A cross-scale visual ultra-precision motion measurement device and method

Abstract

The application discloses a kind of cross-scale visual ultra-precision motion measurement device and method, the device includes image sensor, aperture diaphragm, imaging lens group, coaxial light source, beam splitter, positioning platform, encoding plate and DSP and ARM integrated signal processing system.In the method, by selecting the number of linear feedback shift register and period large λ and carrying out periodic coding, after DSP and ARM signal processing system pre-processing operation to image, static random access memory is driven to carry out image caching;Two-dimensional Fourier transform is executed to cached image, and relevant linear phase processing is carried out;According to the phase data obtained by linear phase processing, the corresponding binary sequence on local image is identified;The identified local binary sequence is looked up to the binary sequence of encoding plate and positioned, and according to the phase constant of periodic phase measurement, fine positioning is carried out.

Description

Technical Field

[0001] This invention relates to the field of visual measurement research, and in particular to a cross-scale visual ultra-precision motion measurement device and method. Background Technology

[0002] Vision-based ultra-precision motion measurement refers to the technology of accurately positioning a target to a reference point in a given coordinate system. Due to its advantages such as non-contact, visualization, and multiple degrees of freedom in measurement, this technology is widely in demand in fields such as precision positioning, precision instruments, and high-end equipment.

[0003] While visual measurement methods offer numerous advantages, they typically rely on images acquired under nominal conditions of optimized contrast, uniform illumination, and occlusion prevention. However, guaranteeing these optimal conditions in industrial environments is challenging, necessitating robust design of computer vision methods. Simultaneously, measurement accuracy and measurement range are two crucial performance indicators for visual perception technologies. However, high accuracy and large range are often contradictory: high accuracy in micro-vision systems often comes at the cost of a small field of view; conversely, expanding the field of view stagnates accuracy. Problems such as small field of view and short depth of field also limit the practical operating range of micro-manipulation robots, posing difficulties for complex micro-manipulation tasks and frequently requiring trade-offs and compromises between achieving large range and high accuracy. Therefore, a method is urgently needed to simultaneously improve both effective measurement range and measurement accuracy.

[0004] Existing visual measurement methods can be categorized into feature-based, region-based, and phase-correlation-based methods. The first method utilizes local image features, such as corners or edges, to retrieve displacement, achieving sub-micron resolution. Local methods can achieve resolutions down to one-tenth of a pixel, but the resolution of optical microscopes is limited by light diffraction, resulting in a final resolution close to 10 nanometers. Using scanning electron microscopy could improve image resolution, but this would affect the measurement range. The second method is region-based. In our previous research, we developed a planar two-degree-of-freedom micro-motion measurement device based on an area array camera and a single-path microscope, and proposed a measurement algorithm based on an inverse acceleration optimization search algorithm (Li H, Zhu B, Chen Z, Zhang X. Realtime in-plane displacements tracking of the precision positioning stage based on computer micro-vision[J]. Mechanical Systems and Signal Processing, 2019, 124:111-123.). In the proposed algorithm, due to the use of sub-pixel interpolation, the measurement resolution is significantly improved (the measurement range and measurement resolution are increased from 10 nanometers to approximately 10 nanometers). 3 Upgraded to 10 4 ~10 5 The third method, the phase-based method, can achieve high-resolution measurements, but it is limited to displacements shorter than the pattern period and cannot achieve large-scale measurements (Guelpa V, Laurent GJ, Sandoz P, et al. Subpixelic Measurement of Large 1D Displacements: Principle, Processing Algorithms, Performances and Software[J]. Sensors, 2014, 14(3):5056-73.). The common problem with these existing vision measurement methods is the contradiction between the measurement range and measurement accuracy of the vision measurement system.

[0005] Visual ultra-precision measurement technology can achieve high-precision measurement, real-time high-speed measurement, non-contact measurement, and stable and reliable measurement results in terms of hardware. However, it requires measurement under conditions of low light, low vibration, and low temperature to avoid environmental interference. It also requires high-precision cameras, light sources, lenses, and other equipment, resulting in higher costs. It can be observed that existing visual ultra-precision measurement hardware mainly includes components such as high-precision cameras, lenses, light sources, and image processing software. Most of these involve capturing images with a camera and transmitting them to image processing software on a computer for processing, with less emphasis on integrated hardware. Regarding the hardware issues mentioned above in visual ultra-precision measurement, adopting an integrated approach can reduce hardware costs, increase system flexibility, and improve system stability.

[0006] There is a need for a method that enables ultra-precise measurement across scales, aiming to achieve a balance between visual measurement range and measurement accuracy in microscopic vision systems, and to effectively perceive both macro- and micro-motion positioning platforms that coordinate macro and micro movements, while also exhibiting good robustness in low-quality images (defocus, poor contrast, uneven lighting, and occlusion). Summary of the Invention

[0007] The purpose of this invention is to provide a cross-scale visual ultra-precision motion measurement device and method, which can effectively solve the problems of low positioning accuracy and small measurement range in the process of precision positioning, and provide new ideas for subsequent research and application of precision positioning guided by cross-scale visual ultra-precision measurement. To achieve the above objective, this invention application proposes a cross-scale visual ultra-precision motion measurement device and method.

[0008] The present invention is achieved by at least one of the following technical solutions.

[0009] A cross-scale visual ultra-precise motion measurement method includes the following steps:

[0010] Step 1: Based on the measurement range and resolution, describe and select the number of bits in the linear feedback shift register (LFSR) to produce the required pseudo-random binary sequence;

[0011] Step 2: Select the period size λ and the encoding pattern, and encode the pseudo-random binary data generated in Step 1 to synthesize the encoding board;

[0012] Step 3: Configure and drive the image sensor to acquire images using the DSP and ARM signal processing system;

[0013] Step 4: After the DSP and ARM signal processing systems perform preprocessing operations on the image, they drive the static random access memory to cache the image.

[0014] Step 5: Perform a two-dimensional Fourier transform on the cached image and perform related linear phase processing;

[0015] Step 6: Based on the phase data obtained from the linear phase processing in Step 5, identify the corresponding binary sequence on the local image;

[0016] Step 7: Locate the local binary sequence identified in Step 6 to the binary sequence of the encoding board by looking up the table, and locate it according to the phase constant measured by the periodic phase.

[0017] Furthermore, step 1 includes the following steps:

[0018] Based on the requirements of measurement range and measurement resolution, the LFSR is described and the number of bits n is selected to produce the required pseudo-random binary sequence.

[0019] A linear feedback shift register (LFSR) generates a pseudo-random binary sequence by feeding its own output value back into its own input. The mathematical representation of the output sequence is as follows:

[0020]

[0021] Where X(m) represents the output value of the LFSR at time m. This represents the XOR operator, where n represents the number of bits in the LFSR.

[0022] Furthermore, the encoding board is based on a pattern of linear shift register sequences, and the pattern is based on a stripe distribution.

[0023] Furthermore, the stripes consist of three sets of periods; the binary value 1 is composed of three existing periods; the binary value 0 has a missing central period; each intersection of the binary line and the binary column generates a coding unit consisting of nine points.

[0024] Furthermore, the position (x, y) on the encoder is subdivided over the entire number of wavelengths plus an additional subwavelength distance. The wavelengths are set on the encoder and provided by the vision system. The parameter δ is identified sequentially through a two-step process. x and δ y and k x and k y Retrieve the coordinates x and y on the encoding board, satisfying the following equation:

[0025] x = k x ·λ x +δ x

[0026] y = k y ·λ y +δ y

[0027] Where δ x and δ y In the laser interferometer, the distance is the remaining subwavelength distance, while on the encoder board, they are the distances in the remaining periods along the X and Y directions, respectively; λ x and λ y In the laser interferometer, it is the wavelength, while on the encoder board, it represents the period size set along the X and Y directions; k x and k y It is an integer representing the wavelength / period.

[0028] Furthermore, in step 5, a bandpass filter based on the Fourier transform applied to the recorded image is used to retrieve two sets of linearly related phases associated with the target in the horizontal and vertical directions, the position of the retrieved point distribution relative to the camera pixel frame, and the parameter δ is identified. x and δ y Specifically, it includes the following steps:

[0029] Step 5.1: Perform a two-dimensional Fourier transform on the acquired image. The following is the two-dimensional Fourier transform function:

[0030]

[0031] Where I(i,j) is the pixel intensity at position (i,j) of the m columns and n rows of the image; the power spectrum generated after transformation is filtered twice by Gaussian bandpass filters centered at frequencies f1 (the second largest frequency value in the horizontal frequency domain) and f2 (the second largest frequency value in the vertical frequency domain) and corresponding to the x and y directions of the pattern, respectively. After inverse Fourier transform, the wrapping phase distribution related to the pattern rows (horizontal direction) and columns (vertical direction) is obtained, and (u,v) represents the position of the image in the frequency domain;

[0032] Step 5.2: Unfold the wrapped phase diagram to generate a continuous phase plane, represented as follows:

[0033]

[0034]

[0035] Where i and j are the positions of the unfolded phase map in the frequency domain, P is the row period in pixels, and k1 and k2 represent unknown integer rows and columns. The period P in pixels is related to the physical distance λ between the rows and columns: P = λ·Mag / PixelSize, where Mag is the optical magnification of the imaging system, and PixelSize is the physical size of a pixel on the image sensor; planarity coefficient. and The coordinates corresponding to the spectral lobe f1 are shown, while the coordinates corresponding to the lobe f2 are shown. and Using the phase plane modulo 2π, we obtain the phase constants K1(i,j) and K2(i,j) corresponding to each pattern line in these images; Φ1(i,j) and Φ2(i,j) represent the unfolded diagrams of the row and column-related wrapped phase maps, respectively; φ x / y , φ y / x These represent the absolute phases in the target directions x and y of the encoding board, respectively;

[0036] The deflection angle of the local image relative to the corrected image is obtained from the plane coefficient of the phase plane formula. The formula is as follows:

[0037]

[0038] Where coefficients a and b correspond to the plane equation coefficients of the expanded phase plane, respectively. and

[0039] Furthermore, in step 6, the correct 2k is determined by identifying the relevant binary words encoded in the missing rows and columns. x π and 2k y π-phase adjustment, to eliminate uncertainties throughout the mode period along both directions, specifically includes the following steps:

[0040] Step 6.1: Based on the phase data and linear phase analysis obtained in step 5, the position of the point center is identified as pixels with a phase close to 0; the image background is identified as pixels with a phase close to π; and the Boolean function B... f (i,j) stores the information of the center position of the positioning point, which is determined by the complementary Boolean function B. b (i,j) stores information about the location of the background image;

[0041] Step 6.2: Convert the obtained Boolean image B f and B b They are used to construct complementary foreground thumbnails T. f and background thumbnail T b And foreground supplement thumbnail N f And background supplement thumbnail N b The phase constant of each element (k1, k2) of the thumbnail is equal to that of all image pixels of k1 and k2.

[0042] Step 6.3: For each pair of phase constants k1 and k2, the average foreground intensity can be calculated from the thumbnails obtained in the previous steps. and average background intensity These represent the average intensity observed at the center of the point and around the point, respectively. For a point where no point exists, the phase constant is... and Use similar values, but use different values ​​where points exist;

[0043] Step 6.4: Quantize the corresponding points by performing the element-to-element ratio calculation of the average thumbnail. The ratio thumbnail is represented by r(i,j):

[0044]

[0045] Where r(k1,k2) represents the ratio of the average thumbnail intensity foreground to background. Indicates average foreground intensity The ratio thumbnail r(i,j) represents the average background intensity. It robustly distinguishes between present and missing points. For non-existent points, the ratio is closer to 1, while for present points, the ratio is further away from 1 than the ratio for non-existent points.

[0046] Step 6.5: Determine the modulo-3 phase constants corresponding to the encoded rows and columns, thereby finding all rows and columns representing the encoded information, and reconstructing a thumbnail G(u,v) containing only the encoded information rows and columns. Finally, the corresponding binary data is identified based on the encoded information in the thumbnail G(u,v).

[0047] The measurement device based on a cross-scale visual ultra-precision motion measurement method includes an image sensor, an aperture stop, an imaging lens group, a coaxial light source, a beam splitter, a positioning platform, an encoding board, and a DSP and ARM integrated signal processing system; the parallel light emitted by the coaxial light source is reflected by the encoding board and imaged onto the image sensor through the imaging lens group and the aperture stop.

[0048] The image sensor is used to acquire an coded image containing the current absolute position and send it to the DSP and ARM integrated signal processing system.

[0049] Furthermore, the DSP and ARM integrated signal processing system includes a bootloader configuration module and an image reading module, an image processing module, an image frequency and phase processing module, a binary recognition module, a shift decoding module, and a VGA display module connected in sequence. The signal output terminal of the bootloader configuration module is connected to the image sensor, and the signal output terminal of the image sensor is connected to the image reading module.

[0050] Furthermore, the DSP signal processing system is used to acquire and preprocess digital images, drive SRAM for image caching, use DMA technology, and transmit data to the ARM signal processing system through a high-speed bus interface. The system then performs image frequency and phase processing, encoding and recognition to obtain binary sequence values, and decodes the data to obtain the current two-dimensional absolute position measurement value.

[0051] Compared with existing technologies, the beneficial effects of the present invention are as follows:

[0052] The encoding method of this invention defines the measurement range not by the camera's observation range, but by the physical extension of the periodic pattern. This makes the measurement range independent of the vision system's magnification, and combined with the high sub-pixel resolution allowed by phase calculation, a wide range resolution ratio can be obtained.

[0053] This invention relates to two phase-based processing steps, linear and binary. The first linear step aims to retrieve the precise location of the point distribution relative to a camera pixel frame. This linear step provides the high resolution of the method and relies on classical spectral filtering in the Fourier domain. This processing produces two unfolded phase maps, associated with two directions of the pattern, and representing the current image's position within the entire pattern modulo an unknown phase constant 2k. x π and 2k y π. The purpose of the binary steps is to determine the correct 2k... x π and 2k y π-phase adjustment is necessary to eliminate uncertainty throughout the entire pattern cycle in both directions by identifying the relevant binary words encoded in the missing rows and dots.

[0054] The present invention is based on the development of a phase-specific position decoding program that enables robust identification, applicable even in low-quality images, for retrieving absolute three-degree-of-freedom positions, while retaining the high resolution and extended measurement range capabilities of the method.

[0055] This invention is based on the encoding program of the design encoding board, and modifies the physical period λ. x and λ y We can design different encoding boards that meet the requirements of measurement range and measurement resolution by taking the effective value and the number of bits n.

[0056] This invention enables ultra-precise cross-scale measurement of coupled multi-degree-of-freedom motion within a plane. By rationally selecting the number of bits n and the period size λ of the linear feedback shift register (LFSR) and performing periodic encoding, an encoding board with adjustable resolution and measurement range can be independently designed. Combined with relevant processing algorithms integrated into the DSP and ARM, cross-scale measurement with both high precision and a large measurement range (measurement range and resolution of 10⁻⁶) can be achieved. 6 above). Attached Figure Description

[0057] Figure 1 This is a schematic diagram of the overall structure of a cross-scale visual ultra-precision motion measurement device and method;

[0058] Figure 2This invention refers to an encoding board produced with 4 bits n and a physical period λ of 9 pixels, and an encoding unit composed of 9 encoding points;

[0059] Figure 3 This is a flowchart illustrating a cross-scale visual ultra-precision motion measurement device and method according to the present invention.

[0060] The figure shows: 1-image sensor, 2-aperture stop, 3-imaging lens group, 4-coaxial light source, 5-beam splitter, 6-positioning platform, 7-encoding board. Detailed Implementation

[0061] To enable those skilled in the art to better understand the present invention, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0062] A measurement device for a cross-scale visual ultra-precision motion measurement method includes an image sensor 1, an aperture stop 2, an imaging lens group 3, a coaxial light source 4, a beam splitter 5, an encoding board 7, and a DSP and ARM integrated signal processing system; the parallel light emitted by the coaxial light source 4 is refracted by the beam splitter 5 and reflected by the encoding board 7, and then imaged onto the image sensor 1 by the imaging lens group 3 and the aperture stop 2; the encoding board 7 is located on the positioning platform 6.

[0063] The image sensor 1 is used to acquire an coded image containing the current absolute position and send it to the DSP and ARM integrated signal processing system;

[0064] The DSP and ARM integrated signal processing system includes a Bootloader configuration module and sequentially connected image reading module, image processing module, image frequency and phase processing module, binary recognition module, shift decoding module, and VGA display module (these modules correspond to their own designed algorithms and tasks, and are then integrated into the DSP and ARM). The signal output terminal of the Bootloader configuration module is connected to the image sensor 1, and the signal output terminal of the image sensor 1 is connected to the image reading module.

[0065] The DSP and ARM integrated signal processing system is configured to drive image sensor 1 for image acquisition;

[0066] The DSP and ARM integrated signal processing system is used to acquire and preprocess digital images, and then drive SRAM for image caching. A digital signal processor (DSP) is a microprocessor specifically designed for digital signal processing, while an ARM is a microprocessor based on the RISC architecture, widely used in various embedded systems. The embedded system designed here integrates the DSP and ARM, allowing them to work together and leverage their respective advantages. Image processing tasks can be assigned to the DSP, while control and communication protocol tasks can be assigned to the ARM. When transferring data between the DSP and ARM, DMA (Direct Memory Access) technology can be used to transfer data to the ARM integrated signal processing system through a high-speed bus interface (such as PCIe, USB, Ethernet, etc.) to improve data transmission efficiency. Then, the ARM integrated signal processing system performs image frequency and phase processing, encoding and recognition to obtain binary sequence values, and decoding processing to obtain the current two-dimensional absolute position measurement value.

[0067] This embodiment of a cross-scale visual ultra-precision motion measurement device and method includes the following steps:

[0068] Step 1: Based on the requirements of measurement range and resolution, describe the LFSR and select the number of bits n to produce the required pseudo-random binary sequence; specifically:

[0069] A linear feedback shift register (LFSR) produces a pseudo-random binary sequence by refeeding its output value back to its input. The mathematical representation of its output sequence is as follows:

[0070]

[0071] Where X(m) represents the output value of the LFSR at time m. This represents the XOR operator, where n represents the number of bits in the LFSR.

[0072] Encryption based on row and column order is based on a linear feedback shift register sequence, requiring a binary track regardless of the number of bits considered. In this encoding scheme, each n-bit word shares n-1 bits with its immediate neighbor, and its absolute position can be retrieved from any sequence of n consecutive bits. Combined with the choice to encode each bit using three cycles, this technique allows for encoding from 3·(2 n The encoding space length of the sequence with units of +n-2) is equal to 3·(2 n -1) cycles, while only 3n consecutive cycles are needed to decode the current position. Therefore, the ratio between the measurement range and the minimum observation range is (2 n-1) / n is given, and increases with n.

[0073] Step 2: Based on the requirements of measurement range and resolution, select the period size λ and the encoding pattern, and encode the pseudo-random binary data generated in Step 1 to synthesize the encoding board; by modifying the physical period λ... x and λ y We can design different encoding boards that meet the requirements of measurement range and measurement resolution by taking the effective value and the number of bits n.

[0074] In a preferred embodiment, the encoding board is a square encoding board, and the background color and pattern color of the square encoding board are significantly different.

[0075] Step 3: Configure and drive the image sensor to acquire images using the DSP and ARM signal processing system.

[0076] Step 4: After the DSP and ARM signal processing system preprocesses the image, it drives the SRAM to perform image buffering; the DSP and ARM signal processing system configures and drives the image sensor to perform image acquisition; specifically: the acquired coded local image can be any local area of ​​the encoding board, but the captured coded local image must be square, and it must appear in the recorded image for at least 3n cycles in both directions, where n is the number of bits selected by the linear feedback shift register.

[0077] After the DSP and ARM signal processing system preprocesses the image, it drives the SRAM to cache the image. The specific preprocessing operation is that, based on the understanding of the pattern structure of the encoder board, the information of the captured local image can be used to self-calibrate the overall encoder board information stored on the computer, so that the size of the captured local image matches the size of the overall encoder board stored on the computer. The purpose is to allow the measurement range to no longer be limited by the observation range of the vision system, so that the precise position of the local image to the overall encoder board can be located.

[0078] Steps 3 and 4 use hardware with integrated DSP and ARM signal processing systems to acquire, store, and process images.

[0079] Step 5: Perform a two-dimensional Fourier transform on the cached image and perform related linear phase processing;

[0080] Step 6: The phase data obtained from the linear phase processing in step 5 is used to identify the corresponding binary sequence on the local image. The phase-based position decoding program can achieve robust recognition, even in low-quality images, and is suitable for retrieving absolute three-degree-of-freedom positions, while retaining the high resolution and extended measurement range of the method.

[0081] This involves two phase-based processing steps: linear and binary. The first linear step aims to retrieve the precise location of the point distribution relative to the camera pixel frame. This linear step provides the high resolution of the method and relies on classical spectral filtering in the Fourier domain. This processing produces two unfolded phase maps, associated with two directions of the pattern, and representing the current image's position within the entire pattern modulo an unknown phase constant 2k. x π and 2k y π. The purpose of the binary steps is to determine the correct 2k... x π and 2k y π-phase adjustment is necessary to eliminate uncertainty throughout the entire pattern cycle in both directions by identifying the relevant binary words encoded in the missing rows and dots.

[0082] Step 7: Locate the local binary sequence identified in Step 6 in the overall encoding board binary sequence by looking up the table, and then locate it precisely based on the phase constant measured by the periodic phase.

[0083] As a preferred embodiment, based on requirements such as measurement range and resolution, a period size λ and an encoding pattern are selected. The encoding board is then synthesized by encoding the pseudo-random binary data generated in the previous step. Specifically, the designed periodic frame is based on a linear shift register sequence. To maintain phase continuity and linearity, the pattern used appears... Figure 2 The pattern is based on a regular stripe distribution, with these stripes consisting of three sets of periods. Thus, by designing suitable two-dimensional feature patterns and combining them with appropriate algorithms, cross-scale visual measurements can be achieved, making the measurement range of the visual measurement system unrestricted by measurement accuracy.

[0084] Furthermore, the binary value "1" consists of three existing periods, while the binary value "0" has a missing central period. This choice maintains a high density of frequency carriers without introducing phase nonlinearity. The QR code is obtained by taking a sequence of identical bit LFSRs along the vertical direction. Then, each intersection of the binary line and the binary column generates a coding unit consisting of nine dots, as shown below. Figure 2 As shown.

[0085] A key feature of the designed encoder and decoding method is that the measurement range is no longer defined by the camera's field of view, but by the physical extension of the periodic pattern. This makes the measurement range independent of the vision system's magnification, and combined with the high sub-pixel resolution allowed by phase calculation, a wide range resolution ratio can be achieved.

[0086] The pattern used consists of a two-dimensional periodic distribution of changing points resulting from the product of one-dimensional lines and columns. The periodic frames are designed to allow for accurate measurement of δ via phase measurements.x and δ y The changes caused by missing rows and columns are encrypted with binary code, ensuring that the parameter k is identified. x and k y The row and column order should be explicitly identified when necessary. Therefore, x and y are obtained through phase measurements: and in and

[0087] Where δ x and δ y On the encoder board, these represent the distances in the remaining periods along the X and Y directions, respectively; λ x and λ y On the encoder board, the period sizes of the designed periodic features are fixed along the X and Y directions, respectively; k x and k y y is the integer of the period on the encoder board, and x and y are the positions on the encoder board.

[0088] The binary code encryption method was chosen to meet three complementary requirements: 1) avoiding the introduction of nonlinear distortion from the periodic distribution of rows and columns; 2) providing information redundancy that is beneficial to decoding robustness; and 3) associating encoded features with phase information. In this way, the binary code reduces the amplitude of the periodic frame spectrum without changing the phase of its encoded rows and columns relative to the positions of the image pixel frames.

[0089] First, a two-dimensional Fourier transform is performed on the acquired image, followed by related linear phase processing. Specifically, a bandpass filter based on the Fourier transform applied to the recorded image is used to retrieve two sets of linearly related phases associated with the vertical direction of the target, to retrieve the precise position of the point distribution relative to the camera pixel frame, and to identify the parameter δ. x and δ y This linear step is key to achieving high resolution using this method. The specific method includes the following steps:

[0090] Step 5.1: Perform a two-dimensional Fourier transform on the acquired image. The following is the two-dimensional Fourier transform function:

[0091]

[0092] Where I(i,j) is the pixel intensity at position (i,j) of the m columns and n rows of the image. The power spectrum generated after the transformation is filtered twice by Gaussian bandpass filters centered at frequencies f1 and f2 and corresponding to the x and y directions of the pattern, respectively. After inverse Fourier transform, the wrap-around phase distribution related to the pattern rows and columns is obtained.

[0093] Step 5.2: Since the lines and columns of the known pattern are regular and continuous, the wrapping phase diagram can be unfolded to generate continuous phase planes, which can be represented as:

[0094]

[0095]

[0096] Where i and j are the pixel coordinates counted from the image center, P is the row period in pixels, and k1 and k2 represent unknown integer rows / columns. The pixel period P is related to the physical distance λ between the rows and columns: P = λ·Mag / PixelSize, where Mag is the optical magnification of the imaging system, and PixelSize is the physical size of a pixel on the image sensor. Planarity coefficient and The coordinates corresponding to the spectral lobe f1 are shown, while the coordinates corresponding to the lobe f2 are shown. and By modulo 2π, we can obtain the phase constants K1(i,j) and K2(i,j) corresponding to each pattern line in these images.

[0097] Step 5.3: The deflection angle of the local image relative to the corrected image can be obtained from the plane coefficient of the phase plane formula. The formula is as follows:

[0098] Where coefficients a and b correspond to the plane equation coefficients of the expanded phase plane, respectively. and The deflection angle sought here is not the absolute deflection angle.

[0099] The phase data obtained from the previous linear phase processing is used to identify the corresponding binary sequence on the local image; specifically: by identifying the relevant binary words encoded in the missing rows and columns, the correct 2k sequence is determined. x π and 2k y π-phase adjustment is used to eliminate uncertainties throughout the entire mode cycle along both directions. This step is crucial for achieving robust position detection. The specific method includes the following steps:

[0100] Step 6.1: Based on the phase data and linear phase analysis obtained in step 5, the center position is identified as pixels with a phase close to 0; the image background can be identified as pixels with a phase close to π. This is determined by the Boolean function B. f (i,j) stores the information of the center position of the positioning point, which is determined by the Boolean function B. b (i,j) stores information about the location of the background image.

[0101] Step 6.2: Convert the obtained Boolean image Bf and B b Used to construct complementary foreground and background thumbnails T, respectively. f and T b And supplementary thumbnail N f and N b Each element (k1, k2) of the thumbnail contains all image pixels with phase constants equal to k1 and k2. This thumbnail transformation reduces memory requirements and computation time.

[0102] Step 6.3: For each pair of phase constants k1 and k2, these thumbnails allow for the calculation of average foreground and background intensities. and These represent the average intensity observed at the center of the point and at locations surrounding the point, respectively.

[0103] For the phase constant of a point that does not exist and Similar values ​​are expected, but in the case of points, they are expected to use significantly different values.

[0104] Step 6.4 Next, this is quantified by performing an average thumbnail on the ratio of elements to elements, which is represented by r(i,j).

[0105] The ratio thumbnail r(i,j) can robustly distinguish between existing and missing points. For non-existent points, the ratio is closer to 1, while for existing points, the ratio is much further away from 1 than the ratio for non-existent points. This method clearly distinguishes between existing and missing points and has good robustness.

[0106] Step 6.5: Determine the modulo-3 phase constants corresponding to the encoded rows and columns, thereby finding all rows and columns representing the encoded information, and reassemble them into a thumbnail G(u,v) containing only the encoded information rows and columns. Finally, the corresponding binary data is identified based on the encoded information in the thumbnail G(u,v).

[0107] Similarly, this thumbnail concentrates all the rows and columns representing the encoded information into a smaller matrix, thus saving computational resources. The thumbnail G(u,v) contains the binary sequence information of the local image.

[0108] First, the locally identified binary sequence from the previous step is roughly located by looking up a table in the overall encoding board binary sequence, and then precise location is achieved based on the phase constant measured by the periodic phase measurement. Specifically, the main purpose of this step is to identify the parameter δ sequentially through the previous steps. x and δ y and k x and k yThis completes the final step of retrieving the x and y coordinates within the plane. These parameters satisfy the following equation:

[0109] x = k x ·λ x +δ x

[0110] y = k y ·λ y +δ y

[0111] Where λ x and λ y These are the known periods of the regular features of the object fixed along the X and Y directions, respectively.

[0112] The precise positioning method in step 7 includes the following steps:

[0113] Step 7.1: Extract n consecutive bits of binary data from the identified local binary sequence and the overall encoding board binary sequence stored on the computer (n bits for the linear shift register), and convert these n bits of binary data into decimal.

[0114] Step 7.2: Compare the local decimal sequence with the overall encoding board decimal sequence one by one (lookup table method). If the sequences are the same, the approximate location is successfully found.

[0115] Step 7.2: Then, based on the first constant of the horizontal phase constant matrix and the vertical phase constant matrix obtained from the linear phase measurement (the first constant from top to bottom for the horizontal phase constant and the first constant from left to right for the vertical phase constant), the upper left corner of the local pattern is precisely located on the overall coding board.

[0116] To enable those skilled in the art to better understand the present invention, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

Claims

1. A cross-scale visual ultra-precise motion measurement method, characterized in that, Includes the following steps: Step 1: Based on the measurement range and resolution, describe and select the number of bits in the linear feedback shift register (LFSR) to produce the required pseudo-random binary sequence; Step 2: Select the period size The encoding pattern is encoded based on the pseudo-random binary generated in step 1 to synthesize the encoding board; Step 3: Configure and drive the image sensor to acquire images using the DSP and ARM signal processing system; Step 4: After the DSP and ARM signal processing systems perform preprocessing operations on the image, they drive the static random access memory to cache the image. Step 5: Perform a two-dimensional Fourier transform on the cached image and perform related linear phase processing. Use a bandpass filter based on the Fourier transform applied to the recorded image to retrieve two sets of linearly related phases associated with the vertical direction of the target. Retrieve the position of the retrieved point distribution relative to the camera pixel frame and identify the parameters. and ,in and In the laser interferometer, it is the remaining subwavelength distance, while on the encoder board, they are the distances in the remaining period along the X and Y directions, respectively; Step 6: Based on the phase data obtained from the linear phase processing in Step 5, identify the corresponding binary sequence on the local image; determine the correct binary sequence by identifying the relevant binary words encoded in the missing rows and columns. and Phase adjustment to eliminate uncertainties throughout the mode cycle in both directions. and It is an integer representing the wavelength / period. Step 7: Locate the local binary sequence identified in Step 6 to the binary sequence of the encoding board by looking up the table, and locate it according to the phase constant measured by the periodic phase.

2. The cross-scale visual ultra-precise motion measurement method according to claim 1, characterized in that: Step 1 includes the following steps: Based on the requirements for measurement range and resolution, the LFSR is described and the number of bits n is selected to produce the required pseudo-random binary sequence. The linear feedback shift register (LFSR) generates a pseudo-random binary sequence by reintroducing its own output value into its own input. The mathematical representation of the output sequence is as follows: in, Indicates at time The output value of the LFSR, This represents the XOR operator. Indicates the number of bits in the LFSR.

3. The cross-scale visual ultra-precise motion measurement method according to claim 1, characterized in that: The encoder board is a pattern based on a linear shift register sequence, and the pattern is based on a stripe distribution.

4. The cross-scale visual ultra-precise motion measurement method according to claim 3, characterized in that: The stripes consist of three sets of cycles; a binary value of 1 consists of three existing cycles; a binary value of 0 has a missing center cycle; each intersection of the binary line and the binary column generates a coding unit consisting of nine points.

5. The cross-scale visual ultra-precise motion measurement method according to claim 1, characterized in that: Position on the encoder board The wavelengths are subdivided within the entire wavelength range plus an additional subwavelength distance. The wavelengths are set on an encoder and provided by the vision system, and the parameters are identified sequentially through a two-step process. and as well as and Retrieve coordinates on the encoding board and It satisfies the following equation: in and In the laser interferometer, it is the remaining subwavelength distance, while on the encoder board, they are the distances in the remaining period along the X and Y directions, respectively; and The wavelength is used in the laser interferometer, while the period size is set along the X and Y directions on the encoding board. and It is an integer representing the wavelength / period.

6. The cross-scale visual ultra-precise motion measurement method according to claim 1, characterized in that: Step 5 specifically includes the following steps: Step 5.1: Perform a two-dimensional Fourier transform on the acquired image. The following is the two-dimensional Fourier transform function: in The positions of the m columns and n rows of the image. Pixel intensity at the location; the power spectrum generated after transformation is obtained by converting the power spectrum at frequencies respectively. That is, the second largest frequency value in the horizontal frequency domain and That is, the second largest frequency value in the vertical frequency domain is centered and corresponds to the pattern. x and y The Gaussian bandpass filter in the direction is used for two filtering operations, and then after inverse Fourier transform, the wrapping phase distribution related to the pattern rows and columns is obtained. Indicates the position of the image in the frequency domain; Step 5.2: Unfold the wrapped phase diagram to generate a continuous phase plane, represented as follows: in and It is the position of the unfolded phase map in the frequency domain. It is the row period measured in pixels. , Represents an unknown integer row or column, with a period in pixels. Related to the physical distance λ between rows and columns: in It is the optical magnification of the imaging system. It refers to the physical size of a pixel on an image sensor; the planarity factor. and Corresponding to spectral lobe The coordinates, and the lobes Located at coordinates and , to phase plane Using the modulus, we obtain the phase constant corresponding to each pattern line in these images. , ; , These are unfolded diagrams representing the row and column-related wrapper phase diagrams, respectively. , These represent the absolute phases in the target directions x and y of the encoding board, respectively; The deflection angle of the local image relative to the corrected image is obtained from the plane coefficient of the phase plane formula. The formula is: Where the coefficient a , The plane equation coefficients corresponding to the expanded phase planes and .

7. The cross-scale visual ultra-precise motion measurement method according to claim 1, characterized in that: Step 6 specifically includes the following steps: Step 6.1: Based on the phase data obtained from the linear phase processing in step 5 and the linear phase analysis, the position of the point center is identified as a pixel with a phase close to 0; the image background is identified as a pixel with a phase close to 0. The pixels; determined by Boolean functions. The information about the center position of the location point is stored by complementary Boolean functions. Stores information about the location of the background image; Step 6.2: Obtain the Boolean image. and They are used to construct complementary foreground thumbnails. and background thumbnail And foreground supplement thumbnail and background supplement thumbnails Each element of the thumbnail The phase constant of the set is equal to and All image pixels; Step 6.3: For each pair of phase constants and The average foreground intensity is calculated from the thumbnails obtained in the previous steps. and average background intensity , representing the average intensity observed at the center of the point and around the point, respectively. For the phase constant where no point exists, and Use similar values, but use different values ​​where points exist; Step 6.4: Quantize the corresponding points by performing the element-to-element ratio calculation of the average thumbnail. The ratio thumbnail uses... To indicate: in This represents the ratio of the foreground to the background intensity in the average thumbnail. Indicates average foreground intensity Indicates average background intensity; ratio thumbnail A robust distinction is made between existing and missing points. For points that do not exist, the ratio is closer to 1, while for points that exist, the ratio is further away from 1 than the ratio for points that do not exist. Step 6.5: Determine the modulo-3 phase constants corresponding to the encoded rows and columns, thereby finding all rows and columns representing the encoded information, and reconstructing a thumbnail containing only the encoded information rows and columns. Finally, based on the thumbnail The encoded information enables the recognition of the corresponding binary data.

8. A measuring device for implementing the cross-scale visual ultra-precision motion measurement method as described in claim 1, characterized in that: It includes an image sensor (1), an aperture stop (2), an imaging lens group (3), a coaxial light source (4), a beam splitter (5), a positioning platform (6), an encoding board (7), and a DSP and ARM integrated signal processing system; the parallel light emitted by the coaxial light source (4) is reflected by the encoding board (7) and imaged onto the image sensor (1) through the imaging lens group (3) and the aperture stop (2); The image sensor (1) is used to acquire an coded image containing the current absolute position and send it to the DSP and ARM integrated signal processing system.

9. The measuring device according to claim 8, characterized in that: The DSP and ARM integrated signal processing system includes a Bootloader configuration module and an image reading module, an image processing module, an image frequency and phase processing module, a binary recognition module, a displacement decoding module, and a VGA display module connected in sequence. The signal output terminal of the Bootloader configuration module is connected to the image sensor (1), and the signal output terminal of the image sensor (1) is connected to the image reading module.

10. The measuring device according to claim 8, characterized in that: The DSP and ARM integrated signal processing system is used to acquire and preprocess digital images, drive SRAM for image caching, use DMA technology, and transmit data to the ARM signal processing system through a high-speed bus interface. The system performs image frequency and phase processing, encoding and recognition to obtain binary sequence values, and then decodes the data to obtain the current two-dimensional absolute position measurement value.

Images