Projection figure distortion prediction compensation correction method of laser scanning projection system

Abstract

The application discloses a projection figure distortion prediction compensation correction method of a laser scanning projection system, and comprises the following steps: carrying out laser scanning projection system projection; acquiring projection figure distortion data, establishing a neural network training data set and a test data set; establishing an initial topological structure, weights and thresholds of a BP neural network; optimizing the weights and thresholds of the BP neural network based on a particle swarm optimization algorithm; carrying out deviation compensation on the coordinate positions of each projection point according to the training output of the PSO-BP neural network; selecting feature points in each part of a to-be-projected plane; projecting the feature points by using a distortion deviation correction projector based on the PSO-BP neural network; acquiring feature point projection coordinate values; subtracting the ideal projection point coordinates without distortion from the corrected feature point projection coordinates; judging whether the difference value meets an error interval; if yes, it is determined that the projection coordinate value distortion error has been corrected and compensated; if no, returning to step three for feedback type multiple iterations until the difference value meets the error interval.

Description

Technical Field

[0001] This invention belongs to the field of laser scanning technology, and specifically relates to a distortion prediction and compensation correction method for projected graphics in a laser scanning projection system. Background Technology

[0002] Laser scanning projection technology can accurately project contour shapes, positional markings, and process text information onto a projection surface based on the CAD model of the part to be projected. Therefore, it is widely used in advanced manufacturing and digital intelligent assembly scenarios. This technology uses a rapid, precise, and continuous deflection of a dual-axis galvanometer to continuously change the position and direction of the laser beam in a specified sequence, thereby forming a clear and stable laser contour pattern on the projection surface or the surface of the workpiece. This laser contour pattern or information has accurate three-dimensional shape and spatial position, effectively assisting workers in completing assembly positioning operations and significantly improving assembly positioning efficiency.

[0003] The dual-axis galvanometer in a laser scanning projector consists of two mutually perpendicular plane mirrors mounted on a coil motor. Due to a slight vertical offset, there is a non-linear mapping relationship between the scanning angle of the galvanometer system and the coordinates of the projection working surface, which seriously affects the accuracy of the scanning projection.

[0004] To achieve error compensation and correction of graphic distortion in a laser scanning projection system, it is necessary to predict the distortion value corresponding to each coordinate, thereby compensating for the corresponding distortion error value.

[0005] Currently, the correction of projection distortion in laser scanning projection systems is achieved by establishing a correction table using linear interpolation and employing f-theta lenses. However, obtaining the distortion correction table through interpolation using this method requires the galvanometer system to correct a standard grid file, and the number of grids to be corrected is relatively large. This method is easily affected by the number of grids, limiting its applicability. Furthermore, when using optical elements such as f-theta lenses to correct laser scanning projection instruments with long focal lengths, the introduction of errors due to the lens imaging principle y'=f·θ further reduces projection accuracy. Summary of the Invention

[0006] To address the aforementioned problems, this invention provides a method for predicting and compensating for projection distortion in a laser scanning projection system. This method can accurately predict the distortion of projection points on the projection surface across the entire field of view, thereby compensating for the corresponding distortion error values ​​and improving the efficiency and accuracy of projection distortion correction.

[0007] The objective of this invention is achieved through the following technical solution:

[0008] A method for predicting and compensating for projection image distortion in a laser scanning projection system includes the following steps:

[0009] Step 1: Import the model of the graphic to be projected and project it using the laser scanning projection system;

[0010] Step 2: Obtain and process the projection distortion data to establish a neural network training dataset and a test dataset;

[0011] Step 3: Establish the initial topology, weights, and thresholds of the BP neural network;

[0012] Step 4: Optimize the weights and thresholds of the BP neural network based on the particle swarm optimization algorithm, and train the PSO-optimized BP neural network using the dataset established in Step 2 to obtain the PSO-BP neural network model. Compensate for the deviation of the coordinate positions of each projection point based on the training output of the PSO-BP neural network.

[0013] Step 5: Divide the plane to be projected into several parts at equal intervals, and select feature points in each part. Use the projector with the PSO-BP neural network to complete the distortion correction and project the feature points to obtain the projected coordinate values ​​of the feature points. Subtract the coordinates of the undistorted ideal projection point from the coordinates of the corrected feature projection point, and determine whether the difference meets the error range. If yes, it is determined that the distortion error of the projection coordinate value has been corrected and compensated. If not, return to step 3 for feedback iteration until the difference meets the error range.

[0014] Further, step one includes:

[0015] S11. Import the CAD model of the graphic to be projected;

[0016] S12. Import the coordinate transformation formula between the projection plane coordinate system and the projector coordinate system for the projection points;

[0017] S13. Based on the coordinate transformation relationship between the projection plane coordinate system and the projector coordinate system, guide the deflection of the dual-axis galvanometer to change the projection direction of the projected light rays;

[0018] S14. The target is calibrated by moving the projection light to illuminate the directional regression reflection cooperative target on the projection plane, thereby obtaining the coordinate value of the directional regression reflection cooperative target. The coordinate value is then substituted into the coordinate transformation formula for projection by the laser scanning projection system.

[0019] Furthermore, in step S12, the coordinate transformation relationship between the projection plane coordinate system and the projection points in the projector coordinate system is as follows:

[0020]

[0021] Where x, y, z are the coordinates of any point in the projector coordinate system; e is the vertical distance between the two-axis galvanometers; and H and V are the deflection angles of the laser beams in the horizontal and vertical directions, respectively.

[0022] Further, step two includes: acquiring and processing projection image distortion data using a binocular vision system to establish a neural network test dataset, including test input and test output; calculating the coordinate values ​​of distorted projection points according to the coordinate transformation formula; obtaining the coordinate values ​​of distorted projection points and ideal undistorted projection points; subtracting the x-coordinate of the undistorted ideal projection point from the x-coordinate of the distorted projection point to obtain the distortion error value, thereby establishing a neural network training dataset, including training input and training output; and normalizing the acquired data.

[0023] Furthermore, step two specifically includes:

[0024] S21. Measure and process the actual projection point coordinates on the projection plane using a binocular vision system to establish a neural network test dataset, including test input and test output;

[0025] S22. Obtain the coordinates of the distortion-free ideal projection point;

[0026] S23. Divide the horizontal deflection angle H into several parts within the full field of view of ±30°;

[0027] S24. Divide the vertical coordinate y into several parts within the full field of view of ±30°;

[0028] S25. Substitute the divided H and y into the coordinate transformation formula between the projection points in the projection plane coordinate system and the projector coordinate system, and set the distance between the two galvanometers to obtain the x-coordinate of each distorted projection point.

[0029] S26. Calculate the distortion value Δx for each projection point. The distortion value Δx is expressed as:

[0030] Δx i =x' i -x i

[0031] Where, x' i x is the x-coordinate of the ideal, distortion-free projection point. i The x-coordinate of the distorted projection point;

[0032] S27. Use (x,y) and their respective Δx as training input and training output, respectively;

[0033] S28. Normalize the data.

[0034] Furthermore, step three includes:

[0035] S31. Establish the initial topology of the BP neural network, treating the weights and thresholds as particles in the defined space;

[0036] S32. Using the training dataset established in step two, set the number of training rounds and iterations, and determine the parameters of the PSO-BP neural network, including the number of input layer nodes, the number of output layer nodes, and the number of hidden layers;

[0037] S33. Calculate the fitness of the PSO algorithm, update the position and velocity of the particles, and update the best position of all particles in the entire swarm by comparing the best position found by each particle individually with the best position found by all particles in the swarm.

[0038] S34. Determine whether the fitness value meets the requirements or whether the number of iterations has reached the set number; if yes, assign the PSO-optimized weights and thresholds to the BP neural network; if no, return to step S33 to repeat the process of optimizing the particle swarm weights and thresholds until the termination condition is met.

[0039] Furthermore, in step S33, the root mean square error is selected as the fitness function:

[0040]

[0041] The prediction accuracy between the predicted and actual values ​​is calculated using a fitness function; where yi represents the true value of the i-th sample. denoted by , where N represents the number of samples; the smaller the RMSE value, the better the prediction model fits and the smaller the difference between it and the true value.

[0042] The present invention has the following advantages:

[0043] This invention optimizes the BP neural network using the particle swarm optimization algorithm. Since conventional BP neural networks are prone to getting trapped in local optima or overfitting due to the use of gradient descent to correct weights and thresholds, thus affecting prediction accuracy, this invention further proposes to use particle swarm optimization to adapt to the prediction of projected image distortion. This can avoid getting trapped in local optima, shorten the convergence time, and improve prediction accuracy.

[0044] This invention can accurately predict the distortion values ​​of each point of the projected image on the projection plane within the entire field of view, thereby compensating for the corresponding distortion error values ​​and improving the efficiency and accuracy of projection image distortion correction. Attached Figure Description

[0045] Figure 1This is an overall flowchart of the projection image distortion prediction, compensation and correction method of the laser scanning projection system described in this invention;

[0046] Figure 2 This is a flowchart of the projection process of the laser scanning projection system in Embodiment 1 of the present invention;

[0047] Figure 3 This is a flowchart of the projection image distortion data processing and the establishment of neural network training and testing datasets in Embodiment 1 of the present invention.

[0048] Figure 4 This is a flowchart of the optimization of the BP neural network using the particle swarm optimization algorithm in Embodiment 1 of the present invention;

[0049] Figure 5 This is a flowchart of the process in Embodiment 1 of the present invention, which involves multiple feedback iterations based on actual results to ensure the accuracy of error compensation and correction.

[0050] Figure 6 This is a schematic diagram of the process of acquiring data using a binocular vision system in Embodiment 1 of the present invention;

[0051] Figure 7 It is the Δx error compensation surface diagram described in Embodiment 1 of the present invention;

[0052] Figure 8 This is a graph showing the change in fitness value of the PSO-BP neural network described in Embodiment 1 of the present invention. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of this invention clearer, this invention provides a method for predicting and compensating for projection image distortion in a laser scanning projection system. The embodiments disclosed in this invention will be described in further detail below with reference to the accompanying drawings. The following description, in conjunction with the embodiments and accompanying drawings, does not constitute any limitation on the scope of protection of this invention. The embodiments and accompanying drawings are only used for illustrative purposes and should not be construed as limiting the scope of protection of this invention. Reasonable modifications and combinations included within the scope of the inventive spirit of this invention fall within the protection scope of this invention.

[0054] Example 1

[0055] like Figure 1 As shown, a method for predicting and compensating for projection distortion in a laser scanning projection system includes the following steps:

[0056] Step 1: Import the model of the graphic to be projected and project it using the laser scanning projection system:

[0057] like Figure 2As shown, the CAD model of the graphic to be projected is imported into the computer, and the coordinate transformation relationship between the projection point in the projection plane coordinate system and the projector coordinate system is imported. The target of the directional regression reflection cooperation is calibrated by moving the projection light to illuminate the projection plane, thereby obtaining the coordinate value of the directional regression reflection cooperation target. The coordinate value is then substituted into the coordinate transformation formula for projection by the laser scanning projection system.

[0058] S11. Import the CAD model of the graphic to be projected into the computer;

[0059] S12. Import the coordinate transformation formula between the projection plane coordinate system and the projector coordinate system for the projection points;

[0060] The coordinate transformation relationship between the projection plane coordinate system and the projection point in the projector coordinate system is as shown in formula (1):

[0061]

[0062] Where x, y, z are the coordinates of any point in the projector coordinate system, e is the vertical distance between the two-axis galvanometers. In this embodiment, e is set to 10 mm and z is set to 2000 mm, that is, the distance between the projection surface and the projector is 2 meters; H and V are the deflection angles of the laser beams in the horizontal and vertical directions, respectively.

[0063] S13. Based on the coordinate transformation relationship between the projection plane coordinate system and the projector coordinate system, guide the vertical and horizontal galvanometers of the dual-axis galvanometers to deflect according to the designed deflection angles, thereby changing the projection direction of the projected light rays;

[0064] S14. The target of the directional regression reflection cooperation is calibrated by moving the projection light to illuminate the projection plane, thereby obtaining the coordinate value of the target of the directional regression reflection cooperation. The coordinate value is then substituted into the coordinate transformation formula for projection by the laser scanning projection system.

[0065] Step 2: Obtain and process the projected image distortion data to establish the neural network training and testing datasets:

[0066] like Figure 3 As shown, a binocular vision system is used to acquire and process the distortion data of the projected image, establishing a neural network test dataset, including test input and test output. The coordinates of distorted projection points are calculated using coordinate transformation formulas. The coordinates of distorted projection points and ideal, distortion-free projection points are obtained. The distortion error value is obtained by subtracting the x-coordinate of the ideal, distortion-free projection point from the x-coordinate of the distorted projection point, thus establishing a neural network training dataset, including training input and training output. The acquired data is then normalized.

[0067] S21. By means of Figure 6The binocular vision system shown measures and processes the actual coordinates of the projection points on the plane to be projected, and establishes a neural network test dataset, including test input and test output. Table 1 shows some of the measured data coordinate values.

[0068] Table 1 shows the coordinates of some measured points.

[0069]

[0070] S22. Obtain the coordinates of the distortion-free ideal projection point;

[0071] S23. Divide the horizontal deflection angle H into 60 parts with a step size of 1° within the full field of view of ±30°;

[0072] S24. Within the full field of view of ±30°, the vertical coordinate y is divided into 201 parts with a step size of 10mm;

[0073] S25. Substitute the divided H and y into the formula in step one, and set the distance between the two galvanometers e = 10mm and z = 2000mm to obtain the x-coordinate of each distorted projection point;

[0074] S26. Calculate the distortion value Δx for each projection point. The distortion value Δx is expressed as:

[0075] Δx i =x' i -x i

[0076] Where, x' i x is the x-coordinate of the ideal, distortion-free projection point. i For the x-coordinate of the distorted projection point, such as Figure 7 The figure shown is a surface plot of Δx error compensation.

[0077] S27. Use (x,y) and their respective Δx as training input and training output, respectively;

[0078] S28. Normalize the results obtained above to avoid affecting the training effect of the network due to the size difference between the data.

[0079] Step 3: Establish the initial topology, weights, and thresholds of the BP neural network:

[0080] like Figure 3As shown, the initial topology of the BP neural network is established, and the weights and thresholds of the BP neural network are initialized. The weights and thresholds are treated as particles in the defined space, and the relevant parameters of the BP neural network are initialized. When establishing the initial topology of the BP neural network, the X and Y coordinates of the coordinate points on the projection plane are selected as the training input dataset, and the error values ​​corresponding to each projection point are used as the training input dataset. Therefore, the number of input layer nodes is 2, the number of output layer nodes is 1, and the empirical formula for determining the number of hidden layer nodes H of the BP neural network is: Where b is a constant between 1 and 10. Therefore, the number of H is 2 to 12.

[0081] S31. Establish the initial topology of the BP neural network, treating the weights and thresholds as particles in the defined space;

[0082] S32. Using the training dataset described in step two, i.e. (x,y) and Δx, set the parameters such as the number of training rounds and the number of iterations, and determine the parameters of the PSO-BP neural network, including the number of input layer nodes, the number of output layer nodes, and the number of hidden layers;

[0083] S33. Calculate the fitness of the PSO algorithm, update the position and velocity of the particles, and update the best position of all particles in the entire swarm by comparing the best position found by each particle individually with the best position found by all particles in the swarm.

[0084] The fitness value of the PSO algorithm is one of the indicators for judging whether the prediction result of the network has reached the expected accuracy. Therefore, the root mean square error as shown in formula (2) is selected as the fitness function.

[0085]

[0086] The prediction accuracy between the predicted value and the actual value is calculated using formula (2); where yi represents the true value of the i-th sample. denoted as the predicted value of the i-th sample, and N represents the number of samples; the smaller the RMSE value, the better the prediction model fits and the smaller the difference between it and the true value; it is usually used as an evaluation criterion in regression models.

[0087] The fitness value change curve of the prediction model based on PSO-BP neural network is as follows: Figure 8 As shown. From Figure 8 It can be seen that as the number of iterations increases, the fitness value drops rapidly from 0.00033 to 0.00014, indicating fast convergence. Furthermore, the prediction results of the PSO-BP neural network prediction model are very close to the expected values, demonstrating significant optimization effects. This indicates that the prediction model based on the PSO-BP neural network is effective and feasible.

[0088] The test set RMSE is minimized, meaning the prediction accuracy is highest, when the number of hidden layers is 5. Therefore, the number of hidden layers H is determined to be 5, and the BP neural network topology is 2-5-1. The activation functions for the hidden and output layers of the BP neural network are the Tansig hyperbolic tangent function and the Purelin linear transfer function, respectively. The Trainlm training function is used, with 1000 training iterations, a learning rate of 0.01, and a target error of 1×10⁻⁶. -6 .

[0089] S34. Determine whether the fitness value meets the requirements or whether the number of iterations has reached the set number; if yes, assign the PSO-optimized weights and thresholds to the BP neural network; if no, return to step S33 to repeat the process of optimizing the particle swarm weights and thresholds until the termination condition is met.

[0090] Step 4: Optimize the weights and thresholds of the BP neural network using the Particle Swarm Optimization (PSO) algorithm, and train the PSO-optimized BP neural network using the dataset established in Step 2 to obtain the PSO-BP neural network model. Then, compensate for the deviation in the coordinate positions of each projection point based on the training output of the PSO-BP neural network.

[0091] S41. Optimize the weights and thresholds of the BP neural network based on PSO, and substitute the data established in step two into the PSO-optimized BP neural network to train the neural network, and finally obtain the trained network model.

[0092] S42. Substitute the test input described in step S21 into the trained PSO-BP neural network, and perform inverse normalization on the network output.

[0093] Step 5, as follows Figure 5 As shown, the plane to be projected is divided into several parts at equal intervals, and feature points are selected in each part. A projector with distortion correction achieved by a PSO-optimized BP neural network is used to project the feature points, and the projected coordinate values ​​of the feature points are obtained. The coordinates of the undistorted ideal projection points are subtracted from the coordinates of the corrected feature projection points, and it is determined whether the difference meets the error range. If so, it is determined that the distortion error of the projection coordinate values ​​has been corrected and compensated. If not, the process returns to step three for feedback iterations until the difference meets the error range.

[0094] S51. Divide the plane to be projected into several parts at equal intervals, select feature points in each part, and use a projector with PSO-optimized BP neural network to complete distortion correction to project the feature points and obtain the projected coordinate values ​​of the feature points.

[0095] S52. Subtract the coordinates of the undistorted ideal projection point from the coordinates of the corrected feature projection point, and determine whether the difference meets the error range. If yes, it is determined that the distortion error of the projection coordinate value has been corrected and compensated. If not, return to step three for feedback iteration until the difference meets the error range.

[0096] Prediction accuracy after multiple iterations

[0097] The difference between the coordinates of the undistorted ideal projection point and the corrected feature projection point is less than 0.05 mm, which meets the error range.

[0098] This invention uses a PSO-BP neural network for accurate distortion prediction, ensuring precise deviation compensation for the coordinate positions of each projection point.

[0099] The above description is only within the scope of the technology disclosed in this invention, but the protection scope of this invention is not limited thereto. Any modifications made by those skilled in the art to the surface of this invention are covered within the protection scope of this invention.

[0100] The contents not described in detail in this invention description are common knowledge to those skilled in the art.

Claims

1. A method for predicting, compensating for, and correcting projection image distortion in a laser scanning projection system, characterized in that, Includes the following steps: Step 1: Import the model of the graphic to be projected and project it using the laser scanning projection system; Step 2: Obtain and process the projected image distortion data to establish a neural network training dataset and a test dataset; Step 2 specifically includes: S21. Measure and process the actual projection point coordinates on the projection plane using a binocular vision system to establish a neural network test dataset, including test input and test output; S22. Obtain the coordinates of the distortion-free ideal projection point; S23. Divide the horizontal deflection angle H into several parts within the full field of view of ±30°; S24. Divide the vertical coordinate y into several parts within the full field of view of ±30°; S25. Substitute the defined H and y values ​​into the coordinate transformation formula between the projection plane coordinate system and the projector coordinate system, and set the distance between the two galvanometers to obtain the x-coordinate of each distorted projection point; the coordinate transformation relationship between the projection plane coordinate system and the projector coordinate system is as follows: ； Where x, y, z are the coordinates of any point in the projector coordinate system; e is the vertical distance between the two-axis galvanometers; H and V are the deflection angles of the laser beams in the horizontal and vertical directions, respectively. S26. Calculate the distortion value at each projection point. Distortion value Represented as: ； in, The x-coordinate of the ideal, distortion-free projection point. The x-coordinate of the distorted projection point; S27. Use (x,y) and their respective Δx as training input and training output, respectively; S28. Normalize the data; Step 3: Establish the initial topology, weights, and thresholds of the BP neural network; Step 4: Optimize the weights and thresholds of the BP neural network based on the particle swarm optimization algorithm, and train the PSO-optimized BP neural network using the dataset established in Step 2 to obtain the PSO-BP neural network model. Compensate for the deviation of the coordinate positions of each projection point based on the training output of the PSO-BP neural network. Step 5: Divide the plane to be projected into several parts at equal intervals, and select feature points in each part. Use the projector with the PSO-BP neural network to complete the distortion correction and project the feature points to obtain the projected coordinate values ​​of the feature points. Subtract the coordinates of the undistorted ideal projection point from the coordinates of the corrected feature projection point, and determine whether the difference meets the error range. If yes, it is determined that the distortion error of the projection coordinate value has been corrected and compensated. If not, return to step 3 for feedback iteration until the difference meets the error range.

2. The projection image distortion prediction and compensation correction method for a laser scanning projection system as described in claim 1, characterized in that, Step one includes: S11. Import the CAD model of the graphic to be projected; S12. Import the coordinate transformation formula between the projection plane coordinate system and the projector coordinate system for the projection points; S13. Based on the coordinate transformation relationship between the projection plane coordinate system and the projector coordinate system, guide the deflection of the dual-axis galvanometer to change the projection direction of the projected light rays; S14. The target is calibrated by moving the projection light to illuminate the directional regression reflection cooperative target on the projection plane, thereby obtaining the coordinate value of the directional regression reflection cooperative target. The coordinate value is then substituted into the coordinate transformation formula for projection by the laser scanning projection system.

3. The projection image distortion prediction and compensation correction method for a laser scanning projection system as described in claim 1, characterized in that, Step two includes: acquiring and processing projection distortion data using a binocular vision system to establish a neural network test dataset, including test input and test output; calculating the coordinate values ​​of distorted projection points according to the coordinate transformation formula; obtaining the coordinate values ​​of distorted projection points and ideal undistorted projection points; subtracting the x-coordinate of the undistorted ideal projection point from the x-coordinate of the distorted projection point to obtain the distortion error value, thereby establishing a neural network training dataset, including training input and training output; and normalizing the acquired data.

4. The projection image distortion prediction and compensation correction method for a laser scanning projection system as described in claim 1, characterized in that, Step three includes: S31. Establish the initial topology of the BP neural network, treating the weights and thresholds as particles in the defined space; S32. Using the training dataset established in step two, set the number of training rounds and iterations, and determine the parameters of the PSO-BP neural network, including the number of input layer nodes, the number of output layer nodes, and the number of hidden layers; S33. Calculate the fitness of the PSO algorithm, update the position and velocity of the particles, and update the best position of all particles in the entire swarm by comparing the best position found by each particle individually with the best position found by all particles in the swarm. S34. Determine whether the fitness value meets the requirements or whether the number of iterations has reached the set number; if yes, assign the PSO-optimized weights and thresholds to the BP neural network; if no, return to step S33 to repeat the process of optimizing the particle swarm weights and thresholds until the termination condition is met.

5. The projection image distortion prediction and compensation correction method for a laser scanning projection system as described in claim 4, characterized in that, In step S33, the root mean square error is selected as the fitness function: ； The prediction accuracy between the predicted and actual values ​​is calculated using a fitness function; where yᵢ represents the true value of the i-th sample. denoted by , where N represents the number of samples; the smaller the RMSE value, the better the prediction model fits and the smaller the difference between it and the true value.

Images