[0043] 4. The SPAD receiving array captures the reflected signal, and the time difference data generated by each pixel is sent in parallel to the ranging calculation module, which calculates the original distance value in the next clock cycle. .

[0044] 5. The temperature compensation module reads the current junction temperature, calculates ΔT, and applies it to each... Multiplicative compensation is obtained .

[0045] 6. The filtering module... Perform statistical filtering (the specific method is the same as described in Example 1) to remove outlier noise and obtain the filtered distance value.

[0046] 7. The point cloud generation module calculates the three-dimensional coordinates of each ranging point in the sensor coordinate system based on the filtered distance value and intrinsic parameter data, and writes the results into a dual-port RAM buffer.

[0047] 8. The coordinate transformation module reads point cloud data from the dual-port RAM through the DMA channel, combines it with the pose matrix fed back by the robot in real time, and performs rigid body transformation to obtain global point cloud data.

[0048] 9. The registration calculation module performs ICP registration between the global point cloud data and the theoretical model of the guide vane to obtain the pose deviation.

[0049] 10. The adjustment calculation module calculates the adjustment of each joint through inverse kinematics based on the pose deviation and the robot's current pose.

[0050] 11. The drive control module converts the adjustment amount into drive commands, controlling the six-degree-of-freedom industrial robot to move the guide plate to the target position.

[0051] 12. Repeat steps 2 to 11 above until the pose deviation is less than the preset threshold.

[0052] It is important to note that many hardware features in this embodiment (such as PWM channel-by-channel duty cycle adjustment, global synchronous clock adaptation, temperature linear compensation, and dual-port RAM+DMA pipeline) are important supplements to the basic system structure defined in claim 6. They respectively solve practical engineering problems in the prior art, such as poor reflectivity adaptation, low timing synchronization accuracy, large temperature drift, and limited data processing throughput. These features are not simply a collection of common knowledge, but rather system-level innovations made to meet the special needs of matrix-type multi-point laser positioning installation. For example, even with PWM adjustment, a traditional single-point laser rangefinder only requires one channel; however, in a 512-channel matrix system, the hardware complexity of independently adjusting each channel is extremely high. This embodiment achieves this function through an integrated PWM drive array. Furthermore, in traditional solutions, the latency jitter of multi-module asynchronous communication may only be at the nanosecond level, but the matrix system requires strict consistency in the flight time measurements of all 512 points. Nanosecond-level jitter will lead to millimeter-level distance errors, therefore, global synchronous triggering must be used.

[0053] Through the above system-level design, the matrix multi-point laser positioning and installation system of this embodiment can achieve a repeatability accuracy of up to ±0.38mm in harsh industrial environments (temperature range 0°-50°C, target reflectivity 3%-95%), with a single measurement cycle of less than 1 millisecond and a complete closed-loop control cycle of about 50 milliseconds, which is significantly better than the existing technology.

[0054] The mathematical formulas not explicitly described in this embodiment (such as the time-of-flight ranging formula, statistical filtering formula, ICP registration formula, PID control law, etc.) are the same as those in Embodiment 1. Their specific expressions and parameter definitions are detailed in Embodiment 1 and will not be repeated here. Those skilled in the art, after reading Embodiments 1 and 2, will be able to fully understand and implement the methods and systems protected by this invention.

[0055] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A matrix-type multi-point laser positioning and installation method, characterized in that, Includes the following steps: Step 1: Perform intrinsic parameter calibration on the all-solid-state area array laser sensor to obtain the mapping relationship between the emitted optical axis direction vector and the receiving pixel coordinates of each laser ranging unit; Step 2: A matrix-shaped laser beam is emitted into the target installation area at one time using an all-solid-state area array laser sensor. The matrix-shaped laser beam consists of M rows and N columns of laser ranging points, where M and N are both integers greater than 1. Step 3: Receive the reflected signals from each laser ranging point after reflection from the target, and calculate the original distance value of each laser ranging point based on the ranging principle; Step 4: Perform statistical filtering on the original distance values ​​to remove outliers and noise, and obtain the filtered distance values; Step 5: Based on the filtered distance value and the intrinsic parameters calibrated in Step 1, calculate the three-dimensional coordinates of each ranging point in the sensor coordinate system to form the original point cloud data; Step 6: Transform the original point cloud data to the global mounting coordinate system using rigid body transformation to obtain global point cloud data; Step 7: Register the global point cloud data with the theoretical model of the guide vane installation, and calculate the pose deviation between the actual shape and the theoretical shape of the installation area. The pose deviation includes rotational deviation components and translational deviation components. Step 8: Based on the aforementioned pose deviation and the current actual pose of the guide vane, generate the adjustment amount for each degree of freedom; Step 9: Convert the adjustment amount into a drive command for the installation actuator. Drive the actuator to adjust the guide plate to the target installation position and return to Step 2 to form a closed-loop control until the deviation is less than the preset threshold.

2. The method according to claim 1, characterized in that, The ranging principle in step three is the direct time-of-flight ranging principle, and the original distance value is given by the formula. The calculation yields the result, where c is the speed of light. This is the time difference between laser emission and reception.

3. The method according to claim 1, characterized in that, The distance measurement principle in step three is the dynamic triangulation principle, and the three-dimensional coordinates corresponding to the original distance value are determined by the following formula: Where L is the baseline length between the laser transmitter and receiver, and f is the focal length of the receiving optical system. These are the pixel coordinates of the reflected light spot on the receiving plane. Let the coordinates of the emitted light spot be on the reference plane. The coordinates of the ranging point in the sensor coordinate system.

4. The method according to claim 1, characterized in that, The statistical filtering in step four specifically involves: for each ranging point, calculating the average distance of all points in its neighborhood. and standard deviation , retain satisfaction The point, among which For preset coefficients, This is the original distance value of the i-th point.

5. The method according to claim 1, characterized in that, The registration in step seven employs the iterative nearest-point algorithm, which minimizes the objective function. Solve for the rotation matrix R and the translation vector t, where For points in the global point cloud data, K represents the corresponding point in the theoretical model, and K is the number of point pairs participating in the registration; the rotational deviation component in the pose deviation is expressed in Euler angles, and the translational deviation component is expressed in three-dimensional vector form; the driving command in step nine is generated by a proportional-integral-derivative controller, and its control law is: ;in: Let be the pose deviation vector at the current moment. These are the proportional, integral, and differential coefficient matrices, respectively.

6. A matrix-type multi-point laser positioning and installation system, characterized in that, include: A fully solid-state area array laser sensor is used to emit a matrix-shaped laser beam into the target installation area at one time and receive the reflected signal. The matrix-shaped laser beam consists of M rows and N columns of laser ranging points, where M and N are both integers greater than 1. The calibration storage module is used to store the intrinsic parameter calibration data of the sensor, which includes the mapping relationship between the emitted optical axis direction vector and the received pixel coordinates of each laser ranging unit. The ranging calculation module is connected to the all-solid-state array laser sensor and is used to calculate the original distance value of each laser ranging point based on the ranging principle. The filtering module is used to perform statistical filtering on the original distance value, remove outlier noise, and obtain the filtered distance value. The point cloud generation module is used to calculate the three-dimensional coordinates of each ranging point in the sensor coordinate system based on the filtered distance value and the intrinsic parameters in the calibration storage module, thus forming the original point cloud data. The coordinate transformation module is used to transform the original point cloud data to the global mounting coordinate system through rigid body transformation to obtain global point cloud data; The registration calculation module is used to register global point cloud data with the theoretical model of the guide vane installation, and calculate the pose deviation between the actual shape and the theoretical shape of the installation area. The pose deviation includes rotational deviation components and translational deviation components. The adjustment calculation module is used to generate the adjustment amount for each degree of freedom based on the pose deviation and the current actual pose of the guide plate. The drive control module is used to convert the adjustment amount into drive commands for the installation actuator. The drive actuator adjusts the guide plate to the target installation position and forms a closed-loop control until the deviation is less than the preset threshold.

7. The system according to claim 6, characterized in that, The all-solid-state area array laser sensor includes a vertical-cavity surface-emitting laser array and a single-photon avalanche diode receiving array. The vertical-cavity surface-emitting laser array is arranged in a matrix, with each laser corresponding to an independent driving channel. The driving channel integrates a pulse width modulation control circuit to adjust the duty cycle of the laser emission pulse to adapt to target surfaces with different reflectivities.

8. The system according to claim 6, characterized in that, It also includes a synchronization trigger controller, which is connected to the all-solid-state area array laser sensor, the ranging calculation module, the filtering module and the point cloud generation module respectively. It is used to generate a global synchronization clock signal so that the laser emission, signal reception, distance calculation and point cloud generation operations are completed under the same clock beat. The frequency of the global synchronization clock signal is adaptively adjusted according to the distance range of the target installation area.

9. The system according to claim 6, characterized in that, It also includes a temperature compensation module, which is embedded in the all-solid-state area array laser sensor. This module monitors the junction temperature of the laser diode in real time and compensates for and corrects the original distance value output by the ranging calculation module based on a pre-stored temperature-wavelength drift curve. The correction formula is as follows: ,in The original distance value before compensation, Δ T This refers to the temperature rise relative to the reference temperature. β This is the coefficient of thermal expansion compensation.

10. The system according to claim 6, characterized in that, A pipelined data buffer is cascaded between the point cloud generation module and the coordinate transformation module. The pipelined data buffer is implemented using a dual-port random access memory to temporarily store the raw point cloud data. At the same time, the coordinate transformation module reads the point cloud data in the buffer through a direct memory access channel, so that the point cloud generation and coordinate transformation operations are executed in parallel.

Images