Capillary aperture positioning method and system

Abstract

The invention provides a capillary aperture positioning method and system, and the method comprises the steps: carrying out the global enhancement preprocessing of a to-be-recognized capillary aperture image, so as to eliminate the noise interference of the image, and obtaining more uniform brightness and a high-contrast visual effect; inputting the image after enhancement preprocessing into a target positioning and measuring network, performing feature extraction on a target aperture region in the image by using an adaptive deformation convolution technology, and extracting feature values of the target aperture region in different sizes and directions; and further calculating the three-dimensional space weight of each pixel point in the feature map, and enhancing the weight of the pixel points in the target aperture region. And finally, reducing an actual error value between a real value and a predicted capillary aperture result through a Wise-IoUv3 loss calculation function. According to the invention, the rapid positioning and diameter detection of the aperture of the capillary tube in industrial production are realized, the interference in a complex environment can be effectively overcome, and the detection speed and accuracy are obviously improved.

Description

Technical Field

[0001] The present invention relates to the field of target detection, and more specifically, to a capillary aperture positioning method and system. Background Art

[0002] The internal environment of a capillary is usually a narrow, tortuous and closed space, which makes it difficult for detection equipment to enter and operate. Especially in capillaries of different shapes, the measurement will be more complicated. There may be tiny defects or changes inside the capillary, which requires high measurement accuracy in order to accurately detect these changes and analyze them. Summary of the invention

[0003] In view of the technical problems existing in the prior art, the present invention provides a capillary aperture positioning method and system, aiming to overcome the problems of capillary aperture shape differences, light influence, and light changes caused by narrow internal space, thereby improving the accuracy and adaptability of measurement results.

[0004] According to a first aspect of the present invention, there is provided a capillary aperture positioning method, comprising: Acquire an image of the capillary aperture to be identified; Locating an overexposed area and an underexposed area in the capillary aperture image to be identified, and performing enhancement preprocessing on the overexposed area and the underexposed area; Inputting the capillary aperture image to be identified after enhancement preprocessing into a target positioning and measurement network to obtain the capillary aperture position area in the capillary aperture image to be identified; The step of inputting the capillary aperture image to be identified after enhancement preprocessing into a target positioning and measurement network to obtain the capillary aperture position area in the capillary aperture image to be identified includes: Inputting the capillary aperture image to be identified after enhancement preprocessing into a target positioning and measurement network, wherein the target positioning and measurement network extracts a feature map of the capillary aperture image to be identified based on adaptive deformation convolution; Calculate the three-dimensional spatial dimension weight of each pixel in the feature map, and mark the three-dimensional spatial dimension weight on each pixel, wherein the three-dimensional spatial dimension weight represents the importance of the pixel in the feature map; The marked feature map is predicted to obtain the capillary aperture position area in the capillary aperture image to be identified.

[0005] According to a second aspect of the present invention, there is provided a capillary aperture positioning system, comprising: An acquisition module, used for acquiring an image of the capillary aperture to be identified; An enhancement processing module, used for locating overexposed areas and underexposed areas in the capillary aperture image to be identified, and performing enhancement preprocessing on the overexposed areas and the underexposed areas; An identification module, used for inputting the capillary aperture image to be identified after enhancement preprocessing into a target positioning and measurement network, and obtaining the capillary aperture position area in the capillary aperture image to be identified; The step of inputting the capillary aperture image to be identified after enhancement preprocessing into a target positioning and measurement network to obtain the capillary aperture position area in the capillary aperture image to be identified includes: Inputting the capillary aperture image to be identified after enhancement preprocessing into a target positioning and measurement network, wherein the target positioning and measurement network extracts a feature map of the capillary aperture image to be identified based on adaptive deformation convolution; Calculate the three-dimensional spatial dimension weight of each pixel in the feature map, and mark the three-dimensional spatial dimension weight on each pixel, wherein the three-dimensional spatial dimension weight represents the importance of the pixel in the feature map; The marked feature map is predicted to obtain the capillary aperture position area in the capillary aperture image to be identified.

[0006] The present invention provides a capillary aperture positioning method and system, which performs global enhancement preprocessing on the capillary aperture image to be identified to eliminate image noise interference and obtain a more uniform brightness and high contrast visual effect; the enhanced preprocessed image is input into the target positioning and measurement network, and the adaptive deformation convolution technology is used to extract the features of the target aperture area in the image, and the characteristic values ​​of the target aperture area are extracted in different sizes and directions; the three-dimensional spatial weight of each pixel in the feature map is further calculated to enhance the weight size of the pixel in the target aperture area. Finally, the actual error value between the true value and the predicted capillary aperture result is reduced by the Wise-IoUv3 loss calculation function. The present invention realizes the rapid positioning and diameter detection of capillary aperture in industrial production, can effectively overcome the interference in complex environments, and has significant improvements in detection speed and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Figure 1 A flow chart of a capillary aperture positioning method provided by the present invention; Figure 2 It is a schematic diagram of the structure of the target positioning and measurement network; Figure 3 A schematic diagram of feature extraction for a capillary aperture image to be identified; Figure 4 A schematic diagram of calculating the 3D spatial dimension weight of each pixel in the feature map; Figure 5 A schematic structural diagram of a capillary aperture positioning system provided by the present invention. DETAILED DESCRIPTION

[0008] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. In addition, the technical features in the various embodiments or single embodiments provided by the present invention can be arbitrarily combined with each other to form a feasible technical solution. This combination is not subject to the constraints of the sequence of steps and / or the structural composition mode, but must be based on the ability of ordinary technicians in this field to achieve. When the combination of technical solutions is contradictory or cannot be achieved, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0009] Figure 1 A flow chart of a capillary aperture positioning method provided by the present invention, such as Figure 1 As shown, the method includes: Step 1: Obtain an image of the capillary aperture to be identified.

[0010] It is understandable that firstly, an industrial camera is used to acquire an image of the capillary aperture under the complex conditions to be processed.

[0011] Step 2: locate the overexposed area and the underexposed area in the capillary aperture image to be identified, and perform enhancement preprocessing on the overexposed area and the underexposed area.

[0012] It is understandable that since the image quality consistency cannot be fully guaranteed during the acquisition of capillary aperture images, the obtained images may contain a large amount of irrelevant noise due to overexposure and underexposure. Therefore, it is necessary to pre-process the acquired capillary aperture images, especially for those capillary images with low quality, blurry or affected by noise, and enhance them using specially designed algorithms. This enhancement process aims to optimize the image quality and improve the visualization and feature clarity of low-quality images. The algorithm comprehensively considers factors such as image clarity, noise level and blurriness to ensure more accurate and reliable target detection results in subsequent steps.

[0013] In a possible implementation manner of the present invention, locating the overexposed area and the underexposed area in the capillary aperture image to be identified includes: locating the overexposed area and the underexposed area in the capillary aperture image to be identified based on an LCDNet bilateral network.

[0014] The method of locating the overexposed area and the underexposed area in the capillary aperture image to be identified based on the LCDNet bilateral network includes: Step 21, dividing the input capillary aperture image to be identified into local areas of different scales, and constructing a pyramid space domain according to the local areas of different scales, wherein each level of the pyramid space domain has local color distribution characteristics of different scales.

[0015] Specifically, the LCDNet bilateral network is used to process images under complex conditions, which may contain a lot of irrelevant noise due to overexposure and underexposure. In order to locate the overexposed and underexposed areas in the image, the LCDNet bilateral network uses the local information prior method to locate. It divides the input image and constructs a pyramid space domain, each level of which contains color distributions of different scales, so as to obtain accurate spatial distribution information. Through this step, the LCDNet bilateral network obtains preliminary color domain spatial information.

[0016] The local color distribution characteristics of different scales are calculated by the following formula:

[0017] , is the local color distribution with a scale of K, (i, j) is the index of the pixel in the image, c is the channel index of the image, and for RGB images, its value is 3, and b represents the histogram index into which the calculated pixel color value falls. For LCDNet bilateral network, is the K on the image K local area, ( ) indicates K K pixels in the local area.

[0018] Step 22: identifying overexposed areas and underexposed areas in the capillary aperture image to be identified based on local color distribution characteristics at different scales.

[0019] It is understandable that, according to the color value of each pixel, the over-exposed area and the under-exposed area in the capillary aperture image to be identified are determined.

[0020] Specifically, the above formula is used to calculate in which color value histogram the pixel value calculated for each pixel point at different scales falls. According to the histogram of the color value in which the pixel value of the pixel point falls, the over-exposed color value range and the under-exposed color value range are set, so that the over-exposed area and the under-exposed area in the image can be located.

[0021] The LCDNet bilateral network performs simple convolution and activation function operations on overexposed and underexposed areas to enhance information and highlight the differences between different areas.

[0022] Specifically, the overexposed area and underexposed area are enhanced by the following formula: ; ; in, is the image after convolution operation, is the original aperture image to be identified, represents the convolution operation, For overexposed areas, The underexposed area Representatives To encode, Get the ratio of the two relationships, Indicates activation operation. It represents the capillary aperture image to be identified after enhancement preprocessing.

[0023] In order to better capture the long-range spatial correlation of pixels in the image, LCDNet uses a fusion network to perform preliminary fusion of images at different levels. This step helps provide high-quality local image enhancement. Considering the complementary relationship between overexposed and underexposed areas, the fusion network ensures that the final enhanced image has a natural and smooth effect. Overall, the LCDNet bilateral network is constructed through prior local information and pyramid space domain, combined with simple and effective operations and fusion networks, to achieve fast and effective adaptive enhancement processing of images under complex conditions.

[0024] Step 3: input the capillary aperture image to be identified after enhancement preprocessing into a target positioning and measurement network to obtain the capillary aperture position area in the capillary aperture image to be identified.

[0025] Among them, see Figure 2 , is a schematic diagram of the structure of the target positioning and measurement network, inputting the capillary aperture image to be identified after enhanced preprocessing into the target positioning and measurement network, and obtaining the capillary aperture position area in the capillary aperture image to be identified, including: Step 31, inputting the capillary aperture image to be identified after enhancement preprocessing into a target positioning and measurement network, wherein the target positioning and measurement network extracts a feature map of the capillary aperture image to be identified based on adaptive deformation convolution.

[0026] It is understandable that the enhanced capillary aperture image to be identified is input into the target positioning and measurement network, and processed by the target positioning and measurement network to obtain the capillary aperture position area and capillary aperture type in the capillary aperture image.

[0027] Among them, on the enhanced pre-processed image, the target positioning and measurement network performs the feature extraction operation of adaptive deformable convolution. Since different optical fiber products have different internal channels and structures, in order to more accurately capture the shape and characteristics of the target area. The feature extraction process takes into account the actual shape of the target and adopts an adaptive area coverage strategy to ensure that the extracted feature map is more in line with the geometric features of the target. Through adaptive deformable convolution, the target positioning and measurement network can flexibly adjust the convolution kernel according to the shape changes of the target area, so as to better capture the details and structure of the target. This adaptability ensures that high-quality feature representation can be obtained in target scenes of different sizes, shapes and directions. The obtained target area feature map not only contains the local information of the target, but also reflects the overall shape characteristics of the target.

[0028] in, Figure 3 A schematic diagram of extracting a feature map of a capillary aperture image to be identified. The feature of each sampling point in the capillary aperture image to be identified is extracted according to the following formula:

[0029] Where x represents the capillary aperture image to be identified after enhancement preprocessing, Represents the sampling point position of the current pixel in the image, is the relative offset of the standard convolution sampling point, is the learning offset, which is used to dynamically adjust the position of the sampling point; is the modulation range; is a fixed weight for the sampling position of each pixel, Represents the range of coordinates of the convolution operation position points; The sampling point of the current pixel The characteristics of and modulation range is the learnable amount, The value range of is [0,1]. and The dynamic adjustment of the deformable convolution operation eliminates the interference of irrelevant background, better fits the target area, ensures accurate perception of the target, and has fast and accurate positioning capabilities, providing a richer information basis and faster positioning speed for subsequent detection output Among them, when extracting the features of each sampling point in the capillary aperture image to be identified, if the offset If it is a decimal, it means that the current sampling point is no longer strictly aligned to a discrete pixel point, but falls on a continuous position between two pixels. To solve this problem, bilinear interpolation is used to supplement the pixel positions of the target area in the image, and the approximate pixel values ​​of the decimal positions are calculated from the surrounding integer pixels to ensure that the feature values ​​at the continuous positions are available.

[0030] The features of each pixel in the capillary aperture image to be identified are extracted by the above formula, and an original feature map fitting the target area is obtained.

[0031] Step 32, calculate the three-dimensional spatial dimension weight of each pixel point in the feature map, and mark the three-dimensional spatial dimension weight on each pixel point, wherein the three-dimensional spatial dimension weight represents the importance of the pixel point in the feature map.

[0032] It can be understood that after extracting the feature map of the capillary aperture image to be identified in step 31, the 3D spatial dimension weight of each pixel in the feature map is calculated by the energy function. The size of the 3D spatial dimension weight represents the importance of the pixel. Figure 4 The calculation process of the 3D spatial dimension weight of the pixel point is shown. The difference between the reference point in the vertical direction and the vertical direction is calculated by introducing an energy function. Each channel in the feature map contains M energy functions, where M is the product of the height and width of the feature map. This ensures that each pixel in the feature map is calculated comprehensively and carefully to better reflect the distribution of the target in 3D space. By learning the importance of each position in the feature map, the target positioning and measurement network can pay more attention to the key dimensions of the target area. The SimAm method ensures more targeted feature extraction of key dimensions by dynamically and adaptively adjusting the feature map. This non-parameterized weight acquisition method helps the model better adapt to changes in different target shapes, sizes, and directions. The obtained 3D dimension weights serve as an important guiding factor in the feature extraction stage, allowing the network to pay more attention to the key features of the aperture area, thereby improving the accuracy of subsequent aperture area measurements.

[0033] Specifically, the three-dimensional spatial dimension weight of each pixel in the feature map is calculated, including: ; in, They represent the weight system of the target neuron, the bias coefficient, the output of the target positioning and measurement network, and the feature of the w-th pixel in the input feature map respectively; and Represent the true output of the target neuron and the true output of other neurons respectively; and They represent the linear transformation outputs of their respective target neurons. Each channel contains M energy functions, where M is equal to the height H of the feature map multiplied by the width W. , is to average the errors of other neurons.

[0034] The present invention ensures that each pixel in the feature map is calculated comprehensively and carefully to more accurately reflect the distribution of the target in three-dimensional space. In order to avoid the energy function value being too large, the sigmoid function is introduced to adjust the 3D spatial dimension weight of each pixel in the feature map to limit its value range to [0,1], thereby improving its controllability. The relative importance of each pixel in the feature map can be explained by the weight size of the three-dimensional spatial dimension. This provides guidance for subsequent adaptive deformable convolution, ensuring that more attention is paid to important areas during the target positioning process, and effectively obtaining a feature map with relatively important pixels.

[0035] After calculating the 3D spatial dimension weight of each pixel in the feature map, the 3D spatial dimension weight is marked on each corresponding pixel in the feature map to mark the importance of each pixel in the feature map, so that the subsequent target positioning and measurement network will focus on the area with larger 3D spatial dimension weight.

[0036] Step 33: predict the marked feature map to obtain the capillary aperture position area in the capillary aperture image to be identified.

[0037] It is understandable that step 32 obtains a weighted feature map of pixels in the 3D spatial dimension, and uses Wise-IoUv3 to perform loss calculation for target detection output and capillary aperture image loss calculation. The present invention introduces the concept of outlier to reflect the quality of the capillary aperture true value, and a smaller outlier indicates a high-quality capillary aperture true value. By matching the high-quality capillary aperture true value with a smaller gradient gain, the model is ensured to pay more attention to high-quality targets. This helps to focus the prediction regression frame more on the anchor frame of ordinary quality, avoiding low-quality images from generating more interference gradients for model prediction, thereby improving the accuracy of model prediction. This process involves iterative adjustment of the candidate frame of the target area of ​​the detection output. The Wise-IOUv3 function combines position information and shape matching by considering the value between the detection output and the ground truth, and provides a comprehensive loss evaluation for the adjustment of the target area. The design of this function not only focuses on the accuracy of the detection output, but also fully considers the matching degree of the shape and position of the target area.

[0038] In the process of training the target positioning and measurement network, the loss value is calculated based on the Wise-10Uv3 function, where:

[0039] ; ; in, is a weighting factor related to the distance to the center point; It is the loss function based on IoU. represents the weighted loss value, represents the final loss value, Represents the center point coordinates of the prediction box, Indicates the real coordinates of the center point of the target area, Represents the square of the Euclidean distance between the center point of the prediction box and the center point of the target box, which is used to measure the geometric proximity between the center points of the prediction box and the target box. and Represent the height and width of the target box respectively, Represents the normalized width and height of the minimum bounding box. Both represent constant coefficients is a non-monotonic coefficient.

[0040] The present invention uses the Wise-IOUv3 function to calculate the loss to guide the iterative adjustment of the capillary aperture area. The weighted IoU loss of the target area is measured. is the weight term of the target position and size. Based on this, the weighted IoU loss is increased. , adapt the outlier degree by adjusting the gradient gain (β) to match the true value of the high-quality capillary aperture. This helps to focus the gradient on the anchor frame of normal quality and avoid harmful gradients caused by low-quality images. According to the gradient information of the loss value, the position, size and other parameters of the target area are fine-tuned. This fast detection iterative optimization mechanism can continuously optimize during the training process. And dynamically adjust various parameters.

[0041] See also Figure 5 , provides a capillary aperture positioning system of the present invention, the system comprising: An acquisition module 501 is used to acquire an image of the capillary aperture to be identified; An enhancement processing module 502 is used to locate the over-exposed area and the under-exposed area in the capillary aperture image to be identified, and perform enhancement pre-processing on the over-exposed area and the under-exposed area; The identification module 503 is used to input the capillary aperture image to be identified after enhancement preprocessing into the target positioning and measurement network to obtain the capillary aperture position area in the capillary aperture image to be identified; The step of inputting the capillary aperture image to be identified after enhancement preprocessing into a target positioning and measurement network to obtain the capillary aperture position area in the capillary aperture image to be identified includes: Inputting the capillary aperture image to be identified after enhancement preprocessing into a target positioning and measurement network, wherein the target positioning and measurement network extracts a feature map of the capillary aperture image to be identified based on adaptive deformation convolution; Calculate the three-dimensional spatial dimension weight of each pixel in the feature map, and mark the three-dimensional spatial dimension weight on each pixel, wherein the three-dimensional spatial dimension weight represents the importance of the pixel in the feature map; The marked feature map is predicted to obtain the capillary aperture position area in the capillary aperture image to be identified.

[0042] It can be understood that the capillary aperture positioning system provided by the present invention corresponds to the capillary aperture positioning method provided in the aforementioned embodiments, and the relevant technical features of the capillary aperture positioning system can refer to the relevant technical features of the capillary aperture positioning method, which will not be repeated here.

[0043] The capillary aperture positioning method and system provided by the embodiments of the present invention have the following beneficial effects: (1) The image enhancement of this method aims to filter out irrelevant information in the image and retain valuable image information in industrial production, laying the foundation for subsequent tasks including detection and recognition. Object detection aims to identify and locate the shape and internal structure of a specific object from an image or video, which will be compared with qualified products to determine whether the product meets the high quality requirements. This task requires the algorithm to not only identify the existing objects in the image and accurately locate them, but also provide a bounding box for each object and accurately determine whether it is qualified or not.

[0044] (2) This method combines SimAm image feature extraction technology, 3D dimension weights, and Wise-lOUv3 functions to achieve target detection and adjustment. It can effectively deal with the challenges of capillary aperture positioning and uneven illumination and internal shape diversity in the measurement environment, thereby improving the accuracy and robustness of the measurement. This method ensures that efficient measurement performance can be maintained and gives the solution wide applicability in this field, providing an innovative solution for the further development of capillary measurement technology.

[0045] It should be noted that in the above embodiments, the description of each embodiment has its own emphasis, and for parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0046] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0047] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0048] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0049] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0050] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0051] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A capillary aperture positioning method, characterized in that: include: Acquire an image of the capillary aperture to be identified; Locating an overexposed area and an underexposed area in the capillary aperture image to be identified, and performing enhancement preprocessing on the overexposed area and the underexposed area; Inputting the capillary aperture image to be identified after enhancement preprocessing into a target positioning and measurement network to obtain the capillary aperture position area in the capillary aperture image to be identified; The step of inputting the capillary aperture image to be identified after enhancement preprocessing into a target positioning and measurement network to obtain the capillary aperture position area in the capillary aperture image to be identified includes: Inputting the capillary aperture image to be identified after enhancement preprocessing into a target positioning and measurement network, wherein the target positioning and measurement network extracts a feature map of the capillary aperture image to be identified based on adaptive deformation convolution; Calculate the three-dimensional spatial dimension weight of each pixel in the feature map, and mark the three-dimensional spatial dimension weight on each pixel, wherein the three-dimensional spatial dimension weight represents the importance of the pixel in the feature map; The marked feature map is predicted to obtain the capillary aperture position area in the capillary aperture image to be identified.

2. The capillary aperture positioning method according to claim 1, characterized in that: The locating the over-exposed area and the under-exposed area in the capillary aperture image to be identified comprises: Based on the LCDNet bilateral network, the over-exposed area and the under-exposed area in the capillary aperture image to be identified are located.

3. The capillary aperture positioning method according to claim 2, characterized in that: The method of locating the overexposed area and the underexposed area in the capillary aperture image to be identified based on the LCDNet bilateral network includes: The input capillary aperture image to be identified is divided into local areas of different scales, and a pyramid space domain is constructed according to the local areas of different scales, wherein each level of the pyramid space domain has local color distribution characteristics of different scales; Based on the local color distribution characteristics at different scales, overexposed areas and underexposed areas in the capillary aperture image to be identified are identified.

4. The capillary aperture positioning method according to claim 3, characterized in that: The local color distribution characteristics of different scales are calculated by the following formula: , is the local color distribution with a scale of K, (i, j) is the index of the pixel in the image, c is the channel index of the image, and for RGB images, its value is 3, and b represents the histogram index into which the calculated pixel color value falls. For LCDNet bilateral network, is the K on the image K local area; ( ) indicates K The indices of the pixels in the K local area; The identifying of overexposed areas and underexposed areas in the capillary aperture image to be identified based on local color distribution features at different scales includes: According to the color value of each pixel, the over-exposed area and the under-exposed area in the capillary aperture image to be identified are determined.

5. The capillary aperture positioning method according to claim 1, characterized in that: performing enhancement preprocessing on the overexposed area and the underexposed area, include: ； ； in, is the image after convolution operation, is the original aperture image to be identified, represents the convolution operation, For overexposed areas, For underexposed areas, Representatives To encode, Get the ratio of the two relationships, Indicates activation operation. It represents the capillary aperture image to be identified after enhancement preprocessing.

6. The capillary aperture positioning method according to claim 1, characterized in that: The method of inputting the capillary aperture image to be identified after enhanced preprocessing into a target positioning and measurement network, wherein the target positioning and measurement network extracts a feature map of the capillary aperture image to be identified based on adaptive deformation convolution, comprises: Where x represents the capillary aperture image to be identified after enhancement preprocessing, Represents the position of the current pixel in the image. is the relative offset of the standard convolution sampling point, is the learning offset, which is used to dynamically adjust the position of the sampling point; is the modulation range; is a fixed weight for the sampling position of each pixel, Represents the range of coordinates of the convolution operation position points; The current pixel The characteristics of and modulation range is the learnable amount, The value range of is [0,1]; The feature of each pixel in the capillary aperture image to be identified is extracted to obtain a feature map of the capillary aperture image to be identified.

7. The capillary aperture positioning method according to claim 6, characterized in that: In the process of extracting the characteristic graph of the capillary aperture image to be identified, if the offset It is a decimal number, indicating the current sampling point If it is not strictly aligned to discrete pixel points but falls on a continuous position between two integer pixel points, bilinear interpolation is used to supplement the pixel position of the target area of ​​the capillary aperture image to be identified, and the approximate pixel point at the decimal position is calculated from the surrounding integer pixel points, and the features of the approximate pixel point are extracted to replace the features of the current sampling point.

8. The capillary aperture positioning method according to claim 1, characterized in that: The calculating the three-dimensional spatial dimension weight of each pixel point in the feature map includes: ； in, They represent the weight system of the target neuron, the bias coefficient, the output of the target positioning and measurement network, and the feature of the w-th pixel in the input feature map respectively; and Represent the true output of the target neuron and the true output of other neurons respectively; and They represent the linear transformation outputs of their respective target neurons. Each channel contains M energy functions, where M is equal to the height H of the feature map multiplied by the width W. , is to average the errors of other neurons.

9. The capillary aperture positioning method according to claim 1, characterized in that: In the process of training the target positioning and measurement network, the loss value is calculated based on the Wise-10Uv3 function, where: ； ； in, is a weighting factor related to the distance to the center point; It is the loss function based on IoU. represents the weighted loss value, represents the final loss value, Represents the center point coordinates of the prediction box, Indicates the real coordinates of the center point of the target area, Represents the square of the Euclidean distance between the center point of the prediction box and the center point of the target box, which is used to measure the geometric proximity between the center points of the prediction box and the target box. and Represent the height and width of the target box respectively, Represents the normalized width and height of the minimum bounding box. Both represent constant coefficients is a non-monotonic coefficient, by adjusting Adjust and optimize the target positioning and measurement network.

10. A capillary aperture positioning system, characterized in that: include: An acquisition module, used for acquiring an image of the capillary aperture to be identified; An enhancement processing module, used for locating overexposed areas and underexposed areas in the capillary aperture image to be identified, and performing enhancement preprocessing on the overexposed areas and the underexposed areas; An identification module, used for inputting the capillary aperture image to be identified after enhancement preprocessing into a target positioning and measurement network, and obtaining the capillary aperture position area in the capillary aperture image to be identified; The step of inputting the capillary aperture image to be identified after enhancement preprocessing into a target positioning and measurement network to obtain the capillary aperture position area in the capillary aperture image to be identified includes: Inputting the capillary aperture image to be identified after enhancement preprocessing into a target positioning and measurement network, wherein the target positioning and measurement network extracts a feature map of the capillary aperture image to be identified based on adaptive deformation convolution; Calculate the three-dimensional spatial dimension weight of each pixel in the feature map, and mark the three-dimensional spatial dimension weight on each pixel, wherein the three-dimensional spatial dimension weight represents the importance of the pixel in the feature map; The marked feature map is predicted to obtain the capillary aperture position area in the capillary aperture image to be identified.

Images