Self-learning method for appearance and size of polygonal section

Abstract

The invention relates to a self-learning method for the appearance and size of a polygonal section. A plurality of sensors are utilized to obtain depth information of different positions of the same cross section of an object, and a cross section appearance graph is formed through conversion of a two-dimensional coordinate system. Dead pixels and interference information are removed through filtering, inflection points are obtained according to the difference between edge inflection point positions, and the number of polygons is obtained through the number of the inflection points; intercepting 60%-90% of data segments in the middle of each segment according to the inflection points for linear fitting, and calculating data intersection points among the linear segments according to spatial positions among the linear segments to obtain intersection point coordinates of each edge of the polygon; calculating difference points needing to be inserted according to the coordinates of the intersection points and actual distance information, performing linear interpolation on each section by using inflection point information and the number of difference values, and finally performing connection after interpolation on each section, so as to obtain the outline graph of the whole polygonal section and the size of a line segment between each inflection point. The polygon section appearance self-learning method provided by the invention is short in measurement process and high in efficiency, and effectively improves the accuracy and real-time performance of measurement.

Description

Technical Field

[0001] The invention relates to the field of automatic detection of material shape and size, and in particular to a self-learning method for polygonal cross-section shape and size. Background Art

[0002] With the continuous improvement of industrial automation technology, the deepening of industrial upgrading, the continuous improvement of material production quality and quality requirements, the data detection and inspection methods in the material production and processing process are becoming more and more important. The original material processing process does not have the function of online dimension collection and learning. It is necessary to complete the dimension measurement of each point manually or offline after the material processing is completed or during the inspection process. The measurement process is time-consuming, inefficient, and has poor accuracy and real-time performance. Summary of the invention

[0003] The present invention provides a self-learning method for polygonal cross-section shape and size, so as to solve the problems of long time period, poor accuracy and poor real-time performance caused by the off-line measurement method existing in the prior art.

[0004] In order to achieve the above object, the technical solution of the present invention is as follows: a self-learning method for the shape and size of a polygonal cross-section, comprising the following steps:

[0005] Step 1: The present invention uses a plurality of line laser sensors evenly distributed on the circumference to obtain the depth information of the same cross-sectional shape of the corresponding object at different positions in real time;

[0006] Step 2: by establishing a two-dimensional spatial coordinate system, the depth information of different sensors is converted into coordinate information at different polar coordinate positions on the two-dimensional spatial plane, thereby forming a cross-sectional appearance diagram on the two-dimensional plane;

[0007] Step 3: Use filtering to remove bad pixels and interference information, and then obtain the inflection point coordinates based on the difference between the differential value of the inflection point position of the polygon edge and the differential value of other positions, and use the number of inflection point coordinates obtained to obtain the number of polygons of the corresponding cross section;

[0008] Step 4: segment the continuous cross-sectional shape data on all transformed two-dimensional planes according to the inflection point coordinates;

[0009] Step 5: intercept the data segment of 60%-90% in the middle of each segment, and perform linear fitting on the curve of each data segment;

[0010] Step 6: According to the spatial position relationship between the linear segments, the intersection points of the fitted linear segments and the linear segment data are calculated, so as to obtain the intersection coordinates of the fitted lines of the polygon;

[0011] Step 7: Based on the intersection coordinates between adjacent points and the actual distance information, calculate the interpolation points to be inserted with a certain determined size as the minimum resolution. Use the inflection point information and the number of interpolation points to perform linear interpolation for each segment. Finally, connect the segments after interpolation to obtain the contour map of the entire polygon cross-section and the dimensions of the line segments between each inflection point.

[0012] Step 8: On the time axis, repeat the processes of the above Steps 1 to 7, combine the obtained multiple groups of two-dimensional arrays to get the three-dimensional standard model corresponding to the shape of the material.

[0013] Further, the above Step 3 specifically includes:

[0014] 3.1: Extract the two-dimensional array R ** in the radius column in polar coordinates, perform low-pass filtering, obtain the concave points of the corresponding waveform, form a new array with the position values of the concave points, extract the first value of the concave point array, take the integer part, denote it as a, and intercept the first ** a values of the R radius array to form a new array. Connect the remaining array after interception and the array composed of the intercepted a values in the front-back order to form a new one-dimensional array;

[0015] 3.2: Obtain the convex points of the waveform of the new one-dimensional array formed in Step 3.1, and form a new array with the position values of the convex points.

[0016] Further, the specific steps of the above Step 4 include:

[0017] 4.1: Calculate the length of the one-dimensional array in Step 3, denote it as b. Add a in Step 3 to the new array formed by the position values of the convex points. For the values in this array that are greater than b, replace them with the values obtained by subtracting b from these values, and keep the other values in the array unchanged, thus forming a new array. Sort this array in ascending order of numerical values to obtain the original waveform R ** array of the convex point positions;

[0018] 4.2: Extract two adjacent values from the array of the convex point positions formed in Step 4.1, and combine the last value and the first value to obtain the array D.

[0019] Further, the above Step 5 specifically includes:

[0020] 5.1: Intercept a segment of 60 - 90% in the middle of the array D respectively according to the ratio. The intercepted arrays are A * , B * , C * , D * ;

[0021] 5.2: For A * , B *, C * , D * Four two-dimensional arrays in polar coordinates are respectively converted from polar coordinates to XY coordinates to form four new two-dimensional arrays A ** , B ** , C ** , D ** ;

[0022] 5.3: For A ** , B ** , C ** , D ** The four arrays are respectively subjected to a linear fitting once. For each line segment after fitting, the first and last two points are extracted, and the coefficient k and the constant b of the line are calculated;

[0023] Calculate the corresponding coefficient k and constant b using the formula;

[0024] k1 = (yy1 - yy2) / (xx1 - xx2)

[0025] b1 = yy1 - k1x1

[0026] Calculate the coefficients k1, b1; k2, b2; k3, b3; k4, b4 of the four line segments respectively.

[0027] Furthermore, the specific steps of the above step seven include:

[0028] 7.1: According to the several intersection points obtained in step six, obtain the number of sides of the polygon cross-section, and use the formula for the distance between two points in XY coordinates to find the side lengths of the corresponding sides;

[0029] 7.2: According to the first and last two points extracted from each line segment after linear fitting in step 7.1, the included angle between the corresponding line segment and the X-axis can be calculated. By extracting the included angle values between two adjacent line segments and the X-axis, the included angle between the two line segments can be calculated;

[0030] 7.3: Use the side lengths of each side obtained in step 7.2 to calculate the number of data points required: Using the number of points required for each side and the intersection point positions obtained in step 9, the standard shape of the polygon is obtained.

[0031] Furthermore, in the above step 2, the depth Z and X information are converted to the XY coordinates on the cross-section of the measured material. Specifically, in the conversion process, the reference coordinate values (x0, y0), (x1, y1), (x2, y2), (x3, y3) of the four sensors and the four angles β0, β1, β2, β3 are used:

[0032] (1) The data of each sensor is normalized according to the reference coordinates;

[0033] (2) Perform the conversion in polar coordinates for the arrays of the respective sensors above;

[0034] (3) According to the actual installation position and the four angle data obtained in the early stage, perform the rotation of the angles of the respective spatial positions;

[0035] (4) Connect the polar coordinate data of the above four sensors according to the spatial position angles to form two arrays;

[0036] (5) According to the value of θ * , for θ * and R * that form a two-dimensional array, perform re-sorting, obtain the centroid of the waveform formed by the corresponding two-dimensional array, and obtain the corrected two-dimensional data of the centroid after sorting.

[0037] Further, in the above step six, according to the parameters of two adjacent line segments, calculate the corresponding intersection points.

[0038] Further, in the above step eight, on the time axis, repeat the processes of the above steps one to seven, combine the obtained multiple groups of two-dimensional arrays, and obtain the three-dimensional standard model of the corresponding material shape.

[0039] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0040] 1. The present invention obtains the material shape data during the production process or inspection process in an online real-time manner, combines the shape data of multiple groups of sensors in terms of spatial position, thereby forming a material shape of a polygon in spatial position. For the material with a polygonal cross-section, according to the differential result, obtain the inflection point information, use the inflection point information to obtain the data of each side, perform linear fitting on the data of each side, find the intersection points of adjacent line segments of the fitted line segments, learn the inclination angles and side lengths of each line segment with the intersection points and perform linear first-order interpolation for each segment. Finally, combine each first-order interpolation line segment to form a standard polygon, which is convenient for calculating the error during flaw detection later.

[0041] 2. In the process of obtaining the shape feature parameters of the present invention, the problems of non-linear intersection, the existence of arcs, and possible defects at the inflection points of the actual material are considered. The combination of preliminary fuzzy positioning and linear intercepting, and the method of precise positioning by quadratic linear fitting is adopted, so that the calculation of the standard inflection points is more accurate, and at the same time, the influence of the arcs at the inflection points of the actual material, material defects, and acquisition bad points on the system analysis results is excluded.

[0042] 3. A learning method for the standard external contour of a material with a polygonal cross-section proposed by the present invention first performs coordinate transformation on a spatial plane using several groups of sensor data distributed in the circumferential direction; that is, by rotating the angles of different sensor data and intercepting the intersection points, the data within a certain arc range of each sensor are combined into a plane graph on the entire circumference. Secondly, based on the formed plane graph, the number of sides, side lengths, rotation angles, and standard external shapes of the polygon are learned. During the process, the problems of non-linear intersection at the inflection points, the existence of arcs, and possible defects of the actual material are considered.

[0043] 4. The measurement process and measurement effect of the present invention can be analyzed in real time, providing a good way to trace the reasons for the size data of materials with polygonal cross-sections in the later stage. The measurement process is short, the efficiency is high, effectively improving the accuracy and real-time performance of the measurement. Therefore, it has a wide range of applications and can calculate and learn the number of sides, side lengths, angles of the sides, and standard external shapes of the external shapes of materials with polygonal cross-sections. Detailed implementation manners

[0044] To facilitate the understanding of the present invention, the present invention will be described more comprehensively below with reference to the embodiments.

[0045] The present invention obtains the external shape data of materials during the production process or inspection process in an online real-time manner, and combines the external shape data of multiple groups of sensors in terms of spatial position. The system adopted by this method includes several sensors evenly distributed in the circumferential direction. The coordinate transformation on the spatial plane is performed using the sensor data, and the number of sides, side lengths, rotation angles, and standard external shapes of the polygon are learned based on the formed plane graph.

[0046] A self-learning method for the external shape and size of a polygonal cross-section provided by the present invention includes the following steps:

[0047] Step 1: The present invention uses several line laser sensors evenly distributed on the circumference to obtain the depth information of different positions of the same cross-section of the corresponding object in real time;

[0048] Step 2: Through the established spatial two-dimensional coordinate system, the depth information of different sensors is transformed into coordinate information at different polar coordinate positions on the spatial two-dimensional plane, thereby forming a cross-section external shape graph on the two-dimensional plane;

[0049] Step 3: Use filtering to remove bad points and interference information, and then obtain the inflection point coordinates based on the difference between the differential values at the inflection point positions of the polygon edge and the differential values at other positions. Use the number of obtained inflection point coordinates to obtain the number of polygons of the corresponding cross-section.

[0050] Step 4: Divide the continuous cross-section external shape data on the entire transformed two-dimensional plane according to the inflection point coordinates.

[0051] Step Five: Intercept the middle 60%-90% data segment of each segmented section, and perform linear fitting on the curves of each data segment.

[0052] Step Six: Calculate the intersection points of the linear segments and the linear segment data after fitting according to the spatial position relationship between the linear segments, so as to obtain the intersection coordinates of each segment of the polygon after fitting.

[0053] Step Seven: According to the intersection coordinates between adjacent ones, and according to the actual distance information, with a certain determined size as the minimum resolution, calculate the difference points to be inserted, and use the inflection point information and the number of differences to perform linear interpolation for each segment. Finally, connect each segment after interpolation to obtain the external shape diagram of the entire polygon cross-section and the size of the line segment between each inflection point.

[0054] Step Eight: On the time axis, repeat the processes of the above Step One to Step Seven, and combine the obtained multiple groups of two-dimensional arrays to obtain a three-dimensional standard model corresponding to the external shape of the material.

[0055] Embodiment: The following takes a quadrilateral material as an example to describe in detail the specific process of obtaining standard parameters and external shapes.

[0056] The sensor system evenly distributed in the circumferential direction learns the standard circular and quadrilateral sample rods, and calibrates to obtain the reference coordinate values (x0, y0), (x1, y1), (x2, y2), (x3, y3) of the four sensors and the four angles β0, β1, β2, β3.

[0057] Based on the quadrilateral material, a self-learning method for the external shape and size of a polygon cross-section provided by the present invention specifically includes the following steps:

[0058] Step One: The present invention uses several line laser sensors evenly distributed on the circumference to real-time obtain the depth information of different positions of the same cross-section of the corresponding object.

[0059] In this embodiment, four line laser sensors evenly placed in the circumferential direction are used to obtain the data in the corresponding depth Z and X directions.

[0060] Step Two: Through the established spatial two-dimensional coordinate system, convert the depth information of different sensors into the coordinate information at different polar coordinate positions on the spatial two-dimensional plane, so as to form the cross-section external shape diagram on the two-dimensional plane; specifically, convert the depth Z and X information into the XY coordinates on the cross-section of the measured material.

[0061] In this embodiment, using the reference coordinate values (x0, y0), (x1, y1), (x2, y2), (x3, y3) of the four sensors and the four angles β0, β1, β2, β3 obtained, the process is as follows:

[0062] (1) The data of each sensor is normalized according to the reference coordinates:

[0063] The array of the first sensor converted to the XY coordinates is: Y0 = Z0 - y0, X0 = X0 - x0;

[0064] The array of the second sensor converted to the XY coordinates is: Y1 = Z1 - y1, X1 = X1 - x1;

[0065] The array of the third sensor converted to the XY coordinates is: Y2 = Z2 - y2, X2 = X2 - x2;

[0066] The array of the fourth sensor converted to the XY coordinates is: Y3 = Z3 - y3, X3 = X3 - x3;

[0067] (2) The arrays of the respective sensors above are converted in polar coordinates:

[0068] The array of the first sensor: R0 is the radius array calculated in polar coordinates, and θ0 is the angle array calculated in polar coordinates;

[0069] The array of the second sensor: R1 is the radius array calculated in polar coordinates, and θ1 is the angle array calculated in polar coordinates;

[0070] The array of the third sensor: R2 is the radius array calculated in polar coordinates, and θ2 is the angle array calculated in polar coordinates;

[0071] The array of the fourth sensor: R3 is the radius array calculated in polar coordinates, and θ3 is the angle array calculated in polar coordinates;

[0072] (3) According to the actual installation position and the four angle data obtained in the early stage, the rotation of the respective spatial position angles is performed:

[0073] The array of the first sensor:

[0074] The array of the second sensor:

[0075] The array of the third sensor:

[0076] The array of the fourth sensor:

[0077] (4) The polar coordinate data of the above four sensors are connected according to the spatial position angles to form two arrays, namely R * and θ * , where R * = R0R1R2R3,

[0078] (5) According to the value of θ * , re - sort the two - dimensional array formed by θ * and R * , obtain the centroid of the waveform formed by the corresponding two - dimensional array, subtract the corresponding centroid value from the array, and obtain the two - dimensional data with the centroid corrected after sorting. The values of each column are respectively represented as θ ** and R ** ;

[0079] Through the above steps, a continuous two - dimensional array of the overall cross - section profile corresponding to the data of different sensors is formed. This array is the profile data of the actually measured cross - section;

[0080] Step Three: Use filtering to remove bad points and interference information, then obtain the inflection point coordinates according to the difference between the differential values at the inflection points of the polygon edge and the differential values at other positions, and use the number of obtained inflection point coordinates to obtain the number of polygons of the corresponding cross - section, specifically including:

[0081] 3.1 Extract the R ** radius column of the two - dimensional array in polar coordinates, perform low - pass filtering, obtain the concave points of the corresponding waveform, obtain the concave point position values to form a new array, extract the first value of the concave point array, take the integer part, denoted as a, and intercept the first a values of this R ** radius array to form a new array, and connect the remaining array after interception and the array composed of the intercepted a values in the front - to - back order to form a new one - dimensional array;

[0082] 3.2: Obtain the convex points of the waveform of the new one - dimensional array formed in step 3.1, and the obtained convex point position values form a new array. The size of this array is the number of sides of the calculated polygon cross - section. The number of sides obtained in this embodiment is 4;

[0083] Step Four: Segment the continuous cross - section profile data on the entire two - dimensional plane after conversion according to the inflection point coordinates. The specific steps include:

[0084] 4.1 Calculate the length of the one - dimensional array in step three, denoted as b, add a in step three to the new array formed by the convex point position values, for the values in this new array that are greater than b, replace them with the value obtained by subtracting b from this value, and keep the other values in this array unchanged, thereby forming a new array. Sort this new array in ascending order of numerical values to obtain the array of the R ** convex point positions;

[0085] 4.2: Extract two adjacent values of the convex point position array formed in step 4.1, and combine the last value and the first value. For example, for the four convex point positions in this embodiment, the first and second position values are used to determine the intercepted original two - dimensional data θ** and R ** for this segment of array A; the second and third position values are used to determine the original two-dimensional data θ to be intercepted ** and R ** for this segment of array B; the third and fourth position values are used to determine the original two-dimensional data θ to be intercepted ** and R ** for this segment of array C; the fourth and first position values are used to determine the original two-dimensional data θ to be intercepted ** and R ** for this segment of array. The source of this segment of data is relatively special. It is formed by connecting two parts of arrays, namely, the values from the fourth position to the end value of this array and the values from the start of the array to the first position to be intercepted, and is denoted as array D;

[0086] Step Five: Intercept the 60%-90% data segment in the middle of each segmented segment, and perform linear fitting on the curve of each data segment. The specific steps are as follows:

[0087] 5.1: In this embodiment, for the obtained array D, intercept 80% in the middle according to the proportion, that is, delete 10% of the data at the start of each array and 10% of the data at the end. This value is determined according to the amount of data and the size of the material and can be adjusted within a certain range. The intercepted arrays are respectively A * 、B * 、C * 、D * ;

[0088] 5.2: For A * 、B * 、C * 、D * the four two-dimensional arrays in polar coordinates obtained in Step 5.1, perform the conversion from polar coordinates to XY coordinates respectively to form four new two-dimensional arrays A ** 、B ** 、C ** 、D ** ;

[0089] 5.3: Perform linear fitting on A ** 、B ** 、C ** 、D ** the four arrays obtained in Step 5.2 respectively once. Extract the first and last two points from each fitted line segment and calculate the coefficient k and constant b of this line.

[0090] For example, extract the first and last two points (xx1, yy1) and (xx2, yy2) of array A ** and calculate the corresponding coefficient k and constant b using the following formula;

[0091] k1 = (yy1 - yy2) / (xx1 - xx2)

[0092] b1 = yy1 - k1x1

[0093] Calculate the coefficients k1, b1; k2, b2, k3, b3, k4, b4 of the four line segments respectively;

[0094] Step 6: According to the spatial position relationship between the linear segments, calculate the intersection points of the fitted linear segments and the linear segment data, so as to obtain the intersection coordinates of each segment of the polygon after fitting; the specific process is as follows:

[0095] Calculate the corresponding intersection points according to the parameters of two adjacent line segments:

[0096] In this embodiment, four intersection points of the four fitted line segments are calculated respectively, and the calculation process is as follows:

[0097] The intersection position of the first line segment and the second line segment

[0098]

[0099] In the same way, the intersection position of the second line segment and the third line segment can be obtained The intersection position of the third line segment and the fourth line segment The intersection position of the fourth line segment and the first line segment

[0100] Step 7: According to the intersection coordinates between adjacent ones, according to the actual distance information, with a certain determined size as the minimum resolution, calculate the difference points to be inserted, use the inflection point information and the number of differences, perform linear interpolation for each segment, and finally connect each segment after interpolation, so as to obtain the external shape of the entire polygon cross-section and the size of the line segment between each inflection point. The specific steps include:

[0101] 7.1 According to the four intersection points of the fitted straight lines obtained in Step 6, it can be obtained that the number of sides of the polygon cross-section is 4. Using the pairwise adjacent intersection values among the four intersection points, and the intersection value of the last and the first, use the formula for the distance between two points in the XY coordinate system to find the side lengths of the corresponding four sides;

[0102] 7.2: According to each line segment after linear fitting in Step 7.1, the first two points extracted from each line segment can be used to calculate the included angle between the corresponding line segment and the X-axis. Calculate the included angles between the four sides and the X-axis in this embodiment respectively. Extract the included angle values between two adjacent line segments and the X-axis, and the included angle between the two line segments can be calculated, so as to know the rotation angles of the corresponding figure in the spatial position with respect to the X-axis and the Y-axis;

[0103] 7.3: Using the side lengths of each edge obtained in step 7.2, with a resolution of 0.1 mm in this embodiment, this parameter is determined according to the size of the material and the design accuracy, calculate the number of data points required: Using the number of points required for each edge and the intersection positions obtained in step 9, extract the values of two adjacent intersections, the intersection values of the last and the first, and combine with the number of points required for each edge to perform linear interpolation, so as to obtain 4 line segments with the number of points basically consistent with the original data. Connect the two-dimensional arrays of the 4 line segments in the XY coordinate in sequence, and the standard shape of the polygon is obtained;

[0104] Step eight: On the time axis, repeat the processes of the above steps one to seven, combine the obtained multiple groups of two-dimensional arrays, and obtain a three-dimensional standard model corresponding to the shape of the material.

[0105] The above embodiments only represent several implementation manners of the present invention. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several deformations and improvements can still be made, and these all belong to the protection scope of the present invention.

Claims

1. A method for self-learning the shape and size of a polygonal cross section, characterized by: The following steps are involved: Step 1: The present invention uses a plurality of line laser sensors evenly distributed on the circumference to obtain the depth information of the same cross-sectional shape of the corresponding object at different positions in real time; Step 2: by establishing a two-dimensional spatial coordinate system, the depth information of different sensors is converted into coordinate information at different polar coordinate positions on the two-dimensional spatial plane, thereby forming a cross-sectional appearance diagram on the two-dimensional plane; Step 3: Use filtering to remove bad pixels and interference information, and then obtain the inflection point coordinates based on the difference between the differential value of the inflection point position of the polygon edge and the differential value of other positions, and use the number of inflection point coordinates obtained to obtain the number of polygons of the corresponding cross section; Step 4: segment the continuous cross-sectional shape data on all transformed two-dimensional planes according to the inflection point coordinates; Step 5: intercept the data segment of 60%-90% in the middle of each segment, and perform linear fitting on the curve of each data segment; Step 6: According to the spatial position relationship between the linear segments, the intersection points of the fitted linear segments and the linear segment data are calculated, so as to obtain the intersection coordinates of the fitted lines of the polygon; Step 7: According to the coordinates of the adjacent intersection points and the actual distance information, with a certain size as the minimum resolution, calculate the difference points to be inserted, use the inflection point information and the number of differences to perform linear interpolation of each segment, and finally connect each segment after interpolation to obtain the outline of the entire polygonal cross section and the size of the line segment between each inflection point; Step 8: Repeat the process from step 1 to step 7 on the time axis, combine the multiple sets of two-dimensional arrays obtained, and obtain a three-dimensional standard model of the corresponding material shape.

2. A method for self-learning the shape and size of a polygonal cross section according to claim 1, characterized in that: The step three specifically includes: 3.1: Extract the two-dimensional array R in polar coordinates ** Radius column, low-pass filter, get the concave point of the corresponding waveform, get the concave point position value to form a new array, extract the first value of the concave point array, find the integer, record it as a, intercept the R ** The first a values ​​of the radius array form a new array, and the remaining array after truncation and the array composed of the truncation a values ​​are newly connected in a back-to-front order to form a new one-dimensional array; 3.2: Obtain salient points from the new one-dimensional array waveform formed in step 3.1, and obtain a new array formed by the salient point position values.

3. A method for self-learning the shape and size of a polygonal cross section according to claim 2, characterized in that: The specific steps of step 4 include: 4.1: Calculate the length of the one-dimensional array in step 3, record it as b, add the new array formed by the convex point position value to a in step 3, and obtain a new array. The values ​​in the array that are greater than b are replaced by the value of b minus the value of b. The other values ​​in the array remain unchanged, thus forming a new array. Sort the array from small to large values ​​to obtain the original waveform R ** An array of bump locations; 4.2: Extract two adjacent values ​​of the salient point position array formed in step 4.1, combine the last value with the first value, and obtain array D.

4. A method for self-learning the shape and size of a polygonal cross section according to claim 3, characterized in that: The step five specifically includes: 5.1: Cut off the middle 60-90% of array D in proportion, and the cut arrays are A * , B * , C * , D * ; 5.2: For A obtained in step 7 * , B * , C * , D * The four two-dimensional arrays in polar coordinates are converted from polar coordinates to XY coordinates to form four new two-dimensional arrays A in XY coordinates. ** , B ** , C ** , D ** ; 5.3: A obtained in step 5.2 ** , B ** , C ** , D ** Each of the four arrays is linearly fitted once, and the first two points of each line segment after fitting are extracted to calculate the coefficient k and constant b of the line; The formula calculates the corresponding coefficient k and constant b; k1=(yy1-yy2) / (xx1-xx2) b1=yy1-k1x1 Calculate the coefficients of the four line segments k1, b1; k2, b2, k3, b3, k4, b4 respectively.

5. A method for self-learning the shape and size of a polygonal cross section according to claim 4, characterized in that: The step seven specifically comprises the following steps: 7.1: According to the number of intersection points obtained in step 6, obtain the number of sides of the polygonal cross section, and use the formula of the distance between two points under the XY coordinates to calculate the length of the corresponding side; 7.2: After linear fitting of each line segment in step 7.1, the first two points extracted from each line segment can be used to calculate the angle between the corresponding line segment and the X-axis. The angle between two adjacent line segments and the X-axis can be extracted to calculate the angle between the two line segments. 7.3: Using the length of each side obtained in step 7.2, calculate the required number of data points: Using the number of points required for each side and the intersection position obtained in step 9, the standard shape of the polygon is obtained.

6. A method for self-learning the shape and size of a polygonal cross section according to claim 5, characterized in that: In step 2, the depth Z and X information are converted to the XY coordinates on the cross section of the material being measured. In the specific conversion process, the reference coordinate values ​​(x0, y0), (x1, y1), (x2, y2), (x3, y3) of the four sensors and the four angles β0, β1, β2, β3 are used: (1) Each sensor data is normalized according to the reference coordinates; (2) converting the arrays of the above sensors into polar coordinates; (3) Rotate the spatial position angles according to the actual installation position and the four angle data obtained in the early stage; (4) Connect the polar coordinate data of the above four sensors according to the spatial position angle to form two arrays; (5) According to θ * The numerical value of θ * and R * The formed two-dimensional array is re-sorted, the center of gravity of the waveform formed by the corresponding two-dimensional array is obtained, and the two-dimensional data with the center of gravity corrected after sorting is obtained.

7. A method for self-learning the shape and size of a polygonal cross section according to claim 6, characterized in that: In the step six, the corresponding intersection point is calculated according to the parameters of two adjacent line segments.

8. A method for self-learning the shape and size of a polygonal cross section according to claim 7, characterized in that: In the step eight, on the time axis, the process from step one to step seven is repeated, and the obtained multiple groups of two-dimensional arrays are combined to obtain a three-dimensional standard model corresponding to the material shape.

Images