Double-sided alignment mounting method for a quartz resonant acceleration sensor

Abstract

The present invention discloses a double-sided alignment and mounting method for a quartz resonant accelerometer. This method, which involves ultra-precision microscopic operations, is applied to the assembly of mid-scale components in quartz resonant accelerometers. Specifically, it involves dual-microscopic vision system camera calibration, component feature recognition, and a low-stress double-sided alignment and mounting method. Unlike the manual press assembly method used in previous experiments, the present invention calibrates the mapping relationship between an industrial area array camera and the world coordinate system, as well as the mapping relationship between the dual-camera coordinate systems, performs coordinate conversion on the feature information collected by the microscopic vision, and utilizes alignment marks on the component surfaces to achieve alignment and assembly of the accelerometer components. This method avoids the impact of large manual assembly errors on the measurement accuracy of quartz resonant frequency misalignment accelerometers.

Description

Technical Field

[0001] The present invention relates to ultra-precision operations in the microscopic field and is applied to the assembly of mid-scale parts in quartz resonant acceleration sensors. It specifically involves camera calibration of a dual-microscopic vision system, feature recognition of parts, and a low-stress double-sided alignment and placement method. Background Art

[0002] Quartz resonant accelerometer (hereinafter referred to as sensor) is one of the key core components of inertial measurement system. It has very important application value in aerospace, inertial navigation and guidance, weapon attitude and aiming stabilization, earthquake monitoring and other high-precision fields that have high requirements for accelerometer accuracy. It has become an important development direction of microsensors. Figure 1 Figure 2 shows the operating principle of a high-precision quartz resonant accelerometer. It employs the resonance principle, consisting of a quartz mass-spring system and a quartz resonator in a differential arrangement. When subjected to acceleration, the quartz sensitive mass displaces, inducing axial tension / compression on the quartz resonator. Under this tensile / compressive stress, one frequency of the differentially arranged quartz resonator increases, while the other decreases. A resonant circuit outputs the resonator's vibration frequency as a periodic electrical signal. A frequency acquisition circuit measures the waveform frequency, achieving frequency-based digital measurement of the acceleration signal with high resolution and stability.

[0003] The accuracy of the sensor depends on the processing accuracy of the components and the integrated packaging accuracy and strength of the components. The processing accuracy of the components can currently reach a relatively high level of technology. The accuracy and strength of the integrated packaging of the components have become the key factors restricting the performance of the sensor.

[0004] At present, the proposed assembly scheme for the quartz pendulum and quartz vibration beam of this sensor is direct bonding of quartz to quartz. This scheme avoids the mismatch of thermal expansion coefficients between heterogeneous materials and the influence of curing stress on the reliability and stability of the sensor. In the direct bonding process, the quartz wafer after surface activation is still manually pressed and assembled, which easily leads to large manual assembly errors, causing the resonant frequency of the quartz resonator to drift, seriously affecting the performance of the sensor. The quartz pendulum and quartz vibration beam are cross-scale parts, and the relatively cumbersome manual alignment and assembly process conditions will also make the sensor assembly efficiency low. Therefore, the development of an assembly method for cross-scale parts of quartz resonant acceleration sensors is of great value and necessity for my country's high-tech fields. Summary of the Invention

[0005] The objectives of the present invention are achieved through the following technical solutions.

[0006] A double-sided alignment mounting method for a quartz resonant acceleration sensor, comprising:

[0007] Use the calibration plate to calibrate the first mapping relationship between the camera coordinate system and the world coordinate system, and the second mapping relationship between the upper and lower camera coordinate systems in the dual-microscope vision system;

[0008] Placing a quartz vibration beam at the bottom of the calibration plate, the upper camera acquires an image of the quartz vibration beam and extracts pixel coordinates of four assembly points in the quartz vibration beam, and calculating the world coordinates of the four assembly points in the quartz vibration beam using a first mapping relationship;

[0009] A quartz pendulum is placed in the field of view of the lower camera, an image of the quartz pendulum is acquired, and pixel coordinates of four assembly points in the quartz pendulum are extracted. The pixel coordinates corresponding to the four assembly points in the quartz pendulum in the coordinate system of the upper camera are calculated through a second mapping relationship, thereby calculating the world coordinates corresponding to the four assembly points in the quartz pendulum;

[0010] According to the correspondence between the four assembly points in the quartz vibration beam and the four assembly points in the quartz pendulum, the posture of the quartz pendulum is adjusted, and the quartz pendulum is placed on the quartz vibration beam;

[0011] The assembly site coordinates of another quartz vibration beam are obtained, its posture is adjusted, and it is placed on the quartz pendulum. Pressure and heating are applied simultaneously on the two quartz vibration beams to realize the assembly of the acceleration sensor.

[0012] Furthermore, the calibration plate is a plane calibration plate with a 27×27 array of calibration points with a diameter of 0.25 mm. The calibration plate has an outer dimension of 150×150 mm and a thickness of 0.5 mm. The calibration point processing accuracy is 1 μm and the collinearity accuracy is 1 μm.

[0013] Furthermore, the plane calibration plate is made of film material, and its thickness is less than the overlapping thickness of the depth of field of the upper and lower cameras.

[0014] Furthermore, the calibration includes angle calibration and distance calibration.

[0015] Furthermore, the angle calibration process is as follows:

[0016] The calibration point images collected by the upper and lower cameras are processed to extract the coordinates of the circle center. The pixel coordinates of the three circle centers in the upper camera pixel coordinate system are set as follows: , , , the coordinates of the center of the circle in the lower camera are , , ; First calculate the mean of the horizontal and vertical coordinates of the point and :

[0017] (3)

[0018] The parameter n represents the number of calibration points in each group selected. X1, Y1, X2, Y2, X3, and Y3 all correspond to the coordinates above. , , , are the coordinates of the same point in the upper and lower camera pixel coordinate systems respectively;

[0019] Recalculate the coefficient 、 、 :

[0020] (4)

[0021] The slope of the regression line is:

[0022] Similarly, the slope of the straight line containing the center points of the circle in the lower camera can be calculated ; The rotation angle between the upper camera pixel coordinate system and the lower camera pixel coordinate system is equal:

[0023] (5).

[0024] Furthermore, the distance calibration process is as follows:

[0025] Pixel distance between two measurement groups on the calibration plate As shown in formula (6):

[0026] (6)

[0027] Where d is the actual physical distance between the two measurement groups, is the camera pixel scale coefficient; the coordinates of the center point of the upper camera and lower camera pixel coordinate systems are established by geometric relationship formula (7), and then the analytical expressions of A and B are derived as formula (8):

[0028] (7)

[0029] (8)

[0030] A and B are the distances between the origins of the two coordinate systems. is the scaling factor, X5, Y5 and are the coordinates of the same point in the upper and lower camera pixel coordinate systems, respectively, where is the angle between the second set of calibration points in the lower camera field of view and the pixel coordinate system.

[0031] Furthermore, the adjusting of the quartz pendulum posture and placing the quartz pendulum on the quartz vibration beam include feature recognition of the quartz vibration beam and the quartz pendulum, and the feature recognition includes image preprocessing, template matching, edge detection, and assembly site extraction.

[0032] Furthermore, the template matching adopts a shape-based template matching algorithm, which uses the gradient correlation of the object edge as the matching criterion, extracts the edge features in the ROI, creates a template in combination with the grayscale information, and generates a multi-level image pyramid model according to the size and clarity requirements of the template; then, the template image is searched layer by layer from top to bottom in the image pyramid model until the bottom layer is searched or a confirmed matching result is obtained.

[0033] Furthermore, the edge detection uses the Canny operator to perform sub-pixel edge detection, obtain the edge image of the quartz vibration beam or the quartz pendulum, and extract the XLD contour of the target feature.

[0034] Furthermore, the assembly site extraction includes: thinning edges larger than 1 pixel, performing corresponding processing on discontinuous edges, including generating contours, merging discontinuous edges, separating backgrounds, and obtaining edges; and then extracting assembly site information through straight line fitting, finding intersection points, and extracting corner points.

[0035] The advantage of this invention lies in that, unlike the manual press-fit assembly method used in previous experiments, this method calibrates the mapping relationship between an industrial area array camera and the world coordinate system, as well as the mapping relationship between the two-camera coordinate systems. This method then performs coordinate conversion on feature information collected by microscopic vision and utilizes alignment marks on the component surfaces to achieve alignment and assembly of accelerometer components. This method avoids the impact of quartz resonant frequency misalignment, which can be caused by large manual assembly errors, on the accelerometer's measurement accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Various other advantages and benefits will become apparent to those skilled in the art upon reading the detailed description of the preferred embodiment below. The accompanying drawings are for illustration purposes only and are not to be considered as limiting the present invention. The same reference symbols are used throughout the drawings to represent the same components. In the drawings:

[0037] Figure 1 This is a schematic diagram of the working principle of a high-precision quartz resonant acceleration sensor in the prior art.

[0038] Figure 2 Schematic diagram of the nine-point calibration method used in the present invention.

[0039] Figure 3 This is a schematic diagram of the calibration of the coordinate mapping relationship between the dual visual coordinate systems of the present invention.

[0040] Figure 4 This is a schematic diagram of the Mark area captured by the upper and lower cameras of the present invention.

[0041] Figure 5 This is a schematic diagram of the correct placement of the calibration plate of the present invention.

[0042] Figure 6 This is a schematic diagram of the quartz vibration beam of the present invention.

[0043] Figure 7 Schematic diagram of the quartz pendulum of the present invention.

[0044] Figure 8 This is a diagram showing the acceleration sensor of the present invention after assembly.

[0045] Figure 9 Schematic diagram of the target area image to be acquired on the quartz vibration beam and the pendulum piece of the present invention.

[0046] Figure 10 Schematic diagram of target features detected by the Canny operator in the present invention.

[0047] Figure 11 Schematic diagram of the XLD profile of the present invention.

[0048] Figure 12 This is a schematic diagram of the assembly sites that need to be extracted on the quartz vibration beam and the pendulum piece of the present invention.

[0049] Figure 13 This is a schematic diagram of the alignment of the assembly sites of the quartz vibration beam and the quartz pendulum piece of the present invention. DETAILED DESCRIPTION

[0050] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments described herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0051] The present invention adopts a nine-point calibration method to calibrate the mapping relationship between the camera coordinate system and the world coordinate system, and designs and adopts a high-precision plane calibration plate to obtain the mapping relationship between the upper and lower camera coordinate systems in the dual-microscope vision system. The present invention does not choose the commonly used gold-sprayed mark auxiliary alignment method, but uses the double-sided symmetrical features on the quartz pendulum for alignment and assembly. The manufacturing precision of this feature is high, and there is no burr under microscopic vision, and the recognition effect is good. Rapid positioning is performed based on template matching, sub-pixel feature extraction is achieved through the Canny edge detection algorithm, sub-pixel feature recognition is achieved through the sub-pixel subdivision method of polynomial interpolation, and image processing algorithms such as straight line fitting and corner point extraction are used to obtain the coordinates of the assembly site.

[0052] The technical solution of the present invention will be developed from the following three aspects:

[0053] 1. Camera calibration

[0054] This invention uses a dual-microscope vision system with upper and lower lenses, requiring calibration of the internal and external parameters of the monocular vision system and the coordinate mapping relationship between the two vision coordinate systems. Currently available industrial cameras exhibit minimal distortion, and this invention is used to capture two-dimensional images, so additional calibration of camera distortion parameters is unnecessary.

[0055] The calibration of the internal and external parameters of monocular vision is actually to obtain the conversion relationship between the pixel coordinate system and the spatial coordinate system. The nine-point calibration method is generally used in industry for calibration. The specific calibration principle is as follows:

[0056] 、 They represent the distances in the u and v directions that the origin of the world coordinate system is offset from the origin of the camera pixel coordinate system. For example, if a point [u,v,1] in the pixel coordinate system corresponds to [x,y,1] in the world coordinate system, the offsets of the two coordinate systems are known. 、 , then u=x+ ,v=y+ , written in matrix form is:

[0057]

[0058] 、 They represent the scaling relationship between the world coordinate system and the camera pixel coordinate system. For example, a point [u,v,1] in the pixel coordinate system corresponds to [x,y,1] in the world coordinate system. 、 , then u= x,v= y, written in matrix form is:

[0059]

[0060] Represents the rotation angle of the camera pixel coordinate system relative to the world coordinate system. For example, the polar coordinates of a point [x, y, 1] in the world coordinate system are , , the rotation angles of the two coordinate systems are known , then the polar coordinates of the point corresponding to the coordinates [u,v,1] in the pixel coordinate system are , ,

[0061] Then we have: Written in matrix form:

[0062]

[0063] Represents the shear angle between the camera pixel coordinate system and the world coordinate system. For example, if the shear angle Theta of the two coordinate systems is known at a point [u,v,1] in the camera pixel coordinate system, and the corresponding coordinate of the point in the world coordinate system is [x,y,1], then , , written in matrix form is:

[0064]

[0065] Thus, through the above derivation, the coordinates of point C1 (u, v) in the existing pixel coordinate system correspond to point C2 in the spatial coordinate system as C2 (x, y). The relationship between C1 and C2 is only translation, scaling, rotation and shearing. The corresponding matrix formulas are:

[0066]

[0067] Multiplying the above four matrices together to form an affine transformation matrix can realize the conversion from pixel coordinate system to space coordinate system.

[0068] = (1)

[0069] The final matrix is ​​obtained by multiplying the first four matrices. It can be seen that there are six parameters a1, b1, c1, a2, b2, and c2. To obtain these six parameters, at least six different equations are required. Each point contains (x, y) coordinates, so each point can produce two different equations. Therefore, we only need three points to solve the final affine transformation matrix. To improve accuracy, we can use nine points to calculate the results in N combinations. Based on these results, we use a method similar to averaging to improve accuracy. Figure 2 shown.

[0070] Since the upper and lower cameras are arranged independently, the coordinate mapping relationship between the dual visual coordinate systems needs to be calibrated. The auxiliary tool used for this calibration is the high-precision flat calibration plate designed by the present invention. The calibration plate has a 27×27 array of calibration points with a diameter of 0.25mm. The calibration plate has an overall size of 150×150mm and a thickness of 0.5mm. The calibration point processing accuracy is 1μm and the collinearity accuracy is 1μm. Figure 3 As shown in the figure, (a) the pixel coordinate mapping relationship between the upper and lower cameras; (b) the high-precision plane calibration plate.

[0071] The coordinate mapping relationship between the two sub-camera coordinate systems mainly involves the calibration of four parameters: the distances A and B between the origins of the two coordinate systems; the angle between the two coordinate systems ; Scaling factor By formula The coordinates of the point in the upper camera pixel coordinate system can be converted to the lower camera coordinate system. 、 is the lower coordinate of the upper camera pixel coordinate system, 、 is the lower coordinate in the camera pixel coordinate system.

[0072] (2)

[0073] The high-precision flat calibration plate used is made of film material, and its thickness is less than the overlap thickness of the depth of field of the upper and lower cameras, so that the upper and lower cameras can simultaneously capture the calibration points. Considering the assembly accuracy, camera field of view, and the distance between the optical axes of the upper and lower cameras, the size of the calibration plate and the calibration points on the plate are designed so that the upper left and lower right mark areas of the calibration plate can appear in the field of view of the upper and lower cameras respectively at the same time. The schematic diagram is as follows Figure 4 As shown. Each Mark area contains 3×3 calibration points, with a total of nine points. The distance between the calibration points is known. By extracting the coordinates of the center of the outermost calibration point in the Mark area and fitting a rectangle, it can be used to determine whether the calibration plate is placed flat, thereby evaluating the reliability of the calibration parameters. Figure 5 The one on the left is incorrectly placed, while the one on the right is correctly placed.

[0074] The machining accuracy of the calibration points is 1μm, and the collinearity accuracy is 1μm, which ensures the parallelism of the lines between the calibration points. The calibration points in the middle column of the Mark area captured by the upper and lower cameras are taken as the measurement group, and the distance between the two groups is The specific calibration principle includes angle calibration and distance calibration. First, perform angle calibration, process the calibration point images collected by the upper and lower cameras, extract the coordinates of the circle center, and set the pixel coordinates of the three circle centers in the upper camera pixel coordinate system to be , , , the coordinates of the center of the circle in the lower camera are , , . The straight lines formed by any two groups of calibration points on the plane calibration plate designed by the present invention are parallel to each other, so ideally, the slopes of the two groups of straight lines obtained by image processing should also be equal, then the difference between the actual slopes of the two groups of straight lines can be made the difference in angles between the two sub-camera coordinate systems. Due to the existence of manufacturing errors and circle fitting errors, the three center points are not completely collinear. By establishing a univariate linear regression equation, the straight line where the three points are located is determined and the slope of the regression line is obtained. Taking the three points in the above camera image as an example, first calculate the mean of the horizontal and vertical coordinates of the points. and :

[0075] (3)

[0076] The parameter n represents the number of calibration points in each group selected, which is 3 in this invention. The uppercase X1, X2, X3, Y1, Y2, Y3 all correspond to the coordinates above , , , are the coordinates of the same point in the upper and lower camera pixel coordinate systems respectively;

[0077] Recalculate the coefficient 、 、 :

[0078] (4)

[0079] The slope of the regression line is:

[0080] Similarly, the slope of the straight line containing the center points of the circle in the lower machine can be calculated. ; The rotation angle between the upper camera pixel coordinate system and the lower camera pixel coordinate system is equal:

[0081] (5)

[0082] The distance calibration method is as follows: According to the characteristics of the plane calibration plate, no matter how the pixel coordinate systems of the two cameras are rotated or translated, the pixel distance between the two measurement groups on the calibration plate is unchanged, as shown in formula (6):

[0083] (6)

[0084] Where d is the actual physical distance between the two measurement groups, is the camera pixel scale factor. The coordinates of the center point of the upper and lower camera pixel coordinate systems can be established by geometric relationship formula (7), and then the analytical expressions of A and B are derived as formula (8).

[0085] (7)

[0086] (8)

[0087] X5, Y5 and are the coordinates of the same point in the upper and lower camera pixel coordinate systems. is the angle between the second set of calibration points in the lower camera field of view and the pixel coordinate system.

[0088] 2. Component feature recognition

[0089] The components of this assembly include quartz vibration beam and quartz pendulum, the structure is as follows Figure 6 、 7 shown.

[0090] The quartz vibration beam needs to be assembled symmetrically on both sides of the quartz pendulum. The acceleration sensor after assembly is as follows Figure 8 shown.

[0091] During the placement process, it is necessary to perform feature recognition on the quartz vibration beam and quartz pendulum. The recognition algorithm process is as follows:

[0092] a. Image preprocessing; this includes image smoothing, denoising, and grayscale conversion. In poor conditions, exponential or logarithmic transformations can also be performed on the image grayscale. The properties of exponential and logarithmic functions indicate that exponential transformations can expand the high-grayscale range and enhance the contrast of bright details, while logarithmic transformations can expand the low-grayscale range and enhance the contrast of dark details.

[0093] b. Template matching; In the present invention, template matching is used for coarse positioning and rapid extraction of RIO area images. Since the feature images captured during each placement do not differ in shape or size, a shape-based template matching algorithm is selected to adapt the recognition accuracy and efficiency by controlling the number of pyramid layers when creating the template. The algorithm uses the gradient correlation of the object edge as the matching criterion. The principle is to extract the edge features in the ROI, create a template based on the grayscale information, and generate a multi-level image pyramid model based on the size and clarity requirements of the template. The template image is then searched layer by layer from top to bottom in the image pyramid layer until the bottom layer is searched or a confirmed matching result is obtained. The implementation steps are as follows: (The operators mentioned below are all operators in existing visual image processing software (such as but not limited to halcon) and can be used directly).

[0094] ① Select the target to be detected from the reference image. Use appropriate shape tools, such as rectangular selection, to select ROI from the reference image, and then use the reduce_domain operator to crop the area into an independent image area. Figure 9 What is displayed is the target area image to be obtained on the quartz vibration beam and the pendulum, which is the independent image area we need to capture, that is, the template image.

[0095] ② Create a template. Create the captured template and save it as a template, so that you can read the template handle ModelID for comparison when searching for images.

[0096] ③ Search for the target. After creating the template, read the image to be detected. Use the find_shape_model operator to search for the best matching area, input the detection image and template handle into the operator, the operator will compare the searched image with the template image, and the similarity of the comparison will be stored in the parameter Score. If all parameters are set appropriately, then at least one area with a score greater than the minimum matching score should be found in the image. The results returned by the find_shape_model operator, in addition to the matching score, also include the coordinates and rotation angle of the target, which can be used to calculate the position and rotation angle. Figure 9 Shown is the image of the target area to be acquired on the quartz vibration beam and the pendulum.

[0097] c. Edge detection: Use the Canny operator for sub-pixel edge detection. The Canny operator is currently the most theoretically complete edge detection algorithm. Its basic idea is to find the local maximum of the gradient. First, use a Gaussian smoothing filter to convolve and reduce noise. Then use a pair of convolution kernels to calculate the edge gradient and direction. Then use non-maximum suppression to remove non-edge lines. Finally, use hysteresis thresholds (high and low thresholds) to detect and connect edges. Obtain the edge image of the quartz vibrating beam or quartz pendulum and extract the XLD (eXtended Line Descriptions) outline of the target feature, such as Figure 10 The image shows the target feature contours detected by the Canny operator on the quartz vibration beam and the pendulum.

[0098] d. Assembly site extraction: Refine edges larger than 1 pixel and perform corresponding processing on discontinuous edges, such as generating contours, merging discontinuous edges, and separating backgrounds, to obtain relatively high-quality edges. Then, assembly site information is extracted through methods such as line fitting, finding intersection points, and corner point extraction. The specific process is as follows:

[0099] Use the segment_contours_xld operator to segment the XLD contour, that is, to split the continuous XLD contour into independent XLD line segments, which is convenient for subsequent screening. Observe the contour features and filter the contours through the select_shape_xld operator parameters, including aspect ratio, length, tilt angle, etc., and finally filter out the following Figure 11 The XLD profile shown in Figure 2 is plotted, and its straight line equation can be obtained.

[0100] Got Figure 11 After finding the straight line equation of the outline inside the box, the intersection point of the intersecting outline can be found. The intersection point is Figure 12Among them, point 1 corresponds to point 5, point 2 corresponds to point 6, point 3 corresponds to point 7, and point 4 corresponds to point 8. After the assembly points are aligned, the assembly of the quartz vibration beam and the quartz pendulum can be realized. Figure 12 Shown are the assembly locations that need to be extracted on the quartz vibration beam and the pendulum.

[0101] 3. Low stress double-sided alignment mounting method

[0102] The quartz vibration beam is 100μm thick and is a cross-scale part. It is prone to breakage due to uneven force and requires simultaneous double-sided pressurization and assembly. The quartz pendulum is 500μm thick and has an obvious double-sided symmetrical structure with high symmetry and good edge properties. There are no burrs under microscopic vision, so no additional gold-sprayed marking is required for auxiliary alignment. Under the circular lighting method, the features on its symmetrical structure can be directly selected for double-sided mounting alignment. The specific process is: first place a quartz vibration beam at the bottom, and the upper camera obtains the assembly position information of its upper end face. Then take the quartz pendulum, and the lower camera obtains the assembly position information of its lower end face. The quartz pendulum and the bottom quartz vibration beam are mounted together according to the obtained assembly information. Similarly, another quartz vibration beam can be stacked on the quartz pendulum. Finally, apply pressure and heat to the upper and lower quartz vibration beams at the same time, and the two quartz vibration beams are firmly assembled with the quartz pendulum. Specific implementation method:

[0104] 1. Calibrate the mapping relationship between the camera coordinate system and the world coordinate system , the mapping relationship between the upper and lower camera coordinate systems in the dual-microscope vision system .

[0105] 2. Place a quartz vibration beam at the bottom, and the upper camera acquires its image and extracts the pixel coordinates of assembly sites 1, 2, 3, and 4 、 、 , through the mapping relationship Calculate the world coordinates of the corresponding assembly site 、 、 .

[0106] 3. Place the quartz pendulum in the field of view of the lower camera to obtain its image and extract the pixel coordinates of assembly sites 5, 6, 7, and 8 、 、 , through the mapping relationship , calculate the pixel coordinates corresponding to the upper camera pixel coordinate system 、 、 , thereby calculating the corresponding world coordinates 、 、 .

[0107] 4. According to the corresponding information 1-5, 2-6, 3-7, and 4-8 of the assembly position, adjust the posture of the quartz pendulum and place the quartz pendulum on the bottom quartz vibration beam.

[0108] 5. Similarly, obtain the assembly coordinates of the other quartz vibration beam, adjust its posture, and place it on the quartz pendulum. Finally, apply pressure and heat to both quartz vibration beams simultaneously to complete the assembly of the acceleration sensor.

[0109] Figure 13 This diagram illustrates the alignment of the assembly points of the quartz vibrating beam and the quartz pendulum according to the present invention. This diagram depicts the assembled quartz vibrating beam and the pendulum. The center portion is the quartz vibrating beam, and the pendulum includes a raised portion and a flat portion. The raised features are used to identify the assembly points during assembly. The quartz vibrating beam has two ends, and the magnified assembly relationship and the raised features on the pendulum are not readily apparent. Therefore, the two sub-images on the left and right show separate screenshots of the two ends of the quartz vibrating beam.

[0110] The foregoing description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be readily conceived by a person skilled in the art within the technical scope disclosed herein are intended to be encompassed within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be subject to the scope of protection of the claims.

Claims

1. A double-sided alignment mounting method for a quartz resonant acceleration sensor, characterized in that: Use the calibration plate to calibrate the first mapping relationship between the camera coordinate system and the world coordinate system, and the second mapping relationship between the upper and lower camera coordinate systems in the dual-microscope vision system; Placing a quartz vibration beam at the bottom of the calibration plate, the upper camera acquires an image of the quartz vibration beam and extracts pixel coordinates of four assembly points in the quartz vibration beam, and calculating the world coordinates of the four assembly points in the quartz vibration beam using a first mapping relationship; A quartz pendulum is placed in the field of view of the lower camera, an image of the quartz pendulum is acquired, and pixel coordinates of four assembly points in the quartz pendulum are extracted. The pixel coordinates corresponding to the four assembly points in the quartz pendulum in the coordinate system of the upper camera are calculated through a second mapping relationship, thereby calculating the world coordinates corresponding to the four assembly points in the quartz pendulum; According to the correspondence between the four assembly points in the quartz vibration beam and the four assembly points in the quartz pendulum, the posture of the quartz pendulum is adjusted, and the quartz pendulum is placed on the quartz vibration beam; The assembly site coordinates of another quartz vibration beam are obtained, its posture is adjusted, and it is placed on the quartz pendulum. Pressure and heating are applied simultaneously on the two quartz vibration beams to realize the assembly of the acceleration sensor.

2. The double-sided alignment mounting method of a quartz resonant acceleration sensor according to claim 1, characterized in that: The calibration plate is a flat calibration plate with a 27×27 array of calibration points with a diameter of 0.25 mm. The calibration plate has an overall size of 150×150 mm and a thickness of 0.5 mm. The calibration point processing accuracy is 1 μm and the collinearity accuracy is 1 μm.

3. The double-sided alignment mounting method of a quartz resonant acceleration sensor according to claim 2, characterized in that: The plane calibration plate is made of film material and has a thickness less than the overlapping thickness of the depth of field of the upper and lower cameras.

4. The double-sided alignment mounting method of a quartz resonant acceleration sensor according to claim 1, characterized in that: The calibration includes angle calibration and distance calibration.

5. The double-sided alignment mounting method of a quartz resonant acceleration sensor according to claim 4, characterized in that: The angle calibration process is as follows: The calibration point images collected by the upper and lower cameras are processed to extract the coordinates of the circle center. The pixel coordinates of the three circle centers in the upper camera pixel coordinate system are set as follows: , , , the coordinates of the center of the circle in the lower camera are , , ; First calculate the mean of the horizontal and vertical coordinates of the point and : （3） The parameter n represents the number of calibration points in each group selected. X1, Y1, X2, Y2, X3, and Y3 all correspond to the coordinates above. , , , are the coordinates of the same point in the upper and lower camera pixel coordinate systems respectively; Recalculate the coefficient 、 、 : （4） The slope of the regression line is: ; Similarly, the slope of the straight line containing the center points of the circle in the lower camera can be calculated ; The rotation angle between the upper camera pixel coordinate system and the lower camera pixel coordinate system is equal: （5）。 6. The double-sided alignment mounting method of a quartz resonant acceleration sensor according to claim 5, characterized in that: The distance calibration process is as follows: Pixel distance between two measurement groups on the calibration plate As shown in formula (6): （6）； In the formula d is the actual physical distance between the two measurement groups, is the camera pixel scale factor; the coordinates of the center point of the upper camera and lower camera pixel coordinate systems are established by geometric relationship formula (7), and then deduced A 、 B The analytical expression of is as follows: （7） （8） A 、 B is the distance between the origins of the two coordinate systems, is the scaling factor, X5, Y5 and are the coordinates of the same point in the upper and lower camera pixel coordinate systems, respectively, where is the angle between the second set of calibration points in the lower camera field of view and the pixel coordinate system.

7. The double-sided alignment mounting method of a quartz resonant acceleration sensor according to claim 1, characterized in that: The adjusting of the quartz pendulum posture and placing the quartz pendulum on the quartz vibration beam include: performing feature recognition on the quartz vibration beam and the quartz pendulum, and the feature recognition includes image preprocessing, template matching, edge detection, and assembly site extraction.

8. The double-sided alignment mounting method of a quartz resonant acceleration sensor according to claim 7, characterized in that: The template matching method uses a shape-based template matching algorithm. The algorithm uses the gradient correlation of object edges as the matching criterion, extracts edge features in the ROI, creates a template based on grayscale information, and generates a multi-level image pyramid model based on the size and clarity requirements of the template. The template image is then searched layer by layer from top to bottom in the image pyramid model until the bottom layer is found or a confirmed matching result is obtained.

9. The double-sided alignment mounting method of a quartz resonant acceleration sensor according to claim 7, characterized in that: The edge detection uses the Canny operator to perform sub-pixel edge detection, obtain the edge image of the quartz vibration beam or the quartz pendulum, and extract the XLD contour of the target feature.

10. The double-sided alignment mounting method of a quartz resonant acceleration sensor according to claim 7, characterized in that: The assembly site extraction includes: thinning edges larger than 1 pixel, performing corresponding processing on discontinuous edges, including generating contours, merging discontinuous edges, separating backgrounds, and obtaining edges; and then extracting assembly site information through straight line fitting, finding intersection points, and extracting corner points.

Images