Microscopic field self-adaptive positioning and high-power imaging guiding method

Abstract

The application discloses a microscopic field self-adaptive positioning and high-magnification imaging guiding method and belongs to the technical field of image processing. Firstly, a training sample image is used to train a Yolov11-FocalModulation network; then, a sample to be detected is loaded to a microscope loader, a low-magnification objective lens is used for scanning and splicing to obtain a panoramic image; then, the panoramic image is input into the trained network to obtain a prediction frame; then, smooth prediction frames are screened out according to texture complexity and gradient change amount; finally, the center coordinates of the smooth prediction frames are converted to obtain target physical coordinates, a high-magnification objective lens is switched to scan the region in the frame after the object table is moved to obtain a high-magnification microscopic image. The application intelligently positions the region of interest, improves the high-magnification imaging efficiency and reduces redundant data.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, and in particular to a microscopic field self-adaptive positioning and high-magnification imaging guiding method. BACKGROUND

[0002] Microscopic imaging technology is widely used in biomedical research (microbial staining, pathological section analysis), material science (such as semiconductor wafer defect detection) and industrial quality inspection fields. With the increasing demand for high-resolution imaging, the use of high-magnification objectives (such as 40x, 60x and above) has become a core means to obtain microscopic details. However, the physical characteristics of high-magnification objectives (small field of view range, shallow depth of field) result in a dual problem of efficiency and data volume in full-sample scanning. Traditional high-magnification objective full-sample scanning needs to be positioned and imaged by mechanically moving the objective stage region by region, which is time-consuming and generates a large number of redundant images (such as blank areas or non-target tissues).

[0003] To solve this problem, the prior art generally adopts two types of schemes:

[0004] 1. Artificially marking local scanning:

[0005] The sample is scanned with a low-magnification objective to generate a sample panoramic image. The operator manually marks the scanning area on the panoramic image, and then uses a high-magnification objective to scan these manually marked areas. This method requires manual intervention for each sample, and the efficiency is difficult to meet the needs of high-throughput business,

[0006] 2. Pre-planning path continuous scanning:

[0007] By pre-planning the scanning path, a high-magnification objective is used to continuously scan high-magnification images in a uniformly specified area. This method cannot accurately scan the truly valuable sample area for the high-magnification objective under the condition of irregular content distribution and large individual differences. SUMMARY

[0008] In view of the above deficiencies in the prior art, the present application provides a microscopic field self-adaptive positioning and high-magnification imaging guiding method to solve the problems of low scanning efficiency and excessive redundant image data caused by the small field of view range in the process of scanning non-homogeneous samples by a high-magnification objective.

[0009] In order to achieve the above-mentioned application purposes, the technical scheme adopted by the present application is as follows: a microscopic field self-adaptive positioning and high-magnification imaging guiding method, comprising the following steps:

[0010] S1, training a Yolov11-FocalModulation network using a training sample image to obtain a trained Yolov11-FocalModulation network;

[0011] S2, loading the sample to be detected into a microscope loader, scanning the sample to be detected with a low-power objective to obtain a plurality of local images, splicing the plurality of local images to obtain a panorama of the sample to be detected;

[0012] S3, inputting the panorama of the sample to be detected into the trained Yolov11-FocalModulation network to obtain a prediction box;

[0013] S4, screening the prediction box according to the texture complexity and the gradient change amount, and retaining a smooth prediction box;

[0014] S5, performing coordinate conversion on the center coordinates of the smooth prediction box to obtain target physical coordinates, moving the stage to the target physical coordinates, switching the high-power objective to scan the area in the smooth prediction box to obtain a high-power microscopic image.

[0015] Further, S1 includes the following steps:

[0016] S11, scanning the slide sample with a low-power microscope to obtain a plurality of local images, and splicing the plurality of local images to obtain a panorama of the slide sample;

[0017] S12, manually labeling the abnormal area on the panorama to obtain a training sample image;

[0018] S13, training the Yolov11-FocalModulation network using the training sample image to obtain the trained Yolov11-FocalModulation network.

[0019] Further, the structure of the Yolov11-FocalModulation network in S1 is that the SPPF unit in the original Yolov11 network architecture is replaced by the FocalModulation unit based on the original Yolov11 network architecture.

[0020] Further, S4 includes the following steps:

[0021] S41, converting the prediction box from the RGB color space to the LAB color space, and extracting the luminance channel in the LAB color space;

[0022] S42, performing detail enhancement on the luminance channel of the prediction box to obtain a detail enhancement box;

[0023] S43, adjusting the luminance of the detail enhancement box to obtain a luminance adjustment box;

[0024] S44, extracting the texture complexity from the luminance adjustment box;

[0025] S45, obtain the X-direction gradient and the Y-direction gradient on the brightness adjustment frame, and calculate the gradient variation;

[0026] S46, calculate the texture smoothness according to the gradient variation and the texture complexity;

[0027] S47, reserve the prediction frame with the texture smoothness greater than the smoothness threshold.

[0028] Further, S42 comprises the following steps:

[0029] S421, perform the open operation on the brightness channel of the prediction frame by using the 51x51 elliptical kernel to obtain the open operation result;

[0030] S422, subtract the open operation result from the brightness channel of the prediction frame to obtain the detail enhancement frame.

[0031] Further, the formula of the brightness adjustment in S43 is:

[0032] ;

[0033] Wherein, L AD is the brightness adjustment frame, L is the brightness channel of the prediction frame, and E is the detail enhancement frame.

[0034] Further, the formula of the texture complexity in S44 is:

[0035] ;

[0036] Wherein, C is the texture complexity, l i is the brightness value of the i-th pixel point in the brightness adjustment frame, N is the number of pixel points in the brightness adjustment frame, and i is the number of the pixel point.

[0037] Further, S45 comprises the following steps:

[0038] S451, slide the 5x5 X-direction Sobel kernel on the brightness adjustment frame to calculate the X-direction gradient at each sliding position;

[0039] S452, slide the 5x5 Y-direction Sobel kernel on the brightness adjustment frame to calculate the Y-direction gradient at each sliding position;

[0040] S453, calculate the standard deviation of the X-direction gradient at each sliding position to obtain the X-direction gradient standard deviation;

[0041] S454, calculate the standard deviation of the Y-direction gradient at each sliding position to obtain the Y-direction gradient standard deviation;

[0042] S455, add the X-direction gradient standard deviation and the Y-direction gradient standard deviation to obtain the gradient variation.

[0043] Further, the formula for calculating the texture smoothness in S46 is:

[0044] ;

[0045] Wherein, S is the texture smoothness, C is the texture complexity, and G is the gradient change.

[0046] Further, S5 includes the following sub-steps:

[0047] S51, obtaining the center coordinates of the smoothed prediction frame:

[0048] ;

[0049] Wherein, X p is the horizontal coordinate of the center coordinates, Y p is the vertical coordinate of the center coordinates, x1 is the horizontal coordinate of the first corner point of the smoothed prediction frame, x2 is the horizontal coordinate of the second corner point of the smoothed prediction frame, x3 is the horizontal coordinate of the third corner point of the smoothed prediction frame, x4 is the horizontal coordinate of the fourth corner point of the smoothed prediction frame, y1 is the vertical coordinate of the first corner point of the smoothed prediction frame, y2 is the vertical coordinate of the second corner point of the smoothed prediction frame, y3 is the vertical coordinate of the third corner point of the smoothed prediction frame, and y4 is the vertical coordinate of the fourth corner point of the smoothed prediction frame.

[0050] S52, coordinate conversion is performed on the center coordinates of the smoothed prediction frame to obtain target physical coordinates:

[0051] ;

[0052] Wherein, X z is the horizontal coordinate of the target physical coordinates, Y z is the vertical coordinate of the target physical coordinates, X s is the physical starting calibration horizontal coordinate of the upper left corner of the sample slide where the sample to be detected is located on the objective table, Y s is the physical starting calibration vertical coordinate of the upper left corner of the sample slide where the sample to be detected is located on the objective table, and P d is the coefficient for converting pixel value to physical distance.

[0053] S53, moving the objective table to the target physical coordinates, switching the high-power objective lens to scan the area within the smoothed prediction frame to obtain a high-power microscopic image.

[0054] The beneficial effects of the present application are: the present application processes the panoramic image of the to-be-detected sample through the trained Yolov11-FocalModulation network, obtains the suspected area where the target is located, that is, the prediction box, and then screens out the smooth prediction box through the texture complexity and the gradient change amount, effectively filters the worthless area, and avoids redundant scanning. The present application utilizes the synergy of panoramic scanning splicing of a low-power objective and precise imaging of a high-power objective. In the detection of non-homogeneous samples, macroscopic information is obtained by quickly covering the whole sample with a low-power objective, and microscopic details are captured by focusing on key areas with a high-power objective. The present application breaks through the field-of-view limitation of a high-power objective, greatly improves the scanning efficiency, and reduces the generation of redundant image data. At the same time, coordinate conversion and precise movement control of the objective stage ensure the accurate focusing of high-power imaging on the target area, realize the accurate shooting of the target, improve the scanning efficiency, and reduce the redundant image data. BRIEF DESCRIPTION OF DRAWINGS

[0055] Figure 1 It is a flowchart of a microscopic field adaptive positioning and high-power imaging guiding method;

[0056] Figure 2 It is a panoramic image of a slide sample in S11;

[0057] Figure 3 It is a schematic diagram of a training sample image;

[0058] Figure 4 It is a target box loss schematic diagram of a training set;

[0059] Figure 5 It is a target box loss schematic diagram of a verification set;

[0060] Figure 6 It is a precision change graph when the training set is used;

[0061] Figure 7 It is a recall rate change graph when the training set is used;

[0062] Figure 8 It is a mean average precision change graph when the verification set is used;

[0063] Figure 9 It is a structural schematic diagram of a Yolov11-FocalModulation network;

[0064] Figure 10 It is a panoramic image of a to-be-detected sample in S2;

[0065] Figure 11 It is a schematic diagram of the screened prediction box;

[0066] Figure 12 It is a high-power microscopic image obtained according to the screened prediction box. DETAILED DESCRIPTION

[0067] The specific embodiments of the present application are described below to enable those skilled in the art to understand the present application, but it should be clear that the present application is not limited to the scope of the specific embodiments, and for those skilled in the art, it is obvious that various changes are within the spirit and scope of the present application defined and determined by the appended claims, and all the inventions utilizing the concept of the present application are within the scope of protection.

[0068] As shown in the figure, a microscopic field adaptive positioning and high magnification imaging guidance method comprises the following steps: Figure 1

[0069] S1, training the Yolov11-FocalModulation network with training sample images to obtain a trained Yolov11-FocalModulation network;

[0070] S2, loading the sample to be detected into the microscope loader, scanning the sample to be detected with a low magnification objective lens to obtain multiple local images, and splicing the multiple local images to obtain a panoramic image of the sample to be detected;

[0071] S3, inputting the panoramic image of the sample to be detected into the trained Yolov11-FocalModulation network to obtain a prediction box;

[0072] S4, filtering the prediction box according to the texture complexity and gradient change amount, and retaining the smooth prediction box;

[0073] S5, performing coordinate conversion on the center coordinates of the smooth prediction box to obtain target physical coordinates, moving the stage to the target physical coordinates, switching the high magnification objective lens to scan the area within the smooth prediction box, and obtaining a high magnification microscopic image.

[0074] In this embodiment, S1 comprises the following steps:

[0075] S11, scanning the slide sample with a low magnification microscope to obtain multiple local images, splicing the multiple local images to obtain a panoramic image of the slide sample, as shown in the figure; Figure 2

[0076] S12, manually labeling the abnormal area on the panoramic image to obtain a training sample image, as shown in the figure; Figure 3

[0077] S13, training the Yolov11-FocalModulation network with the training sample image to obtain a trained Yolov11-FocalModulation network.

[0078] ​​​In step S1, the professional annotated dataset is divided into a training set and a validation set in a 7:3 ratio, and input into the Yolov11-FocalModulation network for training and validation. After multiple rounds of training, the recall rate and precision rate of the Yolov11-FocalModulation network no longer increase, i.e., the optimal model parameters are obtained.

[0079] As shown in Figure 4 and 5 , the number of samples in the training set and the validation set reaches 200-400, and the error no longer decreases significantly (indicating that the fitting of the training data is good). As shown in Figure 6 , the number of samples in the training set reaches 200-400, and the precision rate of the Yolov11-FocalModulation network no longer increases. As shown in Figure 7 , the number of samples in the training set reaches 200-400, and the recall rate of the Yolov11-FocalModulation network no longer increases. As shown in Figure 8 , the number of samples in the validation set reaches 200-400, and the average precision no longer increases.

[0080] In this embodiment, the structure of the Yolov11-FocalModulation network in S1 is as follows: on the basis of the original Yolov11 network architecture, the SPPF unit in the original Yolov11 network architecture is replaced by a FocalModulation unit, as shown in Figure 9 .

[0081] In Figure 9 , the abbreviation FM is the FocalModulation unit, and in the Yolov11 network structure, SPPF is a kind of spatial pyramid pooling, which can extract multi-scale feature information by performing pooling operations on features of different scales. However, SPPF performs the same processing on the features of all regions without distinguishing important regions from non-important regions. Therefore, for the task of finding suitable field of view regions in microscopic images, the effect of SPPF is not very ideal. Therefore, the FocalModulation unit with attention mechanism is used to replace the SPPF unit in Yolov11.

[0082] Focal Modulation is an attention mechanism that can weight each region according to its importance. Focal Modulation first calculates the importance factor of each region, and then weights the features according to the importance factor. In this way, the model can pay more attention to the more important suitable field of view regions in the image, thereby improving the extraction ability of suitable field of view features.

[0083] In the embodiment, the panoramic image of the sample to be detected in S2 is as shown in Figure 10

[0084] In the application, the prediction box obtained in S3 comprises four corner point coordinates.

[0085] In the embodiment, S4 comprises the following steps:

[0086] S41, converting the prediction box from the RGB color space to the LAB color space, and extracting the luminance channel in the LAB color space;

[0087] S42, performing detail enhancement on the luminance channel of the prediction box to obtain a detail enhancement box;

[0088] S43, performing luminance adjustment on the detail enhancement box to obtain a luminance adjustment box;

[0089] S44, extracting the texture complexity of the luminance adjustment box;

[0090] S45, obtaining the X-direction gradient and the Y-direction gradient of the luminance adjustment box, and calculating the gradient change amount;

[0091] S46, calculating the texture smoothness according to the gradient change amount and the texture complexity;

[0092] S47, retaining the prediction box with the texture smoothness greater than the smoothness threshold.

[0093] The application extracts the luminance channel sensitive to illumination change in the LAB color space, eliminates color interference, performs detail enhancement on the luminance channel of the prediction box, and then performs luminance adjustment to obtain a luminance adjustment box with enhanced crack features. On the one hand, the application reflects the texture condition of the luminance adjustment box through the texture complexity of the luminance adjustment box, and on the other hand, the application measures the texture condition in different directions by calculating the gradient change amount through the X-direction gradient and the Y-direction gradient. The application measures the smoothness of the prediction box through the two aspects, and selects the prediction box with the smoothness greater than the smoothness threshold, as shown in Figure 11

[0094] In the embodiment, S42 comprises the following steps:

[0095] S421, performing open operation on the luminance channel of the prediction box by using a 51x51 elliptical kernel to obtain an open operation result;

[0096] S422, subtracting the open operation result from the luminance channel of the prediction box to obtain a detail enhancement box.

[0097] The formula for obtaining the detail enhancement box is:

[0098] ;

[0099] ​​Among them, E is the detail enhancement box, L is the brightness channel of the prediction box, b is the 51x51 elliptical kernel, It is an opening operation.

[0100] In this embodiment, the formula for brightness adjustment in S43 is:

[0101] ;

[0102] Among them, L AD is the brightness adjustment frame, L is the brightness channel of the prediction frame, and E is the detail enhancement frame.

[0103] The present invention uses a 51x51 elliptical kernel to perform an opening operation (erode first and then dilate) on the brightness channel, which can effectively remove the non-uniform lighting background, enhance low-contrast details, and enhance the crack characteristics of low-contrast areas. The 2x detail enhancement frame is then added to the brightness channel of the prediction frame to obtain a brightness adjustment frame, which not only retains the basic brightness information but also amplifies the enhanced details.

[0104] In this embodiment, the calculation formula of texture complexity in S44 is:

[0105] ;

[0106] Among them, C is the texture complexity, l i is the brightness value of the i-th pixel in the brightness adjustment frame, N is the number of pixels in the brightness adjustment frame, and i is the pixel number.

[0107] In this embodiment, S45 includes the following sub-steps:

[0108] S451, slide a 5×5 X-direction Sobel kernel on the brightness adjustment frame and calculate the X-direction gradient of each sliding position;

[0109] S452, slide a 5×5 Y-direction Sobel kernel on the brightness adjustment frame and calculate the Y-direction gradient of each sliding position;

[0110] S453, calculating the standard deviation of the X-direction gradient at each sliding position to obtain the X-direction gradient standard deviation;

[0111] S454, calculating the standard deviation of the Y-direction gradient at each sliding position to obtain the Y-direction gradient standard deviation;

[0112] S455: Add the standard deviation of the gradient in the X direction and the standard deviation of the gradient in the Y direction to obtain the gradient variation.

[0113] Different textures, edges (such as cracks) in the image have different rates of brightness change in the horizontal (X direction) and vertical (Y direction). Taking cracks as an example, cracks can have various orientations, and the X direction gradient can highlight the edge change in the vertical direction (such as horizontally extending cracks, which have obvious vertical brightness changes), and the Y direction gradient can highlight the edge change in the horizontal direction (such as vertically extending cracks, which have obvious horizontal brightness changes). By calculating the gradients in the two orthogonal directions, the directional characteristics of the edges and textures in the region can be comprehensively captured, and key features can be avoided from being missed due to a single direction.

[0114] If the region is smooth, the X and Y direction gradient values are mostly small and concentrated, and the standard deviation tends to 0; if the region has complex textures such as cracks, the gradient values at different positions fluctuate greatly (some places have large gradients, and some places have small gradients), and the standard deviation will obviously increase. Through the standard deviation, the "fluctuation situation" of the gradient can be converted into a numerical index to distinguish between smooth regions and complex texture regions.

[0115] In this embodiment, the formula for calculating the texture smoothness in S46 is:

[0116] ;

[0117] Wherein, S is the texture smoothness, C is the texture complexity, and G is the gradient change amount.

[0118] In this embodiment, S5 includes the following steps:

[0119] S51, obtaining the center coordinates of the smooth prediction frame:

[0120] ;

[0121] Wherein, X p is the horizontal coordinate of the center coordinates, Y p is the vertical coordinate of the center coordinates, x1 is the horizontal coordinate of the first corner point of the smooth prediction frame, x2 is the horizontal coordinate of the second corner point of the smooth prediction frame, x3 is the horizontal coordinate of the third corner point of the smooth prediction frame, x4 is the horizontal coordinate of the fourth corner point of the smooth prediction frame, y1 is the vertical coordinate of the first corner point of the smooth prediction frame, y2 is the vertical coordinate of the second corner point of the smooth prediction frame, y3 is the vertical coordinate of the third corner point of the smooth prediction frame, and y4 is the vertical coordinate of the fourth corner point of the smooth prediction frame.

[0122] S52, performing coordinate conversion on the center coordinates of the smooth prediction frame to obtain target physical coordinates:

[0123] ;

[0124] Wherein, X z is the horizontal coordinate of the target physical coordinates, Yz Y is a physical starting calibration longitudinal coordinate of the upper left corner of the sample slide on which the sample to be detected is located on the stage, s Y is a physical starting calibration longitudinal coordinate of the upper left corner of the sample slide on which the sample to be detected is located on the stage, s Y is a physical starting calibration longitudinal coordinate of the upper left corner of the sample slide on which the sample to be detected is located on the stage, d Y is a physical starting calibration longitudinal coordinate of the upper left corner of the sample slide on which the sample to be detected is located on the stage,

[0125] S53, move the stage to the target physical coordinate, switch the high-power objective lens to scan the region in the smooth prediction frame, and obtain a high-power microscopic image, as shown in Figure 12 .

[0126] The sample on the stage remains stationary, and the stage is moved to the coordinate, the high-power objective lens is switched to scan the target region, and a high-power microscopic image of the effective region of the sample is obtained.

[0127] The present application determines the target region for high-power objective lens scanning through professional annotation and a deep learning model, and finally realizes accurate scanning of the specified region by the high-power objective lens, thereby improving the efficiency and accuracy of microscope sample analysis.

[0128] The present application processes the panoramic image of the sample to be detected through the trained Yolov11-FocalModulation network to obtain a suspected region where the target is located, i.e., a prediction frame, and then filters out the smooth prediction frame through texture complexity and gradient change, thereby effectively filtering out worthless regions and avoiding redundant scanning. The present application utilizes the synergy of panoramic scanning and splicing by a low-power objective lens and accurate imaging by a high-power objective lens, in the detection of non-homogeneous samples, not only obtains macroscopic information by quickly covering the entire sample with a low-power lens, but also captures microscopic details by focusing on key regions with a high-power lens, thereby breaking through the field-of-view limitation of the high-power objective lens, greatly improving the scanning efficiency and reducing the generation of redundant image data. At the same time, coordinate conversion and accurate movement control of the stage ensure accurate focusing of the high-power imaging on the target region, realize accurate shooting of the target, improve the scanning efficiency, and reduce redundant image data.

[0129] The above is only a preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A microscopic field adaptive positioning and high-magnification imaging guidance method, characterized in that: The following steps are involved: S1. Use the training sample graph to train the Yolov11-FocalModulation network to obtain the trained Yolov11-FocalModulation network; S2. Load the sample to be tested into the microscope loader, scan the sample to be tested using a low-magnification objective lens to obtain multiple local images, and stitch the multiple local images together to obtain a panoramic image of the sample to be tested; S3. Input the panorama of the sample to be detected into the trained Yolov11-FocalModulation network to obtain the predicted frame; S4. Filter the prediction frames based on texture complexity and gradient change, and retain smooth prediction frames; S5. Perform coordinate transformation on the center coordinates of the smooth prediction frame to obtain the target physical coordinates, move the stage to the target physical coordinates, switch the high-magnification objective lens to scan the area within the smooth prediction frame, and obtain a high-magnification microscopic image.

2. The microscopic field adaptive positioning and high-magnification imaging guidance method according to claim 1, characterized in that: The S1 comprises the following sub-steps: S11. Scan the entire field of view of the slide sample using a low-power microscope to obtain multiple partial images, and stitch the multiple partial images together to obtain a panoramic image of the slide sample; S12, manually marking abnormal areas on the panoramic image to obtain a training sample image; S13. Use the training sample graph to train the Yolov11-FocalModulation network to obtain a trained Yolov11-FocalModulation network.

3. The microscopic field adaptive positioning and high-magnification imaging guidance method according to claim 1, characterized in that: The structure of the Yolov11-FocalModulation network in S1 is as follows: based on the original Yolov11 network architecture, the SPPF unit in the original Yolov11 network architecture is replaced by the FocalModulation unit.

4. The microscopic field adaptive positioning and high-magnification imaging guidance method according to claim 1, characterized in that: The S4 comprises the following sub-steps: S41, converting the prediction frame from the RGB color space to the LAB color space, and extracting the brightness channel in the LAB color space; S42, performing detail enhancement on the brightness channel of the predicted frame to obtain a detail enhanced frame; S43, adjusting the brightness of the detail enhancement frame to obtain a brightness adjustment frame; S44, extracting texture complexity of the brightness adjustment frame; S45, obtaining the X-direction gradient and the Y-direction gradient of the brightness adjustment frame, and calculating the gradient change; S46, calculating texture smoothness according to the gradient variation and texture complexity; S47. Retain the prediction box whose texture smoothness is greater than the smoothness threshold.

5. The microscopic field adaptive positioning and high-magnification imaging guidance method according to claim 4, characterized in that: The S42 includes the following sub-steps: S421, performing an opening operation on the brightness channel of the prediction frame using a 51x51 elliptical kernel to obtain an opening operation result; S422: Subtract the opening operation result from the brightness channel of the predicted frame to obtain a detail enhancement frame.

6. The microscopic field adaptive positioning and high-magnification imaging guidance method according to claim 4, characterized in that: The formula for brightness adjustment in S43 is: ； Among them, L AD is the brightness adjustment frame, L is the brightness channel of the prediction frame, and E is the detail enhancement frame.

7. The microscopic field adaptive positioning and high-magnification imaging guidance method according to claim 4, characterized in that: The calculation formula of the texture complexity in S44 is: ； Among them, C is the texture complexity, l i is the brightness value of the i-th pixel in the brightness adjustment frame, N is the number of pixels in the brightness adjustment frame, and i is the pixel number.

8. The microscopic field adaptive positioning and high-magnification imaging guidance method according to claim 4, characterized in that: The S45 includes the following sub-steps: S451, slide a 5×5 X-direction Sobel kernel on the brightness adjustment frame and calculate the X-direction gradient of each sliding position; S452, slide a 5×5 Y-direction Sobel kernel on the brightness adjustment frame and calculate the Y-direction gradient of each sliding position; S453, calculating the standard deviation of the X-direction gradient at each sliding position to obtain the X-direction gradient standard deviation; S454, calculating the standard deviation of the Y-direction gradient at each sliding position to obtain the Y-direction gradient standard deviation; S455: Add the X-direction gradient standard deviation and the Y-direction gradient standard deviation to obtain the gradient variation.

9. The microscopic field adaptive positioning and high-magnification imaging guidance method according to claim 4, characterized in that: The formula for calculating texture smoothness in S46 is: ； Among them, S is the texture smoothness, C is the texture complexity, and G is the gradient change.

10. The microscopic field adaptive positioning and high-magnification imaging guidance method according to claim 1, characterized in that: The S5 comprises the following sub-steps: S51. Get the center coordinates of the smoothed prediction box: ； Among them, X p is the horizontal coordinate of the center coordinate, Y p is the ordinate of the center coordinate, x1 is the abscissa of the first corner point of the smooth prediction box, x2 is the abscissa of the second corner point of the smooth prediction box, x3 is the abscissa of the third corner point of the smooth prediction box, x4 is the abscissa of the fourth corner point of the smooth prediction box, y1 is the ordinate of the first corner point of the smooth prediction box, y2 is the ordinate of the second corner point of the smooth prediction box, y3 is the ordinate of the third corner point of the smooth prediction box, and y4 is the ordinate of the fourth corner point of the smooth prediction box; S52. Perform coordinate transformation on the center coordinates of the smoothed prediction frame to obtain the target physical coordinates: ； Among them, X z is the horizontal coordinate of the target physical coordinate, Y z is the vertical coordinate of the target physical coordinate, X s Y is the physical starting point for calibration of the upper left corner of the slide where the sample to be tested is located on the stage. s P is the physical starting calibration vertical coordinate of the upper left corner of the slide sample where the sample to be tested is located on the stage, d The coefficient for converting pixel value to physical distance; S53, moving the stage to the target physical coordinates, switching the high-magnification objective lens to scan the area within the smooth prediction frame, and obtaining a high-magnification microscopic image.

Images