Multi-model hair detection and hair density calculation method based on image input

Abstract

The invention provides a hair detection and hair density calculation method based on image input and multi-model analysis. The method comprises the following steps: acquiring a head image by using common imaging equipment; preprocessing the image; inputting the image into a plurality of analysis models for independent analysis; hair detection and hair quantity density calculation precision is optimized through a self-research fusion algorithm and high-order multi-task analysis; outputting the hair quantity estimation, the partitioned hair density, the scalp coverage rate, the health score, the thermodynamic diagram and the trend curve, and generating personalized nursing suggestions. Further, according to the method, images and related data are periodically acquired and stored as historical data, and the historical data are periodically analyzed to generate a hair health change trend; the system can access an additional analysis model or external sensor data through an expansion interface to realize long-term tracking and comprehensive evaluation, so that the evaluation precision and the use value of a user are improved.

Description

Technical Field

[0001] This invention relates to the field of hair health detection and image analysis technology, and in particular to a hair detection and hair density calculation method based on image input and multi-model analysis. It can use ordinary imaging equipment to acquire images, and complete hair segmentation, hair volume estimation, density calculation and trend analysis through a self-developed multi-model fusion algorithm, and generate a visual report and personalized care suggestions. Background Technology

[0002] Existing hair detection technologies rely on polarized light devices, dermatoscopes, or high-magnification microscopic imaging instruments, which are costly, bulky, and complex to operate, and are usually only used in professional institutions or medical settings.

[0003] Image-based hair analysis methods often rely on a single model and cannot simultaneously achieve multi-dimensional analysis of hair quantity, density, health score, and trend prediction, lacking continuity and scalability.

[0004] With the development of mobile devices and AI technology, ordinary users can easily obtain head images, but there is still a lack of a hair health assessment solution that does not require professional equipment and can be tracked over a long period of time. Summary of the Invention

[0005] To address the problems of high cost, complex operation, and limited functionality in existing technologies, this invention provides a method for hair health assessment that does not require specialized equipment and is based on ordinary images. Furthermore, it improves detection accuracy and reliability through a self-developed fusion algorithm and advanced multi-task analysis. Technical solution

[0006] This invention provides a method for hair detection and hair density calculation based on image input and multi-model analysis, comprising the following steps: S1, Image Acquisition

[0007] Head images can be captured using smartphones, regular cameras, or other devices, and support various lighting conditions. S2, Image Preprocessing

[0008] The acquired images are denoised, enhanced, and normalized, with optional illumination correction and contrast adjustment. S3, Multi-model Analysis

[0009] Images are input into multiple artificial intelligence models for analysis, including: >Hair and scalp segmentation model, used to identify hair and scalp regions; >A hair quantity and density estimation model used to calculate the local and overall hair quantity and distribution density; Hair health scoring and trend prediction models are used to assess hair health status and predict future trends. S4, Self-developed Fusion Algorithm and High-Order Multitask Analysis

[0010] The outputs of each model are weighted and integrated to handle uncertainties and prediction biases, and the final hair health assessment result is generated.

[0011] Advanced multi-task analytics can simultaneously optimize quantity, density, and health scores, improving overall assessment accuracy. S5. Results Output and Visualization

[0012] Output multi-dimensional assessment content, including hair volume estimation, hair density in different areas, scalp coverage, health score, heat map, and trend curve.

[0013] Personalized care recommendations and long-term trend analysis reports are generated based on historical image data.

[0014] The results can be displayed on mobile or web interfaces and can be stored for long-term tracking and periodic comparative analysis. S6

[0015] The system acquires user head images and related data according to a preset period (such as daily, weekly, or monthly) and stores them as historical data.

[0016] Historical data can be recorded through version control to ensure that data at different points in time can be compared.

[0017] The periodic analysis module performs statistical processing on the number, density, and health score of hair in historical data to generate a hair health trend curve.

[0018] The system can access other analysis models or external sensor data through expansion interfaces to enhance long-term tracking capabilities.

[0019] Personalized care recommendations are generated based on historical trends and current test results, and these recommendations are displayed in conjunction with the trend results to improve user experience and reference value. Beneficial effects

[0020] This invention requires no special equipment, is low in cost, and is easy to operate, making it suitable for home self-testing, medical auxiliary diagnosis, and hair care institutions.

[0021] Multi-model fusion and high-order multi-task analysis improve the accuracy and reliability of hair detection.

[0022] It can simultaneously provide information on hair quantity, density, health score, trend prediction, and personalized care recommendations.

[0023] It can track users' historical images over a long period of time, enabling short-term, medium-term, and long-term trend analysis.

[0024] It has strong scalability and can connect to more models or external sensor data to achieve comprehensive hair health management.

[0025] Periodic data collection and historical trend analysis can further improve the accuracy of assessments and the reliability of personalized care recommendations. Detailed Implementation

[0026] Users take images of the front, middle, back, and side of the head using ordinary smart devices.

[0027] The system performs noise reduction, enhancement, and standardization on the images.

[0028] The multi-model analysis module processes images and outputs hair segmentation, quantity and density estimation, health score and trend prediction results.

[0029] The self-developed fusion algorithm performs weighted integration of the outputs of each model to form the final evaluation result.

[0030] The assessment results generate heat maps, trend charts, and personalized care recommendations, which users can view or export on mobile devices.

[0031] The above embodiments are only used to illustrate the technical solutions of the present invention and do not constitute a limitation on the scope of protection of the present invention.

[0032] The system collects user head images and related information at set intervals and stores the data in a historical database.

[0033] The periodic analysis module performs statistical analysis on historical data, generates trend curves for hair quantity, density, and health score, and updates long-term trend reports.

[0034] By extending the interface to access external analysis models or sensor data, data fusion can be achieved, enhancing long-term tracking capabilities.

[0035] Personalized care recommendations are generated by combining historical trends and the latest test results, and the trend changes and care plans are displayed on the user's mobile device or web page. Attached Figure Description Figure 1 This is a schematic diagram of the overall process of a hair detection and hair density calculation method based on image input and multi-model analysis according to the present invention. Figure 2 This is a schematic diagram of the image acquisition and preprocessing process in this invention; Figure 3 This is a schematic diagram of the hair volume feature extraction and multi-model fusion analysis module in this invention; Figure 4 This is a schematic diagram of the logical structure of the self-developed fusion algorithm and multi-task analysis module of this invention; Figure 5This is a schematic diagram of the visualization of the evaluation report results of this invention (mobile interface); Figure 6 This is a schematic diagram illustrating the system expansion and long-term tracking functions of the present invention. Figure 1 This is a schematic diagram of the overall process of the hair volume detection method of the present invention, which specifically includes the following modules: S1: Image acquisition module, used to acquire panoramic and directional images of the head; S2: Image preprocessing module, which performs noise reduction, segmentation and enhancement processing on the acquired images; S3: Multi-model analysis module, which uses convolutional neural networks and object detection models to analyze hair quantity and density; S4: Self-developed fusion algorithm and advanced multi-task analysis module (innovation point), used to fuse results from multiple models and perform advanced analysis, marked in red to highlight the technological innovation; S5: Results Output and Visualization Module, which generates visual reports from the analysis results; S6 (optional): Extended / historical data analysis module, connected to S5 via a dashed arrow, enabling long-term tracking and historical data analysis functions. This diagram fully illustrates the technical process from image acquisition to result visualization and extended analysis, providing a clear and intuitive flow demonstration of the core method of this invention. Figure 2 This is a schematic diagram of the image acquisition and preprocessing process. like Figure 2 As shown, the present invention uses ordinary imaging equipment to acquire images of the head region of the object being detected. The image acquisition does not rely on special equipment or a fixed shooting environment and can adapt to different lighting conditions, shooting angles and differences in terminal devices. The acquired raw images enter the image preprocessing module, where noise reduction, image enhancement, size and resolution normalization, and illumination correction are performed sequentially to eliminate the influence of environmental factors and differences in imaging equipment on the analysis results. The above preprocessing operations ensure that images collected at different time points maintain consistency in both visual and numerical features, providing unified data input conditions for subsequent hair volume feature extraction, long-term tracking and management, and prediction of hair volume change trends. Figure 3 This is a schematic diagram of hair volume feature extraction and multi-model fusion analysis. like Figure 3 As shown, via Figure 2 The preprocessed image is used as input to the hair volume feature extraction module, which performs feature processing on the hair information in the head image to extract multidimensional hair volume feature data, including hair density features, hair thickness features, hair distribution features, and visible scalp proportion features. The hair volume feature data is input to the multi-model fusion analysis module, which performs parallel analysis on the hair volume feature through image analysis model, statistical analysis model and trend analysis model, and fuses the output results of each model to generate a comprehensive analysis result. Based on the comprehensive analysis results, the system outputs the current hair volume assessment results, the historical trend analysis results of hair volume changes, and the prediction results of hair volume changes, thereby achieving long-term tracking and management of hair volume status and predictive analysis of hair volume change trends. Figure 4 This is a schematic diagram of the self-developed fusion algorithm and high-order multi-task analysis. like Figure 4 As shown, this invention uses multiple analysis models for head images to independently analyze hair features, hair structure, and scalp condition, obtaining multiple model output results. The output results of the multiple models are input into the self-developed fusion algorithm module. By weighting, evaluating confidence, and handling uncertainty of the results of each model, a unified and stable fusion analysis result is generated. Based on this, the fusion analysis results are further input into the advanced multi-task analysis module to simultaneously complete multi-dimensional analysis tasks such as hair quantity assessment, hair density analysis, and hair health scoring. The final result is a comprehensive hair health assessment that reflects the current hair status of the tested individual and supports subsequent trend analysis and personalized assessment. Figure 5 A diagram to visualize the results. like Figure 5 As shown, this invention presents the comprehensive hair health assessment results in a visual form, including hair volume estimation results, regional hair density maps, scalp coverage heatmaps, and hair health scores and their trend curves. The visualization results are used to intuitively reflect the hair distribution and overall health level of the tested object in different areas, making it easier for users to understand the test results. Meanwhile, based on the multidimensional assessment results, the system further generates personalized nursing suggestions and displays these suggestions in conjunction with the visualization results to enhance the user experience and practical guidance value. Figure 6 This diagram illustrates system expansion and long-term tracking. like Figure 6 As shown, the system of the present invention includes a data acquisition and extended access layer, a data storage and periodic analysis module, and a long-term tracking and historical comparison module. The data acquisition and extended access layer is used to access user head image data and reserves an extended analysis model interface and an external sensor data interface to support system function expansion. The data storage and periodic analysis module is used to store and manage user historical data, and to analyze and process the data according to a preset period. Based on the periodic analysis results, the long-term tracking and historical comparison module is used to analyze the changing trends of hair health status, realize long-term tracking and comprehensive evaluation of the user's hair health status, thereby improving the practicality and scalability of the method of the present invention.

Claims

1. Claim 1: A method for hair detection and hair density calculation based on image input and multi-model analysis, characterized in that, Includes the following steps: S1. Acquire head images using ordinary imaging equipment; S2. Preprocess the image, including denoising, enhancement and normalization. S3. Input the image into multiple artificial intelligence models for analysis, including a hair and scalp segmentation model, a hair quantity and density estimation model, and a hair health score and trend prediction model. S4. We use a self-developed fusion algorithm to perform weighted integration of the output results of multiple models, and use high-order multi-task analysis to optimize the accuracy of hair detection and hair density calculation. S5. Outputs include multi-dimensional assessments such as hair volume estimation, zoned hair density, scalp coverage, health score, heat map, and trend curve, and generates personalized care recommendations. S6. Periodically acquire the images and related data, store them as historical data, and periodically analyze the historical data to generate hair health change trends. Access additional analysis models or external sensor data through an extended interface to achieve long-term tracking and comprehensive evaluation.

2. The dependent claim, according to the method of main claim 1, wherein the preprocessing of step S2 includes image color normalization, noise reduction, and contrast enhancement.

3. Dependent claim, the method according to main claim 1, wherein the segmentation model in step S3 employs a general convolutional neural network structure for identifying hair regions and scalp regions.

4. Dependent claims, the method according to main claim 1, wherein the quantity and density estimation model of step S3 improves detection accuracy through high-order feature extraction and attention mechanisms.

5. Dependent claim, according to the method of main claim 1, wherein the self-developed fusion algorithm in step S4 performs weighted processing on the uncertainty of each model output to achieve optimal integration of multi-model results.

6. A dependent claim, the method according to main claim 1, wherein the personalized care recommendation in step S5 is generated based on the user's historical image data and multidimensional assessment results.

7. A dependent claim, the method according to main claim 1, wherein the periodic analysis of step S6 comprises: Statistical calculations are performed on historical data of hair quantity, density, and health scores, and trend curves are generated.

8. A dependent claim, the method according to main claim 1, wherein the extended interface is used to access additional analysis models, including but not limited to scalp environment analysis models or wearable sensor data interfaces.

9. A dependent claim, the method according to main claim 1, wherein personalized care recommendations generated from long-term tracking are displayed in association with historical trend results to improve user experience and assessment accuracy.

10. A dependent claim, the method according to main claim 1, wherein historical data storage includes data version management to ensure that periodic analysis can compare data at different points in time.

Images