Online Visual Detection Device and Method

Abstract

The present invention discloses an on-line visual detection device and method. It includes: using a surface structured light scanning camera to scan the coil mold to identify the initial position; winding according to the coil winding process to obtain the first scanning data during the winding process. After the first part of the coil is wound, the mold is turned; the surface structured light scanning camera is used to re-identify the initial position, and the remaining part of the coil is wound according to the coil winding process to obtain the second scanning data; according to the initial position, the first scanning data and the second scanning data, calculate the coil width, segment spacing, and major and minor axes of each section and each layer of the coil to obtain the detection result. Through the on-line visual detection of the high-voltage coil, the measurement results of each section and each layer of the wound coil can be obtained. The structure is simple, the cost is low, and the detection is accurate, realizing fully automated work.

Description

Technical Field

[0001] The present invention relates to the technical field of coil manufacturing, and particularly to an on-line vision detection device and method. Background Art

[0002] Currently, during the winding process of high-voltage coils of transformers, traditional dimension measurement methods mainly rely on manual measurement. After the high-voltage coil is wound by a machine, it is necessary for workers to cooperate with the machine to wind a white winding film on the wound coil. After winding the white winding film, workers use a caliper to measure the long and short axes, width, and segment spacing dimensions of the coil. Since the coil size is too large (the conventional maximum size is 3600 mm (length) * 300 mm (width) * 670 mm (height)) and there are many required detection positions, it is difficult to measure with a manual caliper. Therefore, a vision system is added to take pictures and detect the above dimensions to replace manual detection.

[0003] In the prior art, in the patent with the application number CN202311264670Y and the patent name: An on-line vision detection device and method for process dimensions during the winding process of transformer coils, a vision detection device is provided. This device is a transformer coil winding device integrating a six-axis robot and a tracking three-dimensional scanning system. The main structure includes a gantry, a transformer coil winding mechanism, a support frame, a slider, a six-axis robot, and an optical tracker. The six-axis robot is suspended at the lower end of the slider on the support rod, and a three-dimensional scanning system is installed at the front end for real-time scanning and recognition. The data is transmitted to the industrial control computer display through a PLC to achieve real-time display and operation control. However, the mechanical structure required by the device is very complex and relatively expensive, and the suspended scanning device is prone to interfering with workers winding the coil. Therefore, it is not convenient to use in the process. Summary of the Invention

[0004] Therefore, the purpose of the present invention is to provide an on-line vision detection device and method, which can obtain the edge of the coil through mobile photographing detection and output the measurement of the width, long and short axes of each section and each layer of the wound coil in real time.

[0005] To achieve the above purpose, an on-line vision detection method for process dimensions during the winding process of a high-voltage coil of the present invention includes:

[0006] Using a structured light scanning camera to scan the coil mold and identify the initial point;

[0007] Winding according to the coil winding process to obtain the first scanning data during the winding process. After the winding of the first part of the coil is completed, the mold is rotated;

[0008] Using the structured light scanning camera to re-identify the initial point, winding the remaining part of the coil according to the coil winding process, and obtaining the second scanning data;

[0009] Based on the initial position, the first scan data, and the second scan data, calculate the coil width, segment spacing, and major and minor axes of each segment and each layer of the coil to obtain the detection result.

[0010] Further preferably, before identifying the initial position, it also includes correcting the relative parallelism between the coil winding assembly and the structured light scanning camera, and the correction includes the following process:

[0011] Collect multiple groups of three-dimensional point cloud data when the mold rotates one week;

[0012] Perform cylindrical fitting on each group of three-dimensional point cloud data respectively to obtain multiple axes;

[0013] Optimize the obtained multiple axes, and the optimized one is used as the point cloud axis;

[0014] Calculate the included angle between the point cloud axis and the axis of the structured light scanning camera, and perform compensation to obtain the relative parallelism correction result between the coil winding assembly and the structured light scanning camera.

[0015] Further preferably, when performing cylindrical fitting on each group of three-dimensional point cloud data respectively, the fitting formula is calculated according to the following formula:

[0016]

[0017] Wherein, represents the center of all axes obtained after cylindrical fitting (i.e., the centroid of the cylinder), N is the number of axes, is a point on each cylindrical axis (i.e., the center point of the fitted cylinder); is the axis direction of the fitted cylinder, is the eigenvector corresponding to the minimum eigenvalue; is the average distance from all axes to the centroid, dist i is the distance from the axis of each group of three-dimensional data to the centroid.

[0018] Further preferably, after correction, it also includes verifying the x-direction parallelism of the field of view of the structured light scanning camera according to the following process;

[0019] Within the field of view of the structured light scanning camera, horizontally move the calibration plate to two positions;

[0020] According to the poses of the calibration plate at two positions in the camera coordinate system, calculate the θ x ;

[0021] Judge the included angle θ xWhether it meets the coil process requirement range. If it meets, the parallelism of the surface structured light scanning camera in the x-direction meets the requirement. If it does not meet, readjust and verify.

[0022] Further preferably, after calibration, it further includes verifying the parallelism between the surface structured light scanning camera and the coil mold when the surface structured light scanning camera moves horizontally according to the following process;

[0023] Place the calibration plate at the same position on the square shaft and move the surface structured light scanning camera horizontally to two different positions;

[0024] When the surface structured light scanning camera is at the two positions, transfer the point cloud data obtained by scanning the calibration plate to the same position by matrix transformation;

[0025] Perform difference calculation on the point cloud data of the calibration plate taken at the two converted positions to obtain the X-direction deviation and Z-direction deviation of the parallelism with the coil mold.

[0026] Further preferably, the coil winding process includes the number of coil segments required, the winding sequence of each coil segment, the number of layers required for each coil segment, the width of each coil segment, and the interval between adjacent coil segments; among them, according to the number of coil segments required, the coil as a whole is divided into odd-numbered coil segments and even-numbered coil segments, and the scanning data during the winding process of the odd-numbered coil segments is recorded as the first scanning data; the scanning data during the winding process of the even-numbered coil segments is recorded as the second scanning data.

[0027] Further preferably, when calculating the coil width of each layer of each coil segment according to the initial position, the first scanning data, and the second scanning data, it includes:

[0028] Calculate the center position of the coil to be wound according to the initial position and the coil winding process;

[0029] Move the surface structured light scanning camera to the center position of each layer of the coil to be measured in each segment, and collect the scanning data of the area to be measured within the field of view;

[0030] Extract the grayscale image and depth data of the area where the coil is wound at the current position, use edge extraction to obtain the left and right boundaries, and calculate the distance between the two boundaries as the coil width.

[0031] Further preferably, when calculating the segment spacing, according to the initial positions identified before and after, splice the first scanning data and the second scanning data, and extract the odd-numbered coil segment boundary in the first scanning data and the even-numbered coil segment boundary in the second scanning data;

[0032] Then the segment spacing d = d1 - d2 - d3;

[0033] Among them, d1 is the distance between the center positions of the target odd segment and an adjacent target even segment coil; d2 is the distance between the left boundary line of the target odd segment and the center position of the target odd segment coil; d3 is the distance between the right boundary line of the target even segment and the center position of the target even segment coil.

[0034] Further preferably, when calculating the major and minor axes, according to the grayscale image and depth data of the coil winding area at the current position, calculate the average value of the depth data z1 and the average value of the die depth data z2 when the coil is not wound.

[0035] Major and minor axes = (z1 - z2 + D) * 2; where D is the die radius.

[0036] The present invention also provides an on-line visual inspection device for process dimensions during high-voltage coil winding, including:

[0037] A surface structured light scanning camera for acquiring scanning data during coil winding;

[0038] A horizontal driving component for driving the surface structured light scanning camera to perform translational motion;

[0039] A coil winding component for driving the die to rotate for coil winding;

[0040] A controller for controlling the horizontal driving component and the coil winding component, and calculating the detection result according to the above on-line visual inspection method for process dimensions during high-voltage coil winding.

[0041] The on-line visual inspection device and method disclosed in the present application have at least the following advantages compared with the prior art:

[0042] Through the parallelism correction of the coil winding component and the surface structured light scanning camera, and two accuracy verifications when the calibration plate is fixed and the camera moves alternately, the accuracy of the camera field of view and the parallelism accuracy between the camera and the die are verified; ensuring the accuracy of subsequent calculation of coil data.

[0043] By moving the camera, calculate the center position of the coil to be wound for each segment and each layer, use visual inspection to obtain the edge of the coil, and calculate the measurement and real-time output of the width, segment spacing, major and minor axes of the coil wound for each segment and each layer. The structure is simple and the cost is low, completely replacing manual operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a schematic structural diagram of an on-line visual inspection device provided by the present invention;

[0045] Figure 2 It is a schematic flow diagram of an on-line visual inspection method provided by the present invention;

[0046] Figure 3Schematic diagram of the coil winding process of the present invention;

[0047] Figure 4 Point cloud map for calibrating the front shaft and fitting the mold;

[0048] Figure 5 Point cloud map for calibrating the rear shaft and fitting the mold;

[0049] Figure 6 Image collected during the verification process of the parallelism between the camera X-axis and the square shaft;

[0050] Figure 7 Two sets of point cloud data before and after the camera translation transformation;

[0051] Figure 8 Schematic diagram of the mold after inversion of the present invention.

[0052] In the figure:

[0053] 1. 3D structured light camera; 2. Horizontal movement guide rail device; 3. Optical detection electrical cabinet; 4. Spool positioning wedge; 5. High-voltage coil mold; 6. Square shaft. Specific embodiments

[0054] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0055] As Figure 1 shown, an on-line vision detection device for process dimensions during the winding of a high-voltage coil provided in an embodiment of one aspect of the present invention includes: a structured light scanning camera, a horizontal drive assembly, a coil winding assembly, and a controller; wherein, the structured light scanning camera uses a 3D structured light camera 1; the horizontal drive assembly uses a horizontal movement guide rail device 2, and the coil winding assembly includes a spool positioning wedge 4, a square shaft 6, and a high-voltage coil mold 5;

[0056] The 3D structured light camera 1 is installed on the horizontal movement guide rail device 2 and is driven by the horizontal movement guide rail device 2 to scan the entire high-voltage coil mold, identify the initial points of the high-voltage coil mold, and store the three-dimensional data when the high-voltage coil mold winds the coil. According to the process dimensions issued by MES, the controller calculates the center position of each section and each layer of the coil, providing photographing points for the subsequent measurement of the length and short axes, width, and section spacing of the coil. Among them, the width, length, and short axes of each section and each layer of the wound coil can be output in real time after moving and photographing, and the section spacing of each section and each layer is measured by the section spacing measurement method.

[0057] The spool positioning wedge is used to fix the center of the high-voltage coil. Since there is a gap between the high-voltage coil mold 5 and the central square shaft 6, during the coil winding process, the whole coil will vibrate and shift randomly up, down, left and right. To avoid the detection error caused by this vibration and shift, after passing through the central square shaft, a spool positioning wedge needs to be installed to ensure that the central position of the high-voltage coil does not randomly jump and shift during the winding process.

[0058] The present invention also provides an on-line vision detection method for process dimensions during the winding process of a high-voltage coil, including the following steps:

[0059] S1. Use a surface structured light scanning camera to scan the coil mold and identify the initial point;

[0060] It should be noted that as Figure 3 shown, the initial point is the fixed edge position at one end of the mold. When the mold is flipped, the other end of the mold is at the edge position at this time, and it is necessary to re-identify this as the initial point.

[0061] Since the installation of the guide rail and the camera cannot ensure relative parallelism with the winding device, the angle between the camera and the winding device will affect the vision detection accuracy. Therefore, before identifying the initial point, it also includes correcting the relative parallelism between the coil winding assembly and the surface structured light scanning camera. The correction includes the following process:

[0062] Collect multiple groups of three-dimensional point cloud data when the mold rotates one week;

[0063] Perform cylindrical fitting on each group of three-dimensional point cloud data respectively to obtain multiple axes, and this axis is the central axis of the cylinder obtained by fitting;

[0064] Optimize the obtained multiple axes, and the optimized axis is used as the point cloud axis;

[0065] Calculate the included angle between the point cloud axis and the axis of the surface structured light scanning camera, and perform compensation to obtain the relative parallelism correction result between the coil winding assembly and the surface structured light scanning camera.

[0066] During specific correction, first, place the cylindrical mold on the square shaft, collect the three-dimensional point cloud when the mold rotates one week (calculated according to the field of view of the camera, rotating 30° each time), perform cylindrical fitting on the 12 groups of collected three-dimensional point clouds respectively to obtain 12 groups of cylindrical axes, and perform fitting optimization on the 12 groups of cylindrical axes to obtain the final axis. The fitting optimization of the cylindrical axis is shown in formula (1).

[0067] Formula (1)

[0068] Among them, represents the center of all axes obtained after cylindrical fitting (i.e., the centroid of the cylinder),N is the number of axes, is a point on each cylindrical axis (i.e., the center point of the fitted cylinder); is the axis direction of the fitted cylinder, is the eigenvector corresponding to the minimum eigenvalue; is the average distance from all axes to the centroid, dist i is the distance from the axis of each group of three-dimensional data to the centroid.

[0069] When performing fitting, first calculate the center point c of the starting points of all axes using the xyz coordinates of all three-dimensional data points, and select the eigenvector corresponding to the minimum eigenvalue as the axis direction d of the fitted cylinder; combine the average distance from the axis to the centroid; that is, the final axis obtained by optimizing the axis fitting according to the three parameters c, d, and r

[0070] Cylinder.

[0071] Then, according to the included angle between the obtained point cloud axis and the camera coordinate axes and perform compensation. After compensation, the camera coordinate axes are made parallel to the square axis for more accurate calculation of the long and short axes, width, and segment spacing of the coil.

[0072] The poses of the square axis and the fitted die point cloud before and after calibration in the camera coordinate system are as Figure 4 shown.

[0073] After calibration, two cases are considered for the accuracy verification of the vision detection device: 1) Within the camera's field of view, move the calibration plate parallel to the edge of the square axis to two different positions to ensure the parallelism in the x-axis direction under the camera's field of view; 2) Within the camera's field of view, fix the calibration plate at a specific position and move the camera with the horizontal guide rail to verify the parallelism between the horizontal moving platform and the square axis to ensure the overall measurement accuracy. The following describes the specific operation plans for the above two accuracy verifications. Since the width of the coil mainly depends on the data in the X direction of the camera, only by ensuring the parallelism between the X axis and the square axis can the accuracy of the width and segment spacing data be guaranteed. The test accuracy of the long and short axes of the coil depends on the accuracy in the Z direction. Only by verifying the parallelism between the horizontal moving platform and the square axis (X direction), as well as the measurement accuracy in the XYZ three directions, can the accuracy requirements for the measurement of the long and short axes of the coil be guaranteed later.

[0074] As Figure 6 shown, verify the parallelism in the x-axis direction of the structured light scanning camera's field of view according to the following process;

[0075] Within the field of view of the structured light scanning camera, move the calibration plate horizontally to two positions;

[0076] According to the pose of the calibration plate in the camera coordinate system, calculate the included angle between the x-axes of the calibration plate in the camera coordinate system before and after movementθ x ;

[0077] Determine the included angle to see if it meets the range requirements of the coil process. If it does, the parallelism of the surface structured light scanning camera in the x-direction meets the requirements. If not, readjust and verify.

[0078] That is, attach the calibration board to the edge of the square shaft, move the calibration board to two positions within the camera's field of view, calculate the poses of the calibration board during the two movements in the camera coordinate system, and use formula (2) to obtain the included angle between the two poses of the calibration board before and after the movement . If the included angle between the two movement poses does not exceed 0.085°, the parallelism in the x-direction is good within the camera's field of view, and accurate coil width and segment spacing detection data can be provided.

[0079] Formula (2)

[0080] Wherein is the included angle between the poses during the two movements, is the pose of the calibration board at position 1, is the pose of the calibration board at position 2. According to the requirements of the coil winding process, it is qualified if the included angle in the x-direction between the two movements is within 0.085° after conversion. Therefore, the accuracy requirements for the first case can be calculated using this method.

[0081] Such as Figure 7 shown, for the accuracy requirements of the second case: verify the parallelism between the surface structured light scanning camera and the coil mold during horizontal movement according to the following process;

[0082] Place the calibration board at any fixed position on the square shaft and move the surface structured light scanning camera horizontally to two different positions;

[0083] When the surface structured light scanning camera is at the two positions, transfer the point cloud data obtained by scanning the calibration board to the fixed position using matrix transformation;

[0084] Calculate the difference between the calibrated board point cloud data captured at two positions after transformation to obtain the X-direction deviation and Z-direction deviation of the parallelism between the structured light scanning camera and the coil mold when the camera moves horizontally. That is, after the calibrated board is set at a fixed position, when the camera is at the first position, the first point cloud data of the calibrated board is captured; when the camera is at the second position, the second point cloud data of the calibrated board captured by the camera at this time is subjected to coordinate transformation and transformed into the coordinate system of the first point cloud data. At this time, the coordinate systems of the first point cloud data and the second point cloud data are the same, and the difference calculation can be performed. The difference in the X direction of any corner point of the calibrated board in the two point cloud data can be used as the X-direction deviation of the parallelism of the coil mold; similarly, the difference in the Z direction of any corner point in the two point cloud data is used as the Z-direction deviation of the parallelism of the coil mold.

[0085] Place the calibrated board at the same position on the square shaft, and the camera moves to two different positions through the horizontal guide rail to obtain the point cloud data; only transform the calibrated board point cloud data to the same position through the translation matrix, and calculate the difference between the point cloud data of the calibrated board identification points after transformation. The maximum deviation of the identification points captured by the camera before and after the guide rail movement is within 1 mm in the X direction and within 1.5 mm in the Z direction. Therefore, the width and segment spacing detection accuracy of the vision detection system under ideal imaging is within 1 mm, and the long and short axis accuracy within 1.5 mm is considered qualified.

[0086] S2. Wind according to the coil winding process to obtain the first scan data during the winding process. When the winding of the first part of the coil is completed, turn the mold.

[0087] S3. Use a surface structured light scanning camera to re-identify the initial position, and wind the remaining part of the coil according to the coil winding process to obtain the second scan data.

[0088] The coil winding process includes the requirements for the number of coil segments, the winding order of each coil segment, the requirements for the number of layers of each coil segment, the width of each coil segment, and the interval between adjacent coil segments; among them, according to the requirements for the number of coil segments, the coil is divided into odd-numbered coil segments and even-numbered coil segments as a whole. The scan data during the winding process of the odd-numbered coil segments is recorded as the first scan data; the scan data during the winding process of the even-numbered coil segments is recorded as the second scan data.

[0089] S4. Calculate the coil width, segment spacing, and major and minor axes of each segment and each layer of the coil based on the initial position, the first scan data, and the second scan data to obtain the detection result. The coil width refers to the dimension of each segment and each layer of the coil in the axial direction during the coil winding process. The segment spacing is the distance between adjacent segments of the coil; the major and minor axes represent the diameter of the coil. When the mold is cylindrical according to the coil model requirements, the major axis and the minor axis are equal and are both the diameter of the mold after the coil is wound; when the model requires the mold to be elliptical, the major axis and the minor axis are not equal, and in this case, this method can be used to calculate the major axis and the minor axis separately.

[0090] Further preferably, when calculating the coil width of each segment and each layer of the coil based on the initial position, the first scan data, and the second scan data, it includes:

[0091] Calculate the center position of the coil to be wound according to the initial position and the coil winding process;

[0092] Move the structured light scanning camera to the center position of each segment and each layer of the coil to be measured, and collect the scan data of the area to be measured within the field of view;

[0093] Extract the grayscale image and depth data of the area where the coil is wound at the current position, use edge extraction to obtain the left and right boundaries, and calculate the distance between the two boundaries as the coil width. For example, Figure 8 in, the measurement center position of each layer of segment Ⅰ = the initial position of the mold - (coil insulation height + coil width of each layer of segment Ⅰ / 2). After the camera moves to the position to be measured, according to the camera's horizontal x data image (a certain width is extended outside the coil process width) to completely cover and frame the area to be measured, use the grayscale image and the depth z data image of the area to be measured to accurately extract the area where the coil is wound at the current position, use edge extraction to obtain the left and right boundaries, and calculate the distance between the two boundaries as the coil width.

[0094] Further preferably, when calculating the segment spacing, splice the first scan data and the second scan data according to the initially identified positions before and after, and extract the odd-segment coil boundaries in the first scan data and the even-segment coil boundaries in the second scan data;

[0095] Then the segment spacing d = d1 - d2 - d3;

[0096] where, d1 is the distance between the center positions of the target odd segment and an adjacent target even segment of the coil; d2 is the distance between the left boundary line of the target odd segment and the center position of the target odd segment of the coil; d3 is the distance between the right boundary line of the target even segment and the center position of the target even segment of the coil.

[0097] According to the high-voltage coil winding process, after each layer of the odd or even segments of the coil is wound, reverse the winding mold and wind the remaining segments. The schematic diagram of the reversed mold is as shown in Figure 8As shown in the figure. Taking the example of winding the odd segments first, after the i, ii, and iii layers of segments I and III are all wound, the entire mold is reversed, and the i, ii, and iii layers of the even segments II and IV are wound. Therefore, the measurement of the adjacent segment spacing for each layer can only be obtained after the mold is reversed. The process for measuring the segment spacing is to re-identify the initial position of the mold after reversal, and calculate the center photographing position of the layer to be measured after reversal through the drawing. Retrieve the adjacent images of the current segment and current layer stored before the mold is reversed. For example, when measuring the ii layer of segment II, the x, y, z, and grayscale images of the ii layer of segment I and the ii layer of segment III need to be called. The process for identifying the coil boundary line is the same as that for measuring the coil width to identify the boundary line. Taking the adjacent segment I-II spacing as an example, the actual winding spacing of segments I and II = the spacing between the center photographing positions of segments I and II - the left boundary line of segment I relative to the photographing position - the right boundary line of segment II relative to the photographing position.

[0098] Further preferably, when calculating the major and minor axes, according to the grayscale image and depth data of the coil winding area at the current position, calculate the average value z1 of the depth data and the average value z2 of the mold depth data when the coil is not wound;

[0099] Major and minor axes = (z1 - z2 + D) * 2; where D is the radius of the mold. The structured light (3D) camera used in this article can obtain the depth of the object to be measured (relative to a 2D camera). The depth data refers to the depth data of the coil after winding, that is, the depth of this position relative to the camera. Both the mold depth data and the depth data of the wound coil are the depths relative to the camera.

[0100] The present invention provides a line vision detection device for process dimensions during the winding of a high-voltage coil, and provides a vision detection device calibration method and a calibration verification method for the accuracy of the detection device. Calculate the center position of the coil to be wound for each segment and each layer through the winding process, visually detect the edge of the coil, and the width, major and minor axes of the wound coil for each segment and each layer can be output in real time after moving and photographing. Subtract the center position of the coil after two moves and the edge distance to obtain the segment spacing at this position.

[0101] Obviously, the above embodiments are only examples clearly described and not limitations on the implementation manners. For those of ordinary skill in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation manners here. And the obvious changes or modifications derived therefrom are still within the protection scope of the present invention.

Claims

1. An online visual inspection method, characterized in that, Including: Scanning the coil mold using a structured light scanning camera to identify the initial position; Winding according to the coil winding process to obtain the first scanning data during the winding process. After the first part of the coil is wound, the mold is turned. The coil winding process includes the number of coil segments required, the winding order of each coil segment, the number of layers required for each coil segment, the width of each coil segment, and the spacing between adjacent coil segments. Among them, according to the number of coil segments required, the coil is divided into odd-numbered coil segments and even-numbered coil segments as a whole. The scanning data during the winding process of the odd-numbered coil segments is recorded as the first scanning data; the scanning data during the winding process of the even-numbered coil segments is recorded as the second scanning data; Using the structured light scanning camera to re-identify the initial position, winding the remaining part of the coil according to the coil winding process to obtain the second scanning data. Before identifying the initial position, it also includes correcting the relative parallelism between the coil winding component and the structured light scanning camera. After correction, it also includes verifying the parallelism of the x direction of the field of view of the structured light scanning camera and verifying the parallelism between the structured light scanning camera and the coil mold when the structured light scanning camera moves horizontally; According to the initial position, the first scanning data, and the second scanning data, calculate the coil width, segment spacing, and major and minor axes of each layer of each coil segment to obtain the detection result; When calculating the segment spacing according to the initial position, the first scanning data, and the second scanning data, according to the initial positions identified before and after, splice the first scanning data and the second scanning data, and extract the odd-numbered coil segment boundaries in the first scanning data and the even-numbered coil segment boundaries in the second scanning data; Then the segment spacing d = d1 - d2 - d3; Wherein, d1 is the distance between the center positions of the target odd-numbered coil segment and an adjacent target even-numbered coil segment; d2 is the distance between the left boundary line of the target odd-numbered coil segment and the center position of the target odd-numbered coil segment; d3 is the distance between the right boundary line of the target even-numbered coil segment and the center position of the target even-numbered coil segment.

2. The online visual inspection method according to claim 1, wherein Before identifying the initial position, it also includes correcting the relative parallelism between the coil winding component and the structured light scanning camera. The correction includes the following process: Collecting multiple groups of three-dimensional point cloud data of the mold rotating one week; Performing cylindrical fitting on each group of three-dimensional point cloud data respectively to obtain multiple axes; Optimizing the obtained multiple axes, and using the optimized axes as the point cloud axes; Calculating the axial angle between the point cloud axis and the structured light scanning camera, and performing compensation to obtain the relative parallelism correction result between the coil winding component and the structured light scanning camera.

3. The online visual inspection method according to claim 2, wherein When performing cylindrical fitting on each group of three-dimensional point cloud data respectively, the fitting formula is calculated according to the following formula: Among them, represents the center of all axes obtained after cylindrical fitting, denoted as the centroid of the fitted cylindrical body, N is the number of axes, pi is the center point of the fitted cylinder; is the axis direction of the fitted cylinder, is the eigenvector corresponding to the minimum eigenvalue; is the average distance from all axes to the centroid; dist i is the distance from the axis of each group of three-dimensional data to the centroid.

4. The online visual inspection method according to claim 2, wherein, After correction, it also includes verifying the parallelism of the x direction of the field of view of the structured light scanning camera according to the following process; Horizontally moving the calibration plate to two positions within the field of view of the structured light scanning camera; Calculate the included angle between the two poses of the calibration board according to the poses of the calibration board at two positions in the camera coordinate system θ x ; Determine the included angle θ x Check whether it meets the range requirements of the coil process. If it meets the requirements, the parallelism in the x-direction of the structured light scanning camera meets the requirements. If it does not meet the requirements, readjust and verify.

5. The online visual inspection method according to claim 2, characterized in that, After correction, it also includes verifying the parallelism between the structured light scanning camera and the coil mold when the structured light scanning camera moves horizontally according to the following process; Placing the calibration plate at any fixed position on the square shaft, and horizontally moving the structured light scanning camera to two different positions; When the structured light scanning camera is at two positions, the point cloud data obtained by scanning the calibration plate is transferred to the fixed position by matrix transformation; The point cloud data of the calibration plate taken at the two positions after transformation is subjected to difference calculation to obtain the X-direction deviation and Z-direction deviation of the parallelism with the coil mold.

6. The online visual inspection method according to claim 1, wherein When calculating the coil width of each segment and each layer of the coil according to the initial position, the first scan data, and the second scan data, it includes: Calculating the center position of the coil to be wound according to the initial position and the coil winding process; Moving the structured light scanning camera to the center position of each segment and each layer of the coil to be measured, and collecting the scan data of the area to be measured within the field of view; Extracting the grayscale image and depth data of the area where the coil is wound at the current position, using edge extraction to obtain the left and right boundaries, and calculating the distance between the two boundaries as the coil width.

7. The online visual inspection method according to claim 1, characterized in that, When calculating the major and minor axes, according to the grayscale image and depth data of the area where the coil is wound at the current position, calculating the average value z1 of the depth data and the average value z2 of the mold depth data when the coil is not wound; Major and minor axes = (z1 - z2 + D) * 2; where D is the mold radius.

8. An online visual inspection device, characterized in that, It includes: A structured light scanning camera for obtaining scan data during the coil winding process; A horizontal driving component for driving the structured light scanning camera to perform translational motion; A coil winding component for driving the mold to rotate for coil winding; A controller for controlling the horizontal driving component and the coil winding component, and calculating the detection result according to the on-line vision detection method described in any one of the above claims 1-7.

Images