20-time amplification two-dimensional microscopic scanning control system and method

Abstract

The invention discloses a 20-time magnification two-dimensional microscopic scanning control system and method, and the system comprises an intelligent optical imaging module which is used for shooting a clear image which is free of shadow and is high in depth of field at a designated shooting position; the AI co-processing module is used for analyzing the current image shot by the intelligent optical imaging module and estimating a next shooting position and a focusing surface; and the precise motion control module is used for moving the sample to the next shooting position and performing high-precision focusing according to the pre-estimation result of the AI cooperative processing module. According to the system and the method, the focusing state is judged in real time through a color image, and accurate imaging is realized in combination with motion pre-judgment.

Description

Technical Field

[0001] This invention relates to the field of microscopic imaging and automated scanning technology, and in particular to a 20x magnified two-dimensional microscopic scanning control system and method. Background Technology

[0002] Two-dimensional microscopy is a core technology for achieving full-area and high-resolution imaging of samples, and it is widely used in fields such as digital pathology and semiconductor detection.

[0003] In existing technologies, the focusing and imaging of two-dimensional microscopic scanning often suffer from the following pain points:

[0004] First, most real-time autofocus systems rely on additional hardware such as laser sensors (coaxial laser autofocus sensors), which increases equipment costs and complicates the structure.

[0005] Secondly, focus evaluation based on grayscale images is susceptible to interference from sample impurities, leading to focus shift.

[0006] Third, scanning platforms often adopt an imaging mode of "uniform speed movement - instantaneous start and stop". Mechanical vibration and sudden speed changes during the start and stop process will reduce image clarity, or cause image blurring due to continuous high-speed movement. Summary of the Invention

[0007] To address the aforementioned problems in existing technologies, this invention provides a 20x magnified two-dimensional microscopic scanning control system and method, which uses color images to determine the focus status in real time and combines motion prediction to achieve precise imaging.

[0008] To achieve the above objectives, the present invention adopts the following technical solution:

[0009] In one embodiment of the present invention, a 20x magnification two-dimensional microscopic scanning control system is proposed, the system comprising:

[0010] The intelligent optical imaging module is used to capture clear images with no shadows and high depth of field at a designated shooting position;

[0011] The AI ​​collaborative processing module is used to analyze the current image captured by the intelligent optical imaging module and predict the next shooting position and focus surface.

[0012] The precision motion control module is used to move the sample to the next shooting position and perform high-precision focusing based on the prediction results of the AI ​​collaborative processing module.

[0013] Furthermore, the intelligent optical imaging module includes:

[0014] A zoom objective lens for continuous adjustment of the viewing angle and magnification, supporting up to 20x magnification;

[0015] A color CMOS camera for capturing color images at a frame rate of at least 30fps;

[0016] Dual LED light sources are used to provide the illumination beam;

[0017] A digital micromirror device is disposed in the output light path of the dual LED light source to spatially modulate the illumination beam, eliminate shadows, and enhance sample features;

[0018] A depth-of-field extension mechanism, connected to the color CMOS camera, is used to drive the color CMOS camera to move nanometers along the optical axis within a single exposure cycle, quickly acquire a focal stack image, and fuse it into a panoramic depth-of-field preview image in real time.

[0019] Furthermore, the precision motion control module includes:

[0020] A two-dimensional electric translation stage, driven by a servo motor and a ball screw, is used to carry the sample and perform high-precision displacement in the XY plane;

[0021] The Z-axis voice coil focusing mechanism is used to drive the zoom lens to perform focusing motion with nanometer-level precision in the Z-axis direction.

[0022] A high-precision grating encoder is used to measure the actual position of the two-dimensional electric translation stage in real time and feed the position signal back to the heterogeneous motion controller to form a position closed loop.

[0023] Furthermore, an active vibration isolation platform is integrated between the two-dimensional electric translation stage and the equipment base. The active vibration isolation platform uses its built-in inertial sensor to detect vibration signals in real time in the X / Y / Z axis translation, pitch, roll, and yaw directions. The heterogeneous motion controller drives the electromagnetic actuator built into the active vibration isolation platform to generate a reverse damping force according to the vibration signal, thereby achieving vibration isolation in the 1-10Hz frequency band.

[0024] Furthermore, the AI ​​collaborative processing module includes:

[0025] The heterogeneous motion controller, based on FPGA and MCU, is used to synchronously receive AI commands and feedback signals from the high-precision grating encoder, realize real-time closed-loop position control of the two-dimensional electric translation stage and the Z-axis voice coil focusing mechanism, and perform real-time preprocessing of the current image.

[0026] The edge AI computing unit has a built-in pre-trained convolutional neural network model, which is used to analyze the pre-processed current image, predict the next shooting position and focus area, and dynamically optimize the global scanning path.

[0027] An image stitching engine is used to stitch multiple local high-resolution images into a complete large-scale panoramic image based on the actual position information fed back by the high-precision grating encoder.

[0028] In one embodiment of the present invention, a 20x magnified two-dimensional microscopic scanning control method is also proposed, which adopts the above-mentioned 20x magnified two-dimensional microscopic scanning control system and includes the following steps:

[0029] The system begins scanning, and the dual LED light source and the digital micromirror device work together to generate a modulated illumination beam; within a single exposure cycle, the color CMOS camera is driven by the depth-of-field extension mechanism to move nanometers along the optical axis, quickly acquire the focal stack image and fuse it in real time, and output a panoramic depth-of-field preview image.

[0030] The edge AI computing unit analyzes the pre-processed current image through the pre-trained convolutional neural network model to predict the next shooting position and focus area;

[0031] The heterogeneous motion controller synchronously receives AI commands and the actual position feedback from the high-precision grating encoder, drives the two-dimensional electric translation stage and the Z-axis voice coil focusing mechanism to accurately position the sample and the zoom lens to the target position and the focusing surface, respectively.

[0032] Furthermore, to address the height deviation generated during the placement of the sample carrier, multiple non-collinear reference imaging positions are selected in the initial scanning stage. After performing high-precision displacement and focusing at these positions, the actual Z-axis coordinates and corresponding X / Y coordinates of each position are recorded. A spatial model of the sample carrier height is established by fitting using the least squares method: if it is a planar tilt deviation, the fitted model is a linear model Z = aX + bY + c; during subsequent scanning, the height is adjusted according to the current imaging position (X... i ,Y i The estimated Z-axis position is calculated and derived by substituting it into the model, and is used as the initial adjustment target of the Z-axis voice coil focusing mechanism. Then, the deviation ΔZ between the estimated Z-axis position and the actual Z-axis position is used for fine correction.

[0033] Furthermore, during each scanning movement, when the high-precision grating encoder detects that the remaining distance d ≤ deceleration threshold D, the heterogeneous motion controller performs S-shaped smooth deceleration by adjusting the PWM duty cycle.

[0034] Furthermore, before starting the scan, the system performs a low-magnification pre-scan, and the edge AI computing unit analyzes the pre-scanned image, identifies the effective sample area and the blank area, and automatically generates the optimal scanning path.

[0035] Beneficial effects:

[0036] 1. Improved Precision: Phase-shift AI fusion focusing enables focusing error ≤0.33μm, dynamic deceleration and active vibration isolation work together to achieve image acquisition position accuracy ≤2μm, and imaging resolution reaches 0.5μm under 20x zoom objective lens, meeting the needs of high-precision scenarios such as semiconductor chip inspection.

[0037] 2. Efficiency Improvement: The synergistic effect of single-frame initial focus, AI path optimization and dynamic speed adjustment reduces the scanning time of a 25mm×75mm standard pathological slide from 5 minutes to 2.5 minutes, while avoiding mechanical wear caused by ineffective movement.

[0038] 3. Strong adaptability: Multi-segment cross-focusing supports samples with thicknesses of 0.1-1mm; spatial modeling can adapt to slide placement deviations with tilt angles ≤3° and surface deformations ≤2μm; AI texture recognition adapts to various types of samples such as biological, semiconductor, and material samples, eliminating the need for manual parameter template changes and reducing the operational threshold.

[0039] 4. High stability: The six-degree-of-freedom active vibration isolation technology effectively suppresses environmental and mechanical vibrations, and the imaging pass rate in complex environments (such as next to a workshop assembly line) is increased from 72% to 99%, expanding the application scenarios of the equipment. Attached Figure Description

[0040] Figure 1 This is a schematic diagram of the 20x magnified two-dimensional microscopic scanning control system of the present invention;

[0041] Figure 2 This is a schematic diagram of the 20x magnification two-dimensional microscopic scanning control method of the present invention. Detailed Implementation

[0042] The principles and spirit of the present invention will now be described with reference to several exemplary embodiments. It should be understood that these embodiments are provided merely to enable those skilled in the art to better understand and implement the present invention, and are not intended to limit the scope of the present invention in any way. Rather, these embodiments are provided to make this disclosure more thorough and complete, and to fully convey the scope of this disclosure to those skilled in the art.

[0043] Those skilled in the art will recognize that embodiments of the present invention can be implemented as a system, device, method, or computer program product. Therefore, this disclosure can be specifically implemented in the following forms: entirely hardware, entirely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0044] According to an embodiment of the present invention, a 20x magnified two-dimensional microscopic scanning control system and method are proposed. Innovative solutions are proposed from two dimensions: intelligent focusing mechanism and precise motion control. By integrating technologies such as phase shift detection, AI analysis and active vibration isolation, the system achieves a synergistic improvement in accuracy, efficiency and adaptability.

[0045] The principles and spirit of the present invention will be explained in detail below with reference to several representative embodiments.

[0046] Figure 1 This is a schematic diagram of the 20x magnified two-dimensional microscopic scanning control system of the present invention. Figure 1 As shown, the system includes: an intelligent optical imaging module 101, a precision motion control module 103, and an AI collaborative processing module 102. Each module achieves data interaction and collaborative control through a high-speed bus.

[0047] The intelligent optical imaging module 101 is used to capture clear images with no shadows and high depth of field at a designated shooting position;

[0048] AI collaborative processing module 102 is used to analyze the current image captured by the intelligent optical imaging module 101 and predict the next shooting position and focus surface;

[0049] The precision motion control module 103 is used to move the sample to the next shooting position and perform high-precision focusing based on the prediction result of the AI ​​collaborative processing module 102.

[0050] Furthermore, the intelligent optical imaging module 101 includes:

[0051] A zoom objective lens for continuous adjustment of the viewing angle and magnification, supporting up to 20x magnification;

[0052] A color CMOS camera for capturing color images at a frame rate of at least 30fps;

[0053] Dual LED light sources are used to provide the illumination beam;

[0054] A digital micromirror device is disposed in the output light path of the dual LED light source to spatially modulate the illumination beam, eliminate shadows and enhance sample features, and alleviate the problem of overexposure / underexposure in imaging within a single field of view;

[0055] A depth-of-field extension mechanism, connected to the color CMOS camera, is used to drive the color CMOS camera to move nanometers along the optical axis within a single exposure cycle, quickly acquire a focal stack image, and fuse it into a panoramic depth-of-field preview image in real time.

[0056] Furthermore, the precision motion control module 103 includes:

[0057] The two-dimensional electric translation stage, driven by a servo motor and a ball screw, is used to carry the sample and perform high-precision displacement in the XY plane; the ball screw has a lead of 3mm, and in conjunction with the high-precision grating encoder (5000 lines), it achieves a detection accuracy of 0.6µm and a movement accuracy of 3µm;

[0058] The Z-axis voice coil focusing mechanism is used to drive the zoom lens to perform focusing motion with nanometer-level precision in the Z-axis direction; the Z-axis voice coil focusing mechanism and the high-precision grating encoder (5000 lines) achieve nanometer-level motion precision detection.

[0059] A high-precision grating encoder is used to measure the actual position of the two-dimensional electric translation stage in real time and feed the position signal back to the heterogeneous motion controller to form a position closed loop.

[0060] Furthermore, an active vibration isolation platform is integrated between the two-dimensional electric translation stage and the equipment base. The active vibration isolation platform uses its built-in inertial sensor to detect vibration signals in real time in the X / Y / Z axis translation, pitch, roll, and yaw directions. The heterogeneous motion controller drives the electromagnetic actuator built into the active vibration isolation platform to generate a reverse damping force according to the vibration signal, thereby achieving vibration isolation in the 1-10Hz frequency band.

[0061] Furthermore, the AI ​​collaborative processing module 102 includes:

[0062] The heterogeneous motion controller, based on FPGA and MCU, is used to synchronously receive AI commands and feedback signals from the high-precision grating encoder, realize real-time closed-loop position control of the two-dimensional electric translation stage and the Z-axis voice coil focusing mechanism, and perform real-time preprocessing on the current image, including image segmentation, brightness adjustment and texture feature extraction.

[0063] The edge AI computing unit has a built-in pre-trained convolutional neural network model for analyzing the pre-processed current image, estimating the next shooting position and focus plane, and dynamically optimizing the global scanning path: before scanning the sample, a global sample image is captured by another camera, the edges are extracted, and the scanning area is determined (e.g., the scanning area is 80mm^2). The edge AI computing unit plans the scanning path (in what positional order the microscopic images are acquired) based on the size (e.g. 80mm^2), shape, and size (e.g. 0.5mm^2) of each detection area (the size of one microscopic image).

[0064] The image stitching engine is used to stitch multiple local high-resolution images into a complete large-scale panoramic image based on the actual position information fed back by the high-precision grating encoder; it supports the synthesis of images with hundreds of millions of pixels; the images captured by the above scanning are all adjacent, and due to positional deviations, there will be overlaps and gaps; the image stitching engine stitches the images into a complete large image based on the actual position when the images were acquired and the content of the images; the main tasks of the image stitching engine during stitching are: clearing the overlapping areas at the stitching seams and filling in the missing content of adjacent images.

[0065] It should be noted that although several modules of the 20x magnification two-dimensional microscopic scanning control system are mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, according to embodiments of the present invention, the features and functions of two or more modules described above can be embodied in one module. Conversely, the features and functions of one module described above can be further divided and embodied by multiple modules.

[0066] Figure 2 This is a schematic diagram of the 20x magnification two-dimensional microscopic scanning control method of the present invention. Figure 2 As shown, the method includes the following steps:

[0067] S1. The system begins scanning, and the dual LED light source and the digital micromirror device work together to generate a modulated illumination beam; within a single exposure cycle, the color CMOS camera is driven by the depth-of-field extension mechanism to move nanometers along the optical axis, quickly acquire the focal stack image and fuse it in real time, and output a full depth-of-field preview image.

[0068] S2. The edge AI computing unit analyzes the pre-processed current image through the pre-trained convolutional neural network model to predict the next shooting position and focus area;

[0069] S3. The heterogeneous motion controller synchronously receives AI commands and the actual position feedback from the high-precision grating encoder, drives the two-dimensional electric translation stage and the Z-axis voice coil focusing mechanism to accurately position the sample and the zoom lens to the target position and the focusing surface, respectively.

[0070] 1. Focusing System Innovation: Phase-Shift-AI Fusion Intelligent Focusing Solution

[0071] To address the issues of low focusing accuracy and poor adaptability in traditional autofocus systems, a three-layer focusing mechanism of "phase shift detection + AI prediction + multi-segment adaptation" is proposed, as detailed below:

[0072] (1) Single-frame phase-shift high-precision initial focusing: After the sample is loaded into the device, the dual LED light source (red light 630nm, blue light 450nm) simultaneously emits illumination beams to illuminate the sample. The color CMOS camera captures the image. The edge AI computing unit extracts the RGB three-channel phase difference information in the image through a pre-trained convolutional neural network model, and calculates the peak position of image sharpness using the Tenengrad function (an image sharpness evaluation algorithm). Within 10ms, the sample and the zoom lens are moved to the corresponding positions by the two-dimensional electric translation stage and the Z-axis voice coil focusing mechanism to complete the initial focusing. The focusing error is ≤0.33μm. Compared with the traditional multi-frame focusing scheme, the efficiency is improved by more than 80%, and the motion blur problem of multi-frame acquisition is avoided.

[0073] (2) Spatial modulation of LED light source transmission field: After acquiring a frame of image, the overexposed and underexposed areas of the image are analyzed. The light field is modulated by digital micromirror device (DMD) to reduce the light transmission in overexposed areas and enhance the light transmission in underexposed areas. Highlights are suppressed and dark areas are enhanced to obtain normal brightness image 1 (overexposed areas are artificially reduced and underexposed areas are artificially enhanced). Then, the DMD modulation data and image data are fused to generate the real image 2 (correcting the artificially reduced brightness in image 1).

[0074] (3) AI dynamic prediction focusing: During the scanning process, the edge AI computing unit analyzes the texture features of the image in real time and predicts the surface undulation trend of the sample (such as tissue folds in biological slices and height differences of solder joints in semiconductor chips) through a pre-trained convolutional neural network model. Based on the prediction results, a pre-adjustment command is sent to the Z-axis voice coil focusing mechanism in advance (to move the Z-axis voice coil focusing mechanism to a height that can roughly capture the image so that the focusing position can be corrected according to the image). Combined with the feedback correction of real-time phase shift detection ((1) the position calculated in the single-frame phase shift high-precision initial focusing), the "prediction-correction" closed-loop focusing is realized, with a response time ≤30ms and a focusing accuracy of 99.2% for samples with height differences ≤5μm.

[0075] (4) Multi-segment cross-adaptation focusing: For samples of different specifications (such as 0.17mm standard glass slides and 0.5mm thick glass slides), the system automatically calls the corresponding focusing parameter templates, which can adapt to multiple samples without modifying the hardware, thus solving the adaptation limitation of "one lens, one parameter" in traditional equipment.

[0076] (5) Spatial Modeling-Deductive Focusing: To address the height deviation caused by tilting or raising the slide during placement, the system selects three non-collinear reference imaging positions (such as the upper left, upper right, and lower left corners of the scanning area) at the initial scanning stage. After performing high-precision displacement and focusing at these three positions (corresponding to the processes in (1)-(4)), the actual Z-axis coordinates of each position are recorded. Based on the corresponding X / Y coordinates, a spatial model of the slide height is established using the least squares method: if it is a planar tilt deviation, the fitted model is a linear model Z = aX + bY + c, which is the general expression of the spatial plane equation, where a, b, and c are the coefficients of the spatial plane equation. During subsequent scanning, the model is adjusted according to the current image acquisition position (X...). i ,Y i The model is substituted into the calculated Z-axis estimated position, which is used as the initial adjustment target of the Z-axis voice coil focusing mechanism. Then, the deviation ΔZ between the estimated Z-axis position and the actual Z-axis position is used for fine correction, so as to achieve the dual guarantee of "model deduction + real-time calibration".

[0077] The spatial model and AI-predicted target position are added as a reference offset during the movement process to reduce the deviation between the target position and the actual position. Finally, the accurate position is obtained through high-precision focusing of single-frame phase shift of the image.

[0078] Focusing system innovation:

[0079] This system integrates the high precision of phase-shift detection, the predictive capabilities of AI, and the extrapolative nature of spatial modeling, while also improving scene compatibility through multi-segment adaptation. Spatial modeling can reduce focusing errors caused by slide placement deviations by 60%, improving overall focusing accuracy by 3 times compared to traditional methods. The adaptation rate for complex samples increases from 65% to 98%, and the focusing accuracy for slides with tilt angles ≤3° reaches 99.5%.

[0080] 2. Innovation in Motion System: Dynamic Adaptive-Active Vibration Isolation Coordinated Control Scheme

[0081] To address the issues of low motion accuracy, significant vibration interference, and path redundancy, a three-dimensional motion control system combining dynamic deceleration, active vibration isolation, and intelligent planning is constructed, as detailed below:

[0082] (1) Distance-speed dynamic matching deceleration: Abandoning the fixed threshold deceleration mode, the system dynamically calculates the deceleration threshold D based on the preset scanning movement speed V1 and the target image acquisition accuracy requirements using a PID algorithm. During each scanning movement, when the high-precision grating encoder detects that the remaining distance d≤D, the heterogeneous motion controller achieves S-shaped smooth deceleration by adjusting the PWM duty cycle. For example, at a cruising speed of 200mm / s, D=8mm for high-precision image acquisition (2μm resolution) and D=5mm for fast scanning (5μm resolution). The speed fluctuation during deceleration is ≤5%, avoiding vibration and shock from mechanical start-stop.

[0083] (2) Six-degree-of-freedom active vibration isolation: An active vibration isolation platform is integrated between the two-dimensional electric translation stage and the equipment base. The platform uses three built-in inertial sensors to detect vibration signals in real time in the X / Y / Z axis translation, pitch, roll, and yaw directions (sampling rate 1kHz). The heterogeneous motion controller drives the electromagnetic actuator built into the active vibration isolation platform to generate a reverse damping force based on the vibration signals, achieving a vibration isolation efficiency of 99% in the 1-10Hz frequency band, effectively solving the vibration blurring problem during high-magnification imaging (20x zoom objective).

[0084] (3) AI Intelligent Path Planning: Before scanning, the system first performs a low-magnification pre-scan (1x objective lens). The edge AI computing unit analyzes the pre-scan image, identifies the effective sample area and the blank area, and automatically generates the optimal scanning path. For example, for the circular tissue area of ​​the pathological section, a "spiral" path is used instead of the traditional serpentine path, reducing the invalid movement distance by 40%. At the same time, the scanning speed is dynamically adjusted according to the complexity of the sample texture, with the speed reduced to 50 mm / s in the textured area and increased to 300 mm / s in the blank area.

[0085] Innovative advantages of the motion system: dynamic deceleration reduces the image acquisition position error to ≤2μm, active vibration isolation improves the clarity of high-magnification imaging by 25%, and intelligent path planning improves the efficiency of single-sample scanning by more than 35%.

[0086] The 20x magnification two-dimensional microscopic scanning control system and method proposed in this invention have the following advantages compared with the prior art:

[0087] 1. Break through the precision-efficiency contradiction of traditional focusing algorithms, achieve high-precision focusing at the single-frame image level, and improve the adaptability to complex samples.

[0088] 2. Construct a dynamic adaptive motion control system to solve the adaptability problem of fixed deceleration strategy and suppress the interference of mechanical vibration on imaging.

[0089] 3. Reduce unnecessary movement through intelligent path planning and expand the application scope of the device by combining multi-scenario adaptation design.

[0090] While the spirit and principles of the invention have been described with reference to several specific embodiments, it should be understood that the invention is not limited to the disclosed specific embodiments, and the division of aspects does not imply that features in these aspects cannot be combined for benefit; such division is merely for ease of description. The invention is intended to cover various modifications and equivalent arrangements included within the spirit and scope of the appended claims.

[0091] Regarding the limitation of the scope of protection of this invention, those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solution of this invention are still within the scope of protection of this invention.

Claims

1. A 20x magnified two-dimensional microscopic scanning control system, characterized in that, The system includes: The intelligent optical imaging module is used to capture clear images with no shadows and high depth of field at a designated shooting position; The AI ​​collaborative processing module is used to analyze the current image captured by the intelligent optical imaging module and predict the next shooting position and focus surface. The precision motion control module is used to move the sample to the next shooting position and perform high-precision focusing based on the prediction results of the AI ​​collaborative processing module.

2. The 20x magnification two-dimensional microscopic scanning control system according to claim 1, characterized in that, The intelligent optical imaging module includes: A zoom objective lens for continuous adjustment of the viewing angle and magnification, supporting up to 20x magnification; A color CMOS camera for capturing color images at a frame rate of at least 30fps; Dual LED light sources are used to provide the illumination beam; A digital micromirror device is disposed in the output light path of the dual LED light source to spatially modulate the illumination beam, eliminate shadows, and enhance sample features; A depth-of-field extension mechanism, connected to the color CMOS camera, is used to drive the color CMOS camera to move nanometers along the optical axis within a single exposure cycle, quickly acquire a focal stack image, and fuse it into a panoramic depth-of-field preview image in real time.

3. The 20x magnification two-dimensional microscopic scanning control system according to claim 1, characterized in that, The precision motion control module includes: A two-dimensional electric translation stage, driven by a servo motor and a ball screw, is used to carry the sample and perform high-precision displacement in the XY plane; The Z-axis voice coil focusing mechanism is used to drive the zoom lens to perform focusing motion with nanometer-level precision in the Z-axis direction. A high-precision grating encoder is used to measure the actual position of the two-dimensional electric translation stage in real time and feed the position signal back to the heterogeneous motion controller to form a position closed loop.

4. The 20x magnification two-dimensional microscopic scanning control system according to claim 3, characterized in that, An active vibration isolation platform is integrated between the two-dimensional electric translation stage and the equipment base. The active vibration isolation platform uses its built-in inertial sensor to detect vibration signals in real time in the X / Y / Z axis translation, pitch, roll, and yaw directions. The heterogeneous motion controller drives the electromagnetic actuator built into the active vibration isolation platform to generate a reverse damping force according to the vibration signal, thereby achieving vibration isolation in the 1-10Hz frequency band.

5. The 20x magnification two-dimensional microscopic scanning control system according to claim 1, characterized in that, The AI ​​collaborative processing module includes: The heterogeneous motion controller, based on FPGA and MCU, is used to synchronously receive AI commands and feedback signals from the high-precision grating encoder, realize real-time closed-loop position control of the two-dimensional electric translation stage and the Z-axis voice coil focusing mechanism, and perform real-time preprocessing of the current image. The edge AI computing unit has a built-in pre-trained convolutional neural network model, which is used to analyze the pre-processed current image, predict the next shooting position and focus area, and dynamically optimize the global scanning path. An image stitching engine is used to stitch multiple local high-resolution images into a complete large-scale panoramic image based on the actual position information fed back by the high-precision grating encoder.

6. A 20x magnified two-dimensional microscopic scanning control method, employing the 20x magnified two-dimensional microscopic scanning control system as described in any one of claims 1-5, characterized in that, Includes the following steps: The system begins scanning, and the dual LED light source and the digital micromirror device work together to generate a modulated illumination beam; within a single exposure cycle, the color CMOS camera is driven by the depth-of-field extension mechanism to move nanometers along the optical axis, quickly acquire the focal stack image and fuse it in real time, and output a panoramic depth-of-field preview image. The edge AI computing unit analyzes the pre-processed current image through the pre-trained convolutional neural network model to predict the next shooting position and focus area; The heterogeneous motion controller synchronously receives AI commands and the actual position feedback from the high-precision grating encoder, drives the two-dimensional electric translation stage and the Z-axis voice coil focusing mechanism to accurately position the sample and the zoom lens to the target position and the focusing surface, respectively.

7. The 20x magnification two-dimensional microscopic scanning control method according to claim 6, characterized in that, To address the height deviation caused by the placement of the sample carrier, multiple non-collinear reference imaging positions are selected during the initial scanning stage. After performing high-precision displacement and focusing at these positions, the actual Z-axis coordinates and corresponding X / Y coordinates of each position are recorded. A spatial model of the sample carrier height is then established using the least squares method: if it is a planar tilt deviation, the model is fitted as a linear model Z = aX + bY + c; during subsequent scanning, the height is adjusted according to the current imaging position (X...). i ,Y i The estimated Z-axis position is calculated and derived by substituting it into the model, and is used as the initial adjustment target of the Z-axis voice coil focusing mechanism. Then, the deviation ΔZ between the estimated Z-axis position and the actual Z-axis position is used for fine correction.

8. The 20x magnification two-dimensional microscopic scanning control method according to claim 6, characterized in that, During each scanning movement, when the high-precision grating encoder detects that the remaining distance d ≤ deceleration threshold D, the heterogeneous motion controller performs S-shaped smooth deceleration by adjusting the PWM duty cycle.

9. The 20x magnification two-dimensional microscopic scanning control method according to claim 6, characterized in that, Before starting the scan, the system performs a low-magnification pre-scan. The edge AI computing unit analyzes the pre-scanned image, identifies the effective sample area and the blank area, and automatically generates the optimal scanning path.

Images