Image-driven tick detection comb fused with deep learning models

Abstract

The present invention discloses an image-driven tick detection comb that integrates a deep learning model, which relates to the field of tick detection technology and includes an initial exposure time configuration module, an image data acquisition module, a feature extraction and fuzzy quantization module, a fuzziness prediction module, an image classification module, a normal image processing module, and a fuzzy image exposure adjustment module; the initial exposure time configuration module first determines the initial exposure time that is adapted to the current environment and detection requirements, and performs parameter configuration on the micro camera. The present invention uses deep learning and image processing technology to accurately predict image fuzziness, distinguish between normal and fuzzy images, and adopts adaptive exposure time control for fuzzy images to reduce motion blur, ensure image clarity, improve feature extraction accuracy, reduce the probability of misjudgment, detect and remove ticks in a timely manner, and reduce the risk of host infection with pathogens. Dynamic exposure adjustment enhances the robustness and reliability of the equipment in different environments, significantly improving detection accuracy.

Description

Technical Field

[0001] The present invention relates to the field of tick detection technology, and in particular to an image-driven tick detection comb integrating a deep learning model. Background Art

[0002] The image-driven tick detection comb is an innovative device that integrates computer vision and intelligent detection technologies, specifically designed to efficiently identify and remove ticks from the fur of pets or livestock. This comb incorporates a miniature camera or image sensor within the teeth of a traditional comb, capturing real-time images of the fur surface during the combing process. Using a built-in machine learning algorithm, it analyzes image features to accurately identify the presence of ticks. Once a tick is detected, the system alerts with an LED indicator, vibration, or wirelessly transmits the information to a mobile app, allowing the user to quickly take removal measures. Additionally, some high-end models feature automatic collection or electric shock removal to improve tick removal efficiency and reduce the risk of infection for both humans and pets. Suitable for home pet care, livestock epidemic prevention, and parasite monitoring in the wild, this device significantly improves the convenience and accuracy of tick detection.

[0003] Image-driven tick detection typically involves several key steps. The first is data acquisition, where images of pet or livestock fur are captured in real time using a miniature camera (or image sensor). Next, image preprocessing is performed, including denoising and contrast enhancement, to improve image quality and clarity. Key features in the image are then extracted, such as the shape, size, texture, and color features of the tick. These features are typically extracted using image segmentation techniques, edge detection algorithms, or deep learning models. After feature extraction, machine learning algorithms (such as convolutional neural networks (CNNs)) are used to classify the extracted features to determine whether a tick is present in the image. If a tick is detected, the system generates feedback, such as highlighting the tick's location and alerting the user through vibration, sound, or a display screen.

[0004] Existing technologies face a challenge when extracting tick features:

[0005] When ticks are stimulated by external stimuli, such as contact with comb teeth or hair vibrations, they move rapidly to evade detection. Currently, detection systems typically use a constant exposure time setting for their micro-cameras. This can cause moving ticks to appear blurred in the image when exposure times are long, affecting the accuracy of feature extraction. Consequently, ticks may be misidentified as hair nodules or dirt, leading to detection failure. This situation not only delays the timely removal of ticks but also increases the risk of host infection with pathogens.

[0006] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0007] The purpose of the present invention is to provide an image-driven tick detection comb that integrates a deep learning model. By combining the deep learning model with image processing technology, the image blur is accurately predicted, and normal images are effectively distinguished from blurred images. For blurred images, adaptive exposure time control technology is used to reduce the motion blur effect caused by long exposure, thereby ensuring image clarity and improving the accuracy of feature extraction, reducing the probability of misjudgment, and ensuring that ticks can be detected and removed in a timely manner, thereby effectively reducing the risk of host infection with pathogens. At the same time, dynamic exposure time adjustment enables the image-driven tick detection comb to operate stably under various environmental conditions, enhances the robustness and reliability of the device, solves the problem of image blur caused by the rapid movement of ticks, and greatly improves the overall accuracy of the detection system to solve the problems in the above-mentioned background technology.

[0008] To achieve the above objectives, the present invention provides the following technical solutions: an image-driven tick detection comb integrating a deep learning model, comprising an initial exposure time configuration module, an image data acquisition module, a feature extraction and fuzzy quantization module, a fuzziness prediction module, an image classification module, a normal image processing module, and a fuzzy image exposure adjustment module;

[0009] The initial exposure time configuration module first determines the initial exposure time that is suitable for the current environment and detection requirements, and configures the parameters of the micro camera;

[0010] Image data acquisition module: After taking a photo, it fully obtains all information of the image and provides complete data support for subsequent feature extraction;

[0011] The feature extraction and blur quantification module, after obtaining complete image data, uses image processing technology to extract feature information related to image blur, performs feature engineering on the extracted feature information, and preliminarily quantifies the image distortion caused by motion blur, providing a basis for subsequent blur determination.

[0012] The blur prediction module inputs the blur features processed by feature engineering into a pre-trained deep learning model and uses the deep learning model to predict the blur of the current image;

[0013] The image classification module divides the captured images into normal images and blurred images based on the prediction results of the deep learning model;

[0014] The normal image processing module, for normal images, continues to maintain the initial exposure time for the micro camera to take pictures, and performs subsequent tick feature extraction, identification and positioning;

[0015] The blurred image exposure adjustment module uses a fuzzy logic algorithm to adaptively adjust the initial exposure time for blurred images based on the image blur prediction results, shortening the actual exposure time when the micro camera takes pictures and reducing the blur effect caused by motion. Subsequent tick feature extraction, identification, and positioning are performed only after the image is determined to be a normal image.

[0016] Preferably, the initial exposure time adapted to the current environment and detection needs is determined based on the static detection requirements. The static detection requirements mean that when detecting ticks, the ticks and their surroundings are preset to be in a relatively static state to ensure that the micro camera can capture clear images without motion blur.

[0017] Preferably, feature information related to image blur is extracted through image processing technology, wherein the extracted feature information includes the gradient change of the image edge and the distribution of the image spatial texture. Feature engineering processing is performed on the extracted feature information to generate edge gradient reference values ​​and spatial texture chaos reference values, respectively. The edge gradient reference values ​​and spatial texture chaos reference values ​​are used to preliminarily quantify the image distortion caused by motion blur, providing a basis for subsequent blur judgment.

[0018] Preferably, the edge gradient reference value and the spatial texture chaos reference value after feature engineering processing are input into a pre-trained deep learning model, a fuzzy quantization coefficient is generated based on the deep learning model, and the blurriness of the current image is predicted by the fuzzy quantization coefficient.

[0019] Preferably, the fuzzy quantization coefficient generated when the blurriness of the current image is predicted by the pre-trained deep learning model is compared and analyzed with a pre-set fuzzy quantization coefficient reference threshold, and the captured image is divided. The specific division steps are as follows:

[0020] If the generated fuzzy quantization coefficient is greater than the fuzzy quantization coefficient reference threshold, the captured image is classified as a fuzzy image image;

[0021] If the generated fuzzy quantization coefficient is less than or equal to the fuzzy quantization coefficient reference threshold, the captured image is classified as a normal image.

[0022] Preferably, the specific steps of performing feature engineering processing on the extracted gradient change features of the image edge to generate the edge gradient reference value are as follows:

[0023] The second-order derivative of the Laplace operator is used to extract the gradient information of each point in the image, and the gradient of the image is calculated to capture the local brightness change of each pixel in the image. The extracted expression is as follows: ,in: Representing an image In position The Laplace operator result at represents the change in the brightness change rate of the pixel point, which can describe the curvature of the image in the horizontal and vertical directions, and thus capture the edge information of the image. Indicates that the image is at position The pixel value at and Represent the second-order derivatives in the horizontal and vertical directions respectively;

[0024] Based on the gradient change of each pixel in the image, the edge gradient reference value is defined to quantify the blur of the image. The edge gradient reference value reflects the overall blur of the image by accumulating the intensity of the gradient change and using an exponential decay function. The specific calculation expression is: ,in, represents the edge gradient reference value, represents the set of all pixels in the image, is the absolute value of the gradient.

[0025] Preferably, the specific steps of performing feature engineering processing on the distribution features of the extracted image spatial texture to generate the spatial texture chaos reference value are as follows:

[0026] The image is decomposed into multiple frequency sub-bands using wavelet transform. Local texture features at different levels are obtained through the decomposed frequency sub-bands. The texture information of each local area is represented as a feature vector, which reflects the local texture pattern of the area. In order to further quantify the disorder of these local texture features, a new local feature calculation formula is adopted. The calculation expression is: ,in: Indicates the position in the image The local texture variation at the location reflects the difference between the texture of the surrounding area and the central area. Indicates the position in the image The texture value of Represents the texture value of the center of the local area, Parameters for adjusting sensitivity to texture changes;

[0027] After the local texture features are extracted, the degree of image disorder is quantified by calculating the local texture variation of each local area. Based on the quantified result of the local texture variation, a spatial texture disorder reference value is generated to evaluate the overall texture disorder of the image. The expression for generating the spatial texture disorder reference value is: ,in: is the quantized spatial texture chaos reference value, is the position in the image The local texture features of is the global texture average feature of the image, is a parameter that controls the perception of texture differences, k is the index of the local area currently being processed. As the calculation process proceeds, k It will traverse all local areas, N is the total number of local regions divided in the image.

[0028] Preferably, for blurred images, a fuzzy logic algorithm is used to adaptively control the initial exposure time according to the image blur prediction result of the image. The specific steps are as follows:

[0029] When processing blurred images, the image’s blur quantization coefficient is first calculated using a deep learning model. , quantify the blur degree of the image, in order to effectively adjust the exposure of the image, the blur quantization coefficient is compared with the preset blur quantization coefficient reference threshold Compare the two values ​​and calculate the exposure adjustment factor based on the difference between them based on fuzzy logic algorithm. , the specific exposure adjustment factor calculation formula is: ,in, is the gain coefficient of fuzzy inference, which is used to adjust the sensitivity of exposure time adjustment;

[0030] Based on the generated exposure adjustment factor Adjust the initial exposure time to optimize the shooting quality. The specific adjustment expression is: ,in, is the adjusted exposure time, that is, the exposure time actually used, is the initial exposure time, It is the exposure adjustment coefficient, which controls the amplitude of the exposure time change.

[0031] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0032] By combining deep learning models with image processing techniques, this invention accurately predicts image blur and effectively distinguishes normal from blurred images. For blurred images, adaptive exposure time control reduces the motion blur effect caused by long exposure times, thereby ensuring image clarity and improving the accuracy of feature extraction. This reduces the probability of misidentification and ensures that ticks are detected and removed promptly, effectively reducing the risk of host infection with pathogens. Furthermore, dynamic exposure time adjustment enables the image-driven tick detection comb to operate stably under a variety of environmental conditions, enhancing the device's robustness and reliability, addressing the image blur caused by rapid tick movement, and significantly improving the overall accuracy of the detection system. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction to the drawings required for use in the embodiments will be given below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0034] Figure 1 Schematic diagram of the module of the image-driven tick detection comb integrated with the deep learning model of the present invention.

[0035] Figure 2 This is a schematic diagram of the principle of the image-driven tick detection comb of the present invention. DETAILED DESCRIPTION

[0036] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of this disclosure will be thorough and complete and will fully convey the concepts of the example embodiments to those skilled in the art.

[0037] The present invention provides Figure 1 The image-driven tick detection comb integrated with the deep learning model shown includes an initial exposure time configuration module, an image data acquisition module, a feature extraction and fuzzy quantization module, a fuzziness prediction module, an image classification module, a normal image processing module, and a fuzzy image exposure adjustment module;

[0038] The initial exposure time configuration module first determines the initial exposure time that is suitable for the current environment and detection requirements, and configures the parameters of the micro camera;

[0039] This step ensures the microcamera captures images with appropriate exposure parameters, balancing light and detail, providing a clear foundation for images. Correct initial exposure settings are a prerequisite for subsequent image processing and inspection, and can, to a certain extent, prevent issues such as overly dark or overly bright images caused by improper exposure.

[0040] The initial exposure time is determined based on the static detection requirement, which is tailored to the current environment and detection needs. Static detection requires that the tick and its surroundings remain relatively still during tick detection, ensuring the micro-camera captures clear, motion-blur-free images. This facilitates accurate detection and identification of non-moving ticks.

[0041] Before image acquisition, an appropriate exposure time should be set based on the lighting conditions of the inspection environment, the characteristics of the inspection object, and the hardware parameters of the microcamera to ensure the optimal balance of image quality. Exposure time refers to the time the microcamera's photosensitive element is exposed to light, which directly affects the image's brightness, clarity, and dynamic range. If the exposure time is too short, the image may be too dark, resulting in a loss of detail; if the exposure time is too long, it may cause overexposure or produce smearing or blurring when the inspection object moves slightly. Therefore, under the premise of static detection, it is necessary to select an initial exposure time that can ensure sufficient brightness while minimizing noise and blur. This allows subsequent feature extraction and tick identification to be performed on the basis of high-quality images, improving the accuracy and stability of detection.

[0042] After taking a photo, the image data acquisition module fully captures all image information, including raw pixel data, brightness distribution, color channels, image metadata, and other auxiliary information. This step provides complete data support for subsequent feature extraction, allowing subsequent algorithms to fully utilize every part of the image to determine the degree of blur.

[0043] The feature extraction and blur quantification module, after obtaining complete image data, extracts feature information related to image blur through image processing techniques (such as edge detection, texture analysis, and frequency domain conversion). It then performs feature engineering on the extracted feature information to preliminarily quantify the image distortion caused by motion blur, providing a basis for subsequent blur determination.

[0044] Feature information related to image blur is extracted through image processing technology, where the extracted feature information includes the gradient change of the image edge and the distribution of the image spatial texture. Feature engineering processing is performed on the extracted feature information to generate edge gradient reference values ​​and spatial texture chaos reference values, respectively. The edge gradient reference values ​​and spatial texture chaos reference values ​​are used to preliminarily quantify the image distortion caused by motion blur, providing a basis for subsequent blur judgment.

[0045] When performing tick detection based on image-driven detection, smaller gradient changes at the edges of the image usually mean that the image is more blurry. When there is motion blur or other types of blur in the image, the edges and details that should be clearly visible will become blurred, causing the pixel gradients (i.e., brightness changes) in the image to become gentle. This is because the blurring process makes the edge transition less abrupt, thereby reducing the sharpness of the edge and the intensity of the gradient. In a normal, clear image, the gradient changes at the edge are large, which can clearly distinguish between objects and background. By calculating the gradient changes in each area of ​​the image, the blur of the image can be quantified. Images with smaller edge gradient changes usually indicate higher blur.

[0046] The specific steps for performing feature engineering on the extracted gradient change features of the image edge to generate the edge gradient reference value are as follows:

[0047] First, the image gradient is calculated to capture the local brightness changes of each pixel in the image, especially in the edge areas. High-order partial derivatives or the Laplacian operator are used to extract the gradient information of each point in the image. Here, the second-order derivative of the Laplacian operator is used to extract the gradient information of each point in the image. The extraction expression is as follows: ,in: Representing an image In position The Laplace operator result at is specifically the second-order derivative at that position. In other words, it represents the change in the rate of change of the brightness of the pixel point, which can describe the degree of curvature of the image in the horizontal and vertical directions, and thus capture the edge information of the image. Indicates that the image is at position The pixel value at and Represent the second-order derivatives in the horizontal and vertical directions respectively;

[0048] The purpose of this step is to calculate the brightness change of each point in the image, especially at the edge, where the brightness change is large, thereby helping to detect the edge information of the image.

[0049] Based on the gradient change of each pixel in the image, an edge gradient reference value is defined to quantify the blur of the image, especially the distortion caused by motion blur. The edge gradient reference value can reflect the overall blur of the image by accumulating the intensity of the gradient change and using an exponential decay function. The specific calculation expression is: ,in, represents the edge gradient reference value, represents the set of all pixels in the image, is the absolute value of the gradient. This formula combines the strength of the gradient change with its exponential decay, more sensitively capturing areas of weaker gradient changes caused by motion blur, thereby quantifying the blurriness of the image. A larger edge gradient reference value indicates clearer edges and less blur; conversely, a smaller edge gradient reference value indicates greater image blurriness.

[0050] The Edge Gradient Reference Value (EGR) indicates that a larger value, generated through feature engineering of the gradient change features extracted from image edges, indicates higher image clarity, while a smaller value indicates higher image blur. The Edge Gradient Reference Value quantifies edge clarity by calculating the gradient change of each pixel in the image and combining it with exponential decay. In a clear image, edges are typically sharper, pixel values ​​change more dramatically, and the gradient value is larger, resulting in a higher Edge Gradient Reference Value. In a blurry image, edges are blurred, pixel values ​​change more gradually, and the gradient is smaller, resulting in a lower Edge Gradient Reference Value. Therefore, the Edge Gradient Reference Value effectively reflects the degree of image blur: smaller values ​​indicate a blurrier image, while larger values ​​indicate a clearer image.

[0051] When performing image-driven tick detection, a high degree of disorder in the distribution of spatial texture in an image generally indicates a high degree of image blur. In image processing, texture distribution generally reflects the richness of detail in an image. When an image is clear, details and textures have a relatively regular and distinct distribution, with sharp edges and relatively orderly changes within local areas. In contrast, when an image is blurred, details are smoothed or blurred, and the texture distribution in the image becomes disordered and chaotic, with local changes being relatively smooth and lacking distinct edges and features. Especially in the case of motion blur, the directionality and structure of the texture may disappear, resulting in a significant increase in the disorder of the texture in the image. Therefore, the disorder of spatial texture can be used as an effective indicator of image blur, indicating a high degree of blur in the image and a significant loss of detail.

[0052] The specific steps of performing feature engineering on the distribution features of the extracted image spatial texture to generate spatial texture chaos reference values ​​are as follows:

[0053] First, the image is subjected to multi-scale wavelet transform or local binary pattern (LBP) extraction to obtain the texture features in each local area. These methods can capture subtle texture changes in the image, especially in high-frequency and detail areas. The image is decomposed into multiple frequency sub-bands using wavelet transform (such as Haar wavelet or Daubechies wavelet), and local texture features at different levels are obtained through these frequency sub-bands. The texture information of each local area can be represented as a feature vector, reflecting the local texture pattern of the area. To further quantify the disorder of these local texture features, a new local feature calculation formula is adopted, and the calculation expression is: ,in: Indicates the position in the image The local texture variation at the location reflects the difference between the texture of the surrounding area and the central area. Indicates the position in the image The texture value of Represents the texture value of the local area center, It is a parameter that adjusts the sensitivity of texture changes, and its value is usually 2;

[0054] In this way, the degree of texture change in each local area is calculated, the spatial texture features of the image are extracted from the global image, and the changes in local texture are quantified, thus providing a basis for subsequent fuzziness judgment.

[0055] After the local texture features are extracted, the degree of image disorder is quantified by calculating the local texture variation of each local area. Based on the quantified result of the local texture variation, a spatial texture disorder reference value is generated to evaluate the overall texture disorder of the image. The expression for generating the spatial texture disorder reference value is: ,in: is the quantized spatial texture chaos reference value, is the position in the image The local texture features of is the global texture average feature of the image, It is a parameter that controls the perception of texture difference, usually 3 is selected. k is the index of the local area currently being processed. As the calculation process proceeds, k It will traverse all local areas, N is the total number of local regions divided in the image;

[0056] By calculating the texture differences across all local regions, we obtain a spatial texture chaos reference value for the image. This value indicates the overall texture disorder of the image. As motion blur increases, local texture differences become more blurred and chaotic, increasing the spatial texture chaos reference value. This effectively quantifies the image distortion caused by motion blur, providing a basis for subsequent blur assessment.

[0057] The Spatial Texture Chaos Reference Value (SCR) indicates that a larger performance value for the SCR, generated through feature engineering of the extracted spatial texture distribution features of an image, generally indicates a higher degree of image blur. The SCR reflects the degree of disorder in the image's texture distribution. When an image is blurred, especially due to motion blur, details are lost, the regularity and structure of the texture become blurred, and local texture variations become more irregular and chaotic. Therefore, the SCR will increase significantly, indicating a higher degree of texture disorder. Conversely, when an image is sharp, the texture distribution is more orderly, details are clearly visible, and local texture variations are smoother and more regular, resulting in a lower SCR. Therefore, the SCR can serve as an effective indicator for measuring image blur: a larger performance value indicates a blurred image, while a smaller performance value indicates a sharper image.

[0058] The blur prediction module inputs the blur features processed by feature engineering into a pre-trained deep learning model and uses the deep learning model to predict the blur of the current image;

[0059] The edge gradient reference value and spatial texture chaos reference value after feature engineering are input into a pre-trained deep learning model, and a fuzzy quantization coefficient is generated based on the deep learning model. The blurriness of the current image is predicted by the fuzzy quantization coefficient.

[0060] A pre-trained deep learning model is one that has been trained on a large number of labeled datasets and specific tasks before actual predictions are made. In the context of image blur prediction, deep learning models typically use supervised learning, using labeled training data to learn how to determine the degree of image blur based on image features (such as edge gradient reference values ​​and spatial texture clutter reference values). The model continuously adjusts its internal parameters through backpropagation and optimization algorithms (such as gradient descent) until it can efficiently establish an accurate mapping between the input features and the actual image blur. During training, the model processes a variety of image types, identifies and extracts key features that affect image blur, and gradually learns how to integrate these features to effectively assess image blur in future predictions.

[0061] This "pre-training" process is usually completed during the development phase and is trained using large-scale, diverse image datasets. These datasets include images with varying degrees of blur, along with corresponding true blur labels. These labels can be manually annotated or quantitative values ​​obtained through other reliable methods (such as pixel offsets for motion blur). Through multiple iterations of training, the deep learning model can gradually improve its prediction accuracy and accurately infer the degree of blur in the image from input image features (such as edge gradients, spatial textures, etc.). Once the training process is complete, the model can make real-time predictions based on new image feature inputs. A trained deep learning model is essentially an optimized black box system that can effectively process and infer unknown data based on the knowledge learned during training.

[0062] In the application scenario of image blur prediction, the edge gradient reference value and spatial texture disorder reference value obtained after feature engineering are passed as input to the deep learning model. This model calculates an output value through forward propagation, commonly called the blur quantization coefficient. This blur quantization coefficient is a quantization result of the current image blur, indicating the degree of blur. Through prior training, the deep learning model has mastered how to extract key information about image blur from these feature values ​​and can even identify subtle signs of blur from complex combinations of image features. During the training phase, the model not only learns how to use basic features such as edge gradients and texture disorder to determine image blur, but also, through learning from a large number of samples, grasps the complex relationship between different image types and blur levels.

[0063] Specifically, the model may employ a convolutional neural network (CNN) or other deep learning architecture suitable for processing image data. These models typically extract spatial features from the image through multiple layers of convolution operations, use activation functions to model nonlinear relationships, and integrate information from different levels through fully connected layers. During training, the network continuously adjusts weights and biases to reduce the error between the predicted value and the actual blur. After training, the model is able to quickly and accurately predict blur when receiving new image features. In particular, based on edge gradient reference values ​​and spatial texture clutter reference values, the model can not only identify significant blur features but also identify the image blur type (such as motion blur, focal blur, etc.) based on patterns in historical training data, thereby deriving more accurate blur quantization coefficients.

[0064] This trained deep learning model exhibits high adaptability in practical applications, capable of handling image blur in diverse environments and across various types. For actual detection tasks, users simply input the model's edge gradient reference values ​​and spatial texture clutter reference values ​​for a new image, and the model generates predictions based on this knowledge. This deep learning-based prediction method is highly automated and intelligent, accurately predicting image blur without human intervention and demonstrating strong robustness across a wide range of blur scenarios.

[0065] The deep learning model is not specifically limited here, and can achieve the edge gradient reference value and spatial texture chaos reference values Perform comprehensive analysis to generate fuzzy quantization coefficients In order to realize the technical solution of the present invention, the present invention provides a specific implementation method; fuzzy quantization coefficient The generated expression is:

[0066]

[0067] , where 、 are the edge gradient reference values and spatial texture chaos reference values The preset scaling factor of 、 The preset proportional coefficient refers to the constant coefficient set in advance during the model construction process, which is used to adjust the edge gradient reference value. and spatial texture chaos reference values Fuzzy quantization coefficient These preset proportional coefficients are usually determined through experience or preliminary experiments, with the aim of balancing the influence of these two features in the final fuzzy quantization coefficient. Specifically, and Represents the weighting coefficients of edge gradients and spatial textures, controlling their relative importance in the prediction model. The purpose of setting these coefficients is to ensure that the model can appropriately adjust its calculation process based on different image characteristics. Usually, during the model training phase, these coefficients are further adjusted through optimization algorithms to make them most suitable for blur prediction of actual images.

[0068] It can be seen from the fuzzy quantization coefficient that the larger the performance value of the edge gradient reference value generated after feature engineering processing of the gradient change characteristics of the extracted image edge, the smaller the performance value of the spatial texture chaos reference value generated after feature engineering processing of the distribution characteristics of the extracted image spatial texture. That is, the smaller the performance value of the fuzzy quantization coefficient generated when the blurriness of the current image is predicted by the pre-trained deep learning model, the lower the blurriness of the image, and vice versa.

[0069] The image classification module divides the captured images into normal images and blurred images based on the prediction results of the deep learning model;

[0070] The fuzzy quantization coefficient generated when the blur of the current image is predicted by the pre-trained deep learning model is compared with the pre-set fuzzy quantization coefficient reference threshold to divide the captured image. The specific division steps are as follows:

[0071] If the generated fuzzy quantization coefficient is greater than the fuzzy quantization coefficient reference threshold, the captured image is classified as a fuzzy image; if the generated fuzzy quantization coefficient is less than or equal to the fuzzy quantization coefficient reference threshold, the captured image is classified as a normal image.

[0072] The normal image processing module, for normal images, continues to maintain the initial exposure time for the micro camera to take pictures, and performs subsequent tick feature extraction, identification and positioning;

[0073] The blurred image exposure adjustment module uses a fuzzy logic algorithm to adaptively adjust the initial exposure time for blurred images based on the image blur prediction results. This shortens the actual exposure time of the micro camera when taking pictures and reduces the blur effect caused by motion. Once the image is determined to be a normal image, subsequent tick feature extraction, identification, and positioning are performed.

[0074] For normal images, the initial exposure time is maintained for capturing and subsequent tick feature extraction, identification, and location. This ensures clear image quality with the appropriate exposure time, providing accurate image information for tick detection. A normal image is one that is unaffected by blur, with sharp details and distinct edges and textures. This provides a reliable foundation for tick feature extraction. By performing tick feature extraction on such images, tick features such as morphology, size, and position can be accurately identified, allowing for further localization and subsequent processing. This process ensures the detection system can efficiently and accurately identify ticks, avoiding false positives and missed detections, improving the accuracy and timeliness of detection, and effectively preventing the risk of tick transmission.

[0075] For blurred images, the fuzzy logic algorithm is used to adaptively control the initial exposure time based on the image blur prediction results. The specific steps are as follows:

[0076] When processing blurred images, the image’s blur quantization coefficient is first calculated using a deep learning model. The fuzzy quantization coefficient quantifies the degree of image blur. The larger the value, the higher the image blur. In order to effectively adjust the image exposure, the fuzzy quantization coefficient is compared with the preset fuzzy quantization coefficient reference threshold. Based on the fuzzy logic algorithm, the exposure adjustment factor is calculated according to the difference between the two. In this process, fuzzy reasoning rules are used, such as "the larger the fuzzy quantization coefficient, the shorter the exposure time should be", and the size of the adjustment factor is derived using fuzzy operations. The specific calculation formula for the exposure adjustment factor is: ,in, is the gain coefficient of fuzzy inference, which is used to adjust the sensitivity of exposure time adjustment;

[0077] The role of this step is to calculate the fuzzy quantization coefficients and the fuzzy quantization coefficient reference threshold The difference is used to quantify the degree of image blur and generate the corresponding exposure adjustment factor through fuzzy reasoning. If the image blur is high, the blur quantization coefficient will be greater than the fuzzy quantization coefficient reference threshold , the exposure adjustment factor As it increases, the system will shorten the exposure time to reduce blur.

[0078] Based on the generated exposure adjustment factor Adjust the initial exposure time to optimize the shooting quality. The specific adjustment expression is: ,in, is the adjusted exposure time, that is, the exposure time actually used, is the initial exposure time, Is the exposure adjustment factor, which controls the amplitude of exposure time change. If the image blur is high, the exposure adjustment factor The value of is larger, so the exposure time after adjustment will be significantly reduced;

[0079] This step adaptively adjusts the exposure time to reduce the effects of motion blur, ensuring a clear image even in fast-moving or blurry conditions. This adjustment ensures image quality and enables more precise feature extraction, particularly during the subsequent tick identification and location process, improving image usability and accuracy.

[0080] Image distortion caused by motion blur is reduced by adaptively adjusting the micro-camera's exposure time. When an image is identified as blurry, the system automatically shortens the exposure time based on the image's blur prediction using a fuzzy logic algorithm, thereby reducing the blurring effect caused by excessive exposure and ensuring optimized image clarity. This adjustment ensures that image details are presented more clearly when photographing fast-moving objects or in unstable environments, preventing motion blur from affecting feature extraction. Subsequent tick feature extraction, identification, and location are performed only when the image is confirmed to be clear and blurry, ensuring the accuracy and reliability of tick detection and avoiding misjudgments or missed detections due to blurry images.

[0081] The above solution effectively solves the problem of image blur caused by the rapid movement of ticks and significantly improves the accuracy of the detection system. First, by combining a deep learning model with image processing technology to accurately predict image blur, normal and blurred images can be accurately identified and classified. For blurred images, an adaptive exposure time control method is used to reduce the motion blur effect caused by long exposure, thereby ensuring image clarity and improving the accuracy of feature extraction. This optimization process reduces the occurrence of misjudgments, ensures that ticks can be detected and removed in a timely manner, and effectively reduces the risk of host infection with pathogens. At the same time, by dynamically adjusting the exposure time, the system can operate stably in various environments, enhancing the robustness and reliability of the device in practical applications.

[0082] like Figure 2 As shown:

[0083] The image-driven tick detection comb works by detecting target objects (such as ticks) through a micro camera. When the device is turned on, after pressing the power button, the micro camera begins to capture images of the target area and displays relevant information on the display. If the device detects the presence of the target object, the alarm light will light up, accompanied by an audible warning. The brush handle of the image-driven tick detection comb contains a collection tube for collecting ticks or other tiny objects, which can be adsorbed or collected into the tube through the bristles. During use, the user can keep abreast of the device's working status and detection results according to the prompts on the display. The overall workflow involves the coordination of multiple functions such as image capture, target recognition, alarm prompts, and object collection to achieve the purpose of accurately detecting and removing ticks.

[0084] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0085] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0086] The above description is merely illustrative of certain exemplary embodiments of the present invention. It goes without saying that those skilled in the art will be able to modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and description are illustrative in nature and should not be construed as limiting the scope of protection of the claims.

Claims

1. Image-driven tick detection comb integrated with deep learning model, characterized by: It includes initial exposure time configuration module, image data acquisition module, feature extraction and fuzzy quantization module, fuzziness prediction module, image classification module, normal image processing module, and fuzzy image exposure adjustment module; The initial exposure time configuration module first determines the initial exposure time that is suitable for the current environment and detection requirements, and configures the parameters of the micro camera; Image data acquisition module: After taking a photo, it fully obtains all information of the image and provides complete data support for subsequent feature extraction; The feature extraction and blur quantification module, after obtaining complete image data, uses image processing technology to extract feature information related to image blur, performs feature engineering on the extracted feature information, and preliminarily quantifies the image distortion caused by motion blur, providing a basis for subsequent blur determination. The blur prediction module inputs the blur features processed by feature engineering into a pre-trained deep learning model and uses the deep learning model to predict the blur of the current image; The image classification module divides the captured images into normal images and blurred images based on the prediction results of the deep learning model; The normal image processing module, for normal images, continues to maintain the initial exposure time for the micro camera to take pictures, and performs subsequent tick feature extraction, identification and positioning; The blurred image exposure adjustment module uses a fuzzy logic algorithm to adaptively adjust the initial exposure time for blurred images based on the image blur prediction results. This shortens the actual exposure time of the micro camera when taking pictures and reduces the blur effect caused by motion. Once the image is determined to be a normal image, subsequent tick feature extraction, identification, and positioning are performed. For blurred images, the fuzzy logic algorithm is used to adaptively control the initial exposure time based on the image blur prediction results. The specific steps are as follows: When processing blurred images, the image’s blur quantization coefficient is first calculated using a deep learning model. , quantify the blur degree of the image, in order to effectively adjust the exposure of the image, the blur quantization coefficient is compared with the preset blur quantization coefficient reference threshold Compare the two values ​​and calculate the exposure adjustment factor based on the difference between them based on fuzzy logic algorithm. , the specific exposure adjustment factor calculation formula is: ,in, is the gain coefficient of fuzzy inference, which is used to adjust the sensitivity of exposure time adjustment; Based on the generated exposure adjustment factor Adjust the initial exposure time to optimize the shooting quality. The specific adjustment expression is: ,in, is the adjusted exposure time, that is, the exposure time actually used, is the initial exposure time, It is the exposure adjustment coefficient, which controls the amplitude of the exposure time change.

2. The image-driven tick detection comb integrated with a deep learning model according to claim 1, characterized in that: The initial exposure time adapted to the current environment and detection needs is determined based on the static detection requirements. The static detection requirements mean that when detecting ticks, the ticks and their surroundings are preset to be in a relatively static state to ensure that the micro camera can capture clear images without motion blur.

3. The image-driven tick detection comb integrated with a deep learning model according to claim 1, characterized in that: Feature information related to image blur is extracted through image processing technology, where the extracted feature information includes the gradient change of the image edge and the distribution of the image spatial texture. Feature engineering processing is performed on the extracted feature information to generate edge gradient reference values ​​and spatial texture chaos reference values, respectively. The edge gradient reference values ​​and spatial texture chaos reference values ​​are used to preliminarily quantify the image distortion caused by motion blur, providing a basis for subsequent blur judgment.

4. The image-driven tick detection comb integrated with a deep learning model according to claim 3, characterized in that: The edge gradient reference value and spatial texture chaos reference value after feature engineering are input into a pre-trained deep learning model, and a fuzzy quantization coefficient is generated based on the deep learning model. The blurriness of the current image is predicted by the fuzzy quantization coefficient.

5. The image-driven tick detection comb integrated with a deep learning model according to claim 4, characterized in that: The fuzzy quantization coefficient generated when the blur of the current image is predicted by the pre-trained deep learning model is compared with the pre-set fuzzy quantization coefficient reference threshold to divide the captured image. The specific division steps are as follows: If the generated fuzzy quantization coefficient is greater than the fuzzy quantization coefficient reference threshold, the captured image is classified as a fuzzy image image; If the generated fuzzy quantization coefficient is less than or equal to the fuzzy quantization coefficient reference threshold, the captured image is classified as a normal image.

6. The image-driven tick detection comb integrated with a deep learning model according to claim 3, characterized in that: The specific steps for performing feature engineering on the extracted gradient change features of the image edge to generate the edge gradient reference value are as follows: The second-order derivative of the Laplace operator is used to extract the gradient information of each point in the image, and the gradient of the image is calculated to capture the local brightness change of each pixel in the image. The extracted expression is as follows: ,in: Representing an image In position The Laplace operator result at represents the change in the rate of change of pixel brightness, which can describe the curvature of the image in the horizontal and vertical directions, and thus capture the edge information of the image. Indicates that the image is at position The pixel value at and Represent the second-order derivatives in the horizontal and vertical directions respectively; Based on the gradient change of each pixel in the image, the edge gradient reference value is defined to quantify the blur of the image. The edge gradient reference value reflects the overall blur of the image by accumulating the intensity of the gradient change and using an exponential decay function. The specific calculation expression is: ,in, represents the edge gradient reference value, represents the set of all pixels in the image, is the absolute value of the gradient.

7. The image-driven tick detection comb integrated with a deep learning model according to claim 3, characterized in that: The specific steps of performing feature engineering on the distribution features of the extracted image spatial texture to generate spatial texture chaos reference values ​​are as follows: The image is decomposed into multiple frequency sub-bands using wavelet transform. Local texture features at different levels are obtained through the decomposed frequency sub-bands. The texture information of each local area is represented as a feature vector, which reflects the local texture pattern of the area. In order to further quantify the disorder of these local texture features, a new local feature calculation formula is adopted. The calculation expression is: ,in: Indicates the position in the image The local texture variation at the location reflects the difference between the texture of the surrounding area and the central area. Indicates the position in the image The texture value of Represents the texture value of the center of the local area, Parameters for adjusting sensitivity to texture changes; After the local texture features are extracted, the degree of image disorder is quantified by calculating the local texture variation of each local area. Based on the quantified result of the local texture variation, a spatial texture disorder reference value is generated to evaluate the overall texture disorder of the image. The expression for generating the spatial texture disorder reference value is: ,in: is the quantized spatial texture chaos reference value, is the position in the image The local texture features of is the global texture average feature of the image, is a parameter that controls the perception of texture differences, k is the index of the local area currently being processed. As the calculation process proceeds, k It will traverse all local areas, N is the total number of local regions divided in the image.

Images