Hemispherical resonator gyroscope assembling method and device based on machine vision

Abstract

The invention discloses a hemispherical resonator gyroscope assembly method and device based on machine vision, and belongs to the field of hemispherical resonator gyroscope precision assembly. After the harmonic oscillator in the hemispherical resonator gyroscope and the base are clamped, the fixed coaxial camera lens assembly is used for achieving shaft hole image collection of a to-be-assembled part-hemispherical harmonic oscillator and a base part-electrode base, and shaft hole movement deviation extraction is completed. Meanwhile, two gap cameras with the included angle of 150 degrees are used for obtaining local gap images and calculating rotation pose errors, a micropositioner drives a harmonic oscillator to complete error adjustment, then a temperature controller controls temperature linear change to conduct indium welding, and assembling is completed. According to the invention, the camera vision detection system is utilized to calculate and adjust the relative pose error of the harmonic oscillator and the electrode base in the assembly process, and meanwhile, the temperature control system is utilized to realize heating at a specified temperature gradient, so that the assembly and connection of the hemispherical harmonic oscillator and the electrode base are realized. The assembly precision of the hemispherical resonator and the electrode base is improved, and the current high-precision assembly requirement for the hemispherical resonator gyroscope is met.

Description

TECHNICAL FIELD

[0001] The application relates to a machine vision-based hemispherical resonator gyro assembling method and device and belongs to the field of hemispherical resonator gyro precision assembling. BACKGROUND

[0002] The hemispherical resonator gyro is a solid wave gyroscope based on the Coriolis effect. Compared with traditional gyroscopes, optical fiber gyroscopes and laser gyroscopes, the hemispherical resonator gyro has the advantages of high measurement accuracy, small size, low power consumption, high reliability, large measurement range and short relative preparation time, and is recognized by the domestic and foreign inertial technology field as the most ideal inertial device in the space navigation and control system in the 21st century. Therefore, the hemispherical resonator gyro is increasingly widely applied in the field of space technology, such as attitude and orbit control systems. The resonator and the electrode base are key components of the hemispherical resonator gyro, and the precision assembling of the core components of the hemispherical resonator gyro is the key to guarantee the performance and quality of the whole machine. Therefore, the research on the intelligent assembling method of the hemispherical resonator gyro based on machine vision has important significance for improving the assembling level of the hemispherical resonator gyro in China.

[0003] The hemispherical resonator gyro is a solid vibration gyroscope, and its general structure can be divided into three-piece structure and two-piece structure. The main structure of the three-piece structure hemispherical resonator gyro is composed of an excitation cover, a hemispherical resonator and a detection base. The two-piece structure is an integrated structure formed by combining the excitation cover and the detection base based on the three-piece structure. The hemispherical resonator is the core sensitive element in both structure forms. In addition to the machining precision of the main components, the assembling precision along the circumferential direction and the axial direction of the hemispherical resonator also greatly determines the final performance index of the hemispherical resonator gyro.

[0004] At present, in the field of hemispherical resonator gyro assembling and adjustment in China, the main means for relative position detection in the assembling process of the planar electrode type hemispherical resonator gyro are contact type, laser type and capacitance type. The traditional contact type detection method is easy to damage the gold-plated surface of the resonator and the reading base, and has low detection precision. The laser type is mainly used for spherical electrode type and cannot be directly applied to the flat electrode. The capacitance detection method depends on preliminary rough adjustment, and the capacitance of the flat electrode type is small (only picofarad level), which is more susceptible to environmental interference.

[0005] At present, the rapid development of deep learning technology brings revolutionary breakthroughs to the field of image super-resolution. The convolutional neural network (CNN) in the deep learning model can learn the complex mapping relationship between the low-resolution image and the high-resolution image from a large amount of training data. After training, the model can automatically extract and learn the features in the image, so as to more accurately restore the high-resolution image. Therefore, it is a new direction to use industrial camera images to obtain assembly images, realize fitting calculation of the relative deviation of the assembly pose of the resonator and the electrode base, and realize the relative error adjustment in the assembly process of the hemispherical resonator gyro based on the deviation amount feedback, so as to realize the precise assembly of the hemispherical resonator gyro. SUMMARY

[0006] In order to solve the problem of the non-uniform error of the axial and radial gap between the hemispherical resonator and the electrode base of the flat electrode type hemispherical resonator gyro, the application further provides a hemispherical resonator gyro assembly method and device based on machine vision.

[0007] In order to achieve the above purpose, the application realizes the following technical scheme:

[0008] A hemispherical resonator gyro assembly method based on machine vision, comprising the following steps:

[0009] S1, clamping of the hemispherical resonator and the electrode base. The hemispherical resonator and the electrode base are placed on the positioning tool of the hemispherical resonator and the electrode base from the feeding area, and are clamped by the Y-direction displacement table and the Z-direction displacement table in the clamping system.

[0010] S2, super-resolution after the local gap camera obtains images. Two industrial camera modules in the rotation error detection module obtain local gap images, and an image super-resolution algorithm is used to complete high-resolution reconstruction of the local images.

[0011] S3, calculating the rotation pose error after completing the gap detection. After image preprocessing, the gap edge information is extracted, the PROSAC method is used to fit the straight line where the part edge is located, the local gap extraction is completed, and a BP neural network model is used to construct an adjustment model from the local gap of the hemispherical resonator and the electrode base to the relative pose deviation, so as to guide the micro-motion adjustment system to complete the deviation correction of the rotation error in the assembly error.

[0012] S4, leveling of the resonator and the electrode base. The micro-motion adjustment amount of the hemispherical resonator and the electrode base is calculated according to the relative pose of the hemispherical resonator and the electrode base obtained in S3, and the micro-motion table in the micro-motion adjustment module drives the hemispherical resonator to complete the adjustment of the rotation error.

[0013] S5, the mobile error detection subsystem obtains the axis hole deviation image. The coaxial camera in the mobile error detection subsystem obtains the axis hole deviation image, and after image preprocessing, the axis hole deviation detection algorithm based on RANSAC is used to realize the detection of the axis hole deviation.

[0014] S6, the resonance and electrode base are adjusted coaxially. The relative pose of the hemispherical resonator and the electrode base obtained in S5 is calculated to obtain the fine adjustment amount, and the fine adjustment module drives the hemispherical resonator to complete the adjustment of the mobile error.

[0015] S7, start indium welding. After the alignment of the hemispherical resonator gyroscope is completed, the indium welding control system starts to perform indium welding.

[0016] Preferably, in step S4, the leveling of the hemispherical resonator and the electrode base is completed, and the specific steps are as follows:

[0017] The two industrial cameras of the rotation error measurement module obtain low-resolution images of the local gaps;

[0018] The EDSR algorithm model using fusion space and channel attention mechanism is used to perform high-resolution reconstruction on the obtained resolution images;

[0019] Through image processing and edge fitting, the edges of the parts in the high-resolution image are extracted, and the two local gaps of the resonator and the electrode base are calculated;

[0020] A local gap-rotation error model is established based on a BP neural network, and the establishment process of the model is as follows:

[0021] A mathematical model of rotation error-local gap is established, and the spatial plane equation of the lower lip of the resonator in the ideal pose is:

[0022]

[0023] The spatial plane equation of the upper surface of the electrode base is:

[0024]

[0025] The plane equation of the lower lip plane of the hemispherical resonator after rotating around the x-axis is:

[0026]

[0027] The plane equation of the lower lip plane of the hemispherical resonator after rotating around the x and y axes is:

[0028]

[0029] Wherein, (x, y, z) is the spherical harmonic oscillator center coordinates, R is the spherical harmonic oscillator median radius, r is the electrode base outer diameter, L is the ideal gap between the lower lip of the resonator along the plane and the upper surface of the electrode base, and a and β are the angles of rotation of the lower lip of the spherical harmonic oscillator along the x and y axes.

[0030] The local gaps in the multiple rotation error assembly states are calculated by the above mathematical model, the local gaps in the multiple rotation error assembly states are taken as inputs, the rotation errors in the multiple rotation error assembly states are taken as outputs, the BP neural network is trained, and a local gap-rotation error model is obtained.

[0031] Thus, the local gap-rotation error model based on the BP neural network is established.

[0032] The calculated two local gaps are input into the BP neural network model to obtain the rotation amount that needs to be adjusted, and then the rotation error adjustment of the spherical harmonic resonator and the electrode base is realized.

[0033] Preferably, in step S6, the coaxial adjustment of the spherical harmonic resonator and the electrode base is completed, and the specific steps are as follows:

[0034] The coaxial camera of the translation error measurement module acquires the shaft hole image, and obtains the shaft hole contour point set after image preprocessing and edge detection;

[0035] The RANSAC algorithm is applied to circular fitting of the edge point set contour data points. It is assumed that the number of randomly extracted points is n, the distance error threshold of the circle equation and the contour points is t, the maximum iteration number is T, the data points in the point set are randomly extracted, the circle is fitted by the least square method, the distance and the radius difference of all points in the point set to the center of the circle are calculated according to the returned parameter calculation of the circle, whether it is an inner point is judged by the threshold t, and the number of inner points is recorded, the number of inner points recorded each time is compared, and the number of inner points is taken as the fitting point set of the part edge equation.

[0036] According to the returned parameter calculation of the circle, the distance and the radius difference of all points in the point set to the center of the circle are calculated to divide the point set;

[0037] The least square method is used to fit the corresponding circles of the shaft hole contour point set respectively;

[0038] The shaft hole offset is calculated according to the fitted center position, and then the translation error adjustment of the spherical harmonic resonator and the electrode base is realized.

[0039] Preferably, the machine vision-based spherical harmonic resonator gyro assembly device comprises:

[0040] The rotation error measurement subsystem is composed of two visual detection systems, and the two visual detection systems are arranged in the same horizontal direction of the assembly platform and are distributed at an angle of 150°. Image information of the local gap is obtained by the visual detection system, image feature information is extracted by an image processing algorithm, and the local gap value is calculated.

[0041] The translation error measurement subsystem is mainly composed of a coaxial visual detection system and is located directly below the electrode base platform. The coaxial visual detection system obtains image information of the shaft hole cooperation, and calculates the shaft hole deviation.

[0042] The to-be-assembled part clamping subsystem is mainly composed of an upper center, an upper center connecting device, a lower center, a lower center connecting device, and a Z-direction displacement table. The initial positioning of the hemispherical resonator is completed by the lower center, and the complete positioning and clamping of the hemispherical resonator is completed by the Z-direction displacement table driving the upper center.

[0043] The base part clamping subsystem is mainly composed of a base placement plate, a movable V-shaped block, and an X-direction displacement table. The electrode base is placed on the base placement plate through feeding, and the complete positioning and clamping of the electrode base is completed by the X-direction displacement table driving the movable V-shaped block.

[0044] The fine adjustment subsystem is mainly composed of a fine adjustment table and a pose adjustment calculation module. The relative pose information of the hemispherical resonator and the electrode base is decoupled by the pose adjustment calculation module to obtain the adjustment amount of the hemispherical resonator, and the relative pose error is adjusted by controlling the fine adjustment table.

[0045] The indium soldering control system is composed of a digital intelligent temperature controller and a ceramic heating sheet. The indium soldering connection between the hemispherical resonator and the electrode base is completed by controlling the temperature curve.

[0046] Compared with the prior art, the present application has the following advantages:

[0047] Firstly, the image obtained by the industrial camera is used to detect the pose of the quartz base and the hemispherical resonator. The coaxial camera lens assembly fixed below the workbench realizes the image acquisition of the shaft hole, and the relative movement error is obtained. The two gap cameras with an angle of 150° in the same horizontal direction of the workbench acquire the local gap image, calculate the rotation pose error, and finally the industrial computer controls the multi-degree-of-freedom high-precision displacement table to adjust the relative position of the quartz base and the hemispherical resonator, so as to control the relative pose error of the quartz base and the hemispherical resonator and improve the assembly precision. BRIEF DESCRIPTION OF DRAWINGS

[0048] Figure 1 A hemispherical resonator gyro assembly process flow diagram based on machine vision is provided.

[0049] Figure 2 A schematic diagram of an image super-resolution algorithm process provided by the present application is shown in the figure.

[0050] Figure 3 A schematic diagram of a semi-spherical resonator and electrode base alignment coaxial process provided by the present application is shown in the figure.

[0051] Figure 4 A schematic diagram of a RANSAC algorithm process provided by the present application is shown in the figure.

[0052] Figure 5 A schematic diagram of a semi-spherical resonator and electrode base alignment coaxial process provided by the present application is shown in the figure. DETAILED DESCRIPTION

[0053] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0054] The purpose of the present application is to provide a machine vision-based semi-spherical resonator gyroscope assembly method and device, which solves the problem of uneven errors of the axial and radial gaps between the semi-spherical resonator and the electrode base of the flat-plate electrode type semi-spherical resonator gyroscope.

[0055] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0056] As shown in the figure, the present application provides a machine vision-based semi-spherical resonator gyroscope assembly method, which comprises: Figure 1

[0057] Step S1, place the semi-spherical resonator and the electrode base from the loading area on the positioning tool of the semi-spherical resonator and the electrode base, and clamp by the Y-direction displacement table and the Z-direction displacement table in the clamping system.

[0058] Step S2, rotate the two industrial camera modules in the error detection module to obtain local gap images, and use an image super-resolution algorithm to complete high-resolution reconstruction of the local images, and a schematic diagram of the image super-resolution algorithm process is shown in the figure. Figure 2

[0059] ​​Step S3, after image preprocessing, gap edge information is extracted, PROSAC method is used to fit the straight line where the part edge is located, local gap extraction is completed, and a relative pose deviation adjustment model from the local gap of the hemispherical resonator and the electrode base to the relative pose deviation is constructed by using the BP neural network model, so as to guide the micro-motion adjustment system to complete the deviation correction of the rotation error in the assembly error.

[0060] Step S4, according to the relative pose of the hemispherical resonator and the electrode base, the micro-motion adjustment amount is calculated, and the micro-motion platform in the micro-motion adjustment module drives the hemispherical resonator to complete the leveling of the hemispherical resonator and the electrode base.

[0061] In step S4, the leveling of the hemispherical resonator and the electrode base includes:

[0062] Step S41, two industrial cameras of the rotation error measurement module obtain low-resolution images of the local gap.

[0063] Step S42, the EDSR algorithm model fusing space and channel attention mechanisms is used to perform high-resolution reconstruction on the obtained resolution image.

[0064] Step S43, by image processing and edge fitting, the part edge in the high-resolution image is extracted, and the two local gaps of the resonator and the electrode base are calculated.

[0065] Step S44, a local gap-rotation error model is established based on the BP neural network.

[0066] The local gap-rotation error model includes:

[0067] A mathematical model of rotation error-local gap is established;

[0068] The spatial plane equation of the lower lip of the resonator in the ideal pose is:

[0069]

[0070] The spatial plane equation of the upper surface of the electrode base is:

[0071]

[0072] The plane equation of the lower lip plane of the hemispherical resonator after rotating around the x-axis is:

[0073]

[0074] The plane equation of the lower lip plane of the hemispherical resonator after rotating around the x and y axes is:

[0075]

[0076] Wherein, (x, y, z) is the spherical harmonic oscillator center coordinates, R is the spherical harmonic oscillator median plane radius, r is the electrode base outer diameter, L is the ideal gap between the lower lip of the resonator along the plane and the upper surface of the electrode base, and a and β are the angles of the lower lip of the spherical harmonic oscillator along the plane rotating around the x and y axes.

[0077] The local gaps in the multiple rotational error assembly states are calculated by the above mathematical model, the local gaps in the multiple rotational error assembly states are taken as inputs, the rotational errors in the multiple rotational error assembly states are taken as outputs, the BP neural network is trained, and a local gap-rotational error model is obtained.

[0078] Wherein, since the coordinate information of the relative pose error of the returned part in the visual detection system is based on the pixel coordinate system in the image, and the information required by the part pose adjustment mechanism is the coordinate information in the physical space, it is necessary to establish the mapping relationship between the image coordinate system and the physical space coordinate system, and Zhang Youzheng calibration method is selected because it is simple and has high precision.

[0079] In the visual detection system, the imaging of the camera can be simplified as a linear pinhole model.

[0082] The camera lens has lens distortion, and since the radial distortion is much larger than the tangential distortion, only the radial distortion is considered, and the expression of the radial distortion is:

[0083]

[0084] Wherein x and y are coordinate parameters after distortion correction, x0 and y0 are coordinate parameters on the corresponding distorted image, r is the distance from the point to the lens center (generally the image center), and k1, k2 and k3 are radial distortion coefficients.

[0085] Step S45, input the two calculated local gaps into the BP neural network model to obtain the rotational amount that needs to be adjusted, and then realize the leveling of the spherical harmonic oscillator and the electrode base.

[0086] Step S5, as shown in Figure 3 The coaxial camera in the movement error detection subsystem obtains the shaft hole deviation image, and after image preprocessing, the shaft hole deviation detection algorithm based on RANSAC is used to realize the detection of the shaft hole deviation.

[0087] Step S6, the micro-motion adjustment amount of the spherical harmonic oscillator and the electrode base is calculated according to the obtained relative pose, and the micro-motion stage in the micro-motion adjustment module drives the spherical harmonic oscillator to complete the leveling of the spherical harmonic oscillator and the electrode base.

[0088] In step S6, the leveling of the spherical harmonic oscillator and the electrode base includes:

[0089] Step S61, the coaxial camera of the translation error measurement module acquires the shaft hole image, and obtains the shaft hole contour point set after image preprocessing and edge detection.

[0090] Step S62, the RANSAC algorithm is applied to circular fitting of the edge point set contour data points, and a flowchart of the RANSAC algorithm is shown in Figure 4 The number of random points extracted each time is n, the distance error threshold of the circle equation and the contour points is t, the maximum iteration number is T, the data points in the point set are randomly extracted, the circle is fitted by the least square method, the distance and the radius difference of all points in the point set to the center of the circle are calculated according to the returned parameter calculation of the circle, whether it is an inner point is judged by the threshold t, and the number of inner points is recorded, the number of inner points recorded each time is compared, and the number of inner points is taken as the fitting point set of the part edge equation.

[0091] Step S63, the distance and the radius difference of all points in the point set to the center of the circle are calculated according to the returned parameter calculation of the circle.

[0092] Step S64, the least square method is used to fit the corresponding circles of the shaft hole contour point set.

[0093] Step S65, the shaft hole offset is calculated according to the fitting center position, and the adjustment of the hemispherical resonator and the electrode base is realized.

[0094] Step S7, start indium welding. After the alignment of the hemispherical resonator is completed, the indium welding control system starts to perform indium welding.

[0095] At this point, the process shown in the flowchart is completed. Figure 1

[0096] In addition, as shown in Figure 5 The application also provides a hemispherical resonator assembly device based on machine vision, which comprises:

[0097] The rotation error measurement subsystem is composed of two visual detection systems, and the two visual detection systems are in the same horizontal direction of the assembly platform and are distributed at an angle of 150°. The image information of the local gap is obtained by the visual detection system, the image feature information is extracted by the image processing algorithm, and the local gap value is calculated.

[0098] The translation error measurement subsystem is mainly composed of a coaxial visual detection system and is located directly below the electrode base platform. The shaft hole fitting image information obtained by the coaxial visual detection system is used to calculate the shaft hole deviation.

[0099] ​The to-be-assembled part clamping subsystem is mainly composed of an upper top, an upper top connecting device, a lower top, a lower top connecting device and a Z-direction displacement table; the initial positioning of the hemispherical resonator is completed by the lower top, and the complete positioning and clamping of the hemispherical resonator is completed by the Z-direction displacement table driving the upper top.

[0100] The base part clamping subsystem is mainly composed of a base placement plate, a moving V-shaped block and an X-direction displacement table; the electrode base is placed on the base placement plate through feeding, and the complete positioning and clamping of the electrode base is completed by the X-direction displacement table driving the moving V-shaped block.

[0101] The micro-motion adjustment subsystem is mainly composed of a micro-motion adjustment table and a pose adjustment calculation module; the relative pose information of the hemispherical resonator and the electrode base is decoupled by the pose adjustment calculation module to obtain the adjustment amount of the hemispherical resonator, and the micro-motion adjustment table is controlled to complete the adjustment of the relative pose error.

[0102] The indium soldering control system is composed of a digital intelligent temperature controller and a ceramic heating sheet, and the indium soldering connection between the hemispherical resonator and the electrode base is completed by controlling the temperature curve.

[0103] The present application can be applied to the assembly process of the flat-plate electrode type hemispherical resonator gyroscope, the relative position error of the hemispherical resonator and the flat-plate electrode can be obtained in real time, the relative position of the hemispherical resonator and the flat-plate electrode can be adjusted in the assembly process, the assembly efficiency of the hemispherical resonator gyroscope is improved, the high-quality assembly of the hemispherical resonator gyroscope is ensured, and the comprehensive performance of the hemispherical resonator gyroscope is further improved. Meanwhile, the method provided by the present application is a real-time non-contact measurement method, which can provide data support for the adjustment of the relative position of the hemispherical resonator and the flat-plate electrode, improve the assembly efficiency of the hemispherical resonator gyroscope, improve the assembly quality of the hemispherical resonator gyroscope, and further optimize the comprehensive performance of the hemispherical resonator gyroscope.

Claims

1. A method and apparatus for assembling a hemispherical resonant gyroscope based on machine vision, characterized in that, The system comprises a translational error measurement subsystem, a rotational error measurement subsystem, a part-to-be-assembled clamping system, a base part clamping system, a micro-adjustment system, an assembly control subsystem, and an indium soldering control subsystem. The rotational error detection subsystem, positioned horizontally with the assembly platform, acquires image information of local gaps, extracts image features using image processing algorithms, and calculates the local gap value. The translational error measurement subsystem, located directly below the electrode base platform, acquires image information of the shaft-hole fit and calculates the shaft-hole deviation. The part-to-be-assembled clamping system and the base part clamping system clamp the hemispherical resonator and the electrode base, respectively. The micro-adjustment system, fixed to the part-to-be-assembled clamping system, adjusts the hemispherical resonator's pose to complete the assembly of the hemispherical resonator and the electrode base. The indium soldering control subsystem connects the hemispherical resonator and the electrode base after assembly and positioning.

2. The assembly method and apparatus for a hemispherical resonant gyroscope based on machine vision according to claim 1, characterized in that, The machine vision-based method obtains the pose error of the hemispherical resonator relative to the electrode base, and then completes the assembly. This method includes the following steps: S1. Clamping of the hemispherical resonator and the electrode base. The hemispherical resonator and the electrode base are placed from the loading area onto the positioning fixture of the hemispherical resonator and the electrode base, and clamped by the Y-axis displacement stage and the Z-axis displacement stage in the clamping system. S2. After acquiring images from the local gap camera, super-resolution is performed. The two industrial camera modules in the rotation error detection module acquire local gap images, and the image super-resolution algorithm is used to complete the high-resolution reconstruction of the local images. S3. After completing the gap detection, calculate the rotational pose error. After image preprocessing, extract the gap edge information, use the PROSAC method to fit the straight line where the part edge is located, complete the local gap extraction, and use the BP neural network model to construct an adjustment model from the local gap between the hemispherical harmonic oscillator and the electrode base to the relative pose deviation, so as to guide the micro-motion adjustment system to complete the deviation correction of rotational error in the assembly error. S4. Rotational Error Adjustment. Based on the relative pose of the hemispherical resonator and the electrode base in step 3, the micro-adjustment amount is calculated. The micro-adjustment stage in the micro-adjustment module drives the hemispherical resonator to complete the adjustment of rotational error. This process is also called the leveling of the hemispherical resonator and the electrode base. S5. The motion error detection subsystem obtains the shaft hole deviation image. The coaxial camera in the motion error detection subsystem obtains the shaft hole deviation image, and after image preprocessing, the shaft hole deviation is detected using a RANSAC-based shaft hole deviation detection algorithm. S6. Movement Error Adjustment. Based on the relative pose of the hemispherical resonator and the electrode base in step 5, the micro-adjustment amount is calculated. The micro-adjustment stage in the micro-adjustment module drives the hemispherical resonator to complete the adjustment of movement error. This process is also called the coaxial adjustment of the hemispherical resonator and the electrode base. S7. Start indium soldering. After the alignment of the hemispherical resonant gyroscope is completed, indium soldering is started by the indium soldering control system.

3. The assembly method and apparatus for a hemispherical resonant gyroscope based on machine vision according to claim 1, characterized in that, The rotation error measurement subsystem consists of two vision inspection systems, which are positioned horizontally with the assembly platform at a 150° angle. The vision inspection systems acquire image information of local gaps, and image processing algorithms extract image features to calculate the local gap value.

4. The assembly method and apparatus for a hemispherical resonant gyroscope based on machine vision according to claim 1, characterized in that, The translation error measurement subsystem mainly consists of a coaxial vision inspection system located directly below the electrode base platform. It calculates the shaft hole deviation based on the shaft hole fit image information obtained by the coaxial vision inspection system.

5. The assembly method and apparatus for a hemispherical resonant gyroscope based on machine vision according to claim 1, characterized in that, The part clamping subsystem mainly consists of an upper center, an upper center connecting device, a lower center, a lower center connecting device, and a Z-axis displacement stage; the lower center completes the initial positioning of the hemispherical resonator, and the Z-axis displacement stage drives the upper center to complete the complete positioning and clamping of the hemispherical resonator.

6. The assembly method and apparatus for a hemispherical resonant gyroscope based on machine vision according to claim 1, characterized in that, The base component clamping subsystem mainly consists of a base placement plate, a movable V-block, and an X-axis displacement stage. The electrode base is placed on the base placement plate by feeding, and the X-axis displacement stage drives the movable V-block to complete the complete positioning and clamping of the electrode base.

7. The assembly method and apparatus for a hemispherical resonant gyroscope based on machine vision according to claim 1, characterized in that, The micro-motion adjustment subsystem mainly consists of a micro-motion adjustment stage and a pose adjustment calculation module. The pose adjustment calculation module decouples the relative pose information of the resonator and the electrode base returned by the rotation error detection subsystem and the translation error measurement subsystem to obtain the adjustment amount of the hemispherical resonator, and controls the micro-motion adjustment stage to complete the adjustment of the relative pose error.

8. The assembly method and apparatus for a hemispherical resonant gyroscope based on machine vision according to claim 1, characterized in that, The indium welding control system consists of a digital display intelligent temperature controller and a ceramic heating element. By controlling the temperature curve, it completes the indium welding connection between the hemispherical resonator and the electrode base.

9. The assembly method and apparatus for a hemispherical resonant gyroscope based on machine vision according to claim 1, characterized in that, The leveling of the hemispherical resonator and the electrode base includes: The two industrial cameras in the rotation error measurement module obtain low-resolution images of the local gap; The EDSR algorithm model, which integrates spatial and channel attention mechanisms, performs high-resolution reconstruction on the obtained resolution image; By using image processing and edge fitting methods, the edges of the parts in the high-resolution image are extracted, and the two local gaps between the resonator and the electrode base are calculated. A local gap-rotation error model was established based on a BP neural network. The calculated two local gaps are input into the BP neural network model to obtain the amount of rotation that needs to be adjusted, thereby achieving the leveling of the hemispherical harmonic oscillator and the electrode base.

10. The assembly method and apparatus for a hemispherical resonant gyroscope based on machine vision according to claim 1, characterized in that, The local clearance-rotation error model includes: Establish a mathematical model for rotational error and local clearance; The local gap under multiple rotational error assembly states is calculated using the above mathematical model. The local gap under multiple rotational error assembly states is used as input, and the rotational error under multiple rotational error assembly states is used as output. The BP neural network is trained to obtain the local gap-rotational error model.

11. The assembly method and apparatus for a hemispherical resonant gyroscope based on machine vision according to claim 1, characterized in that, The mathematical model for the rotational error—local clearance—includes: The spatial plane equation of the lower lip of the harmonic oscillator in the ideal pose is: The spatial plane equation of the upper surface of the electrode base is: The equation of the plane of the lower lip of the hemispherical harmonic oscillator after rotation about the x-axis is: The equations of the plane of the lower lip of the hemispherical harmonic oscillator after rotation about the x and y axes are: Where (x,y,z) are the coordinates of the center of the hemispherical harmonic oscillator, R is the mid-surface radius of the hemispherical harmonic oscillator, r is the outer diameter of the electrode base, L is the ideal gap between the plane containing the lower lip of the harmonic oscillator and the plane containing the upper surface of the electrode base, and α and β are the angles of rotation of the lower lip of the hemispherical harmonic oscillator around the x and y axes, respectively.

12. The assembly method and apparatus for a hemispherical resonant gyroscope based on machine vision according to claim 1, characterized in that, The coaxial adjustment of the hemispherical resonator with the electrode base includes: The coaxial camera of the translation error measurement module acquires the shaft hole image, and after image preprocessing and edge detection, the shaft hole contour point set is obtained. The RANSAC algorithm is applied to fit a circle to the contour data points of the edge point set. Let n be the number of random points sampled each time, t be the distance error threshold between the circle equation and the contour points, and T be the maximum number of iterations. Data points are randomly sampled from the point set, and a circle is fitted using the least squares method. Based on the parameters of the returned circle, the difference between the distance from all points in the point set to the center and the radius is calculated. The threshold t is used to determine whether a point is an interior point, and the number of interior points is recorded. The number of interior points recorded each time is compared, and the set with the highest number of interior points is taken as the fitting point set for the part's edge equation. Based on the parameters of the returned circle, calculate the difference between the distance from all points in the point set to the center of the circle and the radius, and then divide the point set. The least squares method is used to fit the circles corresponding to the point sets of shaft hole profiles. The offset of the shaft hole is calculated based on the fitted center position, thereby achieving coaxial adjustment between the hemispherical resonator and the electrode base.

Images