High-precision length measurement method and system based on laser technology

Abstract

The invention discloses a high-precision length measurement method and system based on a laser technology, and the method comprises the steps: obtaining the working data and state data of laser measurement, calculating a laser edge coordinate, a first laser measurement length and a distance measurement scale according to the laser scanning data, processing a camera image to obtain a pixel edge coordinate, and obtaining a second laser measurement length; matching the pixel edge coordinate with the laser edge coordinate to obtain a first edge coordinate, and fusing the first edge coordinate with the second laser measurement length to obtain a laser image length, and constructing a high-precision length measurement model according to the comprehensive measurement length, the distance measurement scale, the transmission impact factor and the first edge coordinate, and inputting to-be-measured working data and state data into the high-precision length measurement model to obtain the measurement length. The method not only can improve the precision of laser length measurement, but also has good interpretability, and can be directly applied to a high-precision length measurement system based on the laser technology.

Description

Technical Field

[0001] The invention relates to the field of length measurement, and in particular to a high-precision length measurement method and system based on laser technology. Background Art

[0002] With the development of modern industry and technology, the requirements for the accuracy of spatial three-dimensional coordinates in major equipment manufacturing, space technology, national defense industry and other fields are increasing. Laser technology plays an increasingly important role in the field of high-precision length measurement due to its unique advantages such as non-contact, high precision, high speed and high reliability.

[0003] Although traditional laser measurement technology has met the accuracy requirements to a certain extent, it still faces many challenges in complex environments, dynamic measurements and long-distance measurements. It needs to be adjusted in combination with effective auxiliary measurement methods. At the same time, laser measurement has a large sampling volume and rich data content, which puts higher requirements on the data processing capabilities of the measurement system. With the rapid development of computer vision technology, image recognition technology is widely used in length measurement. Through the processing of images or videos, it can realize target detection, target tracking, image segmentation, image recognition and other functions, which can be used as an effective auxiliary length measurement method. By fully considering factors such as transmission influence, image conversion and state data, combining laser technology, image recognition technology and machine learning technology, an accurate, efficient and fast high-precision length measurement method and system based on laser technology is designed to overcome the shortcomings of existing length measurement methods and systems, realize efficient and high-precision length measurement, and provide strong technical support for industrial automation, precision manufacturing and other fields. Summary of the invention

[0004] The object of the present invention is to provide a high-precision length measurement method and system based on laser technology.

[0005] To achieve the above object, the present invention is implemented according to the following technical solutions:

[0006] The present invention comprises the following steps:

[0007] Acquire working data and status data of laser measurement, and pre-process the working data and the status data; the working data includes laser scanning data and camera images; the status data includes first status data and second status data;

[0008] Calculate the laser edge coordinates, the first laser measurement length and the distance measurement scale according to the laser scanning data, input the first state data into the transmission influence function to obtain the transmission influence factor, and determine the second laser measurement length according to the transmission influence factor and the first laser measurement length;

[0009] Processing the camera image to obtain pixel edge coordinates, converting the pixel edge coordinates into world edge coordinates according to the second state data, and matching the world edge coordinates with the laser edge coordinates to obtain first edge coordinates;

[0010] The first edge coordinate and the second laser measurement length are combined to obtain a laser image length, and the laser image length is adjusted using the distance measurement scale to obtain a comprehensive measurement length;

[0011] A high-precision length measurement model is constructed according to the comprehensive measurement length, the distance measurement scale, the transmission influence factor and the first edge coordinate, and the working data and status data to be measured are input into the high-precision length measurement model to obtain the measured length.

[0012] Further, the method for calculating the laser edge coordinates, the first laser measurement length and the distance measurement scale includes:

[0013] The laser scanner and the camera are used as a pair of scanning groups, and more than three pairs of scanning groups at different spatial positions are set to obtain the machine working data and status data;

[0014] Determine the reference center and coordinate system of the surface of the measured object, project the spectrum mark as a mark on the reference center and edge of the surface of the measured object during laser scanning, calculate the relative coordinates of the edge point and the reference center as the relative laser edge coordinates, and determine the laser edge coordinates according to the mean, mean square error and skewness of the relative laser edge coordinates of multiple scanning groups;

[0015] The laser scanning data are sorted in time series, the time point when the phase of the laser scanning data changes is set as the edge time point, the interval of the edge time point is calculated to obtain the scanning time, the rotation angle of the laser scanner is determined by the scanning time and the rotation speed of the laser scanner, the edge distance from the scanning point to the edge point is calculated according to the round-trip time of the laser beam corresponding to the edge time point, and the first laser measurement length is calculated according to the edge distance and the rotation angle of the laser scanner;

[0016] The distance measurement scale from the scanning point to the centroid of the reference object surface is calculated according to the round trip time of the laser beam corresponding to the time point when the laser scans to the centroid mark of the reference object surface.

[0017] Furthermore, the method for obtaining the transmission impact factor includes:

[0018] The first state data is input into the transmission influence function to obtain the transmission influence factor, which is expressed as:

[0019]

[0020] Where Q Tis the transmission influence factor, w1, w2, w3 are the transmission influence factor weight coefficients, Δf is the deviation between the actual frequency of the emitted laser and the preset frequency, f0 is the preset frequency of the emitted laser, ΔT is the temperature drift of the laser, k is the temperature wavelength influence coefficient, Δλ max is the deviation between the end value of the laser spectrum and the atmospheric spectrum, Δλ min is the deviation between the laser spectrum and the starting value of the atmospheric spectrum, Δλ is the range of the atmospheric spectrum;

[0021] The second laser measurement length is determined according to the transmission influence factor and the first laser measurement length, and the expression is:

[0022]

[0023] Wherein L2 is the second laser measurement length, L1 is the first laser measurement length, and w4 is the transmission influence factor weight.

[0024] Furthermore, the method for obtaining the pixel edge coordinates includes:

[0025] The OpenCV image processing library is used to read the camera image to be processed, the camera image is grayed to obtain a grayscale image, and the edge detection of the grayscale image is performed to identify the edge position of the grayscale image of the test piece;

[0026] The edge detector uses a Gaussian Laplace filter to perform Gaussian filtering on the grayscale image and perform Laplace transform to obtain the image edge information. The expression is:

[0027]

[0028] in is the Laplace Gaussian operator, is the Laplace operator, G(u,v) is a two-dimensional Gaussian function, σ is the standard deviation of the Gaussian function, g(u,v) is the image after Gaussian Laplace filtering, and f(u,v) is the input grayscale image;

[0029] According to the edge information, a zero-crossing test is performed to obtain the specific edge position. According to the edge position, the eight-connected region is used to trace the image contour in a counterclockwise direction. The expression of the eight-connected rule is:

[0030] N8(p)=H4(p)∪(u+1,v+1),(u+1,v-1),(u-1,v+1),(u-1,v-1)

[0031] Where N8(p) represents the eight-connected region of the starting point p of the boundary tracking, including point p and its corresponding upper, lower, left, right and four diagonal neighboring points; N4(p) represents the four-connected region of point p, including point p and its corresponding upper, lower, left and right neighboring points;

[0032] According to the grayscale calculation value, five distance intervals with different distances are divided to filter the contour points. The screening principle is:

[0033] M1>1-T,M2 <T,M4<T,M5> 1-T

[0034] Where M1, M2, M4, and M5 are the grayscale averages of regions 1, 2, 4, and 5, and T is the grayscale threshold;

[0035] Determine the filtered contour points, use the moment function to calculate the geometric center, and extract the pixel edge coordinates based on the contour data.

[0036] Furthermore, the method of converting the pixel edge coordinates into world edge coordinates includes:

[0037] The second state data includes the camera focal length, the camera pitch angle and the camera translation amount;

[0038] According to the second state data, the camera internal parameter matrix and external parameter matrix are constructed to convert the pixel edge coordinates (u, v) into the camera edge coordinates (X c ,Y c ,Z c ), the camera edge coordinates (X c ,Y c ,Z c ) is converted to world edge coordinates (X w ,Y w ,Z w ).

[0039] Furthermore, the method for obtaining the first edge coordinates includes:

[0040] Calculate the comprehensive similarity between the world edge coordinates and the laser edge coordinates. The expression is:

[0041]

[0042] Where S w,l is the comprehensive similarity between the world edge coordinates and the laser edge coordinates, α and β are weight parameters, is the i-th world edge coordinate point A i and the jth laser edge coordinate point B j The Euclidean distance, n is the number of coordinate points, r is the buffer radius, d A is the distance from the world edge coordinate point to the origin, d B is the distance from the laser edge coordinate point to the origin, is the mean of the vectors corresponding to the world edge coordinate points, is the mean value of the vector corresponding to the laser edge coordinate point;

[0043] The edge coordinate points are matched according to the comprehensive similarity between the world edge coordinates and the laser edge coordinates, and the first edge coordinates are determined according to the mean, variance and skewness of the two edge coordinate points.

[0044] Furthermore, the method for obtaining the length of the laser image includes:

[0045] Calculate the distance and angle between any two first edge coordinate points, and define the two first edge coordinate points and the corresponding distance and angle as a set of edge coordinate data;

[0046] Perform feature fusion on the edge coordinate data and the second laser measurement length to obtain the laser image length;

[0047] The interactive multi-feature fusion method RCI and stacked ensemble classifier are used for feature fusion;

[0048] The interactive multi-feature fusion method RCI includes: calculating the correlation, calculating the connection entropy, calculating the interaction factor and determining the evaluation criteria;

[0049] Determine the evaluation criteria for candidate features, expressed as:

[0050]

[0051] Among them J RCI (L1) is the evaluation criteria for candidate features, F is all features, L2 is the second laser measurement length and edge coordinate data feature c Edge The correlation of is the redundancy, s is the selected feature subset Ω s The characteristics of is the interaction factor, H c (Ω s ,L2,c Edge ) is the selected feature subset Ω when considering the second laser measurement length L2 s and edge coordinate data feature c Edge The joint entropy of c (Ω s ,c Edge ) is to consider only the selected feature subset Ω s and the joint entropy of edge coordinate data feature c;

[0052] The optimal feature subset is obtained according to the evaluation criteria of the candidate features, and the optimal feature subset is input into the stacked ensemble classifier to obtain the laser image length;

[0053] The stacked ensemble classifier includes an input layer, a first classification layer, a second classification layer, and an output layer;

[0054] The first classification layer is composed of a deep neural network DNN and a lightweight gradient boosting machine LightGBM, which performs basic classification on the input data and outputs probability results; the second classification layer is composed of a support vector machine SVM, which performs meta-classification on the input probability results to determine the fusion results of the feature data, and the output layer outputs the comprehensive measurement length; the F1 score, G mean, mean square error MSE and cross validation are used to evaluate the performance of the model.

[0055] Further, the method for obtaining the measured length includes:

[0056] According to the ranging scale, the laser image lengths of the scanning groups at different spatial positions are adjusted to obtain a comprehensive measurement length, and the comprehensive measurement set is formed with the transmission influencing factor, the ranging scale and the first edge coordinates, and the comprehensive measurement set is divided into a training set and a test set;

[0057] Construct a high-precision length measurement model, which includes an input layer, a base model layer, a strategy layer, and an output layer;

[0058] The base model layer is composed of two multi-layer perceptron models and two gradient boosting machine models in parallel;

[0059] In the multi-layer perceptron model: design two hidden layers, use the ReLU function as the activation function to accelerate training and reduce gradient disappearance, use the Cross-Entropy Loss function to solve the problem of the cross-entropy loss function deviating from the evaluation index and overconfidence, and use Adam to optimize the model weight parameters;

[0060] In the gradient boosting machine model: the mathematical model is constructed iteratively, the logarithmic loss function is used to measure the difference between the predicted probability distribution and the true distribution, and the learning rate is adjusted to control the optimization process;

[0061] The strategy layer uses the stacking method to input the prediction structure of the base model into the new model for prediction, obtains the prediction results, and uses the test set to evaluate the high-precision length measurement model;

[0062] The working data and status data of the machine to be measured are input into a high-precision length measurement model to obtain the measured length.

[0063] Second, the high-precision length measurement system based on laser technology, including:

[0064] Data acquisition module: including a laser transmitter, a laser receiver and a high-resolution camera; used to collect machine working data and status data, and pre-process the working data and status data;

[0065] Data processing module: used for calculating the laser edge coordinates, the first laser measurement length and the distance measurement scale according to the laser scanning data, for obtaining the transmission influence factor, for determining the second laser measurement length, and for obtaining the comprehensive measurement length according to the distance measurement scale;

[0066] Fusion model module: used for performing feature fusion on the edge coordinate data and the first laser measurement length by using an interactive multi-feature fusion method RCI and a stacked integrated classifier;

[0067] A measurement model module is used to construct a high-precision length measurement model according to the comprehensive measurement length, the distance measurement scale, the transmission influence factor and the first edge coordinate, and input the working data and status data to be measured into the high-precision length measurement model to obtain the measured length;

[0068] Intelligent supervision module: used for storing, viewing and managing the working data, the status data and the measured length.

[0069] The beneficial effects of the present invention are:

[0070] The present invention is a high-precision length measurement method and system based on laser technology. Compared with the prior art, the present invention has the following technical effects:

[0071] The present invention can improve the accuracy of laser technology length measurement by processing laser scanning data, constructing transmission influencing factors, image recognition, coordinate conversion, data fusion, and model building steps, thereby improving the efficiency and accuracy of high-precision length measurement based on laser technology. The laser technology length measurement is intelligent, which can greatly save resources and improve the length measurement efficiency. It can realize high-precision length measurement of the object to be measured, and quickly perform high-precision length measurement on the object to be measured, providing powerful technical support for modern industrial manufacturing, precision measurement, automatic control and other fields, and is of great significance to high-precision length measurement based on laser technology. It can adapt to the high-precision length measurement needs of different measurement systems and different users based on laser technology, and has certain universality. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 The figure is a flow chart of the steps of the high-precision length measurement method based on laser technology of the present invention. DETAILED DESCRIPTION

[0073] The present invention is further described below by means of specific embodiments. The illustrative embodiments and descriptions of the present invention are used to explain the present invention but are not intended to limit the present invention.

[0074] The high-precision length measurement method and system based on laser technology of the present invention comprises the following steps:

[0075] like Figure 1 As shown, in this embodiment, the following steps are included:

[0076] Acquire working data and status data of laser measurement, and pre-process the working data and the status data; the working data includes laser scanning data and camera images; the status data includes first status data and second status data;

[0077] Calculate the laser edge coordinates, the first laser measurement length and the distance measurement scale according to the laser scanning data, input the first state data into the transmission influence function to obtain the transmission influence factor, and determine the second laser measurement length according to the transmission influence factor and the first laser measurement length;

[0078] Processing the camera image to obtain pixel edge coordinates, converting the pixel edge coordinates into world edge coordinates according to the second state data, and matching the world edge coordinates with the laser edge coordinates to obtain first edge coordinates;

[0079] The first edge coordinate and the second laser measurement length are combined to obtain a laser image length, and the laser image length is adjusted using the distance measurement scale to obtain a comprehensive measurement length;

[0080] A high-precision length measurement model is constructed according to the comprehensive measurement length, the distance measurement scale, the transmission influence factor and the first edge coordinate, and the working data and status data to be measured are input into the high-precision length measurement model to obtain the measured length.

[0081] In this embodiment, the method for calculating the laser edge coordinates, the first laser measurement length and the distance measurement scale includes:

[0082] The laser scanner and the camera are used as a pair of scanning groups, and more than three pairs of scanning groups at different spatial positions are set to obtain the machine working data and status data;

[0083] Determine the reference center and coordinate system of the surface of the measured object, project the spectrum mark as a mark on the reference center and edge of the surface of the measured object during laser scanning, calculate the relative coordinates of the edge point and the reference center as the relative laser edge coordinates, and determine the laser edge coordinates according to the mean, mean square error and skewness of the relative laser edge coordinates of multiple scanning groups;

[0084] The laser scanning data are sorted in time series, the time point when the phase of the laser scanning data changes is set as the edge time point, the interval of the edge time point is calculated to obtain the scanning time, the rotation angle of the laser scanner is determined by the scanning time and the rotation speed of the laser scanner, the edge distance from the scanning point to the edge point is calculated according to the round-trip time of the laser beam corresponding to the edge time point, and the first laser measurement length is calculated according to the edge distance and the rotation angle of the laser scanner;

[0085] The distance measurement scale from the scanning point to the centroid of the reference object surface is calculated according to the round trip time of the laser beam corresponding to the time point when the laser scans to the centroid mark of the reference object surface;

[0086] In the actual evaluation, the length of the drive shaft produced in a factory was measured with high precision based on laser technology. Three pairs of scanning groups (left, middle, and right) were set up to obtain the machine working data and corresponding status data. During the laser scanning, the spectral mark was projected as a mark on the center and edge of the drive shaft axis to obtain the laser edge coordinates (unit: mm): edge point 1 (250.61, 250.19, 0.51), edge point 2 (25.91, 249.87, 1500.49); edge point 1 (250.33, 250.47, 0.02), edge point 2 (249.99, 250.15, 1500.13); edge point 1 (250.18, 250.22, -0.08), edge point 2 (250.03, 250.05, 1499.97), thus the laser edge coordinates are: edge point 1 (250.33, 250.13, -0.10), edge point 2 (250.14, 250.02, 1500.14);

[0087] The scanning rotation angle and the distance from the laser scanner to the edge point calculated according to the laser scanning data are: 45.01°, 1500.12mm, 2121.45mm for the left scanning point; 20°, 4319.38mm, 4319.19mm for the middle scanning point; 30°, 3000.33mm, 2597.84mm for the right scanning point; the corresponding first laser measurement length is: 1500.83mm for the left scanning point, 1499.75mm for the middle scanning point, and 1501.07mm for the right scanning point;

[0088] The distance measurement scales from the three laser scanners to the center point of the drive shaft are: 1677.24mm for the left scanning point, 4253.79mm for the middle scanning point, and 2704.08mm for the right scanning point.

[0089] In this embodiment, the method for obtaining the transmission impact factor includes:

[0090] The first state data is input into the transmission influence function to obtain the transmission influence factor, which is expressed as:

[0091]

[0092] Where Q T is the transmission influence factor, w1, w2, w3 are the transmission influence factor weight coefficients, Δf is the deviation between the actual frequency of the emitted laser and the preset frequency, f0 is the preset frequency of the emitted laser, ΔT is the temperature drift of the laser, k is the temperature wavelength influence coefficient, Δλmax is the deviation between the end value of the laser spectrum and the atmospheric spectrum, Δλ min is the deviation between the laser spectrum and the starting value of the atmospheric spectrum, Δλ is the range of the atmospheric spectrum;

[0093] The second laser measurement length is determined according to the transmission influence factor and the first laser measurement length, and the expression is:

[0094]

[0095] Where L2 is the second laser measurement length, L1 is the first laser measurement length, and w4 is the transmission influence factor weight;

[0096] In the actual evaluation, when high-precision length measurement of the drive shaft produced in a factory was performed based on laser technology, the rated frequency of the laser scanner was 20kHz, the actual frequency was 18kHz, the scanning wavelength was 790-910nm, the atmospheric visible spectrum was 380-780nm, the temperature drift of the laser was 0.5℃, and the temperature-wavelength influence coefficient was 0.1nm / ℃. The transmission influence factor was 0.99944, and the corresponding second laser measurement lengths were 1500.93mm, 1499.84mm, and 1501.16mm.

[0097] In this embodiment, the method for obtaining the pixel edge coordinates includes:

[0098] The OpenCV image processing library is used to read the camera image to be processed, the camera image is grayed to obtain a grayscale image, and the edge detection of the grayscale image is performed to identify the edge position of the grayscale image of the test piece;

[0099] The edge detector uses a Gaussian Laplace filter to perform Gaussian filtering on the grayscale image and perform Laplace transform to obtain the image edge information. The expression is:

[0100]

[0101] in is the Laplace Gaussian operator, is the Laplace operator, G(u,v) is a two-dimensional Gaussian function, σ is the standard deviation of the Gaussian function, g(u,v) is the image after Gaussian Laplace filtering, and f(u,v) is the input grayscale image;

[0102] A zero-crossing test is performed based on the edge information to obtain the specific edge position, a grayscale threshold is set, the absolute value of the difference between the grayscale values ​​of adjacent pixels on the four side lines and the two diagonal lines is calculated, and the pixels whose absolute value is greater than the grayscale threshold are determined to be located on the edge;

[0103] According to the edge position, the eight-connected region is used to trace the contour of the image in a counterclockwise direction. The expression of the eight-connected rule is:

[0104] N8(p)=N4(p)∪(u+1,v+1),(u+1,v-1),(u-1,v+1),(u-1,v-1)

[0105] Where N8(p) represents the eight-connected region of the starting point p of the boundary tracking, including point p and its corresponding upper, lower, left, right and four diagonal neighboring points; N4(p) represents the four-connected region of point p, including point p and its corresponding upper, lower, left and right neighboring points;

[0106] According to the grayscale calculation value, five distance intervals with different distances are divided to filter the contour points. The screening principle is:

[0107] M1>1-T,M2 <T,M4<T,M5> 1-T

[0108] Where M1, M2, M4, and M5 are the grayscale averages of regions 1, 2, 4, and 5, and T is the grayscale threshold;

[0109] Determine the selected contour points, use the moment function to calculate the geometric center, and extract the pixel edge coordinates based on the contour data;

[0110] In the actual evaluation, the camera pixel is 1920*1080, the camera focal length is 50mm, and the camera pitch angle is -10°, 0°, and 10°. After grayscale, edge detection, and contour screening, three groups of images are obtained to obtain three groups of pixel edge coordinates: left scan edge point 1 (650.12, 539.23) and edge point 2 (800.12, 689.23); middle scan edge point 1 (650.09, 539.25) and edge point 2 (800.16, 689.19); right scan edge point 1 (650.15, 539.19) and edge point 2 (800.08, 689.24).

[0111] In this embodiment, the method of converting the pixel edge coordinates into world edge coordinates includes:

[0112] The second state data includes the camera focal length, the camera pitch angle and the camera translation amount;

[0113] Convert pixel edge coordinates (u,v) to camera edge coordinates (X c ,Y c ,Z c ), the expression is:

[0114]

[0115] Where (u0, v0) is the coordinate of the image origin in the pixel coordinate system, is the scaling factor along the axis of the image coordinate system, f is the focal length of the camera, and x and y are the image coordinates;

[0116] The camera edge coordinates (X c ,Y c ,Z c ) is converted to world edge coordinates (X w ,Y w ,Z w ), the expression is:

[0117]

[0118] Where R is the rotation matrix, θ is the camera pitch angle, and T is the translation matrix;

[0119] In the actual evaluation, the world edge coordinates (unit: mm) are calculated according to the intrinsic and extrinsic parameters of the camera: edge point 1 (-155.88, -3.8, 1810.79), edge point 2 (1344.12, -3.8, 1810.79); edge point 1 (-155.90, -3.8, 1809.55), edge point 2 (1344.17, -3.8, 1809.55); edge point 1 (-155.76, -3.8, 1810.11), edge point 2 (1344.06, -3.8, 1810.11).

[0120] In this embodiment, the method for obtaining the first edge coordinates includes:

[0121] Calculate the comprehensive similarity between the world edge coordinates and the laser edge coordinates. The expression is:

[0122]

[0123] Where S w,l is the comprehensive similarity between the world edge coordinates and the laser edge coordinates, α and β are weight parameters, is the i-th world edge coordinate point A i and the jth laser edge coordinate point B j The Euclidean distance, n is the number of coordinate points, r is the buffer radius, d A is the distance from the world edge coordinate point to the origin, d B is the distance from the laser edge coordinate point to the origin, is the mean of the vectors corresponding to the world edge coordinate points, is the mean value of the vector corresponding to the laser edge coordinate point;

[0124] Matching edge coordinate points according to the comprehensive similarity between the world edge coordinates and the laser edge coordinates, and determining the first edge coordinate according to the mean, variance and skewness of the two edge coordinate points;

[0125] In the actual evaluation, the X coordinate of the world edge coordinate corresponds to the Z coordinate of the laser edge coordinate, and the Z coordinate of the world edge coordinate corresponds to the X coordinate of the laser edge coordinate. According to the similarity matching, the first edge coordinates are determined as follows: edge point 1 (1030.24, 123.17, -77.98) and edge point 2 (1030.15, 123.11, 142.13).

[0126] In this embodiment, the method for obtaining the laser image length includes:

[0127] Calculate the distance and angle between any two first edge coordinate points, and define the two first edge coordinate points and the corresponding distance and angle as a set of edge coordinate data;

[0128] Perform feature fusion on the edge coordinate data and the second laser measurement length to obtain the laser image length;

[0129] The interactive multi-feature fusion method RCI and stacked ensemble classifier are used for feature fusion;

[0130] The interactive multi-feature fusion method RCI includes: calculating the correlation, calculating the connection entropy, calculating the interaction factor and determining the evaluation criteria;

[0131] Analyze the correlation between the edge coordinate data and the second laser measurement length, and select the corresponding edge coordinate data, the correlation and the second laser measurement length to form a comprehensive measurement feature according to the correlation;

[0132] Calculate the connection entropy between the comprehensive measurement features and the nonlinear relationship between the measurement variables. The expression is:

[0133]

[0134] in is the connection entropy between comprehensive measurement features, c(·) is the density function, is a random variable, and its marginal function is F i is the candidate feature, n is the number of features;

[0135] The interaction factor is calculated to measure the interaction between the edge coordinate data feature and the second laser measurement length, and the variable with the largest interaction factor is selected. The expression is:

[0136]

[0137] in is the interaction factor, L2 is the second laser measurement length, H c (Ω s ,L2,c Edge) is the selected feature subset Ω when considering the second laser measurement length L2 s and edge coordinate data feature c Edge The joint entropy of c (Ω s ,c Edge ) is to consider only the selected feature subset Ω s and edge coordinate data feature c Edge The joint entropy of

[0138] Determine the evaluation criteria for candidate features, expressed as:

[0139]

[0140] Among them J RCI (L2) is the evaluation criteria for candidate features, F is all features, L2 is the second laser measurement length and edge coordinate data feature c Edge The correlation of is the redundancy, s is the selected feature subset Ω s Features in;

[0141] The optimal feature subset is obtained according to the evaluation criteria of the candidate features, and the optimal feature subset is input into the stacked ensemble classifier to obtain the laser image length;

[0142] The stacked ensemble classifier includes an input layer, a first classification layer, a second classification layer, and an output layer;

[0143] The first classification layer is composed of a deep neural network DNN and a lightweight gradient boosting machine LightGBM, which performs basic classification on the input data and outputs probability results; the second classification layer is composed of a support vector machine SVM, which performs meta-classification on the input probability results to determine the fusion results of the feature data, and the output layer outputs the comprehensive measurement length; the performance of the model is evaluated using F1 score, G mean, mean square error MSE and cross validation;

[0144] In the actual evaluation, the edge coordinates and the second laser measurement length were data fused to obtain the laser image lengths: 1500.52 mm on the left, 1499.97 mm in the middle, and 1500.63 mm.

[0145] In this embodiment, the method for obtaining the measured length includes:

[0146] According to the ranging scale, the laser image lengths of the scanning groups at different spatial positions are adjusted to obtain a comprehensive measurement length, and the comprehensive measurement set is formed with the transmission influencing factor, the ranging scale and the first edge coordinates, and the comprehensive measurement set is divided into a training set and a test set;

[0147] Construct a high-precision length measurement model, which includes an input layer, a base model layer, a strategy layer, and an output layer;

[0148] The base model layer is composed of two multi-layer perceptron models and two gradient boosting machine models in parallel;

[0149] In the multi-layer perceptron model: design two hidden layers, use the ReLU function as the activation function to accelerate training and reduce gradient disappearance, use the Cross-Entropy Loss function to solve the problem of the cross-entropy loss function deviating from the evaluation index and overconfidence, and use Adam to optimize the model weight parameters;

[0150] In the gradient boosting machine model: the mathematical model is constructed iteratively, the logarithmic loss function is used to measure the difference between the predicted probability distribution and the true distribution, and the learning rate is adjusted to control the optimization process;

[0151] The strategy layer uses the stacking method to input the prediction structure of the base model into the new model for prediction, obtains the prediction results, and uses the test set to evaluate the high-precision length measurement model;

[0152] Inputting the working data and status data of the machine to be measured into a high-precision length measurement model to obtain the measured length;

[0153] In the actual evaluation, the laser image length was adjusted according to the ranging scale (1677.24 mm on the left, 4253.79 mm in the middle, and 2704.08 mm on the right) to obtain the comprehensive measurement length: (1500.13 mm on the left, 1500.06 mm in the middle, and 1500.38 mm on the right);

[0154] The working data and status data to be measured in this group are input into the high-precision length measurement model to obtain a measured length of 1500.16mm, the nominal length of the drive shaft is 1500.15mm, the measurement accuracy of the high-precision length measurement method based on laser technology reaches 0.01mm, and the length measurement deviation is 0.00066%.

[0155] Second, the high-precision length measurement system based on laser technology, including:

[0156] Data acquisition module: including a laser transmitter, a laser receiver and a high-resolution camera; used to collect machine working data and status data, and pre-process the working data and status data;

[0157] Data processing module: used for calculating the laser edge coordinates, the first laser measurement length and the distance measurement scale according to the laser scanning data, for obtaining the transmission influence factor, for determining the second laser measurement length, and for obtaining the comprehensive measurement length according to the distance measurement scale;

[0158] Fusion model module: used for performing feature fusion on the edge coordinate data and the first laser measurement length by using an interactive multi-feature fusion method RCI and a stacked integrated classifier;

[0159] A measurement model module is used to construct a high-precision length measurement model according to the comprehensive measurement length, the distance measurement scale, the transmission influence factor and the first edge coordinate, and input the working data and status data to be measured into the high-precision length measurement model to obtain the measured length;

[0160] Intelligent supervision module: used for storing, viewing and managing the working data, the status data and the measured length.

[0161] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A high-precision length measurement method based on laser technology, characterized in that: The following steps are involved: S1, obtaining working data and status data of laser measurement, and preprocessing the working data and the status data; the working data includes laser scanning data and camera images; the status data includes first status data and second status data; S2, calculating the laser edge coordinates, the first laser measurement length and the distance measurement scale according to the laser scanning data, inputting the first state data into a transmission influence function to obtain a transmission influence factor, and determining the second laser measurement length according to the transmission influence factor and the first laser measurement length; S3, processing the camera image to obtain pixel edge coordinates, converting the pixel edge coordinates into world edge coordinates according to the second state data, and matching the world edge coordinates with the laser edge coordinates to obtain first edge coordinates; S4, fusing the first edge coordinate and the second laser measurement length to obtain a laser image length, and adjusting the laser image length using the distance measurement scale to obtain a comprehensive measurement length; S5. Construct a high-precision length measurement model according to the comprehensive measurement length, the distance measurement scale, the transmission influence factor and the first edge coordinate, and input the working data and status data to be measured into the high-precision length measurement model to obtain the measured length.

2. The high-precision length measurement method based on laser technology according to claim 1, characterized in that: The method for calculating the laser edge coordinates, the first laser measurement length and the distance measurement scale comprises: The laser scanner and the camera are used as a pair of scanning groups, and more than three pairs of scanning groups at different spatial positions are set to obtain the machine working data and status data; Determine the reference center and coordinate system of the surface of the measured object, project the spectrum mark as a mark on the reference center and edge of the surface of the measured object during laser scanning, calculate the relative coordinates of the edge point and the reference center as the relative laser edge coordinates, and determine the laser edge coordinates according to the mean, mean square error and skewness of the relative laser edge coordinates of multiple scanning groups; The laser scanning data are sorted in time series, the time point when the phase of the laser scanning data changes is set as the edge time point, the interval of the edge time point is calculated to obtain the scanning time, the rotation angle of the laser scanner is determined by the scanning time and the rotation speed of the laser scanner, the edge distance from the scanning point to the edge point is calculated according to the round-trip time of the laser beam corresponding to the edge time point, and the first laser measurement length is calculated according to the edge distance and the rotation angle of the laser scanner; The distance measurement scale from the scanning point to the centroid of the reference object surface is calculated according to the round trip time of the laser beam corresponding to the time point when the laser scans to the centroid mark of the reference object surface.

3. The high-precision length measurement method based on laser technology according to claim 1, characterized in that: The method for obtaining the transmission impact factor comprises: The first state data is input into the transmission influence function to obtain the transmission influence factor, which is expressed as: Where Q T is the transmission influence factor, w1, w2, w3 are the transmission influence factor weight coefficients, Δf is the deviation between the actual frequency of the emitted laser and the preset frequency, f0 is the preset frequency of the emitted laser, ΔT is the temperature drift of the laser, k is the temperature wavelength influence coefficient, Δλ max is the deviation between the end value of the laser spectrum and the atmospheric spectrum, Δλ min is the deviation between the laser spectrum and the starting value of the atmospheric spectrum, Δλ is the range of the atmospheric spectrum; The second laser measurement length is determined according to the transmission influence factor and the first laser measurement length, and the expression is: Wherein L2 is the second laser measurement length, L1 is the first laser measurement length, and w4 is the transmission influence factor weight.

4. The high-precision length measurement method based on laser technology according to claim 1, characterized in that: The method for obtaining the pixel edge coordinates comprises: The OpenCV image processing library is used to read the camera image to be processed, the camera image is grayed to obtain a grayscale image, and the edge detection of the grayscale image is performed to identify the edge position of the grayscale image of the test piece; The edge detector uses a Gaussian Laplace filter to perform Gaussian filtering on the grayscale image and perform Laplace transform to obtain the image edge information. The expression is: in is the Laplace Gaussian operator, is the Laplace operator, G(u,v) is a two-dimensional Gaussian function, σ is the standard deviation of the Gaussian function, g(u,v) is the image after Gaussian Laplace filtering, and f(u,v) is the input grayscale image; According to the edge information, a zero-crossing test is performed to obtain the specific edge position. According to the edge position, the eight-connected region is used to trace the image contour in a counterclockwise direction. The expression of the eight-connected rule is: N8(p)=N4(p)∪(i+1,v+1),(u+1,v-1),(u-1,v+1),(i-1,v-1) Where N8(p) represents the eight-connected region of the starting point p of the boundary tracking, including point p and its corresponding upper, lower, left, right and four diagonal neighboring points; N4(p) represents the four-connected region of point p, including point p and its corresponding upper, lower, left and right neighboring points; According to the grayscale calculation value, five distance intervals with different distances are divided to filter the contour points. The screening principle is: M1>1-T,M2 <T,M4<T,M5> 1-T Where M1, M2, M4, and M5 are the grayscale averages of regions 1, 2, 4, and 5, and T is the grayscale threshold; Determine the filtered contour points, use the moment function to calculate the geometric center, and extract the pixel edge coordinates based on the contour data.

5. The high-precision length measurement method based on laser technology according to claim 1, characterized in that: The method of converting the pixel edge coordinates into world edge coordinates comprises: The second state data includes the camera focal length, the camera pitch angle and the camera translation amount; According to the second state data, the camera internal parameter matrix and external parameter matrix are constructed to convert the pixel edge coordinates (u, v) into the camera edge coordinates (X c ,Y c ,Z c ), the camera edge coordinates (X c ,Y c ,Z c ) is converted to world edge coordinates (X w ,Y w ,Z w ).

6. The high-precision length measurement method based on laser technology according to claim 1, characterized in that: The method for obtaining the first edge coordinates includes: Calculate the comprehensive similarity between the world edge coordinates and the laser edge coordinates. The expression is: Where S w,l is the comprehensive similarity between the world edge coordinates and the laser edge coordinates, α and β are weight parameters, is the i-th world edge coordinate point A i and the jth laser edge coordinate point B j The Euclidean distance, n is the number of coordinate points, r is the buffer radius, d A is the distance from the world edge coordinate point to the origin, d B is the distance from the laser edge coordinate point to the origin, is the mean of the vectors corresponding to the world edge coordinate points, is the mean value of the vector corresponding to the laser edge coordinate point; The edge coordinate points are matched according to the comprehensive similarity between the world edge coordinates and the laser edge coordinates, and the first edge coordinates are determined according to the mean, variance and skewness of the two edge coordinate points.

7. The high-precision length measurement method based on laser technology according to claim 1, characterized in that: The method for obtaining the laser image length comprises: Calculate the distance and angle between any two first edge coordinate points, and define the two first edge coordinate points and the corresponding distance and angle as a set of edge coordinate data; Perform feature fusion on the edge coordinate data and the second laser measurement length to obtain the laser image length; The interactive multi-feature fusion method RCI and stacked ensemble classifier are used for feature fusion; The interactive multi-feature fusion method RCI includes: calculating the correlation, calculating the connection entropy, calculating the interaction factor and determining the evaluation criteria; Determine the evaluation criteria for candidate features, expressed as: Among them J RCI (L1) is the evaluation criteria for candidate features, F is all features, L2 is the second laser measurement length and edge coordinate data feature c Edge The correlation of is the redundancy, s is the selected feature subset Ω s The characteristics of is the interaction factor, H c (Ω s ,L2,c Edge ) is the selected feature subset Ω when considering the second laser measurement length L2 s and edge coordinate data feature c Edge The joint entropy of c (Ω s ,c Edge ) is to consider only the selected feature subset Ω s and the joint entropy of edge coordinate data feature c; The optimal feature subset is obtained according to the evaluation criteria of the candidate features, and the optimal feature subset is input into the stacked ensemble classifier to obtain the laser image length; The stacked ensemble classifier includes an input layer, a first classification layer, a second classification layer, and an output layer; The first classification layer is composed of a deep neural network DNN and a lightweight gradient boosting machine LightGBM, which performs basic classification on the input data and outputs probability results; the second classification layer is composed of a support vector machine SVM, which performs meta-classification on the input probability results to determine the fusion results of the feature data, and the output layer outputs the comprehensive measurement length; the F1 score, G mean, mean square error MSE and cross validation are used to evaluate the performance of the model.

8. The high-precision length measurement method based on laser technology according to claim 1, characterized in that: The method for obtaining the measured length comprises: According to the ranging scale, the laser image lengths of the scanning groups at different spatial positions are adjusted to obtain a comprehensive measurement length, and the comprehensive measurement set is formed with the transmission influencing factor, the ranging scale and the first edge coordinates, and the comprehensive measurement set is divided into a training set and a test set; Construct a high-precision length measurement model, which includes an input layer, a base model layer, a strategy layer, and an output layer; The base model layer is composed of two multi-layer perceptron models and two gradient boosting machine models in parallel; In the multi-layer perceptron model: design two hidden layers, use the ReLU function as the activation function to accelerate training and reduce gradient disappearance, use the Cross-Entropy Loss function to solve the problem of the cross-entropy loss function deviating from the evaluation index and overconfidence, and use Adam to optimize the model weight parameters; In the gradient boosting machine model: the mathematical model is constructed iteratively, the logarithmic loss function is used to measure the difference between the predicted probability distribution and the true distribution, and the learning rate is adjusted to control the optimization process; The strategy layer uses the stacking method to input the prediction structure of the base model into the new model for prediction, obtains the prediction results, and uses the test set to evaluate the high-precision length measurement model; The working data and status data of the machine to be measured are input into a high-precision length measurement model to obtain the measured length.

9. A high-precision length measurement system based on laser technology, used to perform the method according to any one of claims 1 to 8, characterized in that: include: Data acquisition module: including laser transmitter, laser receiver and high-resolution camera; Used to collect machine working data and status data, and pre-process the working data and status data; Data processing module: used for calculating the laser edge coordinates, the first laser measurement length and the distance measurement scale according to the laser scanning data, for obtaining the transmission influence factor, for determining the second laser measurement length, and for obtaining the comprehensive measurement length according to the distance measurement scale; Fusion model module: used for performing feature fusion on the edge coordinate data and the first laser measurement length by using an interactive multi-feature fusion method RCI and a stacked integrated classifier; A measurement model module is used to construct a high-precision length measurement model according to the comprehensive measurement length, the distance measurement scale, the transmission influence factor and the first edge coordinate, and input the working data and status data to be measured into the high-precision length measurement model to obtain the measured length; Intelligent supervision module: used for storing, viewing and managing the working data, the status data and the measured length.

Images