Multi-plantar wart morphological data acquisition system based on machine vision system

Abstract

The invention relates to the technical field of machine vision, in particular to a multiple verruca plantaris morphological data acquisition system based on a machine vision system, and aims to generate an accurate verruca plantaris three-dimensional morphological graph. The system comprises a plantar wart image acquisition module, a plantar wart area identification module, an image enhancement module, an image restoration module and a plantar wart three-dimensional form reconstruction module. The plantar wart image acquisition module is used for shooting from three different visual angles to obtain a multi-visual-angle plantar wart image; the verruca plantaris area identification module identifies a verruca plantaris area in the multi-view verruca plantaris image and analyzes the verruca plantaris area to obtain key morphological characteristics; the image enhancement module constructs an image enhancement model to perform multi-region enhancement processing on the plantar wart region image to obtain a second enhanced plantar wart image; the image repairing module is used for repairing depth information of the second enhanced plantar wart image to obtain a repaired plantar wart depth map; and the plantar wart three-dimensional shape reconstruction module generates a plantar wart three-dimensional shape graph according to the repaired plantar wart image and the multi-view plantar wart image.

Description

Technical Field

[0001] The invention relates to the technical field of machine vision, and in particular to a multiple plantar wart morphology data acquisition system based on a machine vision system. Background Art

[0002] Plantar warts are skin lesions caused by human papillomavirus infection. They are common on the soles of the feet and other pressure-bearing areas, such as the heels or under the metatarsals, and may cause pain or discomfort when walking or standing. At the same time, plantar warts are highly contagious and can be spread through direct contact, especially in warm and humid environments, such as swimming pools and locker rooms. Therefore, accurate diagnosis and treatment of plantar warts can not only help patients prevent the spread of lesions and complications, but also protect public health.

[0003] However, due to long-term weight pressure, multiple plantar warts usually grow inward, showing characteristics such as rough surface and deep tissue invasion. The lesion morphology is complex, which poses challenges to traditional diagnosis and treatment. Traditional two-dimensional imaging methods cannot accurately capture the depth information of multiple plantar warts, and diagnosis often relies on the doctor's experience and lacks objective quantitative data support. Therefore, a new multiple plantar wart morphology data acquisition technology is urgently needed to overcome the limitations of traditional two-dimensional imaging methods, so as to improve the accurate identification and monitoring of multiple plantar wart lesions.

[0004] Therefore, a multiple plantar warts morphological data acquisition system based on machine vision system was proposed. Summary of the invention

[0005] The object of the present invention is to provide a multiple plantar wart morphological data acquisition system based on a machine vision system, which generates an accurate three-dimensional morphological map of plantar warts through the collaborative work of multiple modules. First, the plantar wart image acquisition module shoots from three different perspectives to obtain a multi-perspective plantar wart image; secondly, the plantar wart region recognition module obtains the key morphological features of the plantar wart region by identifying and analyzing the plantar wart region in the multi-perspective plantar wart image; then, the image enhancement module performs multi-region enhancement processing on the plantar wart region image by constructing an image enhancement model to obtain a second enhanced plantar wart image; then, the image repair module repairs the depth information of the second enhanced plantar wart image to obtain a repaired plantar wart depth map; finally, the plantar wart three-dimensional morphological reconstruction module generates a plantar wart three-dimensional morphological map according to the repaired plantar wart image and the multi-perspective plantar wart image.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A multiple plantar wart morphology data acquisition system based on a machine vision system, comprising:

[0008] The plantar wart image acquisition module is used to capture two-dimensional image data of the plantar wart through a first viewing angle, a second viewing angle, and a third viewing angle to obtain a multi-view plantar wart image;

[0009] A plantar wart region recognition module is used to identify the plantar wart region in the multi-view plantar wart image by constructing a plantar wart region recognition model to obtain a plantar wart region image; and analyze the plantar wart region image to obtain key morphological features of the plantar wart region;

[0010] An image enhancement module is used to construct an image enhancement model to perform multi-region enhancement processing on the plantar wart region image; the image enhancement model includes: a first image enhancement layer, a pressure analysis layer, and a second image enhancement layer; the steps of processing the plantar wart region image are: performing a first enhancement processing on the plantar wart region image through the first image enhancement layer to obtain a first enhanced plantar wart image; performing a pressure analysis on the first enhanced plantar wart image according to the key morphological features through the pressure analysis layer to obtain a pressure value; marking the area where the pressure value is greater than the pressure threshold as a high-pressure area; performing a second enhancement processing on the image of the high-pressure area through the second image enhancement layer to obtain a second enhanced plantar wart image;

[0011] An image restoration module, used for restoring the depth information of the second enhanced plantar wart image to obtain a restored plantar wart depth map;

[0012] The plantar wart three-dimensional morphology reconstruction module is used to construct a plantar wart three-dimensional morphology reconstruction model and generate a plantar wart three-dimensional morphology map according to the repaired plantar wart image and the multi-view plantar wart image.

[0013] Preferably, after the multi-view plantar wart image is acquired by the plantar wart image acquisition module, the multi-view plantar wart image is preprocessed; the preprocessing includes: color space conversion and denoising;

[0014] The color space conversion converts the RGB color space of the multi-view plantar wart image into the HSV color space;

[0015] The denoising method removes the noise generated when the multi-view plantar wart images are collected by a multi-scale denoising algorithm.

[0016] Preferably, the plantar wart region recognition model comprises: a plantar wart region segmentation layer, a segmentation result optimization layer and a key morphological feature extraction layer;

[0017] The plantar wart region segmentation layer uses Faster R-CNN as the main segmentation model to perform plantar wart region recognition on the multi-view plantar wart image. The plantar wart region recognition step is: using a labeling tool to label the plantar wart category of the training set of the multi-view plantar wart image to obtain the labeled training set; the plantar wart categories include: healthy skin, mild plantar warts and fused plantar warts; performing random enhancement processing on the labeled training set through an enhancement strategy to obtain an enhanced training set; training the Faster R-CNN model on the enhanced training set by using a cross entropy loss function and an IOU loss function; using the trained Faster R-CNN model to perform plantar wart region recognition on the test set of the multi-view plantar wart image to generate a preliminary segmentation result of the plantar wart region;

[0018] The segmentation result optimization layer includes: an edge optimization layer and a region filling layer; wherein the edge optimization layer optimizes the edge of the preliminary segmentation result of the plantar wart region by using the Canny edge detection algorithm; the optimization steps are: graying the preliminary segmentation result to obtain a preliminary segmentation result grayscale image; setting high and low thresholds to extract edge information of the preliminary segmentation result grayscale image; merging the preliminary segmentation result and the edge information to obtain an accurate segmentation result of the plantar wart region; the region filling layer fills the holes and cracks in the accurate segmentation result by using morphological operations to obtain an image of the plantar wart region;

[0019] The key morphological feature extraction layer extracts the key morphological features of the plantar wart area image; the key morphological features include: texture features and edge features; wherein the texture features are obtained by performing texture analysis on the plantar wart area image using a gray level co-occurrence matrix; the texture features include: contrast, correlation, entropy and uniformity; the edge features are obtained by analyzing the edge information; the edge features include: plantar wart area area and boundary length.

[0020] Preferably, the first image enhancement layer performs a first enhancement process on the plantar wart area image, including: global contrast enhancement, detail sharpening and denoising; the global contrast enhancement uses CLAHE to adjust the contrast in the plantar wart area to obtain a contrast enhanced plantar wart area; the detail sharpening uses a Laplacian operator to perform high-frequency enhancement on the contrast enhanced plantar wart area to obtain a detail sharpened plantar wart area; the denoising process denoises the detail sharpened plantar wart area through a non-local mean filtering technique to obtain the first enhanced plantar wart image.

[0021] Preferably, the pressure analysis layer performs pressure analysis on the first enhanced plantar wart image, and the steps of the pressure analysis are: using the Sobel operator to calculate the local extraction information of the first enhanced plantar wart image to obtain the gradient amplitude and direction; calculating the gradient change rate, and extracting the area where the gradient change rate is greater than a preset gradient change rate threshold as a candidate high-pressure area; combining the gradient amplitude and texture features to calculate the pressure value; the calculation formula of the pressure value is:

[0022]

[0023] Where P(x,y) is the pressure value of the pixel (x,y); σ is the activation function; ω 1 is the gradient amplitude weight; is the gradient amplitude; α is the gradient amplitude adjustment parameter; ω 2 is the texture feature weight; β 1 is the contrast weight; C is the contrast; β 2 is the correlation weight; Corr is the correlation; β 3 is the entropy weight; E t is entropy; 4 is the uniformity weight; U is the uniformity;

[0024] A pressure threshold is set according to the pressure value, and an area where the pressure value is greater than the pressure threshold is marked as the high-pressure area.

[0025] Preferably, the second image enhancement layer performs a second enhancement process on the image of the high-pressure area, including: a dynamic enhancement parameter layer and a high-pressure area dynamic enhancement layer;

[0026] The dynamic enhancement parameter layer calculates dynamic enhancement parameters for the high-pressure area; the dynamic enhancement parameters include: contrast enhancement factor, multi-scale feature enhancement weight and sharpening intensity coefficient;

[0027] The high-pressure area dynamic enhancement layer performs the second enhancement processing on the high-pressure area by using the dynamic enhancement parameters to obtain a second enhanced plantar wart image.

[0028] Preferably, the step of repairing the depth information of the second enhanced plantar wart image to obtain a repaired plantar wart depth map comprises: a multi-view depth map generating unit and a plantar wart region image repairing unit;

[0029] The multi-view depth map generating unit comprises: calculating the disparity between the multi-view plantar wart images of the first view, the second view and the third view; obtaining a disparity map based on minimizing a cost function; converting the disparity map into a depth map, and aligning the coordinate systems of the depth maps of the first view, the second view and the third view; fusing the depth maps of the first view, the second view and the third view to obtain a multi-view depth map;

[0030] The plantar wart area image restoration unit comprises: an image input layer, a depth map hole detection layer, a hole completion layer and a restored plantar wart depth map output layer; the image input layer inputs the second enhanced plantar wart image and the multi-view depth map into the depth map hole detection layer; the depth map hole detection layer detects missing data in the multi-view depth map through a hole detection algorithm to generate a hole mask; the hole completion layer uses the gradient information of the second enhanced plantar wart image through an interpolation algorithm to complete the hole area in the multi-view depth map; the restored plantar wart depth map output layer repairs the multi-view depth map after hole completion through edge repair and illumination correction to obtain a restored plantar wart depth map.

[0031] Preferably, the plantar wart three-dimensional morphological reconstruction model comprises: an input layer, a camera parameter calibration layer, a point cloud generation and fusion layer, a three-dimensional mesh reconstruction layer, a multi-view image fusion layer and a plantar wart three-dimensional morphological map output layer;

[0032] The input layer inputs the multi-view plantar wart image and the repaired plantar wart depth map into the plantar wart three-dimensional morphological reconstruction model;

[0033] The camera parameter calibration layer uses a chessboard diagram to calibrate the camera's internal and external parameters, and performs coordinate alignment on the multi-view plantar wart images;

[0034] The point cloud generation and fusion layer generates a point cloud through the repaired plantar wart depth map and the camera internal parameters; the multi-view point clouds are mapped to the same coordinate system and merged to obtain a global point cloud;

[0035] The three-dimensional mesh reconstruction layer removes isolated points in the global point cloud by using a SOR algorithm to generate a closed mesh model;

[0036] The multi-view image fusion layer performs weighted fusion on the multi-view images to obtain a fused plantar wart image; the fused plantar wart image is mapped onto the surface of the closed grid model to generate the plantar wart three-dimensional morphological map;

[0037] The plantar wart three-dimensional morphology image output layer outputs the plantar wart three-dimensional morphology image for subsequent diagnosis.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] 1. The present invention proposes an image enhancement model to perform multi-region enhancement processing on the plantar wart area image, focusing on highlighting the key plantar wart morphological features in the high-pressure area, such as edge clarity, texture details and contrast. In the first enhancement stage, the global morphology is preliminarily optimized through the first image enhancement layer to improve the overall visibility; the high-pressure area of ​​the plantar wart area is accurately calibrated through the pressure analysis layer to facilitate subsequent local enhancement of the high-pressure area; in the second enhancement stage, the pressure value and high-pressure area are annotated, and targeted enhancement is implemented for the high-pressure area through the second image enhancement layer to enhance the contrast and texture details of the key areas, while smoothing and protecting the low-pressure area to avoid artifacts introduced by excessive enhancement. This dynamic enhancement strategy can effectively cope with the diversity and complexity of plantar wart morphology, significantly reduce the blur or artifact problems in the lesion area, highlight the detailed information of the plantar wart lesion area, and enable the subsequent three-dimensional morphological reconstruction model to more accurately capture the spatial characteristics and shape changes of the high-pressure area, and generate a more accurate three-dimensional morphological map of plantar warts.

[0040] 2. The present invention proposes a method for repairing a plantar wart depth map based on multi-perspective depth map generation and hole detection and completion, which can obtain a complete repaired plantar wart depth map. In the generation of the multi-perspective depth map, the disparity between the images of the first, second and third perspectives is used to calculate the disparity map through the cost function minimization algorithm, which overcomes the parallax blind area problem existing in the single-perspective depth map and generates a preliminary multi-perspective depth map. In the image repair of the plantar wart area, the missing data area in the multi-perspective depth map is located by the hole detection algorithm, a hole mask is generated, and the hole area of ​​the depth map is accurately completed by the interpolation algorithm, which solves the impact of missing data on the integrity of depth information and detail restoration; the smoothness and lighting consistency of the depth map are further optimized through edge repair and lighting correction, so that the repaired depth map is closer to the depth distribution of the real plantar wart morphology. This depth map repair method effectively solves the problem of missing depth information and ensures that the three-dimensional morphological reconstruction model can accurately reconstruct the three-dimensional morphology of plantar warts based on high-quality depth data.

[0041] 3. The present invention proposes a plantar wart three-dimensional morphological reconstruction model that combines multi-view images and repaired depth maps to generate a high-precision plantar wart three-dimensional morphological map. The reconstruction distortion problem caused by the perspective error is reduced through high-precision camera parameter calibration and multi-view data calibration, and the geometric accuracy of the plantar wart three-dimensional morphological map is improved; the depth information of multiple perspectives is integrated through global point cloud generation and processing, and the integrity of the morphology of the plantar wart area is improved. At the same time, the SOR algorithm is used to effectively filter noise data, further improving the quality of the point cloud data, and laying a solid foundation for three-dimensional grid generation; the realism of the plantar wart three-dimensional morphological map is enhanced through closed grid models and texture mapping, solving the hole problem in traditional point cloud reconstruction. The plantar wart three-dimensional morphological reconstruction model can not only accurately reflect the geometric structure of the plantar wart area, but also enhance its realism and precision, making the final plantar wart three-dimensional morphological map more accurate, and can provide high-precision three-dimensional data support for pathological analysis, treatment plan design, and disease progression tracking. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 A structural diagram of a multiple plantar wart morphology data acquisition system based on a machine vision system provided by an embodiment of the present invention;

[0043] Figure 2 A flow chart of multiple plantar wart morphological data collection provided by an embodiment of the present invention;

[0044] Figure 3 A schematic diagram of an image enhancement model provided by an embodiment of the present invention;

[0045] Figure 4 This is a structural diagram of a three-dimensional morphological reconstruction model of plantar warts provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0046] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of 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.

[0047] Multiple plantar warts are skin lesions caused by human papillomavirus infection and are commonly found in pressure-bearing areas such as the soles of the feet. Due to long-term weight pressure, multiple plantar warts usually grow inward, showing characteristics such as rough surface and deep tissue invasion. The lesion morphology is complex, which poses challenges to traditional diagnosis and treatment. Traditional two-dimensional imaging methods cannot accurately capture the depth information of multiple plantar warts, and diagnosis often relies on the doctor's experience and lacks objective quantitative data support. Therefore, a new multiple plantar wart morphology data acquisition technology is urgently needed to overcome the limitations of traditional two-dimensional imaging methods, thereby improving the accurate identification and monitoring of multiple plantar wart lesions.

[0048] The present invention proposes a multiple plantar wart morphology data acquisition system based on a machine vision system, which realizes the generation of accurate multiple plantar wart three-dimensional morphology images. For a specific system structure diagram, please refer to Figure 1 In order to illustrate that the system of the present invention can accurately generate a three-dimensional morphological image of multiple plantar warts, the effectiveness of the present invention will be described below using two embodiments.

[0049] Embodiment 1

[0050] In the embodiment of the present application, the system proposed by the present invention is used to describe in detail the process of generating a three-dimensional morphological map of multiple plantar warts. In the embodiment of the present application, the generation of a three-dimensional morphological map of multiple plantar warts is aimed at generating a three-dimensional morphological map of multiple plantar warts on the soles of the feet of patient A in the dermatology department of a tertiary hospital. Figure 1 and Figure 2 , Figure 1 The specific structural diagram of the system proposed by the present invention is as follows: Figure 2 The invention is a flow chart for collecting morphological data of multiple plantar warts, comprising: a plantar wart image collection module, a plantar wart region recognition module, an image enhancement module, an image repair module and a plantar wart three-dimensional morphological reconstruction module; the plantar wart image collection module captures two-dimensional image data of plantar warts from three different viewing angles to obtain multi-view plantar wart images; the plantar wart region recognition module identifies plantar wart regions in the multi-view plantar wart images and analyzes them to obtain key morphological features; the image enhancement module constructs an image enhancement model to perform multi-region enhancement processing on the plantar wart region images to obtain a second enhanced plantar wart image; the image repair module repairs the depth information of the second enhanced plantar wart image to obtain a repaired plantar wart depth map; the plantar wart three-dimensional morphological reconstruction module generates a plantar wart three-dimensional morphological map according to the repaired plantar wart image and the multi-view plantar wart image.

[0051] The following is based on Figure 1 The content of this article describes in detail the process of generating the three-dimensional morphological image of multiple plantar warts on the sole of patient A:

[0052] A multiple plantar wart morphology data acquisition system based on a machine vision system, comprising:

[0053] The plantar wart image acquisition module is used to capture two-dimensional image data of the plantar wart through a first viewing angle, a second viewing angle, and a third viewing angle to obtain a multi-view plantar wart image;

[0054] Specifically, a high-resolution camera with more than 5000 pixels is used to capture two-dimensional global images of plantar warts from a first perspective, a second perspective, and a third perspective respectively; the first perspective is a perspective directly below the sole of the foot perpendicular to the sole of the foot; the second perspective is a perspective below the sole of the foot at an angle of 30° to the surface of the sole of the foot; and the third perspective is a perspective below the sole of the foot at an angle of 60° to the surface of the sole of the foot.

[0055] Preferably, after the multi-view plantar wart image is acquired by the plantar wart image acquisition module, the multi-view plantar wart image is preprocessed; the preprocessing includes: color space conversion and denoising;

[0056] The color space conversion converts the RGB color space of the multi-view plantar wart image into the HSV color space;

[0057] The denoising method removes the noise generated when the multi-view plantar wart images are collected by a multi-scale denoising algorithm.

[0058] In an embodiment of the present application, by converting the RGB color space into the HSV color space, the hue, saturation and brightness of the image are separated, making it easier to extract color and morphological features in the subsequent processing process; the multi-scale denoising algorithm is used to remove interference caused by noise, motion blur or equipment limitations in the image during the shooting process, making the image clearer and more realistic. The image after color space conversion and denoising is more suitable for subsequent morphological feature recognition, multi-region enhancement and three-dimensional modeling. These preprocessing measures not only improve the image quality, but also provide high-quality input data for subsequent image processing and analysis (such as point cloud generation and three-dimensional morphological reconstruction).

[0059] Preferably, the plantar wart region recognition module is used to identify the plantar wart region in the multi-view plantar wart image by constructing a plantar wart region recognition model to obtain a plantar wart region image; and analyze the plantar wart region image to obtain key morphological features of the plantar wart region;

[0060] The plantar wart region recognition model comprises: a plantar wart region segmentation layer, a segmentation result optimization layer and a key morphological feature extraction layer;

[0061] The plantar wart region segmentation layer uses the Faster R-CNN model as the main segmentation model to perform plantar wart region recognition on the multi-view plantar wart image. The plantar wart region recognition step is: using a labeling tool to label the plantar wart category of the training set of the multi-view plantar wart image to obtain the labeled training set; the plantar wart categories include: healthy skin, mild plantar warts and fused plantar warts; performing random enhancement processing on the labeled training set through an enhancement strategy to obtain an enhanced training set; training the Faster R-CNN model on the enhanced training set by using a cross entropy loss function and an IOU loss function; using the trained Faster R-CNN model to perform plantar wart region recognition on the test set of the multi-view plantar wart image to generate a preliminary segmentation result of the plantar wart region;

[0062] The segmentation result optimization layer includes: an edge optimization layer and a region filling layer; wherein the edge optimization layer optimizes the edge of the preliminary segmentation result of the plantar wart region by using the Canny edge detection algorithm; the optimization steps are: graying the preliminary segmentation result to obtain a preliminary segmentation result grayscale image; setting high and low thresholds to extract edge information of the preliminary segmentation result grayscale image; merging the preliminary segmentation result and the edge information to obtain an accurate segmentation result of the plantar wart region; the region filling layer fills the holes and cracks in the accurate segmentation result by using morphological operations to obtain an image of the plantar wart region;

[0063] The key morphological feature extraction layer extracts key morphological features of the plantar wart region image; the key morphological features include: texture features and edge features; wherein the texture features are obtained by performing texture analysis on the plantar wart region image using a gray level co-occurrence matrix; the texture features include: contrast, correlation, entropy and uniformity;

[0064] The edge features are obtained by analyzing the edge information; the edge features include: the area of ​​the plantar wart region and the border length.

[0065] In an embodiment of the present application, by constructing a plantar wart region recognition model and using Faster R-CNN as the main segmentation model, combined with an enhancement strategy, a cross entropy loss function, and an IoU loss function, accurate recognition of the plantar wart region is effectively achieved. The segmentation result optimization layer optimizes the segmentation result through Canny edge detection and morphological region filling, ensuring clear boundaries and complete regional images. The key morphological feature extraction layer further analyzes the texture features and edge features of the plantar wart region, including contrast, correlation, entropy, uniformity, area, and boundary length. The model effectively improves the recognition accuracy, detail extraction, and lesion analysis capabilities of the plantar wart region, and provides accurate plantar wart region images and reliable morphological features for subsequent plantar wart three-dimensional morphological modeling.

[0066] Preferably, the image enhancement module is used to construct an image enhancement model to perform multi-region enhancement processing on the plantar wart region image, referring to Figure 3 The schematic diagram of the image enhancement model shown in the figure; the image enhancement model comprises: a first image enhancement layer, a pressure analysis layer, and a second image enhancement layer; the step of processing the plantar wart area image comprises: performing a first enhancement processing on the plantar wart area image through the first image enhancement layer to obtain a first enhanced plantar wart image;

[0067] The first image enhancement layer performs a first enhancement process on the plantar wart area image, including: global contrast enhancement, detail sharpening and denoising; the global contrast enhancement uses CLAHE to adjust the contrast in the plantar wart area to obtain a contrast enhanced plantar wart area; the detail sharpening uses the Laplacian operator to perform high frequency enhancement on the contrast enhanced plantar wart area to obtain a detail sharpened plantar wart area; the denoising process denoises the detail sharpened plantar wart area through a non-local mean filtering technique to obtain the first enhanced plantar wart image.

[0068] The embodiment of the present application proposes a first image enhancement layer to perform a first enhancement process on the plantar wart area image, and significantly improves the quality of the plantar wart area image through global contrast enhancement, detail sharpening and denoising. The CLAHE method enhances the contrast of the image and highlights the details; the Laplacian operator is used for high-frequency enhancement to make the fine textures in the image clearer; the non-local mean filtering technology effectively removes noise and maintains the smoothness and clarity of the image. These processes improve the visualization of the image and provide a high-quality globally enhanced plantar wart image for subsequent pressure analysis and local enhancement of the high-pressure area.

[0069] Preferably, the pressure analysis layer performs pressure analysis on the first enhanced plantar wart image according to the key morphological features to obtain a pressure value; and an area where the pressure value is greater than a pressure threshold is marked as a high-pressure area;

[0070] The pressure analysis layer performs pressure analysis on the first enhanced plantar wart image, and the steps of the pressure analysis are: using the Sobel operator to calculate the local extraction information of the first enhanced plantar wart image to obtain the gradient amplitude and direction; calculating the gradient change rate, and extracting the area where the gradient change rate is greater than the preset gradient change rate threshold as the candidate high-pressure area; combining the gradient amplitude and texture features to calculate the pressure value; the calculation formula of the pressure value is:

[0071]

[0072] Where P(x,y) is the pressure value of the pixel (x,y); σ is the activation function; ω 1 is the gradient amplitude weight; is the gradient amplitude; α is the gradient amplitude adjustment parameter; ω 2 is the texture feature weight; β 1 is the contrast weight; C is the contrast; β 2 is the correlation weight; Corr is the correlation; β 3 is the entropy weight; E t is entropy; 4 is the uniformity weight; U is the uniformity;

[0073] A pressure threshold is set according to the pressure value, and an area where the pressure value is greater than the pressure threshold is marked as the high-pressure area.

[0074] The embodiment of the present application constructs a pressure analysis layer to perform pressure analysis on the first enhanced plantar wart image and mark the high-pressure area. The gradient amplitude and direction of the image are extracted by the Sobel operator, the gradient change rate is calculated, the possible high-pressure area is identified, and the pressure value is calculated by combining the gradient amplitude and texture features (such as contrast, correlation, entropy and uniformity). By setting a pressure threshold, the area with a pressure value greater than the threshold is marked as a high-pressure area. This model effectively identifies and analyzes the possible high-pressure areas in the plantar wart area, and provides an accurate image enhancement area range for the subsequent local enhancement of the high-pressure area through the second image enhancement layer.

[0075] Preferably, the second image enhancement layer performs a second enhancement process on the image of the high-pressure area to obtain a second enhanced plantar wart image; the second image enhancement layer performs a second enhancement process on the image of the high-pressure area, including: a dynamic enhancement parameter layer and a high-pressure area dynamic enhancement layer;

[0076] The dynamic enhancement parameter layer calculates dynamic enhancement parameters for the high-pressure area; the dynamic enhancement parameters include: contrast enhancement factor, multi-scale feature enhancement weight and sharpening intensity coefficient;

[0077] The contrast enhancement factor is dynamically set according to the local contrast distribution in the high-voltage region, and the formula is:

[0078]

[0079] Among them, σ L is the local contrast; N is the number of pixels in the high-voltage area; I(x,y) is the brightness of the pixel (x,y); μ L is the local average brightness; C is the contrast enhancement factor; C 1 is the initial contrast enhancement factor; λ is the contrast enhancement adjustment coefficient;

[0080] The multi-scale feature enhancement weight adjusts the Gaussian weights of different scales according to the gradient amplitude distribution in the high-pressure area, and the formula is:

[0081]

[0082] Among them, ω s Enhance weights for multi-scale features; is the gradient amplitude at the sth scale; S is the total number of scales;

[0083] The sharpening strength coefficient is dynamically adjusted by calculating the mean and standard deviation of the edge response strength, and the formula is:

[0084]

[0085] k=k 1 +η·σ E ;

[0086] Among them, μ E is the mean value of edge response intensity; σ E is the standard deviation of edge response intensity; ΔI is the edge response intensity; k is the sharpening intensity coefficient; k 1 is the initial sharpening strength coefficient; η is the sharpening strength adjustment coefficient;

[0087] The high-pressure area dynamic enhancement layer performs the second enhancement processing on the high-pressure area by using the dynamic enhancement parameters to obtain a second enhanced plantar wart image; the second enhancement processing includes: contrast enhancement, detail enhancement and edge sharpening;

[0088] The formula for contrast enhancement is:

[0089] I enhanced (x,y)=CLAHE(I(x,y),C);

[0090] Among them, I enhanced is contrast enhancement; CLAHE() is the CLAHE algorithm;

[0091] The formula for detail enhancement is:

[0092]

[0093] Among them, R enhanced For detail enhancement; G s is a Gaussian convolution kernel with a scale of s;

[0094] The formula for edge sharpening is:

[0095] I enhanced (x,y)=I(x,y)+k·[I(x,y)-(I(x,y)*G σ )];

[0096] Among them, Ienhanced G is edge sharpening; σ is the Gaussian blur kernel.

[0097] Table 1 shows a comparison of plantar wart repair results obtained by the method of introducing a pressure analysis layer to mark the high-pressure area and dynamically enhancing the high-pressure area and the traditional method.

[0098] Table 1 Comparison of plantar wart repair results based on different methods

[0099] Evaluation Metrics Traditional methods Introducing non-dynamic enhancement of pressure analysis layer Introducing dynamic enhancement of pressure analysis layer Depth information (mm) 2.8 3.0 3.2 Image clarity (%) 67.35 80.42 86.13 Edge accuracy (%) 72.39 82.52 89.68 Detail recovery generally improve Significantly improved

[0100] The embodiment of the present application adopts the idea of ​​allocating rendering resources according to the importance of scene elements in Omniverse, and focuses on enhancing the high-pressure areas in the plantar wart area image marked in the pressure analysis layer through the second image enhancement layer. The dynamic enhancement parameter layer calculates adaptive enhancement parameters, such as contrast enhancement factor, feature enhancement weight and sharpening intensity coefficient, so as to improve the detail clarity and visual performance of the high-pressure area in a targeted manner. The high-pressure area dynamic enhancement layer uses these parameters to optimize the image, making the high-pressure area more prominent, improving the contrast and details of the image, and providing more detailed and clear data support for the subsequent repair of the plantar wart depth map and the reconstruction of the three-dimensional plantar wart morphology through depth information.

[0101] Preferably, the image restoration module is used to restore the depth information of the second enhanced plantar wart image to obtain a restored plantar wart depth map;

[0102] The step of repairing the depth information of the second enhanced plantar wart image to obtain a repaired plantar wart depth map comprises: a multi-view depth map generating unit and a plantar wart region image repairing unit;

[0103] The multi-view depth map generating unit comprises: calculating the disparity between the multi-view plantar wart images of the first view, the second view and the third view; obtaining a disparity map based on minimizing a cost function; converting the disparity map into a depth map, and aligning the coordinate systems of the depth maps of the first view, the second view and the third view; fusing the depth maps of the first view, the second view and the third view to obtain a multi-view depth map;

[0104] The plantar wart area image restoration unit comprises: an image input layer, a depth map hole detection layer, a hole completion layer and a restored plantar wart depth map output layer; the image input layer inputs the second enhanced plantar wart image and the multi-view depth map into the depth map hole detection layer; the depth map hole detection layer detects missing data in the multi-view depth map through a hole detection algorithm to generate a hole mask; the hole completion layer uses the gradient information of the second enhanced plantar wart image through an interpolation algorithm to complete the hole area in the multi-view depth map; the restored plantar wart depth map output layer repairs the multi-view depth map after hole completion through edge repair and illumination correction to obtain a restored plantar wart depth map.

[0105] In the embodiment of the present application, the depth information of the second enhanced plantar wart image is restored through multi-view depth map generation, hole detection and completion, edge repair and illumination correction. The multi-view depth map generation unit provides comprehensive depth data through disparity calculation and depth map fusion, the hole completion layer fills the missing area in combination with gradient information, and edge repair and illumination correction ensure the smoothness and realism of the plantar wart area depth map. Finally, the repaired plantar wart area depth map provides accurate and clear data support for subsequent three-dimensional morphological reconstruction, significantly improving the quality of the three-dimensional plantar wart morphological map.

[0106] Preferably, the plantar wart three-dimensional morphology reconstruction module is used to construct a plantar wart three-dimensional morphology reconstruction model to generate a plantar wart three-dimensional morphology map according to the repaired plantar wart image and the multi-view plantar wart image. Figure 4 The plantar wart 3D morphology reconstruction model structure diagram shown; the plantar wart 3D morphology reconstruction model comprises: an input layer, a camera parameter calibration layer, a point cloud generation and fusion layer, a 3D mesh reconstruction layer, a multi-view image fusion layer and a plantar wart 3D morphology map output layer;

[0107] Among them, the input layer inputs the multi-view plantar wart image and the repaired plantar wart depth map into the plantar wart three-dimensional morphological reconstruction model; the camera parameter calibration layer uses a chessboard diagram to calibrate the camera's internal and external parameters, and aligns the coordinates of the multi-view plantar wart image; the point cloud generation and fusion layer generates a point cloud through the repaired plantar wart depth map and the camera internal parameters; the multi-view point cloud is mapped to the same coordinate system and merged to obtain a global point cloud; the three-dimensional grid reconstruction layer removes isolated points in the global point cloud through the SOR algorithm to generate a closed grid model; the multi-view image fusion layer performs weighted fusion on the multi-view images to obtain a fused plantar wart image; the fused plantar wart image is mapped to the surface of the closed grid model to generate the plantar wart three-dimensional morphological map; the plantar wart three-dimensional morphological map output layer outputs the plantar wart three-dimensional morphological map for subsequent diagnosis.

[0108] Table 2 evaluates the error of the generated plantar wart three-dimensional morphological image through error evaluation indicators, and the error evaluation indicators include: reconstruction error, depth error, surface accuracy error, morphological matching degree and standard deviation.

[0109] Table 2 Error evaluation of plantar wart three-dimensional morphology

[0110] Error evaluation index Error margin Reconstruction error (mm) 0.52±0.07 Depth error (mm) 0.1±0.05 Surface accuracy error (mm) 0.23±0.04 Morphological matching degree (%) 90.4 Standard deviation (mm) 2.85

[0111] The embodiment of the present application proposes a plantar wart three-dimensional morphological reconstruction model that generates a plantar wart three-dimensional morphological map through multi-view images and repaired plantar wart area depth information. The model can combine two-dimensional plantar wart images from different perspectives with the repaired plantar wart area depth map, and through multiple key levels of processing, including camera parameter calibration, point cloud generation and fusion, three-dimensional mesh reconstruction and multi-view image fusion, it realizes accurate three-dimensional morphological reconstruction of the plantar wart area. The SOR algorithm removes isolated points to ensure the closure of the generated three-dimensional mesh model, and the image fusion layer provides high-quality image mapping, so that the final generated three-dimensional morphological map not only has high-precision structural information, but also has a realistic visual effect. The plantar wart three-dimensional morphological reconstruction model greatly improves the generation capability and accuracy of the plantar wart three-dimensional morphological map of the system of the present invention, thereby providing more accurate and comprehensive three-dimensional reference data for medical analysis.

[0112] The multiple plantar wart morphological data acquisition system of the present invention realizes accurate plantar wart three-dimensional morphological reconstruction through multi-module collaboration: the system can obtain two-dimensional image data of plantar warts from multiple angles through the image acquisition modules of the first, second and third perspectives; this multi-perspective image acquisition method can fully display all angles of plantar warts, and provide multi-angle visual information for subsequent plantar wart morphological analysis. Through the plantar wart area recognition module, the plantar wart area in the image can be accurately identified, and the key morphological features of the area can be extracted; this process can provide key information such as the texture, boundary and area of ​​plantar warts, laying the foundation for subsequent multi-region analysis and processing of plantar wart area images. The image enhancement module improves the image quality through the first image enhancement layer, making the details of the plantar wart area clearer; the pressure analysis layer is used to analyze the key morphological features in the image, and the high-pressure area is marked according to the pressure value, and the image details of these high-pressure areas are further enhanced through the second image enhancement layer; this process effectively improves the clarity and details of the image, and provides higher quality input for subsequent deep repair and three-dimensional reconstruction. The image restoration module restores the depth information of the second enhanced image and generates a restored depth map of the plantar wart area; this process ensures the accuracy of the depth information in the reconstructed model and helps to generate a more accurate three-dimensional model. The plantar wart three-dimensional morphology reconstruction module constructs an accurate three-dimensional morphology reconstruction model based on the restored depth map and multi-view images. Through steps such as point cloud generation and fusion and three-dimensional mesh reconstruction, it can merge image information from different perspectives into a complete three-dimensional model to show the three-dimensional morphology of the plantar wart. The three-dimensional morphology map finally generated can intuitively display the three-dimensional structure and surface features of the plantar wart. The multiple plantar wart morphology data acquisition system of the present invention ensures the accuracy and authenticity of the three-dimensional model through a high-precision reconstruction process, thereby providing strong support for medical diagnosis and subsequent treatment.

[0113] Embodiment 2

[0114] In Example 1, the system proposed in the present invention realizes accurate reconstruction and generation of a three-dimensional morphological image of plantar warts. To further verify the effectiveness of the present invention, a three-dimensional reconstruction of a plantar wart morphological image of another patient B is also performed in the present embodiment.

[0115] A multiple plantar wart morphology data acquisition system based on a machine vision system, comprising:

[0116] The plantar wart image acquisition module is used to capture two-dimensional image data of plantar warts through a first viewing angle, a second viewing angle, and a third viewing angle to obtain multi-view plantar wart images.

[0117] Preferably, after the multi-view plantar wart image is acquired by the plantar wart image acquisition module, the multi-view plantar wart image is preprocessed; the preprocessing includes: color space conversion and denoising;

[0118] The color space conversion converts the RGB color space of the multi-view plantar wart image into the HSV color space;

[0119] The denoising method removes the noise generated when the multi-view plantar wart images are collected by a multi-scale denoising algorithm.

[0120] Preferably, the plantar wart region recognition module is used to identify the plantar wart region in the multi-view plantar wart image by constructing a plantar wart region recognition model to obtain a plantar wart region image; and analyze the plantar wart region image to obtain key morphological features of the plantar wart region;

[0121] The plantar wart region recognition model comprises: a plantar wart region segmentation layer, a segmentation result optimization layer and a key morphological feature extraction layer;

[0122] The plantar wart region segmentation layer uses the Faster R-CNN model as the main segmentation model to perform plantar wart region recognition on the multi-view plantar wart image. The plantar wart region recognition step is: using a labeling tool to label the plantar wart category of the training set of the multi-view plantar wart image to obtain the labeled training set; the plantar wart categories include: healthy skin, mild plantar warts and fused plantar warts; performing random enhancement processing on the labeled training set through an enhancement strategy to obtain an enhanced training set; training the Faster R-CNN model on the enhanced training set by using a cross entropy loss function and an IOU loss function; using the trained Faster R-CNN model to perform plantar wart region recognition on the test set of the multi-view plantar wart image to generate a preliminary segmentation result of the plantar wart region;

[0123] The segmentation result optimization layer includes: an edge optimization layer and a region filling layer; wherein the edge optimization layer optimizes the edge of the preliminary segmentation result of the plantar wart region by using the Canny edge detection algorithm; the optimization steps are: graying the preliminary segmentation result to obtain a preliminary segmentation result grayscale image; setting high and low thresholds to extract edge information of the preliminary segmentation result grayscale image; merging the preliminary segmentation result and the edge information to obtain an accurate segmentation result of the plantar wart region; the region filling layer fills the holes and cracks in the accurate segmentation result by using morphological operations to obtain an image of the plantar wart region;

[0124] The key morphological feature extraction layer extracts the key morphological features of the plantar wart area image; the key morphological features include: texture features and edge features; wherein the texture features are obtained by performing texture analysis on the plantar wart area image using a gray level co-occurrence matrix; the texture features include: contrast, correlation, entropy and uniformity; the edge features are obtained by analyzing the edge information; the edge features include: plantar wart area area and boundary length.

[0125] Preferably, the image enhancement module is used to construct an image enhancement model to perform multi-region enhancement processing on the plantar wart region image; the image enhancement model includes: a first image enhancement layer, a pressure analysis layer, and a second image enhancement layer; the step of processing the plantar wart region image is: performing a first enhancement processing on the plantar wart region image by using the first image enhancement model to obtain a first enhanced plantar wart image;

[0126] The first image enhancement layer performs a first enhancement process on the plantar wart area image, including: global contrast enhancement, detail sharpening and denoising; the global contrast enhancement uses CLAHE to adjust the contrast in the plantar wart area to obtain a contrast enhanced plantar wart area; the detail sharpening uses the Laplacian operator to perform high frequency enhancement on the contrast enhanced plantar wart area to obtain a detail sharpened plantar wart area; the denoising process denoises the detail sharpened plantar wart area through a non-local mean filtering technique to obtain the first enhanced plantar wart image.

[0127] Preferably, the pressure analysis layer performs pressure analysis on the first enhanced plantar wart image according to the key morphological features to obtain a pressure value; and an area where the pressure value is greater than a pressure threshold is marked as a high-pressure area;

[0128] The pressure analysis layer performs pressure analysis on the first enhanced plantar wart image, and the steps of the pressure analysis are: using the Sobel operator to calculate the local extraction information of the first enhanced plantar wart image to obtain the gradient amplitude and direction; calculating the gradient change rate, and extracting the area where the gradient change rate is greater than the preset gradient change rate threshold as the candidate high-pressure area; combining the gradient amplitude and texture features to calculate the pressure value; the calculation formula of the pressure value is:

[0129]

[0130] Where P(x,y) is the pressure value of the pixel (x,y); σ is the activation function; ω 1 is the gradient amplitude weight; is the gradient amplitude; α is the gradient amplitude adjustment parameter; ω 2 is the texture feature weight; β 1 is the contrast weight; C is the contrast; β 2 is the correlation weight; Corr is the correlation; β 3 is the entropy weight; E t is entropy; 4 is the uniformity weight; U is the uniformity;

[0131] A pressure threshold is set according to the pressure value, and an area where the pressure value is greater than the pressure threshold is marked as the high-pressure area.

[0132] Preferably, the second image enhancement layer performs a second enhancement process on the image of the high-pressure area to obtain a second enhanced plantar wart image; the second image enhancement layer performs a second enhancement process on the image of the high-pressure area, including: a dynamic enhancement parameter layer and a high-pressure area dynamic enhancement layer;

[0133] The dynamic enhancement parameter layer calculates dynamic enhancement parameters for the high-pressure area; the dynamic enhancement parameters include: contrast enhancement factor, multi-scale feature enhancement weight and sharpening intensity coefficient;

[0134] The contrast enhancement factor is dynamically set according to the local contrast distribution in the high-voltage region, and the formula is:

[0135]

[0136] Among them, σ L is the local contrast; N is the number of pixels in the high-voltage area; I(x,y) is the brightness of the pixel (x,y); μ L is the local average brightness; C is the contrast enhancement factor; C 1 is the initial contrast enhancement factor; λ is the contrast enhancement adjustment coefficient;

[0137] The multi-scale feature enhancement weight adjusts the Gaussian weights of different scales according to the gradient amplitude distribution in the high-pressure area, and the formula is:

[0138]

[0139] Among them, ω s Enhance weights for multi-scale features; is the gradient amplitude at the sth scale; S is the total number of scales;

[0140] The sharpening strength coefficient is dynamically adjusted by calculating the mean and standard deviation of the edge response strength, and the formula is:

[0141]

[0142]

[0143] k=k 1 +η·σ E ;

[0144] Among them, μ E is the mean value of edge response intensity; σ E is the standard deviation of edge response intensity; ΔI is the edge response intensity; k is the sharpening intensity coefficient; k 1 is the initial sharpening strength coefficient; η is the sharpening strength adjustment coefficient;

[0145] The high-pressure area dynamic enhancement layer performs the second enhancement processing on the high-pressure area by using the dynamic enhancement parameters to obtain a second enhanced plantar wart image; the second enhancement processing includes: contrast enhancement, detail enhancement and edge sharpening;

[0146] The formula for contrast enhancement is:

[0147] I enhanced (x,y)=CLAHE(I(x,y),C);

[0148] Among them, I enhanced is contrast enhancement; CLAHE() is the CLAHE algorithm;

[0149] The formula for detail enhancement is:

[0150]

[0151] Among them, R enhanced For detail enhancement; G s is a Gaussian convolution kernel with a scale of s;

[0152] The formula for edge sharpening is:

[0153] I enhanced (x,y)=I(x,y)+k·[I(x,y)-(I(x,y)*G σ )];

[0154] Among them, I enhanced G is edge sharpening; σ is the Gaussian blur kernel.

[0155] Preferably, the image restoration module is used to restore the depth information of the second enhanced plantar wart image to obtain a restored plantar wart depth map;

[0156] The step of repairing the depth information of the second enhanced plantar wart image to obtain a repaired plantar wart depth map comprises: a multi-view depth map generating unit and a plantar wart region image repairing unit;

[0157] The multi-view depth map generating unit comprises: calculating the disparity between the multi-view plantar wart images of the first view, the second view and the third view; obtaining a disparity map based on minimizing a cost function; converting the disparity map into a depth map, and aligning the coordinate systems of the depth maps of the first view, the second view and the third view; fusing the depth maps of the first view, the second view and the third view to obtain a multi-view depth map;

[0158] The plantar wart area image restoration unit comprises: an image input layer, a depth map hole detection layer, a hole completion layer and a restored plantar wart depth map output layer; the image input layer inputs the second enhanced plantar wart image and the multi-view depth map into the depth map hole detection layer; the depth map hole detection layer detects missing data in the multi-view depth map through a hole detection algorithm to generate a hole mask; the hole completion layer uses the gradient information of the second enhanced plantar wart image through an interpolation algorithm to complete the hole area in the multi-view depth map; the restored plantar wart depth map output layer repairs the multi-view depth map after hole completion through edge repair and illumination correction to obtain a restored plantar wart depth map.

[0159] Preferably, a plantar wart 3D morphology reconstruction module is used to construct a plantar wart 3D morphology reconstruction model to generate a plantar wart 3D morphology map according to the repaired plantar wart image and the multi-view plantar wart image; the plantar wart 3D morphology reconstruction model includes: an input layer, a camera parameter calibration layer, a point cloud generation and fusion layer, a 3D mesh reconstruction layer, a multi-view image fusion layer and a plantar wart 3D morphology map output layer;

[0160] Among them, the input layer inputs the multi-view plantar wart image and the repaired plantar wart depth map into the plantar wart three-dimensional morphological reconstruction model; the camera parameter calibration layer uses a chessboard diagram to calibrate the camera's internal and external parameters, and aligns the coordinates of the multi-view plantar wart image; the point cloud generation and fusion layer generates a point cloud through the repaired plantar wart depth map and the camera internal parameters; the multi-view point cloud is mapped to the same coordinate system and merged to obtain a global point cloud; the three-dimensional grid reconstruction layer removes isolated points in the global point cloud through the SOR algorithm to generate a closed grid model; the multi-view image fusion layer performs weighted fusion on the multi-view images to obtain a fused plantar wart image; the fused plantar wart image is mapped to the surface of the closed grid model to generate the plantar wart three-dimensional morphological map; the plantar wart three-dimensional morphological map output layer outputs the plantar wart three-dimensional morphological map for subsequent diagnosis.

[0161] Table 3 evaluates the error of the generated plantar wart three-dimensional morphology image through error evaluation indicators.

[0162] Table 3 Error evaluation of plantar wart three-dimensional morphology

[0163] Error evaluation index Error margin Reconstruction error (mm) 0.54±0.06 Depth error (mm) 0.13±0.04 Surface accuracy error (mm) 0.22±0.05 Morphological matching degree (%) 90.6 Standard deviation (mm) 2.78

[0164] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A multiple plantar wart morphology data acquisition system based on a machine vision system, characterized in that: include: The plantar wart image acquisition module is used to capture two-dimensional image data of the plantar wart through a first viewing angle, a second viewing angle, and a third viewing angle to obtain a multi-view plantar wart image; A plantar wart region recognition module is used to identify the plantar wart region in the multi-view plantar wart image by constructing a plantar wart region recognition model to obtain a plantar wart region image; and analyze the plantar wart region image to obtain key morphological features of the plantar wart region; An image enhancement module, used for constructing an image enhancement model to perform multi-region enhancement processing on the plantar wart region image; The image enhancement model comprises: a first image enhancement layer, a pressure analysis layer, and a second image enhancement layer; the steps of processing the plantar wart region image are: performing a first enhancement process on the plantar wart region image through the first image enhancement layer to obtain a first enhanced plantar wart image; performing a pressure analysis on the first enhanced plantar wart image according to the key morphological features through the pressure analysis layer to obtain a pressure value; marking a region where the pressure value is greater than a pressure threshold as a high-pressure region; performing a second enhancement process on the image of the high-pressure region through the second image enhancement layer to obtain a second enhanced plantar wart image; An image restoration module, used for restoring the depth information of the second enhanced plantar wart image to obtain a restored plantar wart depth map; The plantar wart three-dimensional morphology reconstruction module is used to construct a plantar wart three-dimensional morphology reconstruction model and generate a plantar wart three-dimensional morphology map according to the repaired plantar wart depth map and the multi-view plantar wart image.

2. The multiple plantar wart morphology data acquisition system based on a machine vision system according to claim 1, characterized in that: After the multi-view plantar wart image is acquired by the plantar wart image acquisition module, the multi-view plantar wart image is preprocessed; the preprocessing includes: color space conversion and denoising; The color space conversion converts the RGB color space of the multi-view plantar wart image into the HSV color space; The denoising method removes the noise generated when the multi-view plantar wart images are collected by a multi-scale denoising algorithm.

3. The multiple plantar wart morphology data acquisition system based on a machine vision system according to claim 1, characterized in that: The plantar wart region recognition model comprises: a plantar wart region segmentation layer, a segmentation result optimization layer and a key morphological feature extraction layer; The plantar wart region segmentation layer uses the Faster R-CNN model as the main segmentation model to perform plantar wart region recognition on the multi-view plantar wart image. The plantar wart region recognition step is: using a labeling tool to label the plantar wart category of the training set of the multi-view plantar wart image to obtain the labeled training set; the plantar wart categories include: healthy skin, mild plantar warts and fused plantar warts; performing random enhancement processing on the labeled training set through an enhancement strategy to obtain an enhanced training set; using the enhanced training set to train the Faster R-CNN model by using a cross entropy loss function and an IOU loss function; using the trained Faster R-CNN model to perform plantar wart region recognition on the test set of the multi-view plantar wart image to generate a preliminary segmentation result of the plantar wart region; The segmentation result optimization layer includes: an edge optimization layer and a region filling layer; wherein the edge optimization layer optimizes the edge of the preliminary segmentation result of the plantar wart region by using the Canny edge detection algorithm; the optimization steps are: graying the preliminary segmentation result to obtain a preliminary segmentation result grayscale image; setting high and low thresholds to extract edge information of the preliminary segmentation result grayscale image; merging the preliminary segmentation result and the edge information to obtain an accurate segmentation result of the plantar wart region; the region filling layer fills the holes and cracks in the accurate segmentation result by using morphological operations to obtain an image of the plantar wart region; The key morphological feature extraction layer extracts the key morphological features of the plantar wart area image; the key morphological features include: texture features and edge features; wherein the texture features are obtained by performing texture analysis on the plantar wart area image using a gray level co-occurrence matrix; the texture features include: contrast, correlation, entropy and uniformity; the edge features are obtained by analyzing the edge information; the edge features include: plantar wart area area and boundary length.

4. The multiple plantar wart morphology data acquisition system based on a machine vision system according to claim 1, characterized in that: The first image enhancement layer performs a first enhancement process on the plantar wart area image, including: global contrast enhancement, detail sharpening and denoising; the global contrast enhancement uses CLAHE to adjust the contrast in the plantar wart area to obtain a contrast enhanced plantar wart area; the detail sharpening uses the Laplacian operator to perform high frequency enhancement on the contrast enhanced plantar wart area to obtain a detail sharpened plantar wart area; the denoising process denoises the detail sharpened plantar wart area through a non-local mean filtering technique to obtain the first enhanced plantar wart image.

5. The multiple plantar wart morphology data acquisition system based on a machine vision system according to claim 1, characterized in that: The pressure analysis layer performs pressure analysis on the first enhanced plantar wart image, and the pressure analysis step comprises: using a Sobel operator to calculate local extraction information of the first enhanced plantar wart image to obtain a gradient amplitude and direction; Calculate the gradient change rate, and extract the area where the gradient change rate is greater than the preset gradient change rate threshold as the candidate high-pressure area; calculate the pressure value by combining the gradient amplitude and texture features; the calculation formula of the pressure value is: Among them, P(x,y) is the pressure value of the pixel (x,y); σ is the activation function; ω1 is the gradient amplitude weight; is the gradient amplitude; α is the gradient amplitude adjustment parameter; ω2 is the texture feature weight; β1 is contrast weight; C is contrast; β2 is correlation weight; Corr is correlation; β3 is entropy weight; E t is entropy; β4 is uniformity weight; U is uniformity; A pressure threshold is set according to the pressure value, and an area where the pressure value is greater than the pressure threshold is marked as the high-pressure area.

6. The multiple plantar wart morphology data acquisition system based on a machine vision system according to claim 1, characterized in that: The second image enhancement layer performs a second enhancement process on the image of the high-pressure area, including: a dynamic enhancement parameter layer and a high-pressure area dynamic enhancement layer; The dynamic enhancement parameter layer calculates dynamic enhancement parameters for the high-pressure area; the dynamic enhancement parameters include: contrast enhancement factor, multi-scale feature enhancement weight and sharpening intensity coefficient; The high-pressure area dynamic enhancement layer performs the second enhancement processing on the high-pressure area by using the dynamic enhancement parameters to obtain a second enhanced plantar wart image.

7. The multiple plantar wart morphology data acquisition system based on a machine vision system according to claim 1, characterized in that: The step of repairing the depth information of the second enhanced plantar wart image to obtain a repaired plantar wart depth map comprises: a multi-view depth map generating unit and a plantar wart region image repairing unit; The multi-view depth map generating unit comprises: calculating the disparity between the multi-view plantar wart images of the first view, the second view and the third view; obtaining a disparity map based on minimization of a cost function; converting the disparity map into a depth map; aligning and fusing the depth maps of the first view, the second view and the third view in coordinate systems to obtain a multi-view depth map; The plantar wart area image restoration unit comprises: an image input layer, a depth map hole detection layer, a hole completion layer and a restored plantar wart depth map output layer; the image input layer inputs the second enhanced plantar wart image and the multi-view depth map into the depth map hole detection layer; the depth map hole detection layer detects missing data in the multi-view depth map through a hole detection algorithm to generate a hole mask; the hole completion layer uses the gradient information of the second enhanced plantar wart image through an interpolation algorithm to complete the hole area in the multi-view depth map; the restored plantar wart depth map output layer repairs the multi-view depth map after hole completion through edge repair and illumination correction to obtain a restored plantar wart depth map.

8. The multiple plantar wart morphology data acquisition system based on a machine vision system according to claim 1, characterized in that: The plantar wart 3D morphology reconstruction model comprises: an input layer, a camera parameter calibration layer, a point cloud generation and fusion layer, a 3D mesh reconstruction layer, a multi-view image fusion layer and a plantar wart 3D morphology map output layer; The input layer inputs the multi-view plantar wart image and the repaired plantar wart depth map into the plantar wart three-dimensional morphological reconstruction model; The camera parameter calibration layer uses a chessboard diagram to calibrate the camera's internal and external parameters, and performs coordinate alignment on the multi-view plantar wart images; The point cloud generation and fusion layer generates a point cloud through the repaired plantar wart depth map and the camera internal parameters; the multi-view point clouds are mapped to the same coordinate system and merged to obtain a global point cloud; The three-dimensional mesh reconstruction layer removes isolated points in the global point cloud by using a SOR algorithm to generate a closed mesh model; The multi-view image fusion layer performs weighted fusion on the multi-view images to obtain a fused plantar wart image; the fused plantar wart image is mapped onto the surface of the closed grid model to generate the plantar wart three-dimensional morphological map; The plantar wart three-dimensional morphology image output layer outputs the plantar wart three-dimensional morphology image for subsequent diagnosis.

Images