Nano-scale-oriented metal tip intelligent detector and use and analysis method thereof

Abstract

The invention provides a nanoscale metal tip intelligent detector and a use and analysis method thereof. The device integrates a high-precision imaging and positioning module, an intelligent analysis module, an intelligent control interaction module and an automatic execution mechanism. The imaging and positioning module is composed of an optical microscope, a high-resolution CCD camera and a three-axis precision objective table and supports nanoscale morphology feature capture and micron-scale positioning precision. The intelligent analysis module can automatically extract the curvature radius, the taper angle, the surface roughness and other key parameters. The control interaction module is coupled with the automatic execution mechanism, so that automatic control of the detection process and remote storage and optimization of process data can be realized. The device can automatically complete the detection and intelligent analysis of the morphology of the nanoscale metal tip, is compatible with application scenes such as single nanopore / pore channel preparation, scanning probe tip morphology characterization and nano device detection, and remarkably improves the operation efficiency while maintaining high-precision detection.

Description

Technical Field

[0001] The present invention belongs to the technical field of nanomaterial detection, and specifically relates to a nanoscale-oriented metal tip intelligent detector and a use and analysis method thereof, and is particularly suitable for the automated detection and analysis of metal tip morphology during the preparation of a single nanopore / channel. Background Art

[0002] The preparation and characterization of nanoscale metal tips is the core technology for constructing single nanopores / channels. Their morphological characteristics and physical properties directly determine the accuracy of ion detection, DNA sequencing, molecular sensing and nanofluidic devices. At present, the template method is a common method for preparing single nanopores / channels. The nanopore structure is constructed through the precise coupling of metal needle tips and glass tubes. However, it still faces three major technical bottlenecks during the process of process amplification:

[0003] 1. Lack of evaluation system: Existing detection methods mainly rely on manual interpretation of cutting-edge scanning electron microscope images, lack of quantifiable three-dimensional morphology evaluation standards, insufficient standardization, and characterization accuracy is easily affected by subjective factors;

[0004] 2. Insufficient intelligent detection: The whole process from image acquisition to feature extraction has a low degree of automation. Traditional equipment lacks an intelligent analysis system based on machine vision. The operation process requires professional training and has poor repeatability, making it difficult to achieve high-throughput detection.

[0005] 3. High cost of accurate detection: Nanoscale metal tip detection is highly dependent on high-precision SEM equipment, resulting in high detection costs, cumbersome sample preparation procedures, and irreversible loss of samples during the detection process.

[0006] Therefore, there is an urgent need to develop a detection device that integrates standardized detection and intelligent analysis functions to improve the efficiency and accuracy of nanometal tip morphology detection during the preparation of single nanopores / channels, and to provide technical support for extended fields such as scanning probe tip morphology detection and nanodevice manufacturing. Summary of the invention

[0007] The present invention provides a nanoscale-oriented intelligent metal tip detector and a use and analysis method thereof, which can automatically complete the detection and intelligent analysis of nanoscale metal tip morphology, and is compatible with application scenarios such as single nanopore / channel preparation, scanning probe tip morphology characterization, and nanodevice detection. It can significantly improve operating efficiency while maintaining high-precision detection and effectively reduce the need for manual intervention.

[0008] The technical solution of the present invention is: a nanoscale metal tip intelligent detector, comprising a substrate (12), one side of the substrate (12) is provided with a microscope support (2) and a microscope (1) arranged on the microscope support (2); a light source (3) for illuminating a sample is arranged in the middle of the substrate (12);

[0009] A stage moving device is provided on the other side of the base (12), and the stage moving device has an XYZ three-axis moving mechanism; the X-axis and the Y-axis control the translational movement of the nanometal tip in the microscope, and the Z-axis controls the distance between the sample and the microscope, i.e., the image focus; a placement table is provided on the stage moving device; the placement table comprises a placement platform (4), a connecting bracket (10), and a sample clamping mechanism to be tested; the connecting bracket (10) and the stage moving device are fixedly connected, the placement platform (4) is provided on the connecting bracket (10), and is fixedly connected to the connecting bracket (10); a metal wire clamping shell (5) passes through the placement platform (4) and the sample clamping mechanism to be tested from top to bottom, respectively, and the lower end of the metal wire clamping shell (5) is located in the direction of the lens of the microscope (1);

[0010] A shell (7) is installed above the aforementioned loading platform moving device, and a control module is placed in the shell (7).

[0011] Alternatively, in response to the pain points of detection efficiency bottlenecks, measurement accuracy limitations and over-reliance on manual intervention in the prior art, the present invention provides a nano-metal tip intelligent detection and analysis system that integrates four core components: a high-precision imaging and positioning module, an intelligent analysis module, an intelligent control interaction module and an automated actuator.

[0012] The integrated high-precision imaging and positioning module includes a microscopic imaging unit, an automatic stage and a light source system, wherein:

[0013] The microscopic imaging unit is composed of a microscope and a high-resolution CCD camera, which supports the capture of nano-scale morphological features; the automatic stage is equipped with a three-axis precision motion platform, which is driven by a stepper motor and a ball screw transmission system to achieve micron-level positioning accuracy, wherein the X / Y dual-axis linkage controls the displacement trajectory of the nano-metal tip in the microscope image plane, and the Z axis controls the distance between the objective lens and the sample, i.e., dynamic focusing control; the light source system uses a back-illuminated LED light source to eliminate glare interference from the metal surface.

[0014] The intelligent analysis module is equipped with the Yolov10 real-time detection framework and supports the optimization of morphological processing and image recognition fusion technology to achieve nano-tip target recognition and micron-level positioning; the morphology analysis is based on the improved U-Net architecture to extract the tip curvature radius, cone angle, roughness, stability and other contour shape parameters to improve the detection accuracy.

[0015] The intelligent control interaction module includes a main control unit, a human-computer interaction interface and a cloud data management system, wherein:

[0016] The main control unit adopts the STM32F103RCT6 microcontroller, integrates motor drive and wireless communication modules, and supports multi-device collaborative control; the human-computer interaction interface integrates parameter visualization configuration matrix and real-time panoramic image stream; the cloud data management system realizes synchronous storage of detection data and remote process iterative optimization.

[0017] The automated actuator includes a vertical sample slot and a limit packaging structure, an integrated servo-driven mechanical clamping mechanism, and a rod-shaped metal sample clamping shell. It supports rapid loading and unloading of batch samples and is compatible with a material spectrum including platinum, tungsten, and platinum-iridium alloys. The detection scale covers nanometer to millimeter levels.

[0018] In some of the embodiments, the control and interaction module has a built-in ESP8266 Wi-Fi module, which can realize remote control through the Internet of Things MQTT protocol and upload the detection data to the cloud in real time.

[0019] In some of the embodiments, the light source system can adjust the brightness of the light source through a microcontroller to adapt to a variety of metals and clamping shells.

[0020] In some of the embodiments, the intelligent analysis module integrates an exception handling mechanism, including tip detection failure retry, focus failure retry and contour abnormality prompt.

[0021] In some of the embodiments, the control module includes a microscope, which transmits the captured image of the metal tip to a computer, displays the tip morphology in real time on a human-computer interaction interface, uses an image algorithm combined with deep learning to perform tip target detection and morphology analysis, generates detection results including curvature radius, cone angle, roughness and stability parameters, and pushes them to a cloud database.

[0022] Compared with the existing technical solutions, the present invention achieves breakthrough innovation through the deep integration of multi-dimensional technical systems: relying on the full-process automated detection platform and intelligent morphology analysis algorithm, a high-precision, human-free detection closed loop is constructed; a multi-precision mapping model is adopted to break through the performance limitations of traditional microscope detection and realize standardized quantitative evaluation of nanoscale morphology; through a reconfigurable modular architecture and multi-scenario adaptation strategy, it can be compatible with the entire process of single nanopore / channel preparation, scanning probe tip morphology detection and nanodevice manufacturing, forming a cross-domain, highly robust intelligent detection solution. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0024] Figure 1: Schematic diagram of the overall structure of the detection and analysis instrument, showing the layout of the microscopic imaging unit, automatic stage and control module;

[0025] Figure 2 : Schematic diagram of the clamping mechanism from top view.

[0026] Figure 3 : The third-order function fitting result diagram of the tip edge;

[0027] Figure 4 : Functional block diagram of the control system, describing the logical relationship between data acquisition, algorithm processing and equipment control;

[0028] Figure 5 : Schematic diagram of microscope scanning path;

[0029] Figure 6 :Flowchart of the intelligent analysis algorithm, illustrating the serial architecture of U-Net morphological parameter calculation.

[0030] Reference numerals

[0031] 1. Microscope; 2. Microscope stand; 3. LED light source; 4. Storage platform; 5. Metal wire clamping shell; 6. Y-axis screw slide; 7. Housing; 8. X-axis screw slide; 9. Z-axis screw slide; 10. Connecting bracket; 11. Servo; 12. Base; 13. Clamping platform; 13-1 ridge 14. Servo connecting rod; 14-1 center connecting rod 14-2 side rod; 15. Clamping connecting rod; 16. Clamping claw. DETAILED DESCRIPTION

[0032] The technical solution of the present invention is described in detail below in conjunction with the accompanying drawings, and the implementation details and technical advantages of the present invention are further explained through specific embodiments, which do not constitute a limitation on the scope of protection.

[0033] Example 1: Hardware structure configuration

[0034] like Figure 1 and Figure 2 As shown, the present invention is a nano-scale metal tip intelligent detector, comprising a substrate 12, one side of the substrate 12 (at Figure 1 The front half is provided with a microscope support 2 and a microscope 1 arranged on the microscope support 2. The lens of the microscope 1 is arranged from front to rear.

[0035] The middle of the base 12 (at Figure 1 An LED light source 3 is provided at the center, and the LED light source 3 includes an LED bracket and LED lamp beads, which are perpendicular to the lens direction of the microscope 1.

[0036] The other side of the substrate 12 (at Figure 1The back half is in the middle) and is provided with a stage moving device, which is composed of three axes XYZ, and the Z axis is composed of two parallel front and rear directions ( Figure 1 ) is composed of a slide 9 placed on the X-axis slide 9; the X-axis slide 8 is placed horizontally on the Z-axis slide 9 in the left-right direction and can move forward and backward along the slide 9; the Y-axis slide 6 is placed vertically and connected to the slider of the X-axis slide 8, and the Y-axis slide 6 can move left and right along the X-axis slide 8. The front side of the slider of the Y-axis slide 6 is connected to a storage table, and the storage table can move up and down along the Y-axis slide 6. The three axes XYZ are driven by motors or manually adjusted.

[0037] The storage platform is composed of a storage platform 4, a connecting bracket 10, a steering gear 11, a clamping platform 13, a steering gear connecting rod 14, a clamping connecting rod 15 and a clamping claw 16 (such as Figure 2 As shown in the figure, the connecting bracket 10 is fixedly arranged on the front of the Y-axis slide 6, and the storage platform 4 is fixedly arranged on the upper half of the front of the connecting bracket 10, and is arranged in the horizontal direction. The clamping platform 13 is arranged parallel to the lower side of the storage platform 4 and is fixedly connected to the connecting bracket 10. The steering gear 11 is inserted into the clamping platform 13, and its output shaft is located above the clamping platform 13 and is connected to the steering gear connecting rod 14.

[0038] The steering gear link 14 is composed of a central link 14-1 located in the middle and two side rods 14-2 pivoted at both ends of the central link 14-1. The output shaft of the steering gear 11 is fixedly connected to the middle part of the central link 14-1, so that the central link 14-1 rotates clockwise or counterclockwise. The other end of each side rod 14-2 is pivoted to one end of the clamping link 15. The other end of the clamping link 15 is fixedly connected to the clamping claw 16. The clamping link 15 and the clamping claw 16 each have a pair.

[0039] The edge of the clamping platform 13 ( Figure 1 On the left side, Figure 2 The lower side of the clamping rod 15 is provided with a ridge 13-1 upward, and the back of the clamping rod 15 is provided with a groove corresponding to the ridge 13-1. The ridge 13-1 and the groove cooperate so that the clamping rod 15 can only slide along the ridge 13-1 ( Figure 1 Slide forward and backward. Figure 2 for left and right sliding).

[0040] When in use, the steering gear 11 rotates to drive the central connecting rod 14-1 to rotate with its center point as the center of the circle, thereby driving the two side rods 14-2 and the clamping connecting rod 15 to move. Figure 2 In the embodiment, the two clamping connecting rods can only be relatively close to or relatively far away from each other along the left and right directions of the convex rib 13-1. Then the pair of clamping claws 16 also relatively close to complete clamping or relatively far away to complete separation.

[0041] The wire clamping shell 5 (in which the wire to be tested is clamped) passes through the placement platform 4 from top to bottom and is inserted between the clamping parts of the two clamping jaws 16. The clamping jaws 16 are closed to clamp and fix, restricting their movement. The lower end of the wire clamping shell 5 is located in the direction of the lens of the microscope 1.

[0042] A housing 7 is installed above the aforementioned stage moving device to protect the lead screw slide and to place a control module (microcontroller).

[0043] Furthermore, in the above structure:

[0044] 1. Automatic stage module

[0045] (1) Three-axis displacement mechanism:

[0046] Adopt high-strength aluminum alloy base, X / Y axis travel 0-100mm, Z axis travel 0-50mm, equipped with high-precision ball screw (lead 1mm);

[0047] The drive unit uses a 42 stepper motor (step angle 1.8°) with a 16-subdivision driver, and the repeat positioning error is ≤0.2mm;

[0048] Motion control logic: The motor speed is adjusted by sending pulse signals (frequency 1-10kHz) through the STM32 microcontroller, which can achieve variable speed scanning of 0.1-5mm / s.

[0049] (2) Limiting packaging structure:

[0050] The mechanical clamping device is driven by a single-axis servo, with a maximum clamping force of 2N·m, suitable for wire clamping shells with a diameter of 5-20mm.

[0051] 2. Microscopic imaging module

[0052] The optical microscope is equipped with a 1x-4x objective lens, a CCD camera with a resolution of 1920×1080 pixels and a frame rate of 30fps;

[0053] Light source system: An LED light source (brightness adjustable) is integrated behind the sample, facing the microscope lens, to ensure high contrast in the image and eliminate interference from metal reflections.

[0054] 3. Control and communication module

[0055] Main control chip: STM32F103RCT6 microcontroller, equipped with a four-layer PCB circuit board, integrated motor driver chip, ESP8266 Wi-Fi module;

[0056] Power management: input voltage 12V / 5A, output 5V / 3A through step-down chip, peak power consumption ≤15W;

[0057] Communication protocol: Supports MQTT protocol, and detection data is uploaded to the cloud (Alibaba Cloud IoT platform) in real time, with a transmission delay of <200ms.

[0058] Example 2: Manually acquiring tip morphology features for model training

[0059] Step 1: Image acquisition. Manually adjust the wire position, use a microscope / SEM to acquire an image of the wire tip, and save it as the target image after cropping and binarization preprocessing; the wire position can be adjusted along the XYZ axis and combined with the height of the wire clamping position to make it in the appropriate field of view of the microscope.

[0060] Step 2: Edge extraction. For images acquired by a microscope, the first 5 pixels of the two edges of the metal tip and the first 10 μm (currently 50 pixels) of the edge are extracted according to the magnification, and the first 5 pixels of the tip are used for morphological feature calculation; for images acquired by a SEM, the first 5 pixels of the two edges of the metal tip and the first 1 μm (currently 150 pixels) of the edge are extracted according to the magnification, and used for morphological feature calculation;

[0061] Step 3: Determine the turning point. The pixel points on the upper and lower edges of the tip are fitted with a third-order function to find the curvature turning point. The results are as follows: Figure 3 As shown, the calculation formula is as follows:

[0062] f(x)=ax 3 +bx 2 +cx+d

[0063] f″(x)=6ax+2b

[0064] f″(x)=0

[0065] The point where the second-order derivative is 0 is the point where the curvature changes. Fit the first-order function to all the pixel points between the tip and the turning point, and the calculation formula is as follows:

[0066] f(x)=ax+b

[0067] Where f(x) is the fitting function, and a, b, c, and d are the fitting parameter values ​​of the function. If there is no curvature turning point, it means that the edge is a straight line, and the midpoint of the edge is taken as the curvature turning point.

[0068] Step 4: Calculation of morphological features. Calculate the angle between the first-order function fitting lines of the two edges as the metal tip angle, and calculate the radius of the fitting circle of the first five pixel points of the two edges as the metal tip curvature radius. The first-order function fitting degree R of the above pixel points is 2 As the roughness, the coefficient of variation of the above characteristics is used to characterize the etching stability, and the calculation formula is as follows:

[0069]

[0070] Stability=0.5×anglecv+0.3×radiuscv+0.2×(1-R 2 )

[0071] Among them, std is the standard deviation of multiple experiments, n is the number of samples, yi is the actual observation value, The mean of the true observations, is the predicted value, cv is the coefficient of variation, anglecv is the coefficient of variation of the tip angle, radiuscv is the coefficient of variation of the radius of curvature, R 2 is the function fitting coefficient, and Stability is the comprehensive stability index.

[0072] Step 5: Deep learning model training: Use the acquired images and morphological features to train the deep learning model.

[0073] Example 3: Detection control process

[0074] Step 1: System initialization. The stage performs origin reset, and the X / Y / Z axes touch the limit switches in turn and return to the origin; the microscope switches to the 3× low-magnification objective lens, and the mechanical gripper automatically opens to the maximum opening (20mm) and waits for the sample to be loaded;

[0075] Step 2: Sample loading: Insert the clamping shell holding the wire tip (25 μm in diameter) vertically into the circular groove of the storage plate, and manually trigger the closing of the half-moon groove jaws through the interactive interface or automatically trigger the closing of the half-moon groove jaws by automated equipment after the clamping shell is inserted.

[0076] Step 3: Scan parameter setting. Set the scanning parameters to global positioning + local fine feature extraction through the human-machine interface, X-axis scanning step length 500μm, Y-axis scanning step length 800μm, Z-axis autofocus range ±2mm, and LED light source brightness to 15% brightness.

[0077] Step 4: Local pre-scan. Scan the X / Y axis in a serpentine path ( Figure 5 ), collect images and perform real-time binarization; calculate the area of ​​the black area in each frame, record the corresponding motion ID and area value; after the local pre-scan, the stage moves to the ID corresponding to the maximum area, and starts the autofocus algorithm. The Z axis moves back and forth with a step length of 100μm, and calculates the clarity evaluation function of each frame (the mixed clarity of the Sobel operator, Laplacian operator, and Canny operator). When the clarity reaches the peak, it stops and returns to the last position of the pre-scan.

[0078] Step 5: Metal tip target detection. A three-step positioning strategy is used ( Figure 4 ):

[0079] (1) Cluster denoising: Perform K-means clustering (K=3) on high-magnification images and retain the largest connected domain. Calculate the sum of the squares of the distances from all sample points in a cluster to the centroid using the Euclidean distance, find the solution with the smallest overall sum of squares, and recalculate the center of each cluster. Continuously iterate and optimize the clustering results so that the objects within each cluster are as close as possible, while the objects between different clusters are as separated as possible. The formula is as follows:

[0080]

[0081] Where x is a sample point in the cluster, ci is the index of the cluster to which the data point xi belongs, μj is the centroid of the jth cluster, K is the number of clusters, Sj is the set of data points in the jth cluster, and |Sj| is the number of data points in the set;

[0082] (2) Target recognition: Morphological calculation and Yolov10 model (input size 512×512, confidence threshold 0.5) are used to locate the tip area and output the bounding box coordinates. The confidence calculation formula is as follows:

[0083]

[0084] Confidence=P(Object)*IoU(A,B)

[0085] Among them, A is the target box, B is the predicted box, P is the probability of the object existing in the predicted box, IoU is the intersection over union ratio, and Confidence is the confidence level, which is used to calculate the ratio of the intersection and union of the predicted box and the true box area;

[0086] (3) Accurate positioning: According to the center coordinates of the bounding box, calculate the ΔX and ΔY that the stage needs to move, move the tip to the center of the field of view, perform secondary focusing and collect images. The calculation formula is as follows:

[0087] △X=Xbox_centroid-Xpic_centroid+Xmove

[0088] △Y=Ybox_centroid-Ypic_centroid+Ymove

[0089] Among them, Xbox_centroid and Ybox_centroid are the coordinates of the center point of the predicted bounding box, Xpic_centroid and Ypic_centroid are the coordinates of the center point of the current acquired image, and Xmove and Ymove are the acquisition positions corresponding to the image, which are calculated by the corresponding motion ID;

[0090] Step 6: Output of morphological features. After image preprocessing, input into the improved U-Net model ( Figure 6 ), outputs tip morphology features, including curvature radius, cone angle, roughness and stability), and uploads the data to the cloud.

[0091] Example 4: Batch detection and exception handling

[0092] Batch loading: 12 samples are sequentially grabbed into the vertical sample slot by the robotic arm, and the detection cycle of each sample is ≤30 seconds;

[0093] Tip detection failure retry: If neither the morphology nor the Yolo algorithm detects the metal tip, re-detection is performed, with a maximum of 3 retries;

[0094] Retry after focus failure: When the focus clarity peak value is less than the threshold, the Z-axis scanning range is extended to ±3mm and retries are performed up to 3 times;

[0095] Contour abnormality prompt: If the detected tip is classified as unqualified, the sample will be prompted and recorded as an unqualified sample and will be eliminated in subsequent work.

[0096] Implementation effect:

[0097] 20 sets of metal wire (platinum wire) metal tips were tested and the SEM measurement results were compared:

[0098]

[0099] The results show that in this embodiment, the present invention maintains a high degree of fit under low precision while improving efficiency by more than 10 times compared with traditional SEM, and supports continuous placement and clamping of metal tips by a robotic arm to achieve batch automated operations. It is suitable for scenarios such as single nanopore / channel preparation and scanning probe morphology detection, significantly reducing production costs and improving process consistency.

[0100] The above description is only a preferred embodiment of the present invention, and therefore cannot be used to limit the scope of the present invention. That is, equivalent changes and modifications made according to the patent scope of the present invention and the contents of the specification should still fall within the scope of the present invention.

Claims

1. A nanoscale metal tip intelligent detector, characterized by: It comprises a base (12), one side of which is provided with a microscope stand (2) and a microscope (1) arranged on the microscope stand (2); a light source (3) for illuminating a sample is arranged in the middle of the base (12); A stage moving device is provided on the other side of the base (12), and the stage moving device has an XYZ three-axis moving mechanism; the X-axis and the Y-axis control the translational movement of the nanometal tip in the microscope, and the Z-axis controls the distance between the sample and the microscope, i.e., the image focus; a placement table is provided on the stage moving device; the placement table comprises a placement platform (4), a connecting bracket (10), and a sample clamping mechanism to be tested; the connecting bracket (10) and the stage moving device are fixedly connected, the placement platform (4) is provided on the connecting bracket (10), and is fixedly connected to the connecting bracket (10); a metal wire clamping shell (5) passes through the placement platform (4) and the sample clamping mechanism to be tested from top to bottom, respectively, and the lower end of the metal wire clamping shell (5) is located in the direction of the lens of the microscope (1); A shell (7) is installed above the aforementioned loading platform moving device, and a control module is placed in the shell (7).

2. The nanoscale metal tip intelligent detector according to claim 1, characterized in that: The Z axis of the loading platform moving device comprises two slides (9) placed in parallel in the front-to-back direction; the X axis slide (8) is placed on the Z axis slide (9) in the left-right direction and can move forward and backward along the slide (9); the Y axis slide (6) is placed vertically and connected to the slider of the X axis slide (8), and the Y axis slide (6) can move in the left-right direction along the X axis slide (8); the front side of the slider of the Y axis slide (6) is connected to the loading platform, and the loading platform can move up and down along the Y axis slide (6).

3. The nanoscale metal tip intelligent detector according to claim 1, characterized in that: The sample clamping mechanism comprises a steering gear (11), a clamping platform (13), a steering gear connecting rod (14), a clamping connecting rod (15) and a clamping claw (16); wherein the clamping platform (13) and the connecting bracket (10) are fixedly connected, the steering gear (11) is inserted through the clamping platform (13), and its output shaft is located above the clamping platform (13) and connected to the steering gear connecting rod (14); The steering gear connecting rod (14) comprises a central connecting rod (14-1) located in the middle and two side rods (14-2) respectively pivoted at two ends of the central connecting rod (14-1); the output shaft of the steering gear (11) is fixedly connected to the middle part of the central connecting rod (14-1), so that the central connecting rod (14-1) rotates clockwise or counterclockwise; the other end of each side rod (14-2) is pivotally connected to one end of a clamping connecting rod (15); the other end of the clamping connecting rod (15) is fixedly connected to a clamping claw (16); A limiting structure is provided between the clamping platform (13) and the clamping connecting rod (15) so that the two clamping connecting rods can only slide relative to each other in one direction to move closer or separate.

4. The nanoscale metal tip intelligent detector according to claim 3, characterized in that: The edge of the clamping platform (13) is provided with a convex ridge (13-1) upwardly, and the back of the clamping connecting rod (15) is provided with a groove corresponding to the convex ridge (13-1). The convex ridge (13-1) and the groove cooperate so that the clamping connecting rod (15) can only slide along the convex ridge (13-1).

5. The nanoscale metal tip intelligent detector according to claim 1, characterized in that: It also includes an intelligent analysis module, which is based on a deep learning algorithm and uses the Yolov10 model to achieve tip target detection, combines an improved U-Net architecture to extract tip curvature radius, angle and contour shape parameters, and supports morphological processing and image fusion technology to optimize feature extraction.

6. The nanoscale metal tip intelligent detector according to claim 1, characterized in that , the microscope (1) is composed of a lens and a high-resolution CCD camera supporting nanoscale image acquisition.

7. The nanoscale metal tip intelligent detector according to claim 1, characterized in that ,The XYZ three-axis moving mechanism is driven by a stepper motor.

8. A method for using and analyzing a nanoscale metal tip intelligent detector according to any one of claims 1 to 7, comprising the following steps: Step 1: System initialization, the stage performs origin reset, the X / Y / Z axes touch the limit switches in sequence and then return to the origin; the microscope switches to the 3× low-magnification objective lens, and the mechanical gripper automatically opens to the maximum opening, waiting for the sample to be loaded; Step 2: Sample loading, insert the clamping shell holding the tip of the metal wire vertically into the circular groove of the storage plate, and manually trigger the closing of the half-moon groove jaws through the interactive interface or automatically trigger the closing of the half-moon groove jaws by automated equipment after the clamping shell is inserted; Step 3: Scan parameter setting: set the scanning parameters to global positioning + local fine feature extraction through the human-machine interface, the X-axis scanning step length is 500μm, the Y-axis scanning step length is 800μm, the Z-axis autofocus range is ±2mm, and the LED light source brightness is set to 15% brightness; Step 4: Local pre-scanning, X / Y axis scans in a serpentine path, collects images and performs real-time binarization processing; calculates the area of ​​the black area in each frame of the image, and records the corresponding motion ID and area value; after the local pre-scanning is completed, the stage moves to the ID corresponding to the largest area, and the autofocus algorithm is started; the Z axis moves back and forth with a step length of 100μm, calculates the clarity evaluation function of each frame of the image, stops when the clarity reaches the peak, and returns to the last position of the pre-scan; Step 5: Metal tip target detection; using a three-step positioning strategy: (1) Cluster denoising: Perform K-means clustering (K=3) on high-magnification images and retain the largest connected domain. Calculate the sum of the squares of the distances from all sample points in a cluster to the centroid using the Euclidean distance, find the solution with the smallest overall sum of squares, and recalculate the center of each cluster. Continuously iterate and optimize the clustering results so that the objects within each cluster are as close as possible, while the objects between different clusters are as separated as possible. The formula is as follows: in, x is the sample point in the cluster, c i For data point x i The index of the cluster to which it belongs, μ j is the centroid of the jth cluster, K is the number of clusters, S j is the set of data points of the jth cluster, |S j | is the number of data points in the set; (2) Target recognition: Morphological calculation and Yolov10 model are used to locate the tip area and output the bounding box coordinates. The confidence calculation formula is as follows: Confidence=P(Object)*IoU(A,B) Among them, A is the target box, B is the predicted box, P is the probability of the object existing in the predicted box, IoU is the intersection over union ratio, and Confidence is the confidence level, which is used to calculate the ratio of the intersection and union of the predicted box and the true box area; (3) Accurate positioning: According to the center coordinates of the bounding box, calculate the ΔX and ΔY that the stage needs to move, move the tip to the center of the field of view, perform secondary focusing and collect images. The calculation formula is as follows: △X=X box_centroid -X pic_centroid +X move △And=And box_centroid -AND pic_centroid +Y move Among them, X box_centroid and Y box_centroid is the center point coordinate of the predicted bounding box, X pic_centroid and Y pic_centroid is the center point coordinate of the current image, X move and Y move is the acquisition position corresponding to the image, which is calculated by the corresponding motion ID; Step 6: Output of morphological features.

9. The method for using and analyzing a nanoscale metal tip intelligent detector according to claim 8, characterized in that: It also includes the following batch detection and exception handling steps: Batch loading: Multiple samples are sequentially grabbed into the vertical sample slot by the robotic arm, and the detection cycle of each sample is ≤30 seconds; Tip detection failure retry: If neither the morphology nor the Yolo algorithm detects the metal tip, re-detection is performed, with a maximum of 3 retries; Retry after focus failure: When the focus clarity peak is less than the threshold, the Z-axis scanning range is extended to ±3mm and retries are performed up to 3 times; Contour abnormality prompt: If the detected tip is classified as unqualified, the sample will be prompted and recorded as an unqualified sample and will be eliminated in subsequent work.

10. A method for nanometal tip data acquisition and model training, comprising the following steps: Step 1, image acquisition: manually adjust the position of the platinum wire, use a microscope / SEM to acquire an image of the tip of the platinum wire, and save it as the target image after cropping and binarization preprocessing; wherein, The position of the platinum wire can be adjusted by adjusting the XYZ axis and combining the height of the clamping position of the platinum wire to make it in the appropriate field of view of the microscope; Step 2, edge extraction: for images acquired by microscope, extract the edge of the metal tip at 10 μm, and use it as the morphological feature calculation according to the current magnification of 50 pixels, and extract the first 5 pixels of the tip for the tip curvature radius calculation; for images acquired by SEM, extract the edge of the metal tip at 1 μm, and use it as the morphological feature calculation according to the current magnification of 150 pixels, and extract the first 5 pixels for the tip curvature radius calculation; Step 3: Determine the turning point: Fit the pixel points on the upper and lower edges of the tip with a third-order function to find the curvature turning point. The calculation formula is as follows: f(x)=ax 3 +bx 2 +cx+d f″(x)=6ax+2b f″(x)=0 Solve the point where the second-order derivative is 0, which is the point where the curvature changes; fit the first-order function to all pixel points between the tip and the turning point, and the calculation formula is as follows: f(x)=ax+b Where f(x) is the fitting function, and a, b, c, and d are the fitting parameter values ​​of the function. If there is no curvature turning point, it means that the edge is a straight line, and the midpoint of the edge is taken as the curvature turning point. Step 4, morphological feature calculation: calculate the angle between the two edge fitting lines as the metal tip angle, calculate the radius of the first five pixel fitting circles as the metal tip curvature radius, and the first-order function fitting degree R of the above pixel points 2 As the roughness, the coefficient of variation of the above characteristics is used to characterize the etching stability, and the calculation formula is as follows: Stability=0.5×angle cv +0.3×radius cv +0.2×(1-R 2 ) Among them, std is the standard deviation of multiple experiments, n is the number of samples, yi is the actual observation value, The mean of the true observations, is the predicted value, cv is the coefficient of variation, angle cv is the coefficient of variation of the tip angle, radius cv is the coefficient of variation of the radius of curvature, R 2 is the function fitting coefficient, and Stability is the comprehensive stability index. Step 5: Deep learning model training: Use the acquired images and morphological features to train the deep learning model.

Images