Toothbrush quality assessment method and system based on deep learning and image denoising

Abstract

The present invention provides a toothbrush quality assessment method and system based on the combination of deep learning and image denoising. The simulated tooth images collected during the experiment are processed by tartar simulant detection, and the cleaning power of the toothbrush is measured by comparing the changes in the tartar simulant before and after brushing. The bristle images collected during the experiment are processed by bristle detection, and the durability of the toothbrush is measured by comparing the degree of bristle deformation. A corresponding terminal and computer-readable storage medium are also provided. The present invention adopts a deep learning method to locate the target, and provides a threshold-based segmentation method to further segment the target and automatically extract relevant parameters. A glare elimination method and a memory denoising method are provided to improve the detection accuracy, so that the detection results can be as accurate as possible even in the presence of interference. Therefore, the present invention can well complete the task of toothbrush quality assessment.

Description

Technical Field

[0001] The present invention relates to the field of image processing, and in particular to a toothbrush quality assessment method and system based on a combination of deep learning and image denoising, and also provides a corresponding terminal and computer-readable storage medium. Background Art

[0002] Toothbrushes are an integral part of our lives. Toothbrush quality plays a key role in everyone's oral and dental health. There are many different types of toothbrushes on the market, and the quality of toothbrushes varies greatly from brand to brand. Therefore, it's essential to accurately evaluate toothbrush quality to provide reliable guidance for consumers choosing the right toothbrush and for manufacturers improving their production techniques. Generally speaking, key characteristics of toothbrush quality include its cleaning power and durability. A toothbrush's ability to clean plaque simulant over a given period of time is a natural proxy for its cleaning power. A toothbrush's plaque-cleaning ability can be measured by the change in plaque simulant area during brushing. The durability of a toothbrush is generally measured by the amount of bristle deformation after a specific period of brushing.

[0003] Chinese invention patent CN216411029U designs a bristle strength test device that simulates toothbrush use and collects force data to determine bristle softness and hardness. Chinese invention patent CN113899618A designs a bristle strength test device that measures bristle bundle tension and bending recovery rate through pressure data and manual observation. Chinese patent CN114034475A designs a device and method for testing the cleaning performance of an electric toothbrush, primarily based on changes in the weight and volume of a simulated object. All of the aforementioned patents are hardware-based designs, and currently, no software system exists for analyzing and processing hardware results. Furthermore, existing devices of this type, including those described above, primarily rely on single values ​​such as tension, volume, or weight, and rely primarily on manual reading and recording of results. This results in low efficiency, difficulty improving accuracy, and a lack of standardized evaluation criteria. Using image processing technology to automate the testing process enables more intensive sampling in terms of time and quantity, ensuring comprehensive and accurate test results.

[0004] Image processing technology is increasingly used in industrial production and testing processes. Due to the rapid development of intelligent image processing technology, its application in safety monitoring and quality inspection is becoming more and more extensive, while saving manpower and improving accuracy.

[0005] Threshold segmentation is a widely used traditional image processing algorithm that can use a specified threshold to classify image pixels into several categories. It has the characteristics of simple principle, low computational complexity, and stable performance. In the color space of the image to be segmented, the threshold of the desired area is selected, the pixel blocks in the image are traversed, and the pixel blocks that meet the threshold limit are extracted and merged together, thus segmenting the target area. However, general threshold segmentation methods are easily affected by noise. For example, when performing threshold segmentation based on color features, pixel blocks outside the threshold range may appear in the target area. Although they are outside the threshold, they should obviously belong to the target area. In this case, using simple threshold segmentation is no longer reasonable. For example, the strong light area formed on the surface of the dental plaque simulant due to the influence of light will cover part of the dental plaque simulant, which is called strong light noise.

[0006] Connected domain is generally an image processing algorithm based on binary images. Its purpose is to extract non-adjacent white areas in the binary image separately. The maximum connected domain is to extract the connected white area with the largest number of pixels separately.

[0007] Although the image segmentation algorithm can segment the target area, due to the characteristics of the dental plaque simulant segmentation problem, the dental plaque simulant that has been brushed and separated from the teeth but is still attached to the tooth surface cannot be counted as the target area. Such separated dental plaque simulant is regarded as splash noise, and it is obvious that a simple threshold segmentation algorithm cannot distinguish it. Summary of the Invention

[0008] In response to the above-mentioned deficiencies in the prior art, the present invention provides a toothbrush quality assessment method and system based on the combination of deep learning and image denoising, and also provides a corresponding terminal and computer-readable storage medium to meet the needs of production and life.

[0009] The present invention is achieved through the following technical solutions.

[0010] According to one aspect of the present invention, a toothbrush quality assessment method based on deep learning and image denoising is provided, comprising:

[0011] Acquire original tooth images and original bristle images respectively;

[0012] Based on the original tooth image, a target detection algorithm based on deep learning is used to detect the position of the tooth with the dental plaque simulant attached, so as to obtain a small image of the tooth containing only the dental plaque simulant attached;

[0013] The small tooth image is processed in sequence using a threshold segmentation method, a hole removal and maximum connected domain method, a strong light elimination method, and a memory denoising method to obtain a segmentation result of the dental plaque simulant;

[0014] Based on the low-resolution original bristle image, the bristle positions are detected, and a high-resolution small bristle image containing only the bristles is obtained according to the position information;

[0015] Detecting the small bristle image to obtain the boundary of the bristle area;

[0016] According to the segmentation result of the plaque simulation and the boundary of the bristle area, the area change of the plaque simulation and the deformation of the bristles are obtained respectively, and then the cleaning ability and durability of the toothbrush are obtained respectively, thereby completing the evaluation of the toothbrush quality.

[0017] Optionally, the threshold segmentation method, the hole removal and maximum connected domain method, the strong light elimination method and the memory denoising method are used to process the small tooth image in sequence, and the threshold segmentation method includes:

[0018] Performing threshold segmentation on the small tooth image in color space to obtain a binary image of a normal dental plaque simulant region and a binary image of a dental plaque simulant region with strong light, representing the target region;

[0019] The method for removing holes and finding the maximum connected domain includes:

[0020] performing inversion, maximum connected domain calculation, and inversion processing on the binary image of the normal dental plaque simulant region in sequence to obtain a processed image of the normal dental plaque simulant region;

[0021] The method for eliminating strong light comprises:

[0022] and multiplying the processed normal dental plaque simulant region image with the binary image of the dental plaque simulant region with strong light to obtain a binary image of the dental plaque simulant after removing strong light noise;

[0023] The memory denoising method comprises:

[0024] The binary image of the dental plaque simulation after removing the strong light noise at the current moment is ANDed with the binary image of the dental plaque simulation after removing the strong light noise at the previous moment to obtain a binary image of the dental plaque simulation after removing the simulated object splash noise, which is the segmentation result of the dental plaque simulation.

[0025] Optionally, the detecting of the bristle positions based on the low-resolution original bristle image and obtaining a high-resolution small bristle image containing only the bristles according to the position information includes:

[0026] The threshold segmentation method and the maximum connected domain method are used to process the low-resolution original bristle image in turn, and a high-resolution binary image containing only the bristle area is extracted. The bristle area is intercepted from the binary image containing only the bristle area as a high-resolution small bristle image.

[0027] Optionally, the threshold segmentation method and the maximum connected component method are used to process the original bristle image in sequence:

[0028] The threshold segmentation method comprises:

[0029] Performing threshold segmentation on the original bristle image in color space to obtain a bristle region binary image representing the target region;

[0030] The method for finding the maximum connected domain includes:

[0031] Based on the binary image of the bristle region, all connected regions with the same pixel value are screened out, and the region with the largest number of pixels, ie, the largest connected domain, is used as a high-resolution small bristle image containing only bristles.

[0032] Optionally, detecting the small bristle image to obtain the boundary of the bristle area includes:

[0033] The bristle small images are processed in sequence by adopting a threshold segmentation method and a hole removal and maximum connected domain method to obtain the bristle area and further obtain the bristle area boundary.

[0034] Optionally, the threshold segmentation method and the hole removal and maximum connected component finding method are used to process the bristle small image in sequence:

[0035] The threshold segmentation method comprises:

[0036] Performing threshold segmentation on the small bristle image in color space to obtain a binary image of the bristle region representing the target region;

[0037] The method for removing holes and finding the maximum connected domain includes:

[0038] The binary image of the bristle region is sequentially subjected to the processes of negation, finding the maximum connected domain, and negation to obtain an image of the region where the bristles are located after the processing.

[0039] Optionally, respectively obtaining the cleaning ability and durability of the toothbrush comprises:

[0040] Within a set time, the rate of change in the area of ​​the plaque simulant reflects the cleaning power of the toothbrush; the degree of deformation of the bristles reflects the durability of the toothbrush.

[0041] According to another aspect of the present invention, a toothbrush quality assessment system based on deep learning and image denoising is provided, comprising:

[0042] An image acquisition module, the module is used to respectively acquire an original tooth image and an original bristle image;

[0043] a dental plaque simulant positioning module, which detects the position of the tooth with the dental plaque simulant attached based on the original tooth image and adopts a deep learning-based target detection algorithm to obtain a small image of the tooth containing only the dental plaque simulant attached;

[0044] a dental plaque simulant segmentation module, which sequentially processes the small tooth image containing only the dental plaque simulant by using a threshold segmentation method, a cavity removal and maximum connected domain method, a strong light elimination method, and a memory denoising method to obtain a segmentation result of the dental plaque simulant;

[0045] a bristle positioning module, which detects the position of the bristles based on the low-resolution original bristle image and obtains a high-resolution small bristle image containing only the bristles;

[0046] A bristle segmentation module detects the bristle small image to obtain the boundary of the bristle area;

[0047] The toothbrush quality assessment module obtains the area change of the plaque simulation and the deformation of the bristles based on the segmentation results of the plaque simulation and the boundaries of the bristle area, and then obtains the cleaning ability and durability of the toothbrush respectively, completing the evaluation of the toothbrush quality.

[0048] According to a third aspect of the present invention, a terminal is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor may be configured to execute any one of the above methods when executing the program.

[0049] According to another aspect of the present invention, a computer-readable storage medium is provided, on which a computer program is stored. When the program is executed by a processor, it can be used to perform any of the above methods.

[0050] Due to the adoption of the above technical solution, the present invention has at least one of the following beneficial effects compared with the prior art:

[0051] The present invention adopts a deep learning method to locate the target and provides a threshold-based segmentation method to further segment the target and automatically extract relevant parameters.

[0052] The present invention provides a strong light elimination method and a memory denoising method to improve detection accuracy.

[0053] The present invention ensures that the detection result is as accurate as possible even in the presence of interference. Therefore, the present invention can well complete the task of toothbrush quality assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Other features, objects and advantages of the present invention will become more apparent upon reading the detailed description of non-limiting embodiments with reference to the following drawings:

[0055] Figure 1 This is a workflow diagram of a toothbrush quality assessment method based on the combination of deep learning and image denoising in one embodiment of the present invention.

[0056] Figure 2 This is a flowchart of a toothbrush quality assessment method based on the combination of deep learning and image denoising in a preferred embodiment of the present invention;

[0057] Figure 3 This is a flow chart of dental plaque simulant detection in a preferred embodiment of the present invention;

[0058] Figure 4 This is a flow chart of bristle detection in a preferred embodiment of the present invention.

[0059] Figure 5 This is a schematic diagram of the components of a toothbrush quality assessment system based on the combination of deep learning and image denoising in one embodiment of the present invention. DETAILED DESCRIPTION

[0060] The present invention will be described in detail below with reference to specific embodiments. The following examples will help those skilled in the art to further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that, for those skilled in the art, several changes and improvements can be made without departing from the scope of the present invention. These all fall within the scope of protection of the present invention.

[0061] An embodiment of the present invention provides a toothbrush quality assessment method based on the combination of deep learning and image denoising.

[0062] like Figure 1 As shown, the toothbrush quality assessment method based on the combination of deep learning and image denoising provided in this embodiment may include the following steps:

[0063] S100, acquiring an original tooth image and an original bristle image respectively;

[0064] S200, based on the original tooth image, using a deep learning-based target detection algorithm (e.g., Yolov5) to detect the position of the tooth with the attached plaque simulant, and obtain a small image of the tooth containing only the attached plaque simulant;

[0065] S300, using a threshold segmentation method, a hole removal and maximum connected domain method, a strong light elimination method, and a memory denoising method to sequentially process the small tooth image to obtain a segmentation result of the dental plaque simulant;

[0066] S400, based on the low-resolution original bristle image, detecting the bristle positions to obtain a high-resolution small bristle image containing only the bristles;

[0067] S500, detecting the small bristle image to obtain the boundary of the bristle area;

[0068] S600: Based on the segmentation results of the plaque simulant and the boundaries of the bristle area, the area change of the plaque simulant and the deformation of the bristles are obtained respectively, and then the cleaning ability and durability of the toothbrush are obtained respectively, completing the evaluation of the toothbrush quality.

[0069] In this embodiment, the processing of the tooth image is performed at high resolution. In the processing of the toothbrush image, the first step utilizes a low-resolution image. After obtaining the bristle positions based on the low-resolution image, the second step performs fine segmentation on the high-resolution image. That is, in this embodiment, the original tooth image is a high-resolution image, and the original bristle image is a low-resolution image. High resolution refers to the original image resolution (3840×2748), while low resolution refers to the resolution after reducing the original resolution to one-tenth (384×275). Both resolutions refer to the resolution of the complete image before the bristle subimage is captured.

[0070] In the above embodiments of the present invention:

[0071] S200 is a dental plaque simulant location step: using a deep learning-based target detection algorithm, the position of the tooth to which the dental plaque simulant is attached is detected, and a small image of the tooth containing only the dental plaque simulant is obtained;

[0072] S300 is a dental plaque simulant segmentation step: based on a small image of the tooth containing only the dental plaque simulant, an accurate dental plaque simulant segmentation result is obtained after processing using a threshold segmentation method, a cavity removal and maximum connected domain method, a strong light elimination method, and a memory denoising method;

[0073] S400 is a bristle positioning step: quickly detecting the bristle positions in the original low-resolution image to obtain a high-resolution small bristle image containing only the bristles;

[0074] S500 is the bristle segmentation step: based on the obtained high-resolution small bristle image containing only bristles, the threshold method and the hole removal and maximum connected domain method are used to accurately detect the boundary of the bristle area.

[0075] In a preferred embodiment, the threshold segmentation method in step S300 performs threshold segmentation on the collected color image in the color space to obtain a binary image in which the target area is represented by white. The binary image of the target area includes: a binary image of a normal dental plaque simulant area and a binary image of a dental plaque simulant area with strong light.

[0076] In a preferred embodiment, the method for removing holes and finding the maximum connected component in step S300 is to sequentially perform the processes of inverting, finding the maximum connected component, and inverting on the obtained binary image of the normal dental plaque simulant region representing the target region.

[0077] In a preferred embodiment, before the strong light elimination method in step S300, a threshold segmentation method is used to extract binary images of the normal dental plaque simulant area and the dental plaque simulant area with strong light, respectively. After the normal dental plaque simulant binary image is processed by a hole removal and maximum connected domain method, the processed result is ANDed with the binary image of the dental plaque simulant area with strong light to obtain a binary image of the dental plaque simulant after removing the strong light noise.

[0078] In a preferred embodiment, the memory denoising method in step S300 makes the binary image of the dental plaque simulator after removing the strong light noise at the current moment be ANDed with the binary image of the dental plaque simulator after removing the strong light noise at the previous moment, and obtains the binary image of the dental plaque simulator after removing the simulator splash noise, which is the segmentation result of the dental plaque simulator.

[0079] In a preferred embodiment, the position of the bristles is quickly detected in the low-resolution original image in step S400, and the collected image is processed in sequence using a threshold segmentation method and a maximum connected domain method to extract an image containing only the obtained maximum connected domain, which is a high-resolution small image of the bristles containing only the bristles.

[0080] In a preferred embodiment, the threshold segmentation method in step S400 is: performing threshold segmentation on the original bristle image in the color space to obtain a binary image of the bristle region representing the target region.

[0081] In a preferred embodiment, the method for obtaining the maximum connected domain in step S400 is: based on the binary image of the bristle area, all connected areas with the same pixel value are screened out, among which the area with the largest number of pixels is the maximum connected domain, and the maximum connected domain is used as a high-resolution small bristle image containing only bristles.

[0082] In a preferred embodiment, the bristle small image in step S500 is segmented by a threshold value and a hole removal and maximum connected region method to obtain the bristle area, and then the bristle boundary is accurately detected.

[0083] In a preferred embodiment, the threshold segmentation method in step S500 is: performing threshold segmentation on the small bristle image in the color space to obtain a binary image of the bristle region representing the target region.

[0084] In a preferred embodiment, the method for removing holes and finding the maximum connected component in step S500 is: inverting the binary image of the bristle region, finding the maximum connected component, and inverting the image in sequence to obtain an image of the bristle region after processing.

[0085] In a preferred embodiment, the steps S600 of obtaining the cleaning ability and durability of the toothbrush include:

[0086] Within a set time, the rate of change in the area of ​​the plaque simulant reflects the cleaning power of the toothbrush; the degree of deformation of the bristles reflects the durability of the toothbrush.

[0087] Figure 2 This is a flowchart of a toothbrush quality assessment method based on the combination of deep learning and image denoising provided in a preferred embodiment of the present invention.

[0088] like Figure 2 As shown, the toothbrush quality assessment method based on deep learning and image denoising provided by this preferred embodiment includes the following steps:

[0089] Original image acquisition step: acquiring original tooth images and original bristle images respectively;

[0090] Plaque simulant location step: Using a deep learning-based object detection algorithm, the location of the tooth with the plaque simulant attached is detected, and a small image containing only the tooth with the plaque simulant attached is obtained;

[0091] Plaque simulant segmentation: Based on a small image containing only the tooth with plaque simulant attached, accurate plaque simulant segmentation is obtained through threshold segmentation, cavity removal and maximum connected domain method, highlight removal module, and memory denoising module.

[0092] Bristle positioning step: by quickly detecting the bristle positions in the low-resolution original image, a small high-resolution image containing only the bristles is obtained;

[0093] Brush bristle segmentation step: Based on the obtained high-resolution small image containing only bristles, the threshold segmentation method and the hole removal and maximum connected domain method are used to accurately detect the boundaries of the bristle area;

[0094] Toothbrush quality assessment steps: Based on the segmentation results of the plaque simulant and the boundaries of the bristle area, the area change of the plaque simulant and the deformation of the bristles are obtained respectively, and then the cleaning ability and durability of the toothbrush are obtained respectively to complete the evaluation of the toothbrush quality.

[0095] like Figure 3 As shown, the dental plaque simulant location step utilizes the widely used YOLO object detection algorithm. First, the location of the tooth coated with dental plaque simulant is detected in the original tooth image and outlined with a box. The box is then cropped to create a small image containing only the tooth with the dental plaque simulant attached. The original tooth image specifically refers to a top-down image of the tooth with dental plaque simulant, captured by the experimental equipment. Due to the large size of the original image, the tooth is cropped out using an object detection algorithm. YOLO is a commonly used deep learning-based object detection model, and YOLOv5 is its current, more accurate version. It introduces the SPP-NET architecture and utilizes new data augmentation methods such as mosaic training, SAT, and multi-channel features. The feature pyramid network structure is also replaced with PANet. In this experiment, after collecting data using existing detection equipment, the teeth were manually annotated and trained using this network structure, ultimately resulting in a neural network capable of identifying tooth locations.

[0096] In the dental plaque simulant segmentation step, the obtained small image containing only the teeth with dental plaque simulant attached is first converted into the HSV color feature space, and two threshold segmentation processes are performed on it: T HS and T HV :

[0097]

[0098]

[0099] Among them, T HS and T HV They respectively represent the specific process of threshold segmentation for normal dental plaque simulant and dental plaque simulant with strong light. t1, t2, t3, and t4 represent specific thresholds, which are set based on actual experience. H(x,y), S(x,y), and V(x,y) respectively represent the corresponding values ​​of the three channels of the HSV color features of the image.

[0100] The normal plaque simulant is segmented by threshold value and the strong light elimination method is performed at the same time. HS The segmentation result is treated as a separate branch and processed by the method of removing holes and finding the maximum connected area. HS and T HV The segmentation results are combined as another branch, and the two results are ANDed together. This can fill in the missing parts of the dental plaque simulant caused by strong light during the threshold segmentation process, achieving the purpose of eliminating strong light. The specific AND operation can be expressed as:

[0101] AND(x,y)=Img 1(x,y)&Img 2(x,y)

[0102] Wherein, Img 1 and Img 2 represent two binary images to be ANDed.

[0103] Hole removal and maximum connected component calculation is to find the maximum connected component of the current binary image and make sure there are no missing pixels in the maximum connected component. This is achieved by performing three steps on the current binary image: image inversion, finding the maximum connected component, and image inversion again. Image inversion refers to performing a NOT operation on each pixel in the image:

[0104] NOT(x,y)=!Img(x,y)

[0105] Based on the results after eliminating the glare, the memory denoising method is applied. The purpose of memory denoising is to eliminate splash simulants. Splash simulants refer to plaque simulants that have been brushed and separated from the teeth but are still attached to the tooth surface. The specific method is to record the detection results of consecutive moments in a segmentation result library, and then perform AND operation on the binary image of the final segmentation result after eliminating the glare with the final segmentation result of the previous moment. This ensures that the area of ​​the plaque simulant does not increase unreasonably, thereby achieving the purpose of eliminating surrounding splashes. In particular, in a time series of images, the first image must be specially processed. Specifically, it is subjected to three operations: threshold segmentation, finding the maximum connected domain, and removing holes to find the maximum connected domain.

[0106] like Figure 4 As shown in the figure, the main purpose of the bristle location step is to find the small image where the bristles are located in the original image. First, the image is threshold segmented. The original bristle image refers to the image of the bristles captured by the experimental equipment. The specific shooting direction is from the bristle end of the toothbrush to the bristle-free end. The threshold segmentation can be specifically expressed as:

[0107]

[0108] Among them, T B It represents the specific process of segmenting the bristles, and t5 and t6 represent specific thresholds, which are set based on actual experience.

[0109] The largest connected domain in the segmentation result is used as the location of the bristles. Since the image is relatively complex, reducing the image resolution can greatly improve the detection speed. In addition, the position detection does not require high accuracy. The small image where the detected bristles are located is cropped out at a low resolution.

[0110] The bristle segmentation step refers to the sequential application of the threshold segmentation method and the hole removal and maximum connected domain operations in the cropped high-resolution small image, so that the bristle contour can be obtained.

[0111] Finally, the obtained series of changes in the area of ​​dental plaque simulants, as well as the bristle spreading area and bristle spreading width were used as evaluation indicators for the cleaning ability and durability of the toothbrush for further analysis.

[0112] Figure 5 A schematic diagram of the components of a toothbrush quality assessment system based on deep learning and image denoising provided by one embodiment of the present invention.

[0113] like Figure 5 As shown, the toothbrush quality assessment system based on deep learning and image denoising provided in this embodiment may include the following modules:

[0114] M1, an image acquisition module, which is used to acquire original tooth images and original bristle images respectively;

[0115] M2, dental plaque simulant positioning module, uses a deep learning-based target detection algorithm based on the original tooth image to detect the position of the tooth with dental plaque simulant attached, and obtains a small image of the tooth containing only the dental plaque simulant attached;

[0116] M3, dental plaque simulant segmentation module, uses the threshold segmentation method, the cavity removal and maximum connected domain method, the strong light elimination method, and the memory denoising method to sequentially process the small image of the tooth containing only the dental plaque simulant to obtain the dental plaque simulant segmentation result;

[0117] M4, bristle positioning module, which detects the bristle position based on the low-resolution original bristle image and obtains a high-resolution small bristle image containing only the bristles;

[0118] M5, bristle segmentation module, which detects the bristle small image and obtains the boundary of the bristle area;

[0119] M6, the toothbrush quality assessment module, obtains the area change of the plaque simulant and the deformation of the bristles based on the segmentation results of the plaque simulant and the boundaries of the bristle area, and then obtains the cleaning ability and durability of the toothbrush, completing the evaluation of the toothbrush quality.

[0120] In a preferred embodiment, the threshold segmentation method in the M3 module performs threshold segmentation on the collected color image in the color space to obtain a binary image in which the target area is represented by white. The binary image of the target area includes: a binary image of a normal dental plaque simulant area and a binary image of a dental plaque simulant area with strong light.

[0121] In a preferred embodiment, the method for removing holes and finding the maximum connected component in the M3 module is to perform the processes of inverting, finding the maximum connected component, and inverting the obtained binary image of the normal dental plaque simulation area representing the target area in sequence.

[0122] In a preferred embodiment, before the highlight removal method in module M3, a threshold segmentation method is used to extract binary images of the normal dental plaque simulant region and the dental plaque simulant region with highlight noise. The binary dental plaque simulant image is processed using a cavity removal and maximum connected component method, and the result is ANDed with the binary dental plaque simulant image with highlight noise to obtain a binary dental plaque simulant image after highlight noise removal.

[0123] In a preferred embodiment, the memory denoising method in the M3 module makes the binary image of the dental plaque simulator after removing the strong light noise at the current moment be ANDed with the binary image of the dental plaque simulator after removing the strong light noise at the previous moment, and obtains the binary image of the dental plaque simulator after removing the simulator splash noise, which is the segmentation result of the dental plaque simulator.

[0124] In a preferred embodiment, the position of the bristles is quickly detected in the original low-resolution image in the M4 module, and the collected image is processed in sequence using the threshold segmentation method and the maximum connected domain method to extract an image containing only the obtained maximum connected domain, which is a high-resolution small image of the bristles containing only the bristles.

[0125] In a preferred embodiment, a threshold segmentation method and a hole removal and maximum connected region method are applied to the small bristle image in the M5 module to obtain the area where the bristles are located, and then the bristle boundaries are accurately detected.

[0126] It should be noted that the steps in the method provided by the present invention can be implemented using corresponding modules, devices, units, etc. in the system. Those skilled in the art can refer to the technical solution of the method to implement the composition of the system, that is, the embodiments in the method can be understood as preferred examples of constructing the system, which will not be elaborated here.

[0127] An embodiment of the present invention provides a terminal, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it can be used to execute any one of the methods in the above embodiments.

[0128] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, it can be used to execute any one of the methods in the above embodiments.

[0129] Optionally, the memory is used to store programs; the memory may include volatile memory (English: volatile memory), such as random-access memory (English: random-access memory, abbreviated: RAM), such as static random-access memory (English: static random-access memory, abbreviated: SRAM), double data rate synchronous dynamic random access memory (English: Double Data Rate Synchronous Dynamic Random Access Memory, abbreviated: DDR SDRAM), etc.; the memory may also include non-volatile memory (English: non-volatile memory), such as flash memory (English: flash memory). The memory is used to store computer programs (such as applications, functional modules, etc. that implement the above-mentioned methods), computer instructions, etc., and the above-mentioned computer programs, computer instructions, etc. can be partitioned and stored in one or more memories. In addition, the above-mentioned computer programs, computer instructions, data, etc. can be called by the processor.

[0130] The aforementioned computer programs, computer instructions, etc. may be partitioned and stored in one or more memories, and the aforementioned computer programs, computer instructions, data, etc. may be called by a processor.

[0131] The processor is configured to execute the computer program stored in the memory to implement the various steps of the method involved in the above embodiment. For details, please refer to the relevant description in the above method embodiment.

[0132] The processor and memory can be independent structures or integrated structures. When the processor and memory are independent structures, the memory and processor can be coupled via a bus.

[0133] The toothbrush quality assessment method and system based on the combination of deep learning and image denoising provided by the above embodiment of the present invention processes the simulated tooth images collected during the experiment through plaque simulant detection, and measures the cleaning power of the toothbrush by comparing the changes in plaque simulant before and after brushing; and processes the bristle images collected during the experiment through bristle detection, and measures the durability of the toothbrush by comparing the degree of bristle deformation. The toothbrush quality assessment method and system based on the combination of deep learning and image denoising provided by the above embodiment of the present invention adopts a deep learning method to locate the target, and provides a threshold-based segmentation method to further segment the target, thereby automatically extracting relevant parameters. A glare elimination method and a memory denoising method are provided to improve the detection accuracy, so that the detection results can be as accurate as possible even in the presence of interference. Therefore, the method and system provided by the above embodiment of the present invention can well complete the toothbrush quality assessment task.

[0134] Those skilled in the art will appreciate that, in addition to implementing the system and its various devices, modules, and units provided by the present invention in purely computer-readable program code, it is entirely possible to implement the same functions of the system and its various devices, modules, and units provided by the present invention in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, the system and its various devices, modules, and units provided by the present invention can be considered a hardware component, and the devices, modules, and units included therein for implementing various functions can also be considered as structures within the hardware component; the devices, modules, and units for implementing various functions can also be considered as both software modules implementing the method and structures within the hardware component.

[0135] The above describes specific embodiments of the present invention. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art may make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. The embodiments of this application and the features in the embodiments may be combined with each other in any manner unless there is a conflict.

Claims

1. A toothbrush quality assessment method based on deep learning and image denoising, characterized in that: include: Acquire original tooth images and original bristle images respectively; Based on the original tooth image, a target detection algorithm based on deep learning is used to detect the position of the tooth with the dental plaque simulant attached, so as to obtain a small image of the tooth containing only the dental plaque simulant attached; The small tooth image is processed in sequence using a threshold segmentation method, a hole removal and maximum connected domain method, a strong light elimination method, and a memory denoising method to obtain a segmentation result of the dental plaque simulant; wherein: The threshold segmentation method comprises: Performing threshold segmentation on the small tooth image in color space to obtain a binary image of a normal dental plaque simulant region and a binary image of a dental plaque simulant region with strong light, representing the target region; The method for removing holes and finding the maximum connected domain includes: performing inversion, maximum connected domain calculation, and inversion processing on the binary image of the normal dental plaque simulant region in sequence to obtain a processed image of the normal dental plaque simulant region; The method for eliminating strong light comprises: and multiplying the processed normal dental plaque simulant region image with the binary image of the dental plaque simulant region with strong light to obtain a binary image of the dental plaque simulant after removing strong light noise; The memory denoising method comprises: The binary image of the dental plaque simulant after removing the strong light noise at the current moment is ANDed with the binary image of the dental plaque simulant after removing the strong light noise at the previous moment to obtain a binary image of the dental plaque simulant after removing the simulated object splash noise, which is the segmentation result of the dental plaque simulant; Based on the low-resolution original bristle image, the bristle positions are detected, and a high-resolution small bristle image containing only the bristles is obtained according to the position information; specifically, the method includes: The low-resolution original bristle image is processed in sequence using a threshold segmentation method and a maximum connected component method to extract a high-resolution binary image containing only the bristle region, and the bristle region is intercepted from the binary image containing only the bristle region as a high-resolution small bristle image; The threshold segmentation method comprises: Performing threshold segmentation on the original bristle image in color space to obtain a bristle region binary image representing the target region; The method for finding the maximum connected domain includes: Based on the binary image of the bristle region, all connected regions with the same pixel value are screened out, and the region with the largest number of pixels, i.e., the largest connected domain, is used as a high-resolution small bristle image containing only bristles; Detecting the small bristle image to obtain the boundary of the bristle area; According to the segmentation result of the plaque simulation and the boundary of the bristle area, the area change of the plaque simulation and the deformation of the bristles are obtained respectively, and then the cleaning ability and durability of the toothbrush are obtained respectively, thereby completing the evaluation of the toothbrush quality.

2. The toothbrush quality evaluation method according to claim 1, characterized in that: The detecting of the small bristle image to obtain the boundary of the bristle area includes: The bristle small images are processed in sequence by adopting a threshold segmentation method and a hole removal and maximum connected domain method to obtain the bristle area and further obtain the bristle area boundary.

3. The toothbrush quality evaluation method according to claim 2, characterized in that: The threshold segmentation method and the hole removal and maximum connected domain method are used to process the bristle small image in sequence: The threshold segmentation method comprises: Performing threshold segmentation on the small bristle image in color space to obtain a binary image of the bristle region representing the target region; The method for removing holes and finding the maximum connected domain includes: The binary image of the bristle region is sequentially subjected to the processes of negation, finding the maximum connected domain, and negation to obtain an image of the region where the bristles are located after the processing.

4. The toothbrush quality evaluation method according to claim 1, characterized in that: The cleaning ability and durability of the toothbrush are obtained respectively, including: Within a set time, the rate of change in the area of ​​the plaque simulant reflects the cleaning power of the toothbrush; the degree of deformation of the bristles reflects the durability of the toothbrush.

5. A toothbrush quality assessment system based on deep learning and image denoising, characterized in that: include: An image acquisition module, the module is used to respectively acquire an original tooth image and an original bristle image; a dental plaque simulant positioning module, which detects the position of the tooth with the dental plaque simulant attached based on the original tooth image and adopts a deep learning-based target detection algorithm to obtain a small image of the tooth containing only the dental plaque simulant attached; A dental plaque simulant segmentation module, which uses a threshold segmentation method, a cavity removal and maximum connected domain method, a strong light elimination method, and a memory denoising method to sequentially process the small tooth image containing only the dental plaque simulant to obtain a dental plaque simulant segmentation result; wherein: The threshold segmentation method comprises: Performing threshold segmentation on the small tooth image in color space to obtain a binary image of a normal dental plaque simulant region and a binary image of a dental plaque simulant region with strong light, representing the target region; The method for removing holes and finding the maximum connected domain includes: performing inversion, maximum connected domain calculation, and inversion processing on the binary image of the normal dental plaque simulant region in sequence to obtain a processed image of the normal dental plaque simulant region; The method for eliminating strong light comprises: and multiplying the processed normal dental plaque simulant region image with the binary image of the dental plaque simulant region with strong light to obtain a binary image of the dental plaque simulant after removing strong light noise; The memory denoising method comprises: The binary image of the dental plaque simulant after removing the strong light noise at the current moment is ANDed with the binary image of the dental plaque simulant after removing the strong light noise at the previous moment to obtain a binary image of the dental plaque simulant after removing the simulated object splash noise, which is the segmentation result of the dental plaque simulant; The bristle positioning module detects the bristle positions based on the low-resolution original bristle image to obtain a high-resolution small bristle image containing only the bristles. Specifically, it includes: The low-resolution original bristle image is processed in sequence using a threshold segmentation method and a maximum connected component method to extract a high-resolution binary image containing only the bristle region, and the bristle region is intercepted from the binary image containing only the bristle region as a high-resolution small bristle image; The threshold segmentation method comprises: Performing threshold segmentation on the original bristle image in color space to obtain a bristle region binary image representing the target region; The method for finding the maximum connected domain includes: Based on the binary image of the bristle region, all connected regions with the same pixel value are screened out, and the region with the largest number of pixels, i.e., the largest connected domain, is used as a high-resolution small bristle image containing only bristles; A bristle segmentation module detects the bristle small image to obtain the boundary of the bristle area; The toothbrush quality assessment module obtains the area change of the plaque simulation and the deformation of the bristles based on the segmentation results of the plaque simulation and the boundaries of the bristle area, and then obtains the cleaning ability and durability of the toothbrush respectively, completing the evaluation of the toothbrush quality.

6. A terminal comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, it can be used to perform the method according to any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, it can be used to perform the method according to any one of claims 1 to 4.

Images