A method and system for detecting oral plaque markers

Abstract

The application relates to the technical field of image detection, and particularly discloses a dental plaque marking detection method and system, wherein the method comprises the following steps: acquiring an oral cavity image to be marked; performing quality assessment on the oral cavity image to obtain a quality assessment result; selecting a feature extraction tool combination according to the quality assessment result, and processing the oral cavity image based on the feature extraction tool combination to obtain enhanced image features; identifying a plurality of dental plaque regions based on the enhanced image features; quantifying the confidence degree of each dental plaque region to obtain an uncertainty score; and marking the dental plaque regions in the oral cavity image according to each dental plaque region and the corresponding uncertainty score. The method overcomes the limitations in the prior art that lost details cannot be completely restored, artifacts can be introduced, and identification errors can be aggravated, and the practicability and consistency of the system in actual application are improved.

Description

Technical Field

[0001] This application relates to the field of image detection technology, and more specifically, to a method and system for detecting dental plaque markers in the oral cavity. Background Technology

[0002] In the field of personal health management, home oral care is gradually integrating smart technology, using image recognition to assist in plaque detection, thereby improving the convenience and preventative nature of daily cleaning. This trend aligns with the needs of digital health management, aiming to provide users with intuitive self-monitoring tools to help identify potential oral problems and optimize hygiene habits, thus driving the development of oral health management in the home setting towards automation and personalization.

[0003] However, in a home environment, the instability of handheld device operation by users leads to significant differences in distance, angle, and lighting in the acquired dental images, causing quality issues such as overexposure, shadows, geometric distortion, or blurring. These interfering factors mask the subtle visual features of dental plaque, especially the color and texture changes of early plaque, making it difficult for the recognition module to accurately extract useful information and increasing the risk of missed detections. Image quality issues further reduce the reliability of detection; users may overlook hidden areas, causing cleaning guidance to fail and preventing preventative care goals from being achieved. Although post-processing techniques attempt to correct the images, they cannot fully restore lost details and may even introduce artifacts, exacerbating recognition errors and limiting the system's practicality and consistency in complex home environments.

[0004] There is currently no effective technical solution to the above problems. Summary of the Invention

[0005] The purpose of this application is to provide a method and system for detecting dental plaque marking in the oral cavity, which aims to solve the technical problems in home oral care, such as inaccurate plaque identification, missed detection, and ineffective cleaning guidance caused by image quality issues.

[0006] In a first aspect, this application provides a method for detecting dental plaque markings, the method comprising the following steps:

[0007] S1. Obtain the oral cavity image to be labeled;

[0008] S2. Perform a quality assessment on the oral cavity image to obtain the quality assessment result;

[0009] S3. Based on the quality assessment results, select a combination of feature extraction tools and process the oral cavity image based on the combination of feature extraction tools to obtain enhanced image features;

[0010] S4. Based on the enhanced image features, several dental plaque regions are identified;

[0011] S5. Quantify the confidence level of each dental plaque region to obtain the uncertainty score;

[0012] S6. Mark the dental plaque regions in the oral cavity image according to each plaque region and the corresponding uncertainty score.

[0013] The oral plaque marker detection method, wherein the quality assessment results include global offset quality assessment results and local image degradation assessment results, step S2 includes:

[0014] S21. Perform global color balance analysis based on the average color temperature and color tendency of the oral cavity image to analyze whether there is a global color shift in the oral cavity image and obtain a global shift quality assessment result.

[0015] S22. Perform local image degradation assessment on the oral cavity image to identify local overexposure, shadows, blurring and geometric distortion problems, and obtain local image degradation assessment results.

[0016] The oral plaque marking detection method, wherein step S21 includes:

[0017] S211. Calculate the average brightness value of the RGB three channels of the oral cavity image, and obtain the average color temperature and color tendency of the oral cavity image according to the proportional relationship between the average brightness values ​​of the RGB three channels to obtain the global color characteristics.

[0018] S212. Based on the global color characteristics, determine whether there is a global chromaticity shift in the oral cavity image, and obtain the global shift quality assessment result.

[0019] The oral plaque marker detection method, wherein the localized image degradation assessment includes brightness distribution uniformity assessment, sharpness assessment, and geometric distortion assessment; step S22 includes:

[0020] S221. Divide the oral cavity image into multiple sub-region images based on a preset region division rule;

[0021] S222. Evaluate the uniformity of brightness distribution for each sub-region image. The evaluation of uniformity of brightness distribution includes calculating the brightness distribution parameters of each sub-region image and comparing the brightness distribution parameters with a preset brightness threshold to obtain the evaluation of uniformity of brightness distribution.

[0022] S223. A sharpness assessment is performed on each sub-region image. The sharpness assessment includes calculating the edge feature parameters of each sub-region image, comparing the edge feature parameters with a preset sharpness threshold, and obtaining a sharpness assessment result.

[0023] S224. Identify the tooth contours or gingival margins in the oral cavity image;

[0024] S225. Based on the identified tooth contour or gingival margin, perform a geometric distortion assessment on the oral image. The geometric distortion assessment includes calculating the geometric feature parameters of the tooth contour or gingival margin and comparing the geometric feature parameters with the geometric feature parameters of a standard tooth morphology to obtain a geometric distortion assessment result, which, together with the brightness distribution uniformity assessment and the sharpness assessment result, constitutes a local image degradation assessment result.

[0025] The oral plaque marking detection method, wherein step S3 includes:

[0026] S31. When the global offset quality assessment result indicates that the oral cavity image has a global chromaticity shift, the oral cavity image is globally chromaticity corrected using a preset global color correction tool to obtain a corrected oral cavity image. When the global offset quality assessment result indicates that the oral cavity image does not have a global chromaticity shift, the oral cavity image is regarded as the corrected oral cavity image.

[0027] S32. Based on the local image degradation assessment results, select a feature extraction tool combination and tool execution order from a preset feature extraction tool combination library. The feature extraction tool combination library includes multiple sets of feature extraction tool combinations and tool execution orders that match different local image degradation assessment results.

[0028] S33. The corrected oral cavity image is processed based on the extracted feature extraction tool combination and tool execution order to obtain the enhanced image features.

[0029] The oral plaque marking detection method, wherein step S4 includes:

[0030] S41. Extract local color features, local texture features, and local edge features of multiple local regions from the enhanced image features;

[0031] S42. Obtain the tooth outline or gingival margin in the oral cavity image;

[0032] S43. Based on the tooth contour or the gingival margin, exclude local areas that have a positional matching relationship with preset non-dental plaque structural features;

[0033] S44. Analyze the degree of matching between the local color features, local texture features, and local edge features of the remaining local area and the preset early dental plaque feature pattern to identify and determine the dental plaque area.

[0034] The oral plaque marking detection method, wherein step S44 includes:

[0035] S441. Combine the local color features, local texture features, and local edge features of each local region in the remaining local regions into a local feature vector;

[0036] S442. Calculate the feature distance between each local feature vector and the reference feature vector of the preset early dental plaque feature pattern;

[0037] S443. Compare the feature distance with a preset distance threshold to identify and determine the dental plaque region.

[0038] The oral plaque marking detection method, wherein step S5 includes:

[0039] S51. Calculate the color similarity score, texture similarity score, and edge similarity score between the local color features, local texture features, and local edge features of each dental plaque region and the reference color features, reference texture features, and reference edge features of the preset early dental plaque feature pattern.

[0040] S52. Analyze the consistency among the color similarity score, the texture similarity score, and the edge similarity score to obtain the feature consistency evaluation result;

[0041] S53. Obtain a comprehensive reliability score based on the feature distance and the preset distance threshold;

[0042] S54. Calculate the uncertainty score of the dental plaque region based on the characteristic consistency assessment results and the comprehensive reliability score.

[0043] The oral plaque marking detection method, wherein step S6 includes:

[0044] S61. Based on the uncertainty score corresponding to each dental plaque region, determine the visual attributes of the visual markers of each dental plaque region;

[0045] S62. Based on the visual attributes, add visual markers to each dental plaque region in the oral cavity image using a semi-transparent graphic overlay method. The visual markers include outline drawing or region filling.

[0046] Secondly, this application also provides an oral dental plaque marking detection system, the system comprising:

[0047] The acquisition module is used to acquire images of the oral cavity to be labeled;

[0048] The quality assessment module is used to assess the quality of the oral cavity images and obtain the quality assessment results.

[0049] The processing module is used to select a combination of feature extraction tools based on the quality assessment results, and process the oral cavity image based on the combination of feature extraction tools to obtain enhanced image features;

[0050] The recognition module is used to identify several dental plaque regions based on the enhanced image features;

[0051] The quantification module is used to quantify the confidence level of each dental plaque region and obtain an uncertainty score;

[0052] A labeling module is used to label the dental plaque regions in the oral cavity image according to each plaque region and its corresponding uncertainty score.

[0053] As described above, this application provides a method and system for detecting dental plaque marking. The method acquires an oral image to be marked, performs a quality assessment, and selects an appropriate feature extraction tool combination for processing based on the assessment results to obtain enhanced image features. Based on this, plaque regions are identified and their confidence levels are quantified. Finally, the plaque regions are marked according to their confidence levels. This method effectively solves the problems of inaccurate plaque recognition, missed detection, and ineffective cleaning guidance caused by unstable image quality in home environments (such as overexposure, shadows, blur, geometric distortion, etc.). By introducing an image quality assessment step, this application can specifically optimize the image processing flow, ensuring high-quality plaque features can be extracted even in complex and changing home environments, significantly improving the accuracy and reliability of recognition. Simultaneously, the quantification of the confidence level of plaque regions makes the marking results more instructive. Users can perform more precise oral cleaning based on the differences in marking intensity, thereby achieving effective preventative care. This overcomes the limitations of existing technologies, such as the inability to fully recover lost details and the potential introduction of artifacts that exacerbate recognition errors, improving the system's practicality and consistency in real-world applications. Attached Figure Description

[0054] Figure 1 A flowchart of the oral plaque marking detection method provided in the embodiments of this application.

[0055] Figure 2 This is a schematic diagram of the structure of the oral plaque marking detection system provided in the embodiments of this application.

[0056] Reference numerals: 201, Acquisition module; 202, Quality assessment module; 203, Processing module; 204, Identification module; 205, Quantification module; 206, Tagging module. Detailed Implementation

[0057] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0058] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0059] Firstly, please refer to Figure 1 This application provides a method for detecting dental plaque marking in the oral cavity, which includes the following steps:

[0060] S1. Obtain the oral cavity image to be labeled;

[0061] S2. Perform quality assessment on the oral cavity images and obtain the quality assessment results;

[0062] S3. Based on the quality assessment results, select a combination of feature extraction tools and process the oral cavity image based on the combination of feature extraction tools to obtain enhanced image features;

[0063] S4. Based on enhanced image features, several dental plaque regions were identified;

[0064] S5. Quantify the confidence level of each dental plaque region to obtain the uncertainty score;

[0065] S6. Mark the dental plaque regions in the oral cavity image according to each plaque region and the corresponding uncertainty score.

[0066] Specifically, this application provides a method for detecting dental plaque marking in the oral cavity, aiming to assist users in oral health management through automated and intelligent means. The core of this method lies in the refined processing and analysis of oral images to accurately identify and mark plaque areas.

[0067] More specifically, oral images can be photographs or video frames of the inside of the mouth captured by smartphones, tablets, or other handheld devices. These images typically include oral structures such as teeth, gums, and tongue, and serve as the foundational data for plaque detection. A plaque area refers to a specific set of pixels or image region in an oral image that is identified as containing plaque. These areas may exhibit different color, texture, or morphological characteristics compared to the surface of healthy teeth. This method is typically used in home or personal care settings, where users capture oral images themselves using a handheld device. After image acquisition, the system automatically processes and analyzes the images, providing the user with plaque marking results to guide targeted oral cleaning.

[0068] Specifically, step S2 aims to ensure that the image quality in subsequent processing meets the requirements. Image quality assessment can employ various techniques. For example, parameters such as overall brightness, contrast, and sharpness of the image can be calculated and compared with preset thresholds. In one implementation, histogram analysis can be used to determine whether overexposure or underexposure exists by analyzing the brightness distribution of image pixels. In another implementation, frequency domain analysis methods such as Fourier transform or wavelet transform can be used to assess the degree of blurring and loss of detail in the image. Step S3 is crucial for dynamically adjusting the feature extraction strategy based on image quality. For example, if the quality assessment results show that the image has a global color cast, a combination including color correction tools can be selected. If the image is locally blurred, a combination including image sharpening or super-resolution reconstruction tools can be selected. In one implementation, a feature extraction tool library can be preset, containing various image processing algorithms, such as Gaussian filtering, median filtering, edge detection operators (such as Sobel and Canny), color space conversion (such as RGB to HSV), and contrast enhancement (such as CLAHE). Based on the quality assessment results, the system intelligently selects one or more tools from the library and determines their execution order to form a customized feature extraction tool combination. Step S4, identifying dental plaque regions, can utilize various image recognition techniques. For example, a threshold-based segmentation method can be used to divide regions based on the difference in color or brightness between dental plaque and healthy teeth. In one implementation, a deep learning model (such as U-Net or Mask R-CNN) can be trained, which can learn the visual patterns of dental plaque from enhanced image features and directly output a pixel-level mask of the dental plaque region. In another implementation, traditional image processing methods, such as morphological operations and region growing, can be combined to optimize the initially segmented regions to improve recognition accuracy. Step S5 aims to provide a reliability measure for the recognition results. The quantification of confidence can be based on various factors. For example, it can be comprehensively evaluated based on the probability values ​​output by the recognition model, the matching degree between region features and typical dental plaque features, and the size and shape of the region. In one implementation, for deep learning models, the class probability output by the model can be directly used as the confidence score and converted into an uncertainty score (e.g., 1 minus the confidence probability value). In another implementation, a multi-feature fusion scoring system can be designed, comprehensively considering features such as color saturation, texture roughness, and edge sharpness of the plaque region, to calculate a comprehensive score representing its confidence level. Step S6 is the method of presenting the detection results to the user. The labeling method can be dynamically adjusted based on the uncertainty score. For example, regions with high confidence can be labeled with solid lines, while regions with low confidence can be labeled with dashed lines or semi-transparent lines. In one implementation, graphic markers can be directly overlaid on the oral cavity image, such as drawing the outline of the plaque region or filling the region with semi-transparent material.In another implementation, the color, transparency, or flashing frequency of the marker can be adjusted based on the uncertainty score. For example, areas with higher uncertainty scores can have lighter colors, higher transparency, or slower flashing frequencies to prompt the user that the recognition results for that area may require further attention.

[0069] More specifically, the oral plaque marking detection method of this application first acquires an oral image in step S1, and then performs a quality assessment on the image in step S2 to obtain a quality assessment result. This assessment result provides a key basis for subsequent processing. For example, if the image has a global color cast, the system will select a feature extraction tool combination that includes color correction function in step S3; if the image is locally blurred, it will select a tool that includes sharpening or deblurring function. This method of dynamically adjusting the feature extraction strategy according to the image quality ensures that high-quality enhanced image features can be obtained regardless of the quality of the original image. After obtaining the enhanced image features, step S4 identifies the plaque region based on these features, and step S5 quantifies the confidence level of each region to obtain an uncertainty score. This uncertainty score not only provides a measure of the reliability of the recognition result, but also provides personalized guidance for the final marking. In step S6, the plaque region in the oral image is marked according to the plaque region and the corresponding uncertainty score. For example, for plaque regions with low confidence, softer or more suggestive visual markings (such as semi-transparent, dashed lines, or specific colors) can be used to remind the user that the region may need further examination.

[0070] Compared with existing technologies, the core innovation of this application lies in its adaptive image processing workflow and the quantification and visualization of uncertainty in recognition results. Existing technologies typically employ fixed image processing workflows, which struggle to cope with the complex and varied image quality issues in a home environment, resulting in limited recognition accuracy and user experience. This application, by introducing an image quality assessment and a feature extraction tool selection mechanism based on the assessment results, can optimize processing according to the actual image conditions, thereby obtaining more accurate dental plaque features. Furthermore, the quantification and visualization of the confidence level of dental plaque areas allows users to more intuitively understand the reliability of the detection results and perform targeted oral cleaning based on the intensity of the prompts, effectively avoiding missed detections and misjudgments caused by image quality issues in traditional methods, significantly improving the efficiency and effectiveness of home oral care.

[0071] In some preferred embodiments, the quality assessment results include global offset quality assessment results and local image degradation assessment results, and step S2 includes:

[0072] S21. Perform global color balance analysis based on the average color temperature and color tendency of the oral cavity image to analyze whether there is a global color shift in the oral cavity image and obtain the global shift quality assessment result.

[0073] S22. Perform localized image degradation assessment on oral cavity images to identify localized overexposure, shadows, blurring, and geometric distortion problems, and obtain localized image degradation assessment results.

[0074] Specifically, the aforementioned global offset quality assessment results refer to the evaluation of the overall color balance of the oral cavity image, mainly used to determine whether the image has a general tonal deviation, such as being too yellow, too blue, or too red. This global color shift may be caused by improper camera white balance settings or abnormal ambient light color temperature. The aforementioned local image degradation assessment results focus on quality problems in local areas of the image, such as excessive brightness (local overexposure), insufficient brightness (shadows), blurred details (blur), or deformed tooth structure (geometric distortion) in specific areas of the image. These local problems may stem from shooting distance, inaccurate focus, camera shake, or lens distortion.

[0075] More specifically, step S21 performs a global color balance analysis on the average color temperature and color tendency of the oral cavity image, effectively detecting whether there is a global color shift in the image. The average color temperature reflects the overall warm or cool tone of the image, while the color tendency indicates the degree of deviation of the image in a specific color channel. By comprehensively analyzing these two parameters, it is possible to accurately determine whether there is a global color imbalance in the image. Step S22 performs a local image degradation assessment to identify potential problems such as local overexposure, shadows, blur, and geometric distortion in the image. Local overexposure and shadows can cause the loss of details in some areas of the image, affecting feature extraction; blur directly leads to unclear image details, making it difficult to accurately identify the texture and edges of dental plaque; geometric distortion may change the true shape of teeth and dental plaque, introducing recognition errors. By evaluating these local problems, more refined guidance can be provided for subsequent image enhancement and feature extraction.

[0076] This application's solution refines the quality assessment of oral images into global offset quality assessment and local image degradation assessment, enabling a more comprehensive and accurate capture of various quality issues present in the image. Specifically, global color balance analysis can effectively identify and quantify the overall color deviation of the image, avoiding inaccurate extraction of dental plaque color features due to global color cast. For example, if the image is generally yellowish, the yellow features of dental plaque may be masked or enhanced by the ambient hue, thus affecting the accuracy of recognition. By identifying global color offset, a basis can be provided for subsequent color correction, ensuring that the color features of dental plaque are accurately extracted in a standardized color space. Meanwhile, local image degradation assessment focuses on quality defects in local image regions, such as local overexposure, shadows, blurring, and geometric distortion. These local problems often directly affect the extraction of dental plaque morphology, texture, and edge features. For example, local blurring may result in unclear plaque edges, making it difficult to accurately delineate their contours; local overexposure or shadows may distort the brightness information of the dental plaque area, affecting the extraction of its texture features. By identifying these localized issues, guidance can be provided for selecting subsequent local image enhancement or feature extraction strategies for specific regions, effectively avoiding missed or false detections of dental plaque due to local quality problems. Therefore, the above scheme comprehensively evaluates oral image quality from both global and local dimensions, laying a solid foundation for subsequent image processing and dental plaque recognition.

[0077] Through the aforementioned technical solution, the method of this application enables refined and multi-dimensional evaluation of oral image quality. Compared to basic methods that only perform a single quality assessment, this application can more comprehensively identify and quantify global color shifts and local image degradation issues in oral images. This detailed quality assessment result allows for more targeted selection of subsequent feature extraction tools and image processing, thereby significantly improving the quality and accuracy of enhanced image features. Ultimately, this helps improve the robustness and accuracy of plaque region identification, effectively reducing the misdiagnosis and missed diagnosis rates caused by image quality issues, and providing more reliable technical support for oral health diagnosis.

[0078] In some preferred embodiments, step S21 includes:

[0079] S211. Calculate the average brightness value of the RGB three channels of the oral cavity image, and obtain the average color temperature and color tendency of the oral cavity image based on the proportional relationship between the average brightness values ​​of the RGB three channels to obtain the global color characteristics.

[0080] S212. Based on the global color characteristics, determine whether there is a global chromaticity shift in the oral cavity image, and obtain the global shift quality assessment result.

[0081] Specifically, step S211 aims to quantify the overall color representation of the oral cavity image. Specifically, the RGB three channels refer to the three color channels: Red, Green, and Blue, which are the most basic color components of a digital image. The average brightness value of each channel can be calculated by summing the corresponding channel values ​​of all pixels in the image and dividing by the total number of pixels. For example, for an 8-bit image, the brightness value of each channel ranges from 0 to 255. By comparing the proportional relationship between the average brightness values ​​of these three channels, the average color temperature and color tendency of the image can be inferred. This inference process is a common technique in image processing and will not be elaborated upon here. The average color temperature is usually expressed in Kelvin (K), reflecting the overall warm or cool tone of the image; the color tendency indicates that the image leans towards a certain color, such as reddish, bluish, or greenish. This information collectively constitutes the global color characteristics of the oral cavity image, providing basic data for subsequent color shift judgment. Step S212 uses the obtained global color characteristics to determine whether there is a global color shift in the oral cavity image. Global chromatic aberration refers to a situation where the overall color of an image deviates from the true or standard colors, such as the image appearing yellowish, bluish, or reddish. The determination process involves comparing the acquired global color characteristics with preset standard color characteristics (e.g., the color characteristics of a healthy oral cavity image taken under standard lighting conditions). When the difference between the two exceeds a preset threshold, a global chromatic aberration is determined to exist in the oral cavity image. This yields a global aberration quality assessment result, which indicates the image's quality in terms of color accuracy.

[0082] This application's solution refines global color balance analysis into calculating the average luminance values ​​of the RGB three channels to obtain global color characteristics, and based on this, determines whether there is a global chromaticity shift, thereby systematically achieving global color quality assessment of oral images. Specifically, step S211 quantifies the most basic color components of the image, revealing the image's inherent color attributes, namely average color temperature and color tendency, at the data level. These attributes are macroscopic indicators of image color performance, effectively reflecting whether the image was affected by inappropriate light sources or white balance settings during the shooting process. Subsequently, step S212 utilizes these quantified global color characteristics, comparing them with a standard reference, to objectively and accurately identify whether there is an overall color deviation in the image. This step-by-step analysis method makes global color balance analysis no longer a vague concept, but is achieved through quantifiable indicators and clear judgment logic, thereby ensuring the reliability and repeatability of the quality assessment results.

[0083] The above technical solution provides a specific, systematic, and quantifiable method for performing global color balance analysis of oral images. This method reliably determines whether a global chromaticity shift exists in an oral image, thereby obtaining accurate global shift quality assessment results. This not only improves the objectivity and accuracy of oral image quality assessment but also provides more reliable image input for subsequent image enhancement and plaque recognition, avoiding misjudgments or omissions caused by overall image color deviation, thus enhancing the robustness of the entire plaque marker detection method.

[0084] In some preferred embodiments, the localized image degradation assessment includes brightness distribution uniformity assessment, sharpness assessment, and geometric distortion assessment; step S22 includes:

[0085] S221. Divide the oral cavity image into multiple sub-region images based on a preset region division rule;

[0086] S222. Evaluate the uniformity of brightness distribution for each sub-region image. The evaluation of uniformity of brightness distribution includes calculating the brightness distribution parameters of each sub-region image and comparing the brightness distribution parameters with a preset brightness threshold to obtain the evaluation of uniformity of brightness distribution.

[0087] S223. Perform a sharpness assessment on each sub-region image. The sharpness assessment includes calculating the edge feature parameters of each sub-region image, comparing the edge feature parameters with a preset sharpness threshold, and obtaining the sharpness assessment result.

[0088] S224. Identify tooth outlines or gingival margins in oral images;

[0089] S225. Based on the identified tooth contour or gingival margin, perform geometric distortion assessment on the oral cavity image. The geometric distortion assessment includes calculating the geometric feature parameters of the tooth contour or gingival margin and comparing the geometric feature parameters with the geometric feature parameters of the standard tooth shape to obtain the geometric distortion assessment result, which, together with the brightness distribution uniformity assessment and sharpness assessment results, constitutes the local image degradation assessment result.

[0090] Specifically, the preset region segmentation rule can be understood as dividing the oral cavity image into several sub-region images of equal or unequal size, such as rectangles, circles, or irregular shapes. The purpose is to perform independent quality assessments on different local regions of the oral cavity image to more accurately identify local image degradation problems. For example, grid division or adaptive segmentation based on image content can be used.

[0091] More specifically, brightness distribution uniformity assessment aims to detect localized overexposure or shadowing issues. Brightness distribution parameters may include, but are not limited to, the average brightness of a sub-region image, the standard deviation of brightness, and the brightness histogram distribution. A preset brightness threshold is used to determine whether the brightness distribution is within an acceptable range. For example, when the average brightness of a sub-region image is too high or too low, or the standard deviation of brightness is too large, it can be considered that there is a problem with uneven brightness distribution.

[0092] More specifically, sharpness assessment is used to identify localized blur problems. Edge feature parameters can include gradient magnitude, Laplacian operator response, wavelet transform coefficients, etc., which reflect the detail and sharpness of the image. A preset sharpness threshold is used to distinguish between sharp and blurry regions. For example, when the edge feature parameters are below a certain threshold, it indicates that the image in that sub-region may be blurry.

[0093] More specifically, identifying tooth contours or gingival margins is fundamental to assessing geometric distortion. This can be achieved through image segmentation algorithms, deep learning models, or morphology-based image processing techniques.

[0094] More specifically, geometric distortion assessment aims to detect whether the shape of teeth or gums in an image is distorted. Geometric feature parameters can include the length, width, curvature, and angle of the teeth. By comparing these parameters with the geometric feature parameters of a standard tooth morphology, the degree of distortion can be quantified. For example, if the aspect ratio of the teeth deviates significantly from the standard value, geometric distortion may be present. Finally, the results of brightness distribution uniformity assessment, sharpness assessment, and geometric distortion assessment are combined to form a complete assessment of localized image degradation.

[0095] This application's solution refines the assessment of local image degradation into brightness distribution uniformity assessment, sharpness assessment, and geometric distortion assessment, and designs specific detection methods for each assessment dimension. First, by dividing the oral cavity image into multiple sub-region images, the detection of local problems becomes more refined. Second, by performing brightness distribution uniformity and sharpness assessments on each sub-region image, common image degradation problems such as local overexposure, shadows, and blurring can be effectively identified. Finally, by identifying tooth contours or gingival margins and performing geometric distortion assessment, potential structural distortions in the image can be discovered, thus comprehensively and accurately assessing the local quality of the oral cavity image, providing a reliable image foundation for subsequent feature extraction and plaque recognition.

[0096] In some preferred embodiments, step S3 includes:

[0097] S31. When the global offset quality assessment result indicates that there is a global chromaticity shift in the oral cavity image, the oral cavity image is globally chromaticity corrected using a preset global color correction tool to obtain a corrected oral cavity image. When the global offset quality assessment result indicates that there is no global chromaticity shift in the oral cavity image, the oral cavity image is regarded as the corrected oral cavity image.

[0098] S32. Based on the local image degradation assessment results, select a feature extraction tool combination and tool execution order from the preset feature extraction tool combination library. The feature extraction tool combination library includes multiple sets of feature extraction tool combinations and tool execution orders that match different local image degradation assessment results.

[0099] S33. Based on the extracted feature extraction tool combination and tool execution order, the corrected oral cavity image is processed to obtain enhanced image features.

[0100] Specifically, in step S31, the global color correction tool can be understood as an algorithm or module used to adjust the overall color balance of an image, aiming to eliminate global color shifts caused by factors such as lighting conditions and camera settings. For example, this tool can employ white balance algorithms, histogram equalization, or color enhancement algorithms based on Retinex theory. When the global offset quality assessment result indicates that the oral cavity image has a global color shift, such as an overall yellowish or bluish tint, the corresponding global color correction tool will be activated to process the oral cavity image to obtain a color-balanced corrected oral cavity image. Conversely, if the global color balance of the oral cavity image is good, no additional correction is required, and the original oral cavity image can be directly used as the corrected oral cavity image.

[0101] More specifically, in step S32, the feature extraction toolkit library is a pre-built collection containing various feature extraction toolkits designed for different local image degradation problems (such as local overexposure, shadows, blur, geometric distortion, etc.). Each toolkit may contain different image processing algorithms; for example, it may contain deblurring filters and edge enhancement algorithms for blurred areas, and local contrast enhancement algorithms and brightness adjustment tools for shadowed areas. The library also stores the preferred execution order of these toolkits to ensure optimal processing results. Based on the local image degradation evaluation results, the system intelligently selects the feature extraction toolkit and its execution order that best match the current image's local quality problem from the library.

[0102] More specifically, in step S33, the selected feature extraction tool set will process the globally color-corrected oral cavity image according to its preset execution order. For example, if the evaluation results show that the image has local blurring and shadows, the selected tool set may first apply a deblurring algorithm, then perform local contrast enhancement, and finally extract clearer and more recognizable enhanced image features.

[0103] The proposed solution first performs global color correction on the oral cavity image, eliminating overall color deviation and providing a standardized color foundation for subsequent feature extraction. Based on this, and according to the results of local image degradation assessment, it adaptively selects and applies the most suitable feature extraction tool combination for the current local quality problem, thereby specifically addressing issues such as local overexposure, shadows, blurring, or geometric distortion in the image. This phased, adaptive processing strategy ensures that regardless of the quality defects in the original oral cavity image, it can be effectively preprocessed and enhanced, resulting in higher quality and robustness of the final extracted enhanced image features.

[0104] In some embodiments, assuming an oral cavity image is acquired, its global offset quality assessment indicates a significant yellow chromaticity shift, while the local image degradation assessment shows blurring in the upper left corner and shadowing in the lower right corner. First, in step S31, the system invokes a preset global color correction tool, such as a white balance module based on a gray-world algorithm, to process the entire oral cavity image to eliminate the yellow chromaticity shift, resulting in a color-balanced corrected oral cavity image. Next, in step S32, based on the local image degradation assessment results, the system selects a tool combination from a feature extraction tool combination library that includes a deblurring filter and a local contrast enhancement algorithm. The execution order of this combination is set to perform deblurring first, followed by local contrast enhancement. Finally, in step S33, the selected tool combination processes the corrected oral cavity image in a predetermined order. Specifically, the deblurring filter is applied to the blurred area in the upper left corner of the image, while the local contrast enhancement algorithm is applied to the shadowed area in the lower right corner, ultimately resulting in enhanced image features with clear edges, good contrast, and accurate color. These features will be used for subsequent plaque recognition.

[0105] In some preferred embodiments, step S4 includes:

[0106] S41. Extract local color features, local texture features, and local edge features from multiple local regions of the enhanced image features;

[0107] S42. Obtain the tooth outline or gingival margin in the oral cavity image;

[0108] S43. Based on the tooth contour or gingival margin, exclude local areas that have a positional matching relationship with the preset non-dental plaque structural features;

[0109] S44. Analyze the degree of matching between the local color features, local texture features, and local edge features of the remaining local areas and the preset early dental plaque feature patterns to identify and determine the dental plaque areas.

[0110] Specifically, step S41 is preferably performed by dividing the enhanced image features into regions based on the aforementioned preset region division rules and then extracting local color features, local texture features and local edge features of multiple local regions.

[0111] More specifically, the local color features extracted from the local region in step S41 may include average RGB values, HSV values, Lab values, or color histograms, used to describe the color information of the region. Local texture features may include gray-level co-occurrence matrix (GLCM) parameters (such as contrast, energy, homogeneity), local binary mode (LBP) features, or Gabor filter responses, used to describe the surface roughness, regularity, and other texture information of the region. Local edge features may include edge intensity, direction, or density extracted by the Canny edge detector, Sobel operator, or Laplacian operator, used to describe the boundary information of the region.

[0112] More specifically, in step S42, obtaining the tooth contours or gingival edges in the oral cavity image can be achieved through various image segmentation techniques, such as deep learning-based semantic segmentation models (e.g., U-Net, Mask R-CNN), traditional image processing methods (e.g., active contour models, thresholding combined with morphological operations), or region growing algorithms incorporating prior knowledge. The aim is to provide accurate anatomical references for subsequent exclusion of non-plaque structures.

[0113] More specifically, in step S43, the preset non-plaque structural features can refer to common structures in the oral cavity other than dental plaque, such as healthy enamel surfaces, gingival tissue, tartar, food debris, or restorations. These structures typically have different color, texture, and morphological characteristics than dental plaque. The location matching relationship can be understood as being determined by calculating the spatial relationship between a local region and the tooth contour or gingival margin (e.g., whether it is located on the tooth surface, whether it overlaps with or is adjacent to the gingival margin) and its similarity to non-plaque structural features. For example, if a local region is outside the tooth contour, or its features highly match known tartar features, then that region will be excluded. The aim is to effectively reduce false identification and improve the specificity of dental plaque recognition.

[0114] More specifically, in step S44, the preset early dental plaque feature pattern refers to a reference pattern that represents the visual characteristics of early dental plaque, obtained through training and summarizing a large amount of clinical data. This may include a specific yellowish tint, slightly rough surface texture, irregular edge shape, etc. The matching degree analysis can be achieved by calculating the similarity or distance between the feature vector of the local region and the reference feature vector of the early dental plaque feature pattern, for example, using Euclidean distance, cosine similarity, or classifiers such as support vector machine (SVM) and random forest. The purpose is to accurately identify regions with early dental plaque features, even if their visual manifestation is not yet obvious.

[0115] This application's solution effectively addresses the accuracy and misidentification issues that may exist in basic plaque recognition methods by introducing multi-dimensional local feature extraction, combining anatomical structures for non-plaque region exclusion, and matching analysis of early plaque feature patterns. Specifically, step S41 extracts local color, texture, and edge features, providing richer and more detailed visual information for plaque recognition, enabling subsequent analysis to be based on more comprehensive data. Steps S42 and S43 allow the system to actively exclude local areas that have positional matching relationships with non-plaque structural features using important anatomical information such as tooth contours or gingival margins. This mechanism significantly reduces the possibility of misidentifying tartar, food debris, or healthy gums as plaque, thereby improving the specificity of recognition. Finally, step S44 matches the features of the remaining local areas with preset early plaque feature patterns, enabling more accurate identification of areas with early plaque characteristics, even if the visual appearance of these areas is subtle, thus improving the detection capability of early plaque. Specifically, by comprehensively analyzing local color, texture, and edge features, and combining this with tooth contours or gingival margins to exclude non-plaque areas, the method effectively avoids misidentifying other oral structures as plaque, significantly reducing the false positive rate. Furthermore, by matching with preset early plaque feature patterns, this application can more sensitively detect plaque in its early stages, which is of great significance for the early prevention and intervention of oral diseases. Compared to basic schemes that rely solely on enhanced image features for general recognition, this application's scheme significantly improves both recognition accuracy and specificity, providing a more reliable basis for clinical diagnosis.

[0116] In some preferred embodiments, step S44 includes:

[0117] S441. Combine the local color features, local texture features, and local edge features of each local region in the remaining local regions into a local feature vector;

[0118] S442. Calculate the feature distance between each local feature vector and the reference feature vector of the preset early dental plaque feature pattern;

[0119] S443. Compare the feature distance with the preset distance threshold to identify and determine the dental plaque area.

[0120] Specifically, in step S441, for each remaining local region after initial screening, the extracted local color features, local texture features, and local edge features are integrated into a unified local feature vector. Combining different types of features into a single local feature vector aims to comprehensively and multidimensionally characterize the visual properties of the local region, providing a rich data foundation for subsequent dental plaque recognition.

[0121] More specifically, in step S442, the formed local feature vector is compared with a preset reference feature vector of the early dental plaque feature pattern to calculate the feature distance between them. The preset reference feature vector of the early dental plaque feature pattern is obtained by training and learning from a large number of labeled early dental plaque images and represents a typical feature set of early dental plaque. Various metrics can be used to calculate the feature distance, such as Euclidean distance, cosine similarity, Mahalanobis distance, or Manhattan distance. Choosing an appropriate distance metric can effectively quantify the degree of similarity or difference between the region to be detected and the standard dental plaque pattern.

[0122] More specifically, in step S443, the calculated feature distance is compared with a preset distance threshold. This distance threshold, obtained based on empirical data or trained through a machine learning model, is used to distinguish between plaque regions and non-plaque regions. When the feature distance between the local feature vector and the reference feature vector is less than or equal to the distance threshold, it indicates that the features of the local region highly match the feature patterns of early plaque, and thus the local region is identified as a plaque region. Conversely, if the feature distance is greater than the threshold, the region is not considered a plaque region.

[0123] This application's solution integrates multiple visual features (color, texture, edges) of a local region into a unified local feature vector, and then quantitatively compares it with a reference feature vector of a preset early plaque feature pattern, thereby achieving plaque region identification. This method overcomes the limitations of single-feature recognition, as the visual appearance of plaque may vary due to individual differences, lighting conditions, or developmental stages. Through multi-feature fusion, the essential features of plaque can be captured more comprehensively and robustly. The calculation of feature distance provides an objective quantitative standard, avoiding the bias of subjective judgment. By comparing with a preset distance threshold, potential plaque regions can be effectively distinguished from the background, ensuring the accuracy and consistency of identification.

[0124] In some preferred embodiments, step S5 includes:

[0125] S51. Calculate the color similarity score, texture similarity score, and edge similarity score between the local color features, local texture features, and local edge features of each dental plaque region and the reference color features, reference texture features, and reference edge features of the preset early dental plaque feature pattern.

[0126] S52. Analyze the consistency among color similarity scores, texture similarity scores, and edge similarity scores to obtain feature consistency evaluation results;

[0127] S53. Obtain a comprehensive reliability score based on the feature distance and the preset distance threshold;

[0128] S54. Calculate the uncertainty score of the dental plaque area based on the characteristic consistency assessment results and the comprehensive reliability score.

[0129] Specifically, in step S51, for each identified dental plaque region, its local color features, local texture features, and local edge features are compared with the reference color features, reference texture features, and reference edge features corresponding to a preset early dental plaque feature pattern, respectively. Through this comparison, color similarity scores, texture similarity scores, and edge similarity scores can be calculated respectively. These scores aim to quantify the degree of matching between the dental plaque region and the standard early dental plaque pattern across different feature dimensions, providing multi-dimensional matching information. For example, the color similarity score can be obtained by calculating the intersection of color histograms or the reciprocal of the Euclidean distance; the texture similarity score can be calculated using the similarity measure of gray-level co-occurrence matrix (GLCM) or local binary pattern (LBP) features; and the edge similarity score can be determined by the matching degree of edge density, edge orientation histogram (HOG), or Canny edge detection results.

[0130] More specifically, in step S52, the consistency among the obtained color similarity scores, texture similarity scores, and edge similarity scores is analyzed to obtain a feature consistency evaluation result. Consistency analysis can be understood as assessing whether the degree of matching across these different feature dimensions is mutually supportive. For example, if a dental plaque area exhibits high similarity across all three dimensions—color, texture, and edge—its feature consistency evaluation result will be high; conversely, if the similarity is high in one dimension but low in others, the feature consistency evaluation result will be low. The aim is to identify regions that exhibit stable matching across multiple features, thereby improving the reliability of identification. In practical applications, the feature consistency evaluation result can be obtained by calculating the standard deviation, coefficient of variation, or weighted average of these similarity scores, or by setting rules, such as a high feature consistency evaluation result when all scores are above a certain threshold.

[0131] More specifically, in step S53, a comprehensive reliability score can be obtained based on the aforementioned feature distance and a preset distance threshold. This comprehensive reliability score aims to reflect the overall recognition reliability of the plaque region, and its purpose is to incorporate the initial feature matching degree into the calculation of the uncertainty score. For example, the smaller the feature distance and the much lower it is than the distance threshold, the higher the comprehensive reliability score, indicating that the region is more likely to be plaque. Alternatively, the ratio of the feature distance to the preset distance threshold can be calculated as the comprehensive reliability score.

[0132] More specifically, in step S54, an uncertainty score for the dental plaque region is calculated based on the feature consistency assessment results and the overall reliability score. This uncertainty score is a result that comprehensively considers the consistency of multi-dimensional feature matching and the overall recognition reliability, aiming to provide a more comprehensive and accurate confidence measure. For example, the uncertainty score can be obtained by weighted summation of the feature consistency assessment results and the overall reliability score, or by mapping using a preset lookup table.

[0133] This application's solution effectively addresses the limitations of the basic solution in quantifying the confidence level of dental plaque regions by introducing multi-dimensional feature similarity assessment and consistency analysis. Specifically, step S51 provides matching details of dental plaque regions across different visual dimensions by calculating the similarity between local color features, local texture features, and local edge features and the reference pattern. This makes the understanding of dental plaque features more comprehensive and avoids misjudgments caused by poor matching of a single feature. Furthermore, step S52 performs consistency analysis on these multi-dimensional similarity scores, which can identify regions that show stable matching across all key features, thereby effectively filtering out false positives caused by accidental local similarities and significantly improving the robustness of recognition. Simultaneously, step S53 integrates the preliminary feature distance assessment results into a comprehensive reliability score, ensuring that the overall recognition strength of dental plaque regions is considered. Finally, step S54 combines the feature consistency assessment results and the overall reliability score to calculate the uncertainty score, which not only reflects the overall matching degree between the dental plaque area and the reference pattern, but also includes the degree of mutual support between different feature dimensions, thus providing a more refined and reliable confidence measure.

[0134] In some preferred embodiments, step S54 includes:

[0135] S541. Compare the characteristic consistency assessment results with the preset consistency threshold;

[0136] S542. Compare the overall reliability score with the preset reliability threshold;

[0137] S543. When the feature consistency assessment result does not meet the consistency threshold or the comprehensive reliability score does not meet the reliability threshold, the uncertainty score is set to the preset upper limit of the uncertainty score.

[0138] S544. When the feature consistency assessment result meets the consistency threshold and the comprehensive reliability score meets the reliability threshold, calculate the uncertainty score based on the feature consistency assessment result and the comprehensive reliability score.

[0139] Specifically, in step S541, the preset consistency threshold is a pre-set value used to determine whether the feature consistency evaluation result reaches an acceptable level. For example, this threshold can be set based on a large amount of training data and expert experience to ensure that only sufficiently consistent features are considered reliable. In step S542, the preset reliability threshold is another pre-set value used to determine whether the overall reliability score reaches an acceptable level. This threshold can also be determined based on experience or statistical analysis to filter out recognition results with low reliability.

[0140] More specifically, step S543 is designed to address situations with low recognition quality. When the feature consistency assessment result does not meet a preset consistency threshold, or the overall reliability score does not meet a preset reliability threshold, it indicates that the currently identified plaque area has high uncertainty. In this case, to avoid misleading labeling, the uncertainty score is directly set to a preset upper limit for uncertainty. This upper limit typically represents the highest or most severe uncertainty; for example, it can be set to 1.0 to clearly indicate that the recognition result for this area is highly unreliable.

[0141] More specifically, step S544 aims to handle cases with high recognition quality. When the feature consistency assessment result meets a preset consistency threshold, and the overall reliability score also meets a preset reliability threshold, it indicates that the currently identified plaque area has high confidence. At this point, the uncertainty score is calculated based on the feature consistency assessment result and the overall reliability score. For example, a basic reliability score can be obtained by weighting the numerical data of the feature consistency assessment result and the overall reliability score based on preset weighting coefficients, and then the difference between the preset maximum score and the basic reliability score is calculated as the uncertainty score. In this way, the confidence level of the plaque area can be quantified more accurately, providing a more refined basis for subsequent labeling.

[0142] Through the above technical solution, this application can effectively improve the accuracy and robustness of plaque region uncertainty score calculation. Specifically, by introducing consistency and reliability thresholds, the quality of the recognition results can be initially screened, avoiding misleading low uncertainty scores under low-quality data or low-reliability recognition conditions. When the recognition quality is poor, setting the uncertainty score as an upper limit can promptly alert the user to the high uncertainty of the region's recognition, thereby effectively avoiding potential misdiagnosis or misjudgment. When the recognition quality meets the requirements, the uncertainty can be quantified more precisely through calculation, making the final plaque marking more accurate and reliable. This segmented calculation strategy significantly enhances the system's adaptability to complex oral images and improves the clinical practical value of plaque marking.

[0143] In some preferred embodiments, step S6 includes:

[0144] S61. Based on the uncertainty score corresponding to each dental plaque region, determine the visual attributes of the visual markers of each dental plaque region.

[0145] S62. Based on visual attributes, add visual markers to each dental plaque region in the oral cavity image using a semi-transparent graphic overlay method. The visual markers include outline drawing or region filling.

[0146] Specifically, visual attributes include color, transparency, line thickness, or flashing frequency. The higher the uncertainty score, the weaker the cue strength of the corresponding visual attribute.

[0147] More specifically, in step S61, the determination of visual attributes aims to transform the quantified uncertainty score into intuitive visual cues. For example, a high uncertainty score indicates a low confidence level in the system's recognition of the plaque area, in which case the visual attributes can be set to a weaker cue intensity, such as using a lighter color, higher transparency, thinner lines, or a lower flashing frequency. Conversely, a low uncertainty score indicates a high confidence level in the system's recognition of the plaque area, in which case the visual attributes can be set to a stronger cue intensity, such as using a darker color, lower transparency, thicker lines, or a higher flashing frequency. This mapping relationship allows the visual markers to dynamically reflect the reliability of the recognition results. The adjustment of visual attributes can be based on a preset mapping function or lookup table to ensure a reasonable correspondence between the uncertainty score and the visual cue intensity.

[0148] More specifically, in step S62, the visual markers can be added using a semi-transparent graphic overlay. This means that the added markers will not completely obscure the original oral cavity image, but will be presented in a way that allows simultaneous observation of the original image details and the marker information. The semi-transparent overlay can be achieved by adjusting the alpha channel value of the marker graphic. Specific forms of visual markers can include contour drawing, i.e., drawing lines along the edges of the identified plaque area; or area filling, i.e., coloring within the identified plaque area. Both methods can be adjusted in conjunction with the aforementioned visual attributes to achieve differentiated visual cues. For example, for high-confidence plaque areas, a solid fill with a clear outline can be used; while for low-confidence areas, only a semi-transparent contour drawing may be used.

[0149] The proposed solution links the uncertainty score of a dental plaque region with the visual attributes of the visual markers, enabling the marking results to dynamically and intuitively reflect the confidence level of recognition. Specifically, step S61 maps the uncertainty score to visual attributes such as color, transparency, line thickness, or flashing frequency, achieving a quantitative visual expression of recognition reliability. For example, when the system has high uncertainty in recognizing a certain dental plaque region, its marker color may be lighter, its transparency higher, or its lines thinner, or even flashing slowly to avoid overemphasizing or misjudging uncertain areas. Subsequently, step S62 presents these markers with differentiated visual attributes on the oral image using a semi-transparent graphic overlay, ensuring the integrity of the original image information while clearly highlighting the dental plaque region and its recognition confidence level. This mechanism allows dentists to quickly determine which dental plaque regions are highly certain and which require further manual review based on the visual intensity of the markers, thereby improving the accuracy and efficiency of diagnosis.

[0150] Through the above technical solution, this application provides a more intelligent and user-friendly method for dental plaque marking. Compared with traditional single marking methods, this application introduces uncertainty scores to dynamically adjust visual attributes, allowing doctors or users to intuitively understand the recognition confidence of each plaque area. This not only improves the richness and accuracy of marking information but also helps avoid incorrect diagnoses or treatment plans due to misjudgment or overconfidence. Furthermore, the semi-transparent overlay method ensures the visibility of original image details, further enhancing its practical value in clinical applications. This allows doctors to clearly observe the original morphology of teeth and gums while reviewing the marking results, thus making a more comprehensive and reliable judgment.

[0151] Secondly, please refer to Figure 2 Some embodiments of this application also provide an oral dental plaque marker detection system, the system comprising:

[0152] The acquisition module 201 is used to acquire the oral cavity image to be labeled;

[0153] The quality assessment module 202 is used to assess the quality of oral images and obtain quality assessment results;

[0154] The processing module 203 is used to select a combination of feature extraction tools based on the quality assessment results, and process the oral cavity image based on the combination of feature extraction tools to obtain enhanced image features;

[0155] The recognition module 204 is used to identify several dental plaque regions based on enhanced image features;

[0156] The quantification module 205 is used to quantify the confidence level of each dental plaque region and obtain an uncertainty score;

[0157] The labeling module 206 is used to label the dental plaque regions in the oral cavity image according to each plaque region and the corresponding uncertainty score.

[0158] Through the aforementioned technical solution, the system of this application can achieve refined and multi-dimensional evaluation of oral image quality. Compared to basic solutions that only perform a single quality assessment, this application can more comprehensively identify and quantify global color shifts and local image degradation problems in oral images. This detailed quality assessment result allows for more targeted selection of subsequent feature extraction tools and image processing, thereby significantly improving the quality and accuracy of enhanced image features. Ultimately, this helps improve the robustness and accuracy of plaque region identification, effectively reducing the misdiagnosis and missed diagnosis rates caused by image quality issues, and providing more reliable technical support for oral health diagnosis.

[0159] Furthermore, the units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0160] Furthermore, the functional modules in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0161] In this document, relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, without necessarily requiring or implying any such actual relationship or order between these entities or operations.

[0162] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for detecting oral plaque markers, characterized by, The method comprises the following steps: S1, acquiring an oral image to be marked; S2, performing quality assessment on the oral image to obtain a quality assessment result; S3, selecting a feature extraction tool combination according to the quality assessment result, and processing the oral image based on the feature extraction tool combination to obtain enhanced image features; S4, identifying a plurality of plaque regions based on the enhanced image features; S5, quantifying the confidence degree of each plaque region to obtain an uncertainty score; S6, marking the plaque regions in the oral image according to each plaque region and the corresponding uncertainty score; The quality assessment result comprises a global offset quality assessment result and a local image degradation assessment result, and the local image degradation assessment comprises brightness distribution uniformity assessment, sharpness assessment and geometric distortion assessment; Step S2 comprises: S21, performing global color balance analysis based on the average color temperature and color inclination of the oral image to analyze whether the oral image has global color offset, and obtaining a global offset quality assessment result; S22, performing local image degradation assessment on the oral image to identify local overexposure, shadow, blur and geometric distortion problems, and obtaining a local image degradation assessment result; Step S3 comprises: S31, when the global offset quality assessment result indicates that the oral image has global color offset, performing global color correction on the oral image using a preset global color correction tool to obtain a corrected oral image, and when the global offset quality assessment result indicates that the oral image does not have global color offset, regarding the oral image as the corrected oral image; S32, selecting a feature extraction tool combination and tool execution order from a preset feature extraction tool combination library according to the local image degradation assessment result, the feature extraction tool combination library comprising a plurality of feature extraction tool combinations and tool execution orders matched with different local image degradation assessment results; S33, processing the corrected oral image based on the extracted feature extraction tool combination and tool execution order to obtain the enhanced image features.

2. The oral plaque marker detection method according to claim 1, characterized in that, Step S21 comprises: S211, calculating the average brightness values of the RGB three channels of the oral image, and obtaining the average color temperature and color inclination of the oral image according to the proportional relationship between the average brightness values of the RGB three channels to obtain global color characteristics; S212, judging whether the oral image has global color offset according to the global color characteristics to obtain a global offset quality assessment result.

3. The oral plaque marker detection method according to claim 1, characterized in that, Step S22 comprises: S221, dividing the oral image into a plurality of sub-region images based on a preset region division rule; S222, performing brightness distribution uniformity assessment on each sub-region image, the brightness distribution uniformity assessment comprising calculating the brightness distribution parameters of each sub-region image and comparing the brightness distribution parameters with a preset brightness threshold to obtain brightness distribution uniformity assessment; S223, performing sharpness evaluation on each of the sub-region images, the sharpness evaluation comprising calculating an edge feature parameter of each of the sub-region images, comparing the edge feature parameter with a preset sharpness threshold, and obtaining a sharpness evaluation result; S224, identifying a tooth contour or a gum edge in the oral image; S225, performing geometric distortion evaluation on the oral image based on the identified tooth contour or gum edge, the geometric distortion evaluation comprising calculating a geometric feature parameter of the tooth contour or gum edge, and comparing the geometric feature parameter with a geometric feature parameter of a standard tooth shape, and obtaining a geometric distortion evaluation result, to form a local image degradation evaluation result together with the brightness distribution uniformity evaluation result and the sharpness evaluation result.

4. The oral plaque marker detection method according to claim 1, characterized by, Step S4 comprises: S41, extracting local color features, local texture features and local edge features of a plurality of local regions from the enhanced image features; S42, obtaining a tooth contour or a gum edge in the oral image; S43, excluding a local region that has a position matching relationship with a preset non-plaque structure feature from the tooth contour or the gum edge; S44, analyzing matching degrees of the local color features, the local texture features and the local edge features of the remaining local regions with a preset early plaque feature pattern, to identify and determine the plaque region.

5. The oral plaque marker detection method according to claim 4, characterized in that, Step S44 comprises: S441, combining the local color features, the local texture features and the local edge features of each of the remaining local regions into a local feature vector; S442, calculating a feature distance between each local feature vector and a reference feature vector of a preset early plaque feature pattern; S443, comparing the feature distance with a preset distance threshold, to identify and determine the plaque region.

6. The oral plaque marker detection method according to claim 5, wherein, Step S5 comprises: S51, calculating color similarity scores, texture similarity scores and edge similarity scores between the local color features, the local texture features and the local edge features of each plaque region and reference color features, reference texture features and reference edge features of a preset early plaque feature pattern; S52, analyzing consistency among the color similarity scores, the texture similarity scores and the edge similarity scores, to obtain a feature consistency evaluation result; S53, obtaining a comprehensive reliability score according to the feature distance and the preset distance threshold; S54, calculating an uncertainty score of the plaque region according to the feature consistency evaluation result and the comprehensive reliability score.

7. The oral plaque marker detection method according to claim 1, wherein, Step S6 comprises: S61, determining visual attributes of visual marks of each of the plaque regions according to the uncertainty score corresponding to each of the plaque regions; S62, adding visual marks to each plaque region in the oral image in a semi-transparent graphical overlay manner according to the visual attributes, the visual marks comprising contour drawing or region filling.

8. An oral plaque marker detection system, comprising: The system comprises: an acquisition module configured to acquire an oral image to be marked; a quality evaluation module configured to perform quality evaluation on the oral image, and obtain a quality evaluation result; The processing module is configured to select a feature extraction tool combination according to the quality evaluation result, and process the oral cavity image based on the feature extraction tool combination to obtain enhanced image features; The recognition module is configured to recognize a plurality of dental plaque regions based on the enhanced image features; The quantification module is configured to quantify a confidence degree of each dental plaque region to obtain an uncertainty score; The marking module is configured to mark the dental plaque regions in the oral cavity image according to each dental plaque region and the corresponding uncertainty score; The quality evaluation result includes a global offset quality evaluation result and a local image degradation quality evaluation result, and the local image degradation evaluation includes brightness distribution uniformity evaluation, sharpness evaluation, and geometric distortion evaluation. The step of performing quality evaluation on the oral cavity image to obtain the quality evaluation result includes: S21. Perform global color balance analysis based on the average color temperature and color inclination of the oral cavity image to analyze whether the oral cavity image has global color offset, and obtain a global offset quality evaluation result; S22. Perform local image degradation evaluation on the oral cavity image to identify local overexposure, shadow, blur, and geometric distortion problems, and obtain a local image degradation quality evaluation result; The step of selecting a feature extraction tool combination according to the quality evaluation result, and processing the oral cavity image based on the feature extraction tool combination to obtain enhanced image features includes: S31. When the global offset quality evaluation result indicates that the oral cavity image has global color offset, perform global color correction on the oral cavity image using a preset global color correction tool to obtain a corrected oral cavity image; when the global offset quality evaluation result indicates that the oral cavity image does not have global color offset, regard the oral cavity image as the corrected oral cavity image; S32. According to the local image degradation quality evaluation result, select a feature extraction tool combination and a tool execution order from a preset feature extraction tool combination library, and the feature extraction tool combination library includes a plurality of feature extraction tool combinations and tool execution orders matched with different local image degradation evaluation results; S33. Process the corrected oral cavity image based on the extracted feature extraction tool combination and tool execution order to obtain the enhanced image features.

Images