Humanoid model posture detection system and method based on image analysis
Through the humanoid model posture detection system based on image analysis, the deep learning model is used to automatically identify and calibrate model postures, and the problems of high cost of traditional clothing photography, large time consumption and poor posture consistency are solved, and efficient and flexible clothing photography process and professionalism are achieved. Improved degree.
Patent Information
- Application Number
- CN202411490290.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-24
- Publication Date
- 2024-11-29
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Traditional clothing photography relies on live models and professional photographers, resulting in high costs, high time consumption, poor posture repetition, difficulty in post-processing and limited model availability, affecting shooting efficiency and flexibility.
The humanoid model pose detection system based on image analysis is adopted, including image acquisition, preprocessing, key point detection and pose estimation modules, to identify model key points and calculate pose parameters through deep learning models to realize automated pose detection and calibration.
It significantly improves the accuracy of model pose recognition, reduces manual intervention, shortens shooting cycle, reduces labor costs, ensures posture consistency and professionalism, improves shooting flexibility and display effects, and adapts to changing shooting environments and clothing styles.
Smart Images

Figure CN119049132A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of clothing photography, and in particular to a humanoid model posture detection system and method based on image analysis. Background Art
[0002] Fashion photography is an important means for the fashion industry to showcase design works, in which the model's posture is crucial to convey the style and characteristics of clothing. Traditional fashion photography relies on professional photographers and real models, and manually adjusts the model's posture to capture the best visual effect. However, this shooting method has some limitations: Cost and time consumption: The cost of real models and professional photography teams is high, and the shooting process is time-consuming, which is not conducive to rapid response to market changes. Pose Repeatability: Maintaining pose consistency is challenging when there are many repetitions of the same garment at different angles or poses.
[0003] Difficulty of post-processing: In real-life photography, post-processing requires a lot of work to modify the model's posture or background, which increases the workload.
[0004] Limited by model availability: The shooting plan is limited by the model's schedule, affecting the shooting progress and flexibility. Summary of the invention
[0005] The main purpose of the present invention is to provide a humanoid model posture detection system and method based on image analysis, aiming to solve the technical problems of cost and time consumption, difficulty in post-processing of posture repeatability and limitation of model availability.
[0006] In order to achieve the above-mentioned invention object, the first aspect of the present invention proposes a humanoid model posture detection system based on image analysis, including: an image acquisition module, a preprocessing module, a key point detection module, a posture estimation module and a result output and feedback module; The image acquisition module is used to acquire the original model image; The pre-processing module is used to perform preliminary processing on the collected model image; The key point detection module is used to identify key parts of the model based on the processed model image; The posture estimation module is used to calculate the posture parameters of the model according to the position information of the key parts and convert the posture parameters into a standard posture; The result output and feedback module is used to display the result of posture detection to the user.
[0007] Furthermore, the image acquisition module includes a camera control unit and a shooting environment setting unit, wherein the camera control unit is responsible for controlling the parameter settings of the camera, and the shooting environment setting unit is used to adjust the shooting angle, lighting conditions and background arrangement; Furthermore, the preprocessing module includes an image denoising unit, an image scaling unit, an image standardization unit, an edge detection unit and a background separation unit. The image denoising unit is used to remove noise interference in the image and improve the image quality. The image scaling unit is used to adjust the image to a suitable size to facilitate the efficient operation of the subsequent algorithm. The image standardization unit is used to make different images consistent in color and brightness. The edge detection unit is used to extract the image edge. The background separation unit is used to separate the image background. Furthermore, the key point detection module includes a deep learning model training unit, a key point prediction unit and a multi-task learning coordination unit, wherein the deep learning model training unit is used to train the key point detection model, the key point prediction unit is used to predict the key point position, and the multi-task learning coordination unit is used to coordinate multi-task learning; Further, the posture estimation module includes a posture calculation unit and a standard posture conversion unit, the posture calculation unit is used to obtain posture parameters, and the standard posture conversion unit is used to convert into a standard posture model; Furthermore, the result output and feedback module includes a visualization display unit, a feedback information generation unit and a parameter adjustment unit, wherein the visualization display unit is used to intuitively present posture information, the feedback information generation unit is used to generate feedback prompts, and the parameter adjustment unit is used to automatically adjust relevant parameters; In a second aspect, a method for detecting a humanoid model's posture based on image analysis is provided, the method comprising: Collect original model images; Performing preliminary processing on the collected model image; identifying key parts of the model based on the processed model image; Calculating the posture parameters of the model according to the position information of the key parts, and converting the posture parameters into a standard posture; Display the results of gesture detection to the user.
[0008] The method specifically comprises the following steps: Step 1: Image acquisition: The original model image data is obtained through the image acquisition module to ensure that the image covers the complete posture of the model and the resolution and clarity of the image meet the basic requirements of subsequent processing; Step 2: Preprocessing: After completing the image acquisition, the image denoising unit is used to remove noise through filtering methods, the image scaling unit adjusts the image size, the image standardization unit unifies the relevant attributes of the image, and the edge detection unit extracts the contour, and the background separation unit implements the background separation operation; Step 3: Key point detection: After preprocessing, the deep learning model training unit uses a large amount of labeled data to train the model, and then the key point prediction unit predicts the key point position. The multi-task learning coordination unit continuously optimizes the effect of multi-task learning during the training and prediction process. Step 4: Posture estimation: Based on the results of key point detection, the posture calculation unit calculates the posture parameters according to the key point positions, and the standard posture conversion unit converts the posture parameters into the standard posture; Step 5: Result output and feedback: Finally, the visualization display unit clearly presents the posture information to the user. Once the detection result does not meet the requirements, the feedback information generation unit will generate a prompt, and the parameter adjustment unit will automatically adjust the relevant parameters and templates accordingly.
[0009] Furthermore, the step of identifying key parts of the model based on the processed model image includes: Obtain multiple images containing various model poses, and annotate the model poses in each image, including marking the coordinate positions of key body parts; Select a deep learning model architecture suitable for key point detection and pose estimation tasks, normalize the image, normalize the pixel values to a specific range, and perform data augmentation operations; Divide the labeled data into training set, validation set and test set, select the Adagrad optimization algorithm, and set the initial learning rate and learning rate decay strategy parameters; Iteratively input the images in the training set into the model, obtain the predicted key point coordinates through forward propagation operation, use the loss function to calculate the deviation between the predicted value and the actual annotation, perform back propagation based on the loss value, calculate the gradient of the model parameters, and use the optimization algorithm to update the model parameters according to the gradient to gradually reduce the loss; Based on the performance on the monitoring validation set, the early stopping method is used to terminate the training to avoid overfitting, and the model with the best performance on the validation set is selected as the final training model; Analyzing the processed model image based on the final training model to identify key parts of the model; In a third aspect, the invention provides the following technical solution: a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, a humanoid model posture detection method based on image analysis is implemented; In a fourth aspect, the present invention provides the following technical solution: a readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the above-mentioned human model posture detection method based on image analysis. Beneficial Effects
[0010] 1. The present invention significantly improves the recognition accuracy of model postures under complex backgrounds by utilizing deep learning algorithms, realizes a fully automated processing flow from image acquisition to posture analysis, reduces manual intervention, shortens the shooting cycle through automated posture detection and calibration, and speeds up the work efficiency of clothing photography.
[0011] 2. The present invention reduces the reliance on professional photographers and post-editors, effectively reduces labor costs, ensures the consistency and standardization of model postures in different shoots, and improves the professionalism of clothing display. The user-customized interface enables users to easily set and adjust model postures, improves user experience, and can adapt to changing shooting environments and different clothing styles.
[0012] 3. The present invention allows users to instantly obtain detection results and make adjustments through a real-time feedback mechanism, thereby improving the flexibility of shooting. The system continuously optimizes the posture detection algorithm through continuous learning, thereby improving the accuracy and efficiency of detection. The built-in data security measures ensure the protection of image data and comply with privacy protection standards. The scalable design of the system facilitates the integration and upgrading of future technologies and ensures long-term technological leadership. Through precise posture detection and automated adjustment, the online and offline display effects of clothing are enhanced. BRIEF DESCRIPTION OF THE DRAWINGS
[0013] Figure 1 It is a schematic diagram of a framework of a humanoid model posture detection system based on image analysis of the present invention; Figure 2 The present invention is a flowchart of a method for detecting a humanoid model's posture based on image analysis.
[0014] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION
[0015] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.
[0016] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first" and "second" are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined.
[0017] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection, a direct connection, or an indirect connection through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.
[0018] In the present invention, unless otherwise clearly specified and limited, a first feature being "above" or "below" a second feature may include that the first and second features are in direct contact, or may include that the first and second features are not in direct contact but are in contact through another feature between them. Moreover, a first feature being "above", "above" and "above" a second feature includes that the first feature is directly above and obliquely above the second feature, or simply indicates that the first feature is higher in level than the second feature. A first feature being "below", "below" and "below" a second feature includes that the first feature is directly below and obliquely below the second feature, or simply indicates that the first feature is lower in level than the second feature.
[0019] Reference Figure 1 , an embodiment of the present invention is a humanoid model posture detection system based on image analysis, including: an image acquisition module, a preprocessing module, a key point detection module, a posture estimation module and a result output and feedback module; An image acquisition module, used to acquire original model images; A preprocessing module, used for preliminary processing of the collected model images; A key point detection module, used to identify key parts of the model based on the processed model image; The posture estimation module is used to calculate the posture parameters of the model based on the position information of key parts and convert the posture parameters into standard posture; The result output and feedback module is used to display the results of posture detection to the user.
[0020] The image acquisition module includes a camera control unit and a shooting environment setting unit. The camera control unit is responsible for controlling the camera parameter settings, adjusting the focal length, aperture size, sensitivity and shutter speed to ensure that the acquired images have clear details and appropriate brightness. The shooting environment setting unit is used to adjust the shooting angle, lighting conditions and background layout, so as to create a rich and diverse shooting environment that meets actual needs and acquire a large number of multi-dimensional and multi-angle model images; The preprocessing module includes an image denoising unit, an image scaling unit, an image standardization unit, an edge detection unit and a background separation unit. The image denoising unit is used to remove noise interference in the image and improve the image quality. The image scaling unit is used to adjust the image to a suitable size to facilitate the efficient operation of the subsequent algorithm. The image is intelligently adjusted to the most suitable size to facilitate the efficient operation of the subsequent algorithm without losing key information. The image standardization unit is used to make different images consistent in color and brightness to reduce errors caused by image differences. The edge detection unit is used to extract image edges. The background separation unit is used to separate the image background. The Canny edge detection algorithm is used to accurately extract image edges, highlight the model's contour, and efficiently separate the image background to ensure the clarity and prominence of the model's main body. The key point detection module includes a deep learning model training unit, a key point prediction unit and a multi-task learning coordination unit. The deep learning model training unit is used to collect massive and diverse annotated image data, covering images of models with various postures, body shapes and clothing styles, and accurately annotate the key points in these images using professional annotation tools. Subsequently, the architecture of the residual network or the deep learning of the network architecture based on the attention mechanism is used to train the model. The key point prediction unit is used to predict the key point position. The multi-task learning coordination unit is used to coordinate multi-task learning. By intelligently allocating the weights of different tasks, optimizing the design of the loss function, and effectively coordinating the relationship between different subtasks, the effect of multi-task learning is continuously optimized, and the overall performance and generalization ability of the model are improved. The posture estimation module includes a posture calculation unit and a standard posture conversion unit. The posture calculation unit is used to obtain posture parameters and convert key points into posture parameters. The posture calculation unit uses complex geometric algorithms and machine learning techniques to accurately calculate the angle, direction and relative position relationship posture parameters of each part of the model's body based on the coordinate position of the key points. The standard posture conversion unit is used to convert to a standard posture model. The result output and feedback module includes a visualization display unit, a feedback information generation unit and a parameter adjustment unit. The visualization display unit is used to intuitively present posture information, the feedback information generation unit is used to generate feedback prompts, and the parameter adjustment unit is used to automatically adjust relevant parameters; The second aspect refers to Figure 2 , a humanoid model posture detection method based on image analysis, the method comprising: Collect original model images; Perform preliminary processing on the collected model images; Identify key parts of the model based on the processed model image; Calculate the model's posture parameters based on the position information of key parts and convert the posture parameters into standard posture; Display the results of gesture detection to the user.
[0021] The above method specifically comprises the following steps: Step 1: Image acquisition: The original model image data is obtained through the image acquisition module to ensure that the image covers the complete posture of the model and the resolution and clarity of the image meet the basic requirements of subsequent processing; Step 2: Preprocessing: After completing the image acquisition, the image denoising unit is used to remove noise through filtering methods, the image scaling unit adjusts the image size, the image standardization unit unifies the relevant attributes of the image, and the edge detection unit extracts the contour, and the background separation unit implements the background separation operation; Step 3: Key point detection: After preprocessing, the deep learning model training unit uses a large amount of labeled data to train the model, and then the key point prediction unit predicts the key point position. The multi-task learning coordination unit continuously optimizes the effect of multi-task learning during the training and prediction process. Step 4: Posture estimation: Based on the results of key point detection, the posture calculation unit calculates the posture parameters according to the key point positions, and the standard posture conversion unit converts the posture parameters into the standard posture; Step 5: Result output and feedback: Finally, the visualization display unit clearly presents the posture information to the user. Once the detection result does not meet the requirements, the feedback information generation unit will generate a prompt, and the parameter adjustment unit will automatically adjust the relevant parameters and templates accordingly.
[0022] The steps of identifying key parts of the model based on the processed model image include: Obtain multiple images containing various model poses, and annotate the model poses in each image, including marking the coordinate positions of key body parts; Select a deep learning model architecture suitable for key point detection and pose estimation tasks, normalize the image, normalize the pixel values to a specific range, and perform data augmentation operations; Divide the labeled data into training set, validation set and test set, select the Adagrad optimization algorithm, and set the initial learning rate and learning rate decay strategy parameters; Iteratively input the images in the training set into the model, obtain the predicted key point coordinates through forward propagation operation, use the loss function to calculate the deviation between the predicted value and the actual annotation, perform back propagation based on the loss value, calculate the gradient of the model parameters, and use the optimization algorithm to update the model parameters according to the gradient to gradually reduce the loss; Based on the performance on the monitoring validation set, the early stopping method is used to terminate the training to avoid overfitting, and the model with the best performance on the validation set is selected as the final training model; Analyze the processed model images based on the final training model to identify the key parts of the model; A third embodiment of the present invention is based on the same inventive concept and proposes a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, a method for detecting a humanoid model posture based on image analysis in the above embodiment is implemented; The fourth embodiment of the present invention is based on the same inventive concept. The present invention proposes a terminal, which includes: a processor and a memory; the processor and the memory communicate with each other; the memory is used to store instructions; the processor is used to execute the instructions in the memory to execute a human model posture detection method based on image analysis of the above embodiment.
[0023] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.
Claims
1. A humanoid model posture detection system based on image analysis, characterized in that: include: Image acquisition module, preprocessing module, key point detection module, posture estimation module and result output and feedback module; The image acquisition module is used to acquire the original model image; The pre-processing module is used to perform preliminary processing on the collected model image; The key point detection module is used to identify key parts of the model based on the processed model image; The posture estimation module is used to calculate the posture parameters of the model according to the position information of the key parts and convert the posture parameters into a standard posture; The result output and feedback module is used to display the result of posture detection to the user.
2. The humanoid model posture detection system based on image analysis according to claim 1, characterized in that: The image acquisition module includes a camera control unit and a shooting environment setting unit. The camera control unit is responsible for controlling the parameter settings of the camera, and the shooting environment setting unit is used to adjust the shooting angle, lighting conditions and background arrangement.
3. The humanoid model posture detection system based on image analysis according to claim 1, characterized in that: The preprocessing module includes an image denoising unit, an image scaling unit, an image standardization unit, an edge detection unit and a background separation unit. The image denoising unit is used to remove noise interference in the image, the image scaling unit is used to adjust the image to a suitable size, the image standardization unit is used to make different images consistent in color and brightness, the edge detection unit is used to extract image edges, and the background separation unit is used to separate the image background.
4. The humanoid model posture detection system based on image analysis according to claim 1, characterized in that: The key point detection module includes a deep learning model training unit, a key point prediction unit and a multi-task learning coordination unit. The deep learning model training unit is used to train the key point detection model, the key point prediction unit is used to predict the key point position, and the multi-task learning coordination unit is used to coordinate multi-task learning.
5. The humanoid model posture detection system based on image analysis according to claim 1, characterized in that: The posture estimation module includes a posture calculation unit and a standard posture conversion unit. The posture calculation unit is used to obtain posture parameters, and the standard posture conversion unit is used to convert into a standard posture model.
6. The humanoid model posture detection system based on image analysis according to claim 1, characterized in that: The result output and feedback module includes a visualization display unit, a feedback information generation unit and a parameter adjustment unit. The visualization display unit is used to present posture information, the feedback information generation unit is used to generate feedback prompts, and the parameter adjustment unit is used to automatically adjust relevant parameters.
7. A method for detecting posture of a humanoid model based on image analysis, characterized in that: The humanoid model posture detection system based on image analysis according to any one of claims 1 to 6, the method comprising: Collect original model images; Performing preliminary processing on the collected model image; identifying key parts of the model based on the processed model image; Calculating the posture parameters of the model according to the position information of the key parts, and converting the posture parameters into a standard posture; Display the results of gesture detection to the user.
8. The method for detecting posture of a humanoid model based on image analysis according to claim 7, characterized in that: The step of identifying key parts of the model based on the processed model image comprises: Obtain multiple images containing various model poses, and annotate the model poses in each image, including marking the coordinate positions of key body parts; Select a deep learning model architecture suitable for key point detection and pose estimation tasks, normalize the image, normalize the pixel values to a specific range, and perform data augmentation operations; Divide the labeled data into training set, validation set and test set, select the Adagrad optimization algorithm, and set the initial learning rate and learning rate decay strategy parameters; Iteratively input the images in the training set into the model, obtain the predicted key point coordinates through forward propagation operation, use the loss function to calculate the deviation between the predicted value and the actual annotation, perform back propagation based on the loss value, calculate the gradient of the model parameters, and use the optimization algorithm to update the model parameters according to the gradient to gradually reduce the loss; Based on the performance on the monitoring validation set, the early stopping method is used to terminate the training to avoid overfitting, and the model with the best performance on the validation set is selected as the final training model; The processed model image is analyzed based on the final training model to identify key parts of the model.
9. A computer device 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 computer program, the method for detecting the posture of a human model based on image analysis as described in any one of claims 7 to 8 is implemented.
10. A readable storage medium, characterized in that: The readable storage medium stores a computer program, and when the computer program is executed by the processor, the method for detecting the posture of a human model based on image analysis as claimed in claim 7 is implemented.
Citation Information
Patent Citations
Posture estimation and human body analysis system based on multi-task deep learning
CN111738091A
Multi-person body posture estimation method based on deep learning
CN117831121A
Human body posture recognition method and recognition system based on TensorFlow.js
CN118247810A
Gaze estimation system and method thereof
EP4213105A1
System and method for reconstructing a 3D human body under clothing
US20210375045A1
