Multi-dimensional deformation measurement method and device for cryogenic equipment, cryogenic equipment and computer readable storage medium

Abstract

The application relates to the technical field of low-temperature equipment, and discloses a multi-dimensional deformation measurement method for low-temperature equipment, which comprises the following steps: acquiring reference three-dimensional data of the low-temperature equipment in a non-deformation state and collecting two-dimensional images of a plurality of feature points on key components of the low-temperature equipment by using a monocular vision system in real time; extracting real-time two-dimensional pixel coordinates of the plurality of feature points based on the two-dimensional images; performing reverse calculation by using a perspective geometry algorithm based on the reference three-dimensional data, imaging geometric parameters of the monocular vision system pre-calibrated, and the real-time two-dimensional pixel coordinates, so as to determine real-time three-dimensional space coordinates of the plurality of feature points; and determining multi-dimensional deformation parameters of the key components based on the real-time three-dimensional space coordinates and ideal three-dimensional space coordinates. In this way, the multi-dimensional geometric deformation of the key components caused by thermal expansion and cold contraction can be measured at low cost and high precision by using an existing monocular system. The application further discloses a multi-dimensional deformation measurement device for low-temperature equipment, low-temperature equipment and a computer readable storage medium.

Description

Technical Field

[0001] This application relates to the field of cryogenic equipment technology, and for example to a method, apparatus, cryogenic equipment, and computer-readable storage medium for multidimensional deformation measurement of cryogenic equipment. Background Technology

[0002] With the rapid development of the biomedical and precision manufacturing fields, automated operations involving extreme environments such as ultra-low temperature cold storage are becoming increasingly common. The operational precision of cryogenic equipment directly determines the safety of biological sample storage and processing. To achieve precise grasping and manipulation of target objects, real-time monitoring of the geometric deformation of key equipment structures, such as storage unit support mechanisms, has become a core element in ensuring the stable operation of cryogenic storage systems. Traditional geometric measurement techniques mainly rely on high-precision 3D sensors, such as laser scanners, which can acquire three-dimensional point cloud data of the target, thereby enabling three-dimensional reconstruction and deformation analysis.

[0003] Currently, relevant technologies have disclosed schemes for monitoring the three-dimensional deformation of large-scale civil structures using binocular vision principles. For example, by deploying feature points at key locations such as reservoir dams and acquiring video sequences using binocular cameras, and after monocular tracking and matching and stereo matching, the three-dimensional spatial coordinates of the feature points are calculated by combining the camera's intrinsic and extrinsic parameters, thereby achieving quantitative monitoring of structural displacement. This scheme can effectively solve the problem of three-dimensional deformation variation of large-scale structures in complex outdoor environments, acquiring depth information through stereo vision technology, and providing data support for structural safety assessment.

[0004] In the process of implementing the embodiments of this disclosure, at least the following problems were found in the related art: While related technologies have achieved the capture of the target's three-dimensional spatial coordinates, they still have significant limitations when applied to automated equipment in cryogenic environments. These solutions rely on dedicated binocular imaging hardware, which is not only costly but also requires specific installation space and baseline distance, making them difficult to deploy flexibly within the compact structure of cryogenic storage equipment. Furthermore, the stereo matching algorithm of binocular systems is prone to matching errors in complex lighting and texture environments such as low temperatures and frost, affecting measurement accuracy. For critical components in cryogenic equipment, such as storage locations, the multidimensional deformation caused by thermal expansion and contraction, especially parameters like changes in vertical spacing and planar tilt angles, it remains difficult to achieve low-cost, high-precision quantitative measurement without increasing hardware burden. Therefore, how to utilize a monocular vision system in cryogenic equipment to accurately analyze the multidimensional geometric deformation parameters of critical components, replacing expensive dedicated 3D sensors, has become a pressing technical problem to be solved.

[0005] It should be noted that the information disclosed in the background section above is only used to enhance the understanding of the background of this application, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] To provide a basic understanding of some aspects of the disclosed embodiments, a brief summary is given below. This summary is not intended as a general commentary, nor is it intended to identify key / important components or describe the scope of protection of these embodiments, but rather as a prelude to the detailed description that follows.

[0007] This disclosure provides a method, apparatus, cryogenic device, and computer-readable storage medium for measuring multidimensional deformation in cryogenic equipment, so as to accurately analyze the multidimensional geometric deformation parameters of key components using a monocular vision system in the cryogenic equipment.

[0008] In some embodiments, the multidimensional deformation measurement method for cryogenic equipment includes: acquiring reference three-dimensional data of the cryogenic equipment in a deformation-free state and real-time acquisition of two-dimensional images of multiple feature points on key components of the cryogenic equipment using a monocular vision system, wherein the reference three-dimensional data includes the ideal three-dimensional spatial coordinates of the multiple feature points; extracting the real-time two-dimensional pixel coordinates of the multiple feature points based on the two-dimensional images; determining the real-time three-dimensional spatial coordinates of the multiple feature points by inverse calculation using a perspective geometry algorithm based on the reference three-dimensional data, pre-calibrated imaging geometric parameters of the monocular vision system, and the real-time two-dimensional pixel coordinates; and determining the multidimensional deformation parameters of the key components based on the real-time three-dimensional spatial coordinates and the ideal three-dimensional spatial coordinates; wherein the multidimensional deformation parameters include the vertical spacing change, planar offset, and / or planar tilt angle of the key components.

[0009] In some embodiments, the multidimensional deformation measurement device for cryogenic equipment includes a processor and a memory storing program instructions, the processor being configured to execute the aforementioned multidimensional deformation measurement method for cryogenic equipment when running the program instructions.

[0010] In some embodiments, the cryogenic device includes: a cryogenic device body; and the aforementioned multidimensional deformation measuring device for cryogenic devices, which is installed on the cryogenic device body.

[0011] In some embodiments, the computer-readable storage medium stores program instructions that, when executed, cause a computer to perform the aforementioned multidimensional deformation measurement method for cryogenic equipment.

[0012] The multidimensional deformation measurement method, apparatus, cryogenic equipment, and computer-readable storage medium for cryogenic equipment provided in this disclosure can achieve the following technical effects: This solution uses the ideal three-dimensional spatial coordinates of key components in a deformation-free state of cryogenic equipment as a benchmark, and then uses the equipment's existing monocular vision system to acquire two-dimensional images in real time. Based on pre-calibrated imaging geometric parameters, a perspective geometry algorithm is used to inversely calculate the real-time two-dimensional pixel coordinates of the extracted feature points, thereby accurately determining the real-time three-dimensional spatial coordinates of the feature points. Finally, by comparing the real-time coordinates with the ideal coordinates, multi-dimensional deformation parameters, including changes in vertical spacing, planar offset, and planar tilt angle, can be calculated simultaneously. This solution eliminates the need for costly binocular vision hardware that is difficult to deploy in compact cryogenic environments, effectively avoiding the technical bottleneck of stereo matching algorithms prone to errors in cryogenic frost environments. Without increasing the hardware burden, it achieves low-cost, high-precision quantitative measurement of multi-dimensional geometric deformation of key components caused by thermal expansion and contraction using the existing monocular system, effectively solving the technical challenge of multi-dimensional deformation measurement in cryogenic equipment.

[0013] The above general description and the description below are exemplary and illustrative only and are not intended to limit this application. Attached Figure Description

[0014] One or more embodiments are illustrated by way of example with reference to the accompanying drawings. These illustrations and drawings do not constitute a limitation on the embodiments. Elements having the same reference numerals in the drawings are shown as similar elements. The drawings are not to be scaled. And wherein: Figure 1 This is a schematic diagram of a multidimensional deformation measurement method for cryogenic equipment provided in an embodiment of this disclosure; Figure 2 This is a schematic diagram of a method for obtaining real-time two-dimensional pixel coordinates provided in an embodiment of this disclosure; Figure 3 This is a schematic diagram of a method for determining the vertical spacing variation of key components according to an embodiment of this disclosure; Figure 4 This is a schematic diagram of a method for determining the tilt angle of a plane provided in an embodiment of this disclosure; Figure 5 This is a schematic diagram of a method for determining the real-time three-dimensional spatial coordinates of multiple feature points provided in an embodiment of this disclosure; Figure 6 This is a schematic diagram of an apparatus for a multidimensional deformation measurement method for cryogenic equipment provided in an embodiment of this disclosure. Detailed Implementation

[0015] To provide a more detailed understanding of the features and technical content of the embodiments of this disclosure, the implementation of the embodiments of this disclosure will be described in detail below with reference to the accompanying drawings. The accompanying drawings are for illustrative purposes only and are not intended to limit the embodiments of this disclosure. In the following technical description, for ease of explanation, several details are used to provide a full understanding of the disclosed embodiments. However, one or more embodiments may still be implemented without these details. In other cases, well-known structures and devices may be simplified in their depiction to simplify the drawings.

[0016] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this disclosure are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of this disclosure described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion.

[0017] Unless otherwise stated, the term "multiple" means two or more.

[0018] In this embodiment of the disclosure, the character " / " indicates that the objects before and after it are in an "or" relationship. For example, A / B means: A or B.

[0019] The term "and / or" describes an association between objects, indicating that three relationships can exist. For example, A and / or B means: A or B, or A and B.

[0020] The term "correspondence" can refer to an association or binding relationship. The correspondence between A and B means that there is an association or binding relationship between A and B.

[0021] Combination Figure 1 As shown, optionally, this disclosure provides a method for multidimensional deformation measurement of cryogenic equipment, the cryogenic equipment being equipped with a monocular vision system, including: S11, the cryogenic equipment acquires the reference three-dimensional data of the cryogenic equipment in an undeformed state and the monocular vision system acquires the two-dimensional images of multiple feature points on the key components of the cryogenic equipment in real time, wherein the reference three-dimensional data includes the ideal three-dimensional spatial coordinates of multiple feature points.

[0022] S12, the cryogenic equipment extracts the real-time two-dimensional pixel coordinates of multiple feature points based on a two-dimensional image.

[0023] S13, the cryogenic equipment uses perspective geometry algorithm to back-calculate and determine the real-time three-dimensional spatial coordinates of multiple feature points based on reference three-dimensional data, pre-calibrated imaging geometric parameters of monocular vision system, and real-time two-dimensional pixel coordinates.

[0024] S14, the cryogenic equipment determines the multidimensional deformation parameters of key components based on real-time three-dimensional spatial coordinates and ideal three-dimensional spatial coordinates; among which, the multidimensional deformation parameters include the vertical spacing change, planar offset and / or planar tilt angle of the key components.

[0025] In this solution, cryogenic equipment refers to automated storage equipment capable of stable operation in ultra-low temperature environments, such as -80°C and below. Specific forms include, but are not limited to, automated sample storage systems, intelligent access devices within cryogenic cold storage facilities, or biological sample cryopreservation equipment. In this embodiment, the cryogenic equipment integrates a monocular vision system for visual perception and a correction control system for performing grasping actions, enabling real-time adjustment of the actuator's posture after acquiring deformation data.

[0026] Before the equipment is officially put into operation, an initialization operation is required to establish the baseline needed for subsequent calculations. This step acquires baseline three-dimensional data of the cryogenic equipment in a deformation-free state, which is the fundamental basis for determining whether deformation will occur subsequently. Specifically, under ideal conditions where the equipment has been assembled and debugged and has not yet experienced any temperature fluctuations or mechanical stress, the ideal three-dimensional spatial coordinates of multiple feature points located on key components, such as the storage support mechanism, are determined using high-precision measurement methods. These feature points can be speckles sprayed on the surface of the support mechanism or pre-pasted circular reflective markers; their precise spatial positions under ideal conditions constitute the baseline three-dimensional data. During the real-time monitoring phase, the monocular vision system installed on the cryogenic equipment begins operation, acquiring two-dimensional images containing the aforementioned multiple feature points in real time. The monocular vision system is usually fixedly installed on the equipment's actuator or fixed frame, and its field of view can cover the area of ​​the key components that need to be monitored. By continuously acquiring image sequences, the system can capture component displacements caused by thermal expansion and contraction due to the cryogenic environment or mechanical fatigue caused by long-term operation. In practical applications, taking an automated sample storage system as an example, this system is equipped with a monocular vision system and a deviation correction control system. During the system initialization phase, the storage location support structure under no-load conditions is first scanned to obtain the ideal three-dimensional spatial coordinates of eight preset circular reflective markers on the support structure as reference values. Subsequently, during long-term system operation, whenever the ambient temperature changes significantly or the equipment reaches a preset operating cycle, the monocular vision system acquires the storage location image at the current moment, providing raw data for subsequent feature point extraction and multi-dimensional deformation calculation. Through the above method, this solution lays a reliable data foundation for the subsequent accurate calculation of the real-time three-dimensional spatial coordinates and multi-dimensional deformation parameters of key components.

[0027] As an optimization scheme, when the monocular vision system acquires two-dimensional images of multiple feature points on key components of the cryogenic equipment in real time, it also includes: the cryogenic equipment evaluating the imaging quality indicators of the two-dimensional images in real time, the imaging quality indicators including at least image contrast, sharpness or signal-to-noise ratio; if the imaging quality indicators are lower than a preset quality threshold, an image quality enhancement mechanism is activated; the image quality enhancement mechanism includes at least one of the following operations: adjusting the exposure parameters of the monocular vision system, turning on or adjusting the illumination parameters of the supplementary lighting device equipped with the individual vision system, or triggering the defrosting / defogging device set in front of the lens of the monocular vision system to work.

[0028] Understandably, low temperatures easily cause lens frost and fogging, severely impacting image quality. This solution introduces closed-loop image quality control, which evaluates image quality in real time and intervenes proactively, such as with automatic defrosting and dynamic illumination. This ensures that even in harsh environments, the system can still acquire high-quality raw image data, providing a fundamental guarantee for subsequent high-precision inverse calculations.

[0029] In this scheme, the cryogenic equipment can process 2D images acquired by a monocular vision system based on a pre-trained deep learning model to quickly determine the target region containing multiple feature points. This deep learning model can employ an object detection network, pre-trained on a large number of sample images containing feature points, enabling it to accurately identify the approximate locations of feature points in the image, thus narrowing the search range of subsequent processing from the entire image to a smaller candidate region. This process is equivalent to coarse localization, effectively eliminating background interference and improving the algorithm's processing speed and robustness. Subsequently, within the determined target region, a sub-pixel image processing algorithm is used to precisely locate the feature points, thereby extracting the real-time 2D pixel coordinates of multiple feature points. Specifically, within the target region, edge detection operators can be combined with gray-level centroid methods or interpolation fitting methods to perform sub-pixel level coordinate calculations on the corners or center points of feature points, obtaining coordinate values ​​with higher precision than integer pixels. This scheme, through the aforementioned hierarchical localization strategy, balances the real-time performance of image processing with measurement accuracy, providing high-precision input data for subsequent perspective geometry inverse calculations. In practical applications, taking an automated sample storage system as an example, after the monocular vision system acquires images of the storage support structure in real time, it first calls a pre-trained YOLO object detection model to perform forward inference on the images, quickly selecting multiple target regions containing eight circular reflective markers. Subsequently, within each target region, a sub-pixel edge detection algorithm based on Zernike moments is used to fit the edges of the markers, accurately calculating the sub-pixel coordinates of the center of each marker. Through this method, this solution can still stably obtain high-precision feature point pixel coordinates in complex environments such as low-temperature frost or uneven lighting, laying a solid foundation for subsequent inverse calculation of three-dimensional spatial coordinates.

[0030] In this scheme, the cryogenic device can establish a perspective projection relationship equation between real-time two-dimensional pixel coordinates and the real-time three-dimensional spatial coordinates to be solved, based on the pre-calibrated imaging geometric parameters of the monocular vision system. The imaging geometric parameters include camera intrinsic parameters, extrinsic parameters, and distortion coefficients. The camera intrinsic parameters describe the camera's internal optical characteristics such as focal length and principal point coordinates; the extrinsic parameters describe the camera's rotation and translation relative to the world coordinate system; and the distortion coefficients are used to correct the influence of lens distortion on the imaging position. By substituting these parameters into the collinearity equation, a mathematical mapping model between the two-dimensional pixel coordinates and the three-dimensional spatial coordinates can be constructed. Subsequently, the ideal three-dimensional spatial coordinates of multiple feature points are used as initial values ​​and input into the perspective projection relationship equation for iterative solution. During the solution process, the real-time two-dimensional pixel coordinates of the feature points are used as observations, and the real-time three-dimensional spatial coordinates to be solved are continuously optimized by minimizing the reprojection error. When the solution result of the perspective projection relationship equation meets the preset convergence conditions, such as the reprojection error being less than a set threshold or the coordinate change between two adjacent iterations being less than a preset value, the iteration stops, and the real-time three-dimensional spatial coordinates of multiple feature points are output. In practical applications, taking an automated sample storage system as an example, the system pre-obtains the intrinsic parameter matrix, distortion coefficients, and extrinsic parameter matrix of the monocular camera relative to the storage location coordinate system using the Zhang Zhengyou calibration method. After feature point extraction from the real-time acquired image, the sub-pixel coordinates of eight circular reflective markers are obtained. The system calls the SolvePnP algorithm in OpenCV, using the ideal three-dimensional spatial coordinates of the eight markers as initial inputs, and iteratively solves the problem by combining the real-time two-dimensional pixel coordinates with the pre-calibrated imaging geometric parameters. After multiple iterations, when the reprojection error converges to within 0.1 pixels, the algorithm outputs the real-time three-dimensional spatial coordinates of the eight markers at the current moment. Through this method, this scheme can accurately calculate the real-time three-dimensional spatial positions of feature points on key components without relying on binocular hardware, providing reliable data support for subsequent calculations of multi-dimensional deformation parameters.

[0031] In an optimized scheme, after the cryogenic device determines the real-time three-dimensional spatial coordinates of multiple feature points, the following steps are also included: the cryogenic device acquires real-time sensing data from at least one strain sensor or displacement sensor installed on a key component; the cryogenic device fuses and compares the real-time three-dimensional spatial coordinates with the real-time sensing data; if the fusion comparison result meets a preset consistency threshold, the real-time three-dimensional spatial coordinates are confirmed to be valid; otherwise, a sensor abnormality alarm is triggered or a recalibration process is initiated.

[0032] Understandably, pure visual measurements may introduce sporadic errors in extreme environments due to factors such as lighting, frost, and fog. This solution introduces traditional sensors, such as strain gauges and LVDTs, as redundancy checks. By heterogeneously fusing and cross-checking visual and physical sensor data, a dual-protection mechanism is constructed, solving the problem of insufficient reliability of single visual sensing and significantly improving the system's fault self-diagnosis capability and the reliability of measurement data.

[0033] In this scheme, critical components refer to structural units in cryogenic equipment that support the target object, play a decisive role in the operational accuracy of the equipment, and require real-time monitoring of their geometric deformation, such as the storage support mechanism. Vertical spacing variation characterizes the relative positional change of the critical component in the vertical direction, planar offset characterizes the translational displacement of the critical component in the horizontal plane, and planar tilt angle characterizes the attitude change of the plane containing the critical component. These three parameters complement each other, comprehensively reflecting the complex deformations caused by thermal expansion and contraction or mechanical fatigue.

[0034] Specifically, when determining the vertical spacing variation of key components based on real-time and ideal 3D spatial coordinates, a first feature point and a second feature point are first identified from multiple feature points. These two feature points are located at different positions on the key components, such as the upper and lower surfaces of the storage support mechanism, respectively. Then, based on the real-time and ideal 3D spatial coordinates of the first and second feature points, their Z-axis components in the vertical direction are extracted, and the real-time vertical spacing between the first and second feature points is calculated. Simultaneously, based on the ideal 3D spatial coordinates of the first and second feature points, the standard vertical spacing between them is calculated. Finally, the vertical spacing variation is determined based on the difference between the real-time and standard vertical spacing.

[0035] Optionally, when determining the plane tilt angle based on real-time and ideal three-dimensional spatial coordinates, the cryogenic equipment first identifies three non-collinear feature points from multiple feature points. Based on the real-time three-dimensional spatial coordinates of these three non-collinear feature points, a real-time spatial plane is fitted to the location of the key component at the current moment, and the real-time normal vector of this plane is calculated. Simultaneously, based on the ideal three-dimensional spatial coordinates, the ideal normal vector of the key component in its undeformed state is obtained. This ideal normal vector can be obtained by performing plane fitting on the three corresponding non-collinear feature points in the ideal three-dimensional spatial coordinates, or by directly storing pre-calculated ideal normal vector values. Finally, the angle between the real-time normal vector and the ideal normal vector is calculated; this angle is the plane tilt angle.

[0036] Optionally, when determining the planar offset based on real-time and ideal three-dimensional spatial coordinates, the cryogenic equipment first obtains the real-time X-axis and Y-axis components of at least one of multiple feature points based on their real-time three-dimensional spatial coordinates. Simultaneously, it obtains the ideal X-axis and Y-axis components of the feature point based on its ideal three-dimensional spatial coordinates. Then, it calculates the difference between the real-time and ideal X-axis components to obtain the X-direction offset; and calculates the difference between the real-time and ideal Y-axis components to obtain the Y-direction offset. Finally, based on the X-direction and Y-direction offsets, it determines the planar offset, which can be in the form of a two-dimensional vector containing component displacements in both the X and Y directions.

[0037] In practical applications, taking an automated sample storage system as an example, eight circular reflective markers are set on the support structure of the storage location, with two markers located at the upper and lower ends of the support column, respectively. After the system calculates the real-time three-dimensional spatial coordinates of these two markers, it calculates the distance between them in the Z-axis direction and subtracts it from the standard vertical spacing stored during initial calibration to obtain the change in vertical spacing Δh caused by low-temperature shrinkage of the support column. Simultaneously, the real-time three-dimensional spatial coordinates of three non-collinear markers on the support platform are selected, the plane of the current platform is fitted, its normal vector is calculated, and the angle between this normal vector and the ideal normal vector in the initial state is calculated to obtain the tilt angle Δθ caused by uneven force on the platform. Furthermore, the real-time X and Y coordinates of a marker at the center of the support platform are selected and subtracted from the ideal X and Y coordinates in the initial state to obtain the offset Δx and Δy of the platform in the horizontal plane. Through the above methods, this scheme can comprehensively and quantitatively obtain the multi-dimensional deformation parameters of key components, providing accurate data input for subsequent corrective control and predictive maintenance.

[0038] The multidimensional deformation measurement method for cryogenic equipment provided in this disclosure uses the ideal three-dimensional spatial coordinates of key components in a deformation-free state as a reference, and then uses the existing monocular vision system of the equipment to acquire two-dimensional images in real time. Based on pre-calibrated imaging geometric parameters, a perspective geometry algorithm is used to inversely calculate the real-time two-dimensional pixel coordinates of the extracted feature points, thereby accurately determining the real-time three-dimensional spatial coordinates of the feature points. Finally, by comparing the real-time coordinates with the ideal coordinates, multidimensional deformation parameters, including vertical spacing changes, planar offsets, and planar tilt angles, can be calculated simultaneously. This approach eliminates the need for costly binocular vision hardware that is difficult to deploy in compact cryogenic environments, effectively avoiding the technical bottleneck of stereo matching algorithms prone to errors in cryogenic frost environments. Without increasing the hardware burden, it achieves low-cost, high-precision quantitative measurement of multidimensional geometric deformation of key components caused by thermal expansion and contraction using an existing monocular system, effectively solving the technical challenge of multidimensional deformation measurement in cryogenic equipment.

[0039] In one optimized approach, before acquiring baseline 3D data, the cryogenic equipment establishes a unified world coordinate system, which provides a consistent measurement benchmark for all ideal 3D spatial coordinates and real-time 3D spatial coordinates. This eliminates systematic errors caused by inconsistencies in coordinate systems and ensures the accuracy and reliability of subsequent multidimensional deformation parameter calculations.

[0040] Combination Figure 2 As shown, optionally, in step S12, the cryogenic device extracts the real-time two-dimensional pixel coordinates of multiple feature points based on a two-dimensional image, including: S21, the cryogenic equipment processes two-dimensional images based on a pre-trained deep learning model to determine the target region where multiple feature points are located.

[0041] S22, within the target area, the cryogenic equipment uses a sub-pixel image processing algorithm to extract the real-time two-dimensional pixel coordinates of multiple feature points.

[0042] In this scheme, the cryogenic equipment can process the two-dimensional images acquired by the monocular vision system based on a pre-trained deep learning model to determine the target region where multiple feature points are located. This deep learning model can employ a convolutional neural network-based target detection architecture, such as YOLO or Faster R-CNN. By pre-training on a large number of sample images containing feature points, it can quickly and accurately identify the approximate locations of feature points in complex backgrounds such as low temperatures, frost, or uneven lighting. This efficiently narrows the search scope for subsequent fine-tuning from the entire image to one or more smaller candidate target regions. This process is equivalent to coarse localization, effectively eliminating the influence of background noise and interference, significantly improving the processing speed and robustness of the algorithm. Subsequently, within the determined target region, a sub-pixel image processing algorithm is used to extract the real-time two-dimensional pixel coordinates of multiple feature points. Since the target region has been significantly reduced, a computationally more complex but more accurate algorithm can be used for precise localization. Specifically, edge detection operators can be combined with gray-level centroid methods, Gaussian surface fitting methods, or sub-pixel localization methods based on Zernike moments within the target area to perform sub-pixel-level coordinate calculations on the corners or center points of feature points. This yields coordinate values ​​with precision higher than integer pixels, typically reaching 0.1 pixels or even higher. This coarse-to-fine hierarchical localization strategy ensures both real-time image processing and provides a high-precision data foundation for subsequent perspective geometry inverse calculations.

[0043] In practical applications, taking an automated sample storage system as an example, after the monocular vision system acquires images of the storage support structure in real time, it first calls a pre-trained YOLOv5 target detection model to perform forward inference on the entire image, quickly selecting eight rectangular target regions containing eight circular reflective markers. Subsequently, within each target region, a weighted calculation is performed on all pixels in the marker region using a bilinear interpolation-based gray-level centroid method to accurately fit the sub-pixel coordinates of the center of each marker, achieving a coordinate accuracy of 0.1 pixels. Through this method, this solution can stably and efficiently acquire high-precision feature point pixel coordinates even in complex environments such as low-temperature frost, uneven lighting, or partial occlusion, laying a solid foundation for subsequent inverse calculation of three-dimensional spatial coordinates.

[0044] As an alternative to this step, the coarse localization process can also use traditional image processing algorithms instead of deep learning models. For example, adaptive thresholding based on grayscale histograms combined with morphological filtering can be used to initially extract candidate regions for feature points; or HSV space thresholding based on feature point colors can be used to quickly locate the possible positions of feature points. The fine localization process can also use other sub-pixel localization algorithms. For example, for circular markers, the center coordinates can be extracted using a least-squares ellipse fitting method; for checkerboard corner points, sub-pixel precise localization can be achieved using Harris corner detection combined with the Forstner localization operator. Thus, by using traditional image processing algorithms for coarse localization and combining multiple sub-pixel localization algorithms as alternatives, accurate feature point extraction can be flexibly achieved under different application scenarios and feature point types, effectively ensuring the robustness and adaptability of this method in diverse industrial environments.

[0045] In one optimized scheme, after extracting the real-time 2D pixel coordinates, the pre-calibrated distortion parameters of the monocular vision system are first obtained. These distortion parameters typically include radial and tangential distortion coefficients, used to describe the nonlinear imaging distortion caused by lens manufacturing or assembly errors. Subsequently, based on the distortion parameters, the extracted real-time 2D pixel coordinates are distorted. By mapping the actual imaging points back to the ideal imaging position, the pixel position offset caused by lens distortion is eliminated, thus obtaining the corrected real-time 2D pixel coordinates. Correspondingly, during perspective geometry inverse calculation, the original extracted pixel coordinates are no longer used. Instead, the corrected real-time 2D pixel coordinates are used as input, combined with pre-calibrated imaging geometry parameters and reference 3D data, and a perspective geometry algorithm is used for inverse calculation to finally determine the real-time 3D spatial coordinates of multiple feature points. Through this distortion correction process, the influence of lens defects on the imaging position can be effectively eliminated, ensuring that the pixel coordinates input to the inverse calculation model strictly conform to the pinhole imaging model, thereby significantly improving the accuracy of real-time 3D spatial coordinate calculation and providing a reliable guarantee for the accurate calculation of subsequent multi-dimensional deformation parameters.

[0046] In one optimized scheme, the cryogenic device uses a sub-pixel image processing algorithm within the target area to extract the real-time two-dimensional pixel coordinates of multiple feature points, including: The cryogenic device acquires the historical pixel coordinates of multiple feature points in the previous frame image.

[0047] Based on historical pixel coordinates and combined with the preset motion trajectory or real-time motion parameters of the cryogenic equipment actuator, the cryogenic equipment predicts the predicted pixel coordinates of multiple feature points in the current frame image.

[0048] The cryogenic equipment dynamically adjusts the target area based on predicted pixel coordinates to narrow down the target area.

[0049] Within the dynamically adjusted target area, the cryogenic equipment uses a sub-pixel image processing algorithm to extract real-time two-dimensional pixel coordinates.

[0050] Specifically, the cryogenic device first acquires the historical pixel coordinates of multiple feature points in the previous frame image, recording the precise location of the feature points at the previous moment. Then, based on these historical pixel coordinates and combined with the preset motion trajectory or real-time motion parameters of the cryogenic device's actuator, a motion prediction model estimates the possible locations of multiple feature points in the current frame image, thus obtaining predicted pixel coordinates. Next, based on the predicted pixel coordinates, the initially determined target area is dynamically adjusted to further narrow its range, ensuring it closely surrounds the expected locations of the feature points. Finally, within the dynamically adjusted, smaller target area, a sub-pixel image processing algorithm extracts the real-time two-dimensional pixel coordinates of multiple feature points. This approach, by introducing motion prediction-based dynamic ROI technology combined with the kinematic model of the actuator, can more accurately predict the possible locations of feature points, thereby further narrowing the search range. This not only significantly improves the real-time performance of image processing and reduces computational consumption but also effectively avoids target loss due to rapid movement or brief occlusion, improving tracking robustness.

[0051] Combination Figure 3 As shown, optionally, the cryogenic equipment determines the vertical spacing changes of key components based on real-time three-dimensional spatial coordinates and ideal three-dimensional spatial coordinates, including: S31, the cryogenic equipment determines a first feature point and a second feature point from multiple feature points; wherein the first feature point and the second feature point are respectively set at different positions of the key components.

[0052] S32, the cryogenic equipment determines the real-time vertical distance between the first feature point and the second feature point based on the real-time three-dimensional spatial coordinates of the first feature point and the second feature point.

[0053] S33, the cryogenic equipment determines the standard vertical distance between the first feature point and the second feature point based on the ideal three-dimensional spatial coordinates of the first feature point and the ideal three-dimensional spatial coordinates of the second feature point.

[0054] S32, the cryogenic equipment determines the change in vertical spacing based on the real-time vertical spacing and the standard vertical spacing.

[0055] In this scheme, the cryogenic equipment first determines a first feature point and a second feature point from multiple feature points. The first and second feature points are respectively located at different positions on the critical component; for example, one is located on the upper surface of the storage support mechanism, and the other on the lower surface of the support mechanism, or they are respectively located on two relatively moving components, so that the geometric deformation of the critical component in the vertical direction can be characterized by the change in their relative positions. Subsequently, the cryogenic equipment extracts the Z-axis components of the first and second feature points in the vertical direction based on their real-time three-dimensional spatial coordinates. By calculating the difference between the Z-axis components of the first and second feature points, the real-time vertical distance between the two feature points is obtained, which reflects the actual distance between the two feature points in the height direction at the current moment. Simultaneously, the ideal Z-axis components of the first and second feature points in the vertical direction are extracted based on their ideal three-dimensional spatial coordinates. By calculating the difference between the ideal Z-axis components of the first and second feature points, the standard vertical distance between them is obtained. This standard vertical distance represents the reference distance between the two feature points when the equipment is in a deformation-free state. Finally, the difference between the real-time vertical distance and the standard vertical distance is calculated; this difference represents the change in the vertical distance of the critical component. When this value is positive, it indicates that the vertical distance between the two feature points has increased, meaning that the critical component has undergone tensile deformation in the vertical direction; when this value is negative, it indicates that the vertical distance has decreased, meaning that compressive deformation has occurred.

[0056] In practical applications, taking an automated sample storage system as an example, a circular reflective marker is set at the top and bottom of the support column of the storage location as a feature point. During system initialization, the ideal three-dimensional spatial coordinates of the upper marker are obtained through high-precision measurement as 100.00, 200.00, 500.00, and the ideal three-dimensional spatial coordinates of the lower marker are 100.00, 200.00, 0.00, calculating a standard vertical spacing of 500.00 mm. After operating in a low-temperature environment for a period of time, the system calculates the real-time three-dimensional spatial coordinates of the upper marker as 100.02, 199.98, 499.85, and the real-time three-dimensional spatial coordinates of the lower marker as 100.01, 200.01, 0.02, calculating a real-time vertical spacing of 499.83 mm. Subtracting the standard vertical spacing from the real-time vertical spacing yields a change in vertical spacing of -0.17 mm, indicating that the vertical height of the support column decreased by 0.17 mm due to low-temperature contraction. Through the above methods, this solution can accurately quantify the degree of deformation of key components in the vertical direction, providing crucial data support for subsequent grasping posture adjustment and structural health assessment.

[0057] As an alternative to this step, the vertical spacing change can also be calculated using multi-point fitting. For example, multiple feature points can be selected on the upper surface of the key component to fit an upper plane, and multiple feature points on the lower surface to fit a lower plane. The vertical distance between the upper and lower planes is then calculated as the real-time vertical spacing, which is compared with the vertical spacing under standard conditions. This reduces single-point measurement errors through multi-point redundancy information, further improving the stability and accuracy of the vertical spacing change calculation. Furthermore, for key components with symmetrical structures, multiple pairs of feature points can be selected to calculate the vertical spacing change separately, and the average value is taken as the final measurement result. By employing multi-point fitting or symmetrical point pair averaging as alternatives, the redundant information of feature points can be fully utilized to effectively suppress single-point measurement noise and random errors, thereby significantly improving the anti-interference capability of the vertical spacing change calculation and the stability and accuracy of the measurement results.

[0058] In one optimized scheme, the cryogenic device determines a first feature point and a second feature point from multiple feature points, including: the cryogenic device acquiring the encoded identification information of each feature point among the multiple feature points; and based on the encoded identification information, the cryogenic device identifying the feature point that matches a preset code as the first feature point or the second feature point.

[0059] In this scheme, the cryogenic equipment first acquires pre-set coded identification information on each feature point. This coded identification information can be the feature point's own geometric shape code, reflectivity code, or an attached QR code, colored ring, or other identifiable marker. Subsequently, the cryogenic equipment parses the coded identification information of each feature point based on image processing or decoding algorithms and compares it with a pre-set coding library, thereby accurately identifying feature points that match the pre-set codes and designating them as either the first or second feature point. This scheme, by introducing a coding identification mechanism, enables rapid and accurate location of target feature points in large-scale, multi-feature-point scenarios, effectively avoiding mismatches caused by similar appearances or symmetrical positions of feature points, significantly improving the reliability and automation of feature point recognition. In practical applications, for example, different QR codes with different patterns are pasted on eight circular reflective markers on a storage support mechanism. After acquiring the image, the monocular vision system uses a decoding algorithm to identify two points marked A01 and B02, which are used as the first and second feature points for calculating the vertical spacing change, respectively.

[0060] In an optimized scheme, the cryogenic device determines a first feature point and a second feature point from multiple feature points, including: the cryogenic device acquiring the spatial position distribution of each feature point among the multiple feature points; and the cryogenic device matching two feature points corresponding to a preset positional relationship from the spatial position distribution based on prior information of the geometric structure of key components, which are respectively used as the first feature point and the second feature point.

[0061] In this scheme, the cryogenic device first acquires the spatial distribution of multiple feature points obtained through image analysis within the current field of view, including the two-dimensional pixel coordinates or preliminary calculated three-dimensional relative positions of each feature point. Subsequently, the cryogenic device calls pre-stored prior information on the geometric structure of key components. This prior information describes the inherent positional relationships between feature points on the key components, such as vertical correspondence, left-right symmetry, or specific distance constraints. By matching the measured spatial distribution of feature points with the preset geometric prior information, two feature points corresponding to preset positional relationships, such as vertical correspondence or horizontal alignment, are identified and designated as the first and second feature points, respectively. Thus, through this geometric prior-based matching method, automatic identification and pairing can be achieved using the inherent spatial relationships between feature points without additional coding or labeling, effectively simplifying system complexity and improving the accuracy and robustness of identification in scenarios with similar feature point appearances. In practical applications, for example, the four corner points of the upper platform and the four corner points of the lower platform of the storage support mechanism have a vertically corresponding geometric relationship. By analyzing the spatial distribution of eight feature points, the system automatically matches two pairs of points corresponding to the upper and lower positions, which are respectively used as the first feature point and the second feature point for calculating the change of vertical spacing.

[0062] Combination Figure 4 As shown, optionally, the cryogenic device determines the plane tilt angle based on real-time three-dimensional spatial coordinates and ideal three-dimensional spatial coordinates, including: S41, the cryogenic equipment identifies three non-collinear feature points from multiple feature points.

[0063] S42, the cryogenic equipment fits the real-time spatial plane where the key component is located at the current moment based on the real-time three-dimensional spatial coordinates of three non-collinear feature points, and determines the real-time normal vector of the real-time spatial plane.

[0064] S43, the cryogenic equipment obtains the ideal normal vector of the key component in the formless deformation state based on the ideal three-dimensional spatial coordinates.

[0065] S44, the cryogenic equipment calculates the angle between the real-time normal vector and the ideal normal vector to obtain the plane tilt angle.

[0066] In this scheme, three non-collinear feature points are first determined from multiple feature points. Selecting three non-collinear points is the minimum requirement for determining a spatial plane, ensuring a unique definition of the plane. These three feature points are typically located on the same rigid surface of the key component, such as the upper or lower surface of a storage support platform, to ensure that the fitted plane accurately reflects the actual orientation of the surface. Subsequently, based on the real-time 3D spatial coordinates of the three non-collinear feature points, the real-time spatial plane of the key component at the current moment is fitted. The spatial plane fitting can employ a geometric method of determining a plane using three points, i.e., calculating the plane equation using the coordinates of the three points. After determining the real-time spatial plane, the real-time normal vector of the plane is further calculated. This normal vector is perpendicular to the plane, and its direction uniquely represents the spatial orientation of the plane at the current moment. Simultaneously, based on the ideal 3D spatial coordinates, the ideal normal vector of the key component in its undeformed state is obtained. The ideal normal vector can be obtained in a way that corresponds to the real-time normal vector. Specifically, three non-collinear feature points with the same position or geometric relationship as those used in real-time calculations are selected from the ideal 3D spatial coordinates, fitted to the ideal spatial plane, and then the normal vector is calculated. Alternatively, the pre-calculated ideal normal vector value can be stored directly during system initialization for direct retrieval during real-time calculations. Finally, the angle between the real-time normal vector and the ideal normal vector is calculated; this angle represents the plane tilt angle of the critical component. This angle can be calculated using the dot product formula of two vectors, and its magnitude directly reflects the degree of deflection of the critical component's current posture relative to its initial reference posture in space. It can be used to determine whether the component has tilted and the direction and magnitude of the tilt. In practical applications, taking an automated sample storage system as an example, four circular reflective markers are set on the upper surface of its storage support platform. During system initialization, three non-collinear marker points are selected, with ideal 3D spatial coordinates of A (0.00, 0.00, 100.00), B (200.00, 0.00, 100.00), and C (0.00, 200.00, 100.00). The fitted ideal spatial plane is a horizontal plane, and the ideal normal vector is 0.00, 0.00, 1.00. After running in a low-temperature environment for a period of time, the system calculates the real-time 3D spatial coordinates of the three marker points as follows: A (0.10, -0.05, 99.90), B (200.05, 0.02, 100.02), and C (-0.03, 200.08, 99.95). The fitted real-time spatial plane is then calculated, and its normal vector is 0.0015, -0.0012, 1.0000. By calculating the angle between the real-time normal vector and the ideal normal vector, the current tilt angle of the platform relative to the horizontal plane is approximately 0.09 degrees. Through this method, the solution can accurately quantify the tilt attitude of key components in space, providing crucial data support for subsequent attitude compensation of the grasping mechanism and structural health assessment.

[0067] As an alternative to this step, the plane tilt angle can also be calculated using the least squares method based on multi-point fitting. When there are more than three feature points, spatial plane fitting can be performed using all feature points located on the same rigid surface. The optimal plane equation is solved by minimizing the sum of squared distances from each point to the fitted plane, thereby improving fitting accuracy and noise resistance. This approach can effectively suppress the influence of single feature point positioning errors on the plane fitting results. Furthermore, the plane tilt angle can also be characterized by tilt components around the X-axis and Y-axis, i.e., calculating the deflection angles of the real-time plane relative to the ideal plane in the pitch and roll directions, respectively, thus providing more intuitive guidance for subsequent attitude adjustments.

[0068] Combination Figure 5 As shown, optionally, the cryogenic device, based on reference 3D data, pre-calibrated imaging geometric parameters of a monocular vision system, and real-time 2D pixel coordinates, uses a perspective geometry algorithm to perform inverse calculations to determine the real-time 3D spatial coordinates of multiple feature points, including: S51, the cryogenic equipment establishes the perspective projection relationship equation between the real-time two-dimensional pixel coordinates and the real-time three-dimensional spatial coordinates to be solved based on the imaging geometric parameters.

[0069] S52, the cryogenic equipment inputs the ideal three-dimensional spatial coordinates of multiple feature points into the perspective projection relationship equation for solution.

[0070] S53, when the solution result of the perspective projection relationship equation meets the preset convergence condition, the cryogenic equipment outputs the real-time three-dimensional spatial coordinates of multiple feature points.

[0071] In this scheme, the cryogenic equipment first establishes a perspective projection relationship equation between real-time 2D pixel coordinates and the real-time 3D spatial coordinates to be solved, based on the pre-calibrated imaging geometric parameters of the monocular vision system. These imaging geometric parameters include camera intrinsic parameters, extrinsic parameters, and distortion coefficients. The camera intrinsic parameters describe the camera's internal optical characteristics such as focal length and principal point coordinates; the extrinsic parameters describe the camera's rotation matrix and translation vector relative to the world coordinate system; and the distortion coefficients are used to correct the influence of lens distortion on the imaging position. Substituting these parameters into the collinearity equation or the direct linear transformation formula, a mathematical relationship describing how 3D spatial points are projected onto the 2D image plane can be constructed. This relationship establishes a constraint equation between the real-time 2D pixel coordinates and the real-time 3D spatial coordinates to be solved. Subsequently, the ideal 3D spatial coordinates of multiple feature points are used as initial values ​​and input into the perspective projection relationship equation for iterative solution. Understandably, since the real-time 2D pixel coordinates have already been obtained through image extraction and the imaging geometric parameters are known, only the real-time 3D spatial coordinates remain unknown in the equation. Nonlinear optimization algorithms, such as the Levenberg-Marquardt algorithm or the Gauss-Newton method, are employed. Starting with ideal 3D spatial coordinates, the iterations are continuously adjusted by minimizing the reprojection error—the error between the pixel coordinates obtained by reprojecting the currently solved real-time 3D spatial coordinates onto the image plane and the actual extracted real-time 2D pixel coordinates—to gradually approximate the true values. When the solution to the perspective projection equation meets preset convergence conditions, iteration stops and the real-time 3D spatial coordinates of multiple feature points are output. Preset convergence conditions may include a reprojection error less than a preset threshold (e.g., 0.1 pixels), a coordinate change less than a preset value (e.g., 0.01 millimeters) between two consecutive iterations, or reaching a preset maximum number of iterations (e.g., 100 iterations). When any convergence condition is met, the solution is considered to have reached the required accuracy, and the output coordinates represent the precise 3D spatial position of the feature points at the current moment.

[0072] In practical applications, taking an automated sample storage system as an example, the system pre-observes the intrinsic parameters of the monocular camera using the Zhang Zhengyou calibration method: fx=3500, fy=3500, u0=640, v0=512, distortion coefficients k1=-0.12, k2=0.05, and the extrinsic parameters of the camera relative to the storage location coordinate system. After feature point extraction of the real-time acquired image, the sub-pixel coordinates of eight circular reflective markers are obtained. The system calls the SolvePnP algorithm in OpenCV, taking the ideal three-dimensional spatial coordinates of the eight markers as initial values, and iteratively solves the problem by combining the real-time two-dimensional pixel coordinates with the pre-calibrated imaging geometric parameters. After multiple iterations, when the reprojection error converges to 0.08 pixels, which is lower than the preset threshold of 0.1 pixels, the algorithm stops iterating and outputs the real-time 3D spatial coordinates of eight marker points at the current moment. For example, the coordinates of point 1 are 100.23, 199.87, 499.65, and the coordinates of point 2 are 200.15, 300.02, 499.58, etc. Through this method, this solution can accurately calculate the real-time 3D spatial positions of feature points on key components without relying on binocular hardware.

[0073] As an alternative to this step, the perspective projection relation equations can also be solved using a direct linear transformation method. For a sufficient number of feature points, the analytical solution for real-time 3D spatial coordinates can be obtained directly by solving the linear equation system without iterative optimization. This alternative is computationally fast and suitable for applications with extremely high real-time requirements and a sufficient number of feature points. Alternatively, an optimization framework based on bundle adjustment can be used to simultaneously optimize the 3D coordinates of multiple feature points and camera pose parameters, further improving the solution accuracy through global optimization. This is particularly suitable for large-scale scenes or joint solution of multiple frames. When the number of feature points is small, a closed-loop solution based on geometric constraints can be used, utilizing known distance constraints or coplanar constraints to assist in the solution. This approach allows for flexible selection of the optimal solution strategy based on the number of feature points, accuracy requirements, and computational resource limitations in the actual application scenario. This effectively balances computational efficiency and scene adaptability while ensuring solution accuracy, meeting the real-time 3D coordinate measurement needs under different working conditions.

[0074] In one optimized scheme, before the cryogenic device performs inverse calculations using a perspective geometry algorithm based on reference 3D data, pre-calibrated imaging geometry parameters of a monocular vision system, and real-time 2D pixel coordinates, it also includes: The cryogenic equipment obtains the current ambient temperature of the cryogenic equipment.

[0075] The cryogenic equipment uses a preset temperature-deformation compensation model to perform temperature compensation correction on the imaging geometric parameters or ideal three-dimensional spatial coordinates to obtain the compensated imaging geometric parameters or compensated ideal three-dimensional spatial coordinates.

[0076] The cryogenic equipment uses perspective geometry algorithms to perform inverse calculations based on reference 3D data, compensated imaging geometric parameters or compensated ideal 3D spatial coordinates, and real-time 2D pixel coordinates.

[0077] Understandably, in ultra-low temperature environments, the thermal expansion and contraction of camera lens materials may cause slight changes in internal parameters, thus introducing measurement errors. Therefore, by introducing a temperature sensor and a compensation model, temperature drift is suppressed, significantly improving the robustness and accuracy of the measurement system under extreme temperature fluctuations.

[0078] Optionally, after determining the multidimensional deformation parameters of key components, the method further includes: The cryogenic equipment outputs multidimensional deformation parameters to the web correction control system, so that the web correction control system can adjust the gripping posture of the cryogenic equipment's actuators according to the multidimensional deformation parameters.

[0079] In this scheme, the cryogenic equipment outputs the calculated multidimensional deformation parameters to its connected correction control system in real time. This correction control system is the core unit responsible for the precise motion control of the actuators in the cryogenic equipment, typically integrated into the main controller or a dedicated motion control module. The multidimensional deformation parameters include vertical spacing changes, planar offset, and planar tilt angle. These parameters comprehensively describe the geometric deviations of key components relative to their undeformed state at the current moment. Upon receiving these deformation parameters, the correction control system uses them as input commands to calculate in real time the displacement and attitude adjustment amounts that the actuators need to compensate for. This drives the actuators, such as robotic arms or grippers, to make precise adjustments in the horizontal, vertical, and angular orientations, thereby compensating for positioning errors caused by cryogenic deformation and ensuring the accuracy and reliability of the gripping operation. Through this real-time feedback and compensation mechanism, a closed-loop integration of measurement and control is achieved, enabling the cryogenic equipment to maintain high-precision automated operation capabilities even under complex deformation conditions.

[0080] In practical applications, taking an automated sample storage system as an example, after long-term low-temperature operation, the support platform of the storage location experiences slight tilting and height changes. A monocular vision system acquires images in real time and calculates the current vertical spacing change of the storage location to be -0.15 mm, the planar offset to be 0.08 mm in the X direction and -0.12 mm in the Y direction, and the planar tilt angle to be 0.05 degrees. The system outputs these multi-dimensional deformation parameters to the correction control system in real time. Based on a preset compensation algorithm, the correction control system calculates that when the robotic arm grasps the cryopreservation tubes, it needs to increase the Z-axis descent height by 0.15 mm, move the horizontal position 0.08 mm in the positive X direction and 0.12 mm in the negative Y direction, and simultaneously tilt the end effector 0.05 degrees around the corresponding axis for attitude compensation. Subsequently, the correction control system drives the servo motor to adjust the robotic arm's trajectory according to the calculated compensation amounts, ensuring that the gripper can accurately insert the cryopreservation tubes into the storage location, effectively avoiding the risk of grasping failure or sample collision due to deformation.

[0081] In an alternative approach, multidimensional deformation parameters can also be output to the equipment's human-machine interface or monitoring system to display the deformation status of key components in real time in a visual manner, allowing operators to determine whether manual intervention or equipment maintenance is required. Furthermore, deformation parameters can also be output to the equipment's data logging system, stored in conjunction with equipment operation logs and ambient temperature data for subsequent fault tracing and deformation pattern analysis. For intelligent equipment with predictive maintenance capabilities, multidimensional deformation parameters can also be input to a health assessment module, combining historical data for trend analysis. When deformation parameters exceed a preset safety threshold, an early warning signal is automatically triggered, prompting maintenance personnel to perform inspection or calibration. This approach, by outputting multidimensional deformation parameters to the human-machine interface, data logging system, or health assessment module, enables visualized monitoring of deformation status, traceable analysis of historical data, and proactive early warning of abnormal trends, thereby significantly improving the intelligent operation and maintenance level and fault response capabilities of cryogenic equipment.

[0082] Optionally, after determining the multidimensional deformation parameters of key components, the method further includes: Cryogenic equipment constructs a multidimensional deformation time series dataset for key components based on multidimensional deformation parameters; the multidimensional deformation time series dataset is used to train a deformation prediction model to achieve predictive calibration of the deformation trend of key components.

[0083] In this solution, the cryogenic equipment stores and organizes multidimensional deformation parameters obtained from each measurement in chronological order to construct a multidimensional deformation time-series dataset for key components. This time-series dataset records multidimensional deformation information of key components at different time points, such as changes in vertical spacing, planar offset, and planar tilt angle. Each data record is associated with a corresponding timestamp, thus forming time-series data that reflects the evolution of deformation over time. The completed multidimensional deformation time-series dataset can then be used to train a deformation prediction model. This model can be a machine learning model based on recurrent neural networks, long short-term memory networks, or time series analysis algorithms. By learning from historical deformation data, it can uncover the evolutionary patterns of deformation with temperature changes, operating time, and other factors, thereby predicting the future deformation trend of key components. Based on the prediction results, the cryogenic equipment can perform preventative calibration before the deformation exceeds the allowable range, adjust the operating parameters of the actuators in advance, or arrange maintenance plans to avoid capture failures or equipment malfunctions caused by excessive deformation, thus realizing a shift from reactive response maintenance to proactive predictive maintenance.

[0084] In practical applications, taking an automated sample storage system as an example, this system triggers a multidimensional deformation measurement whenever there is a significant temperature change or the equipment reaches a preset operating cycle during its one-year continuous operation. The changes in vertical spacing, planar offset, and planar tilt angle of the storage support platform obtained from each measurement are recorded and stored in a database, with each record accompanied by a precise timestamp and ambient temperature data. After a year of data accumulation, a multidimensional deformation time-series dataset containing hundreds of deformation records is formed. The system then uses a Long Short-Term Memory (LSTM) network to train this dataset, learning the correlation between deformation, temperature changes, and operating time. After training, the system can predict the potential deformation of the storage support platform based on temperature forecast data for the next three days, and automatically issue a calibration command before the predicted deformation exceeds a safety threshold, notifying maintenance personnel to conduct inspections or triggering an automatic compensation mechanism for preventative adjustments. This extends simple deformation measurement to deformation trend prediction and preventative maintenance, significantly improving the operational reliability and intelligence level of cryogenic equipment.

[0085] In an alternative approach, the multidimensional deformation time-series dataset can also be used for other analytical purposes. For example, this dataset can be correlated with equipment fault records to uncover the correlation between deformation parameters and specific fault types, and to establish fault warning rules. When the real-time measured deformation parameters approach the deformation characteristics at the time of historical faults, the system automatically triggers a potential fault warning. Furthermore, this dataset can also be used to predict the remaining service life of equipment. By analyzing the cumulative trend of deformation over time and combining it with material fatigue characteristic models, the remaining service life of critical components can be estimated, providing a basis for spare parts procurement and replacement planning. For application scenarios lacking machine learning modeling capabilities, a simple trend analysis method based on statistical thresholds can be used. An alarm threshold for the deformation change rate can be set, and an alarm can be triggered when the deformation change rate measured multiple times consecutively exceeds the set value. This approach, by combining the multidimensional deformation time-series dataset with fault record correlation analysis, remaining service life prediction, or statistical threshold warnings, can deeply mine the intrinsic value of historical deformation data, achieving early warning of potential faults in critical components, quantitative assessment of remaining service life, and real-time monitoring of abnormal deformation trends. This provides multidimensional data decision support for predictive maintenance and intelligent operation and maintenance of cryogenic equipment.

[0086] In one optimized approach, after constructing the multidimensional deformation time-series dataset, the following steps are also included: The cryogenic equipment inputs a multidimensional deformation time series dataset into a pre-trained equipment health assessment model to obtain the current health status score and / or remaining service life prediction value of key components.

[0087] The cryogenic equipment determines whether its health status score is lower than a preset health threshold, or whether its remaining service life prediction is lower than a preset service life threshold.

[0088] If so, the cryogenic equipment will generate maintenance warning information or spare parts replacement prompts.

[0089] In this solution, after constructing a multidimensional deformation time series dataset, it is input into a pre-trained equipment health assessment model. The model outputs a health status score or a predicted remaining service life value, which is then compared with a preset threshold. When the assessment result is lower than the threshold, maintenance warnings or spare parts replacement prompts are automatically generated. This enables quantitative assessment of the health status of key components and proactive warnings of abnormal conditions, effectively improving the intelligence level and fault response efficiency of predictive maintenance for cryogenic equipment.

[0090] It should be noted that although this solution uses cryogenic equipment as an exemplary application scenario for detailed description, those skilled in the art will understand that the core of the multidimensional deformation measurement method provided by this invention lies in perspective geometry inverse calculation based on monocular vision, and it is also applicable to other automated or industrial equipment requiring high-precision deformation monitoring. All equivalent substitutions or adaptive adjustments to application scenarios made using the concept of this invention should be included within the scope of protection of this invention.

[0091] Combination Figure 6 As shown, this disclosure provides a multidimensional deformation measurement device 300 for cryogenic equipment, including a processor 301 and a memory 302. Optionally, the device 300 may further include a communication interface 303 and a bus 304. The processor 301, communication interface 303, and memory 302 can communicate with each other via the bus 304. The communication interface 303 can be used for information transmission. The processor 301 can call logical instructions in the memory 302 to execute the multidimensional deformation measurement method for cryogenic equipment described in the above embodiment.

[0092] Furthermore, the logic instructions in the aforementioned memory 302 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium.

[0093] The memory 302, as a computer-readable storage medium, can be used to store software programs and computer-executable programs, such as program instructions / modules corresponding to the methods in the embodiments of this disclosure. The processor 301 executes functional applications and data processing by running the program instructions / modules stored in the memory 302, thereby implementing the multidimensional deformation measurement method for cryogenic equipment in the above embodiments.

[0094] The memory 302 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the terminal device. Furthermore, the memory 302 may include high-speed random access memory and may also include non-volatile memory.

[0095] This disclosure provides a cryogenic device, including a cryogenic device body and the aforementioned multidimensional deformation measuring device 300 for the cryogenic device. The multidimensional deformation measuring device 300 for the cryogenic device is mounted on the cryogenic device body. The mounting relationship described herein is not limited to placement inside the cryogenic device body, but also includes mounting connections with other components of the cryogenic device, including but not limited to physical connections, electrical connections, or signal transmission connections. Those skilled in the art will understand that the multidimensional deformation measuring device 300 for the cryogenic device can be adapted to feasible cryogenic device bodies, thereby realizing other feasible embodiments.

[0096] This disclosure provides a computer-readable storage medium storing computer-executable instructions configured to perform the above-described multidimensional deformation measurement method for cryogenic equipment.

[0097] The technical solutions of this disclosure can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes one or more instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in this disclosure. The aforementioned storage medium can be a non-transitory storage medium, such as a USB flash drive, external hard drive, read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk, etc., and other media capable of storing program code.

[0098] The foregoing description and accompanying drawings fully illustrate embodiments of this disclosure to enable those skilled in the art to practice them. Other embodiments may include structural, logical, electrical, procedural, and other changes. The embodiments represent only possible variations. Individual components and functions are optional unless explicitly required, and the order of operation may vary. Parts and features of some embodiments may be included in or replace parts and features of other embodiments. Moreover, the terminology used in this application is for describing embodiments only and is not intended to limit the claims. As used in the description of embodiments and claims, the singular forms “a,” “an,” and “the” are intended to equally include the plural forms unless the context clearly indicates otherwise. Similarly, the term “and / or” as used in this application means including one or more of the associated listed items and all possible combinations thereof. Additionally, when used in this application, the term "comprise" and its variations "comprises" and / or "comprising" refer to the presence of stated features, integrals, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or groups thereof. Without further limitations, an element defined by the phrase "comprises a..." does not exclude the presence of other identical elements in the process, method, or apparatus that includes said element. In this document, each embodiment may focus on the differences from other embodiments, and similar or identical parts between embodiments can be referred to mutually. For methods, products, etc., disclosed in the embodiments, if they correspond to the method section disclosed in the embodiments, the relevant parts can be referred to the description of the method section.

[0099] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the embodiments of this disclosure. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0100] The methods and products disclosed in the embodiments herein (including but not limited to devices and equipment) can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For instance, the division of units may be merely a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces, and the indirect coupling or communication connection of devices or units may be electrical, mechanical, or other forms. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to implement this embodiment according to actual needs. In addition, the functional units in the embodiments of this disclosure may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0101] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than that shown in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. In the descriptions corresponding to the flowcharts and block diagrams in the accompanying drawings, the operations or steps corresponding to different blocks may also occur in a different order than disclosed in the description, and sometimes there is no specific order between different operations or steps. For example, two consecutive operations or steps may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. Each block in a block diagram and / or flowchart, and combinations of blocks in a block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

Claims

1. A method for multidimensional deformation measurement in cryogenic equipment, wherein the cryogenic equipment is equipped with a monocular vision system, characterized in that, include: The reference three-dimensional data of the cryogenic equipment in an undeformed state is acquired, and the two-dimensional images of multiple feature points on the key components of the cryogenic equipment are acquired in real time by a monocular vision system. The reference three-dimensional data includes the ideal three-dimensional spatial coordinates of the multiple feature points. Based on the two-dimensional image, extract the real-time two-dimensional pixel coordinates of the multiple feature points; Based on the reference 3D data, the pre-calibrated imaging geometric parameters of the monocular vision system, and the real-time 2D pixel coordinates, the real-time 3D spatial coordinates of the multiple feature points are determined by inverse calculation using a perspective geometry algorithm. Based on the real-time three-dimensional spatial coordinates and the ideal three-dimensional spatial coordinates, the multidimensional deformation parameters of the key components are determined; wherein, the multidimensional deformation parameters include the vertical spacing change, planar offset and / or planar tilt angle of the key components.

2. The method according to claim 1, characterized in that, Based on the two-dimensional image, the real-time two-dimensional pixel coordinates of the multiple feature points are extracted, including: The two-dimensional image is processed based on a pre-trained deep learning model to determine the target region where the multiple feature points are located. Within the target area, a subpixel image processing algorithm is used to extract the real-time two-dimensional pixel coordinates of the multiple feature points.

3. The method according to claim 1, characterized in that, Based on the real-time three-dimensional spatial coordinates and the ideal three-dimensional spatial coordinates, the vertical spacing change of the key components is determined, including: A first feature point and a second feature point are determined from the plurality of feature points; wherein the first feature point and the second feature point are respectively located at different positions of the key component; Based on the real-time three-dimensional spatial coordinates of the first feature point and the real-time three-dimensional spatial coordinates of the second feature point, determine the real-time vertical distance between the first feature point and the second feature point; Based on the ideal three-dimensional spatial coordinates of the first feature point and the ideal three-dimensional spatial coordinates of the second feature point, determine the standard vertical distance between the first feature point and the second feature point; The change in vertical spacing is determined based on the real-time vertical spacing and the standard vertical spacing.

4. The method according to claim 1, characterized in that, Based on the real-time three-dimensional spatial coordinates and the ideal three-dimensional spatial coordinates, the plane tilt angle is determined, including: From the plurality of feature points, determine three non-collinear feature points; Based on the real-time three-dimensional spatial coordinates of the three non-collinear feature points, fit the real-time spatial plane where the key component is located at the current moment, and determine the real-time normal vector of the real-time spatial plane. Based on the ideal three-dimensional spatial coordinates, obtain the ideal normal vector of the key component in the undeformed state; The angle between the real-time normal vector and the ideal normal vector is calculated to obtain the plane tilt angle.

5. The method according to claim 1, characterized in that, Based on the reference 3D data, the pre-calibrated imaging geometric parameters of the monocular vision system, and the real-time 2D pixel coordinates, a perspective geometry algorithm is used to inversely calculate and determine the real-time 3D spatial coordinates of the multiple feature points, including: Based on the imaging geometric parameters, establish the perspective projection relationship equation between the real-time two-dimensional pixel coordinates and the real-time three-dimensional spatial coordinates to be solved; The ideal three-dimensional spatial coordinates of the multiple feature points are input into the perspective projection relationship equation for solution; When the solution to the perspective projection relationship equation satisfies the preset convergence condition, the real-time three-dimensional spatial coordinates of the multiple feature points are output.

6. The method according to any one of claims 1 to 5, characterized in that, After determining the multidimensional deformation parameters of the key component, the method further includes: The multidimensional deformation parameters are output to the correction control system so that the correction control system can adjust the gripping posture of the actuator of the cryogenic equipment according to the multidimensional deformation parameters.

7. The method according to any one of claims 1 to 5, characterized in that, After determining the multidimensional deformation parameters of the key component, the method further includes: Based on the multidimensional deformation parameters, a multidimensional deformation time-series dataset of the key component is constructed; wherein, the multidimensional deformation time-series dataset is used to train a deformation prediction model to achieve predictive calibration of the deformation trend of the key component.

8. A multidimensional deformation measurement device for cryogenic equipment, comprising a processor and a memory storing program instructions, characterized in that, The processor is configured to execute, when running the program instructions, the multidimensional deformation measurement method for cryogenic equipment as described in any one of claims 1 to 7.

9. A cryogenic device, characterized in that, include: The body of the cryogenic equipment; The multidimensional deformation measuring device for cryogenic equipment as described in claim 8 is installed on the body of the cryogenic equipment.

10. A computer-readable storage medium storing program instructions, characterized in that, When the program instructions are executed, they cause the computer to perform the multidimensional deformation measurement method for cryogenic equipment as described in any one of claims 1 to 7.

Images