A crowd density recognition method and system based on a dynamic optimization diffusion model

Abstract

The application discloses a kind of crowd density identification method and system based on dynamic optimization diffusion model, method includes S1, layer-by-layer diffusion simulation: by gradually increasing the noise of image, simulate the process of image data gradually decaying, obtain noisy image;S2, enhanced learning correction: using denoising process and image quality maintenance strategy, restore high-quality crowd density map from noisy image;S3, latent space dynamic optimization: based on crowd density map, dynamically adjust the parameters of crowd density identification model in latent space, obtain crowd density identification model under optimal parameter state.The advantage is: improve the accuracy and reliability of crowd density identification method and system under complex environmental conditions.Reduce the dependence on high-performance computing resources, so that crowd density identification method and system can be more efficiently run in resource-constrained environment.

Description

Technical Field

[0001] This invention relates to the fields of image processing and machine learning technology, and in particular to a method and system for identifying crowd density based on a dynamic optimization diffusion model. Background Technology

[0002] In the field of crowd density recognition, existing methods primarily rely on traditional image processing and machine learning, which typically require high-quality input images. In practical applications, such as urban surveillance or public safety monitoring systems, image quality often degrades due to environmental factors like changes in lighting, occlusion, or camera movement. These variations severely impact the accuracy and reliability of the recognition. Furthermore, existing methods often have high computational resource requirements when processing large amounts of real-time video data, which is particularly pronounced in resource-constrained environments.

[0003] The main problems faced by existing methods include:

[0004] 1. Under complex environmental conditions (such as unstable lighting, high dynamic background, etc.), image quality degradation can lead to inaccurate crowd density estimation.

[0005] 2. Existing systems are difficult to operate efficiently in resource-constrained environments because they typically require a large amount of computing resources to process and analyze image data. Summary of the Invention

[0006] The purpose of this invention is to provide a method and system for identifying crowd density based on a dynamic optimization diffusion model, thereby solving the aforementioned problems existing in the prior art.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0008] A crowd density identification method based on a dynamic optimization diffusion model includes the following steps:

[0009] S1. Layer-by-layer diffusion simulation: By gradually increasing the noise in the image, the process of image data gradually decaying is simulated to obtain a noisy image;

[0010] S2, Enhanced Learning Correction: Utilizes denoising processes and image quality preservation strategies to recover high-quality crowd density maps from noisy images;

[0011] S3. Dynamic optimization of latent space: Based on the crowd density map, the parameters of the crowd density recognition model are dynamically adjusted in the latent space to obtain the crowd density recognition model under the optimal parameter state.

[0012] Preferably, step S1 specifically involves gradually introducing noise into the image data to simulate image quality degradation in a real-world environment, thereby training a dynamically optimized diffusion model to adapt to the impact of interference factors that may be encountered in actual monitoring scenarios on image data quality; the calculation formula is as follows:

[0013]

[0014] Among them, I t and I t-1 These are the noisy images at time steps t and t-1, respectively; N(0,1) is a random variable following a standard normal distribution with a mean of 0 and a standard deviation of 1, used to simulate the random noise added in each step; k t dt is the diffusion coefficient at time step t; dt is the time increment.

[0015] Preferably, in the layer-by-layer diffusion simulation, starting from the initial clear image, random noise is gradually introduced, and the noise at each step is calculated based on the diffusion coefficient, which is dynamically adjusted according to the environmental changes in the actual monitoring scenario.

[0016] Preferably, step S2 specifically involves determining the total loss function based on the denoising function and the regularization function, and using the total loss function to recover a high-quality, denoised crowd density map from the noisy image; the calculation formula is as follows:

[0017]

[0018] Where L is the total loss function; D(I) t ) is the denoising function; ||I t -D(I t )‖ 2 R(I) represents the mean square error between the crowd density map and the noisy image; λ is the regularization coefficient; R(I) t ) is the regularization function; T is the total number of time steps.

[0019] Preferably, the denoising function is obtained by training a deep learning model on a corresponding training dataset; the denoising function can learn to identify and eliminate noise components in an image while maintaining the original features and quality of the image.

[0020] Preferably, the regularization function is obtained by training a deep learning model on the corresponding training dataset; the regularization function can be automatically adjusted based on the content complexity of the image, and it provides additional constraints to maintain the important details and quality of noisy images.

[0021] Preferably, step S3 specifically involves optimizing the parameters of the crowd density identification model in the latent space using the total reward function of the latent space, making it adaptable to various monitoring environments and changes in crowd density; the calculation formula is as follows:

[0022]

[0023] Where R is the total reward function of the potential space; x j Let θ be the j-th feature in the latent space, which is the data extracted from the crowd density map; θ is the parameter of the crowd density recognition model; f is the reward function, which is used to adjust the behavior of the crowd density recognition model in the latent space to maximize the accuracy and efficiency of the model recognition; N is the total number of features in the latent space.

[0024] Preferably, the latent space dynamic optimization utilizes the feature learning theory in machine learning to extract deep features from the crowd density map through autoencoders or generative adversarial networks. These features are transformed into a multi-dimensional latent space, where each dimension captures certain key aspects of the crowd density map. In the latent space, the model does not operate directly on the crowd density map, but rather on higher-level abstract features, thereby enabling the model to learn and optimize more flexibly and effectively.

[0025] Preferably, step S3 is followed by:

[0026] S4. Crowd density estimation: Crowd density is identified using a dynamic optimization diffusion model under optimal parameter conditions, and the final crowd density estimation result is obtained.

[0027] The present invention also aims to provide a crowd density identification system based on a dynamic optimization diffusion model. This identification system can implement the methods described above. The identification system includes...

[0028] Layer-by-layer diffusion simulation module: By gradually increasing the noise in the image, it simulates the gradual decay process of image data to obtain a noisy image;

[0029] Enhanced learning correction module: Utilizes denoising processes and image quality preservation strategies to recover high-quality crowd density maps from noisy images;

[0030] Latent Space Dynamic Optimization Module: Based on the crowd density map, the parameters of the crowd density recognition model are dynamically adjusted in the latent space to obtain the crowd density recognition model under the optimal parameter state.

[0031] The beneficial effects of this invention are as follows: 1. Through layer-by-layer diffusion simulation, this invention simulates various image quality degradation factors that may be encountered in real-world environments, such as changes in illumination and occlusion. This simulation enables the recognition system to better adapt to these changes in practical applications, thereby maintaining high accuracy. In experiments, this invention improved the accuracy of crowd density recognition by approximately 10% to 15% compared to existing technologies in environments with unstable illumination and occlusion. 2. Enhanced learning correction and latent space dynamic optimization steps significantly reduce the demand for computing resources by optimizing the model processing flow and parameter adjustments. This makes this invention suitable for environments with limited computing power, such as achieving efficient operation on low-cost hardware. Compared to existing technologies, this invention reduces computing resource consumption by approximately 20% when processing the same amount of data. 3. Latent space dynamic optimization enables the system to dynamically adjust its behavior according to different monitoring scenarios, a flexibility not possessed by existing technologies. This invention can automatically adjust parameters according to environmental changes, making it suitable for various monitoring scenarios from indoors to outdoors and from day to night, greatly expanding its application scope. 4. In testing, this invention evaluated the stability and accuracy of the model through a series of crowd density recognition experiments under different illumination and occlusion conditions. Test results show that the recognition accuracy of the present invention remains above 85% under extreme lighting conditions, while the accuracy of the prior art is only about 70% under the same conditions. Attached Figure Description

[0032] Figure 1 This is a flowchart of the identification method in an embodiment of the present invention. Detailed Implementation

[0033] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0034] This embodiment provides a crowd density identification method based on dynamic optimization diffusion simulation, aiming to improve existing crowd monitoring systems and enhance their accuracy and reliability in complex environments through a series of innovative technical steps. First, a layer-by-layer diffusion simulation is used to simulate various factors that may lead to image quality degradation in real-world monitoring environments, such as changes in illumination and occlusion, gradually introducing image noise to train the model to adapt to these adverse conditions. Next, reinforcement learning correction, through a complex denoising process and image quality preservation strategy, recovers a high-quality crowd density map from the noisy image, ensuring data accuracy and image usability. Finally, dynamic optimization of the latent space optimizes the model's response to different monitoring scenarios by dynamically adjusting model parameters in the latent space, significantly improving the model's adaptability and recognition accuracy. Overall, this invention significantly improves the performance of crowd density identification technology in variable and complex real-world environments, making it suitable for fields such as public safety and urban monitoring. It reduces reliance on high-performance computing resources, enabling the deployment of efficient monitoring systems even in resource-constrained environments. Figure 1 As shown, the identification method specifically includes the following parts:

[0035] I. Layer-by-layer diffusion simulation

[0036] Layer-by-layer diffusion simulation provides a method for simulating image quality degradation in real-world environments. By progressively introducing noise, it effectively trains the model to adapt to various image degradation scenarios that may be encountered in actual surveillance situations, such as changes in lighting and occlusion. This simulation provides an ideal training basis for subsequent reinforcement learning correction, ensuring that the denoising function can operate effectively under complex real-world conditions.

[0037] Layer-by-layer diffusion simulation specifically involves simulating the impact of various interference factors (such as changes in lighting, occlusion, and dynamic background lighting) on ​​image data quality in a real monitoring environment. This process gradually increases the noise in the image, enabling the model to more robustly handle image quality degradation in practical applications. The relevant formula is...

[0038]

[0039] Among them, I t and I t-1 These are the noisy images at time steps t and t-1, respectively; N(0,1) is a random variable following a standard normal distribution with a mean of 0 and a standard deviation of 1, used to simulate the random noise added in each step; k t dt is the diffusion coefficient at time step t, a key parameter used to control the intensity of noise increase. Its value can be dynamically adjusted according to environmental changes in the actual monitoring scenario. dt is the time increment, which defines the interval of each time step and determines the rate at which noise is added.

[0040] The core idea of ​​layer-by-layer diffusion simulation originates from diffusion processes in physics and chemistry, analogous to thermal diffusion or the random motion of particles in a fluid. In this method, this concept is applied to image processing to simulate the gradual degradation of image quality in the real world. In practice, starting with an initial clear image, random noise is gradually introduced, with the noise at each step calculated based on the diffusion coefficient k. t This calculated coefficient can be adjusted based on actual environmental conditions (such as reduced visibility at night or in foggy weather). t The dynamic adjustment is based on environmental monitoring data (such as light sensor readings, weather information, etc.). For example, under poor lighting conditions, k... t The value is set relatively high to simulate strong image noise. This method allows the recognition system to take various complex factors into account during training, thus providing better adaptability and robustness to these non-ideal conditions in practical applications.

[0041] In practical applications, layer-by-layer diffusion simulation can significantly improve the performance of crowd density recognition systems under various environmental conditions. For example, in a large music festival or sporting event, monitoring systems often face complex environmental conditions such as rapid changes in lighting and fast-moving crowds. Traditional crowd density recognition systems may fail due to sudden changes in image quality. By using layer-by-layer diffusion simulation, the system can handle these sudden changes more effectively because it has been sufficiently trained to simulate such harsh conditions. Furthermore, this technology is also applicable to traffic monitoring and urban security, especially under severe weather conditions such as fog or heavy rain, which significantly affect the image quality captured by cameras. By simulating these conditions during model training, the applicability and accuracy of the system can be significantly improved. In summary, layer-by-layer diffusion simulation not only provides an effective means of simulating real-world monitoring environments but also provides strong support for the practical deployment of crowd density recognition technology, enabling it to maintain high efficiency and accuracy in variable and complex real-world environments.

[0042] II. Enhanced Learning Correction

[0043] Enhanced learning-based correction improves the ability to recover high-quality images from noisy images generated by layer-by-layer diffusion simulations through refined denoising and image quality preservation. This not only improves image quality but, more importantly, increases the accuracy of crowd density estimation. This step directly utilizes data generated by layer-by-layer diffusion simulations, demonstrating how noise can be reduced while preserving image detail.

[0044] Enhanced learning correction specifically involves recovering a high-quality crowd density map from noisy images using an advanced model training strategy. This process relies on a carefully designed loss function that emphasizes not only noise removal but also image quality maintenance to ensure the accuracy of the crowd density estimate in the recovered image. The relevant formula is as follows:

[0045]

[0046] Where L is the total loss function, used as the optimization objective during training; D(I t ) is a denoising function that processes the input noisy image I. t The function outputs the denoised image. This function is learned and can identify and eliminate noise components in an image; ||I t -D(I t )‖ 2 R(I) represents the mean squared error between the crowd density map and the noisy image, used to evaluate the quality of the denoising effect; λ is the regularization coefficient, a hyperparameter used to adjust the balance between denoising effect and image detail preservation. Higher λ values ​​enhance the preservation of image details, while lower values ​​emphasize denoising; t ) is the regularization function, which provides additional constraints to maintain important details and quality of the image. This is to prevent excessive denoising, which could erase details in the image that are crucial for crowd density estimation; T is the total number of time steps.

[0047] In practical implementation, the denoising function D(I) t This is typically achieved through deep learning models, such as convolutional neural networks, which can learn complex image features and noise patterns. Using a training dataset, the network learns to identify and remove noise while preserving as much of the image's original features and quality as possible. The regularization function R(I) t It is also obtained through training, and it can automatically adjust based on the complexity of the image content. For example, it can strengthen the protection of image details in densely populated areas, while it may be less strict in background areas.

[0048] In practical applications of crowd density recognition, especially in public safety and urban surveillance, image quality is affected by various factors such as weather conditions and lighting changes. In these situations, reinforcement learning correction ensures that the system can accurately estimate crowd density even under less-than-ideal image quality conditions. For example, during a large outdoor event, images captured by surveillance cameras may become blurry due to day-night cycles or sudden weather changes. In this case, reinforcement learning correction dynamically adjusts the model's denoising and detail-preserving strategies, enabling the system to operate stably under various environmental conditions and provide accurate crowd statistics. Furthermore, this technology is also well-suited for use in traffic monitoring systems, helping to accurately identify and count the number of vehicles and pedestrians in complex traffic flows, maintaining high accuracy even in low-visibility environments such as rain or fog. In conclusion, reinforcement learning correction not only improves the accuracy of crowd density recognition systems but also significantly enhances their applicability and robustness in various environments, making it an indispensable component of modern surveillance systems.

[0049] III. Dynamic Optimization of Potential Space

[0050] Latent space dynamic optimization improves the overall system response by adjusting the model's performance in the latent space. This step dynamically adjusts model parameters based on the outputs of the first two steps, optimizing the detection accuracy of crowd density. It leverages the data features processed in the first two steps to further enhance the model's adaptability and accuracy in complex scenarios.

[0051] The dynamic optimization of the latent space specifically involves redesigning the model's learning process within a dynamically optimized latent space to achieve high adaptability and recognition accuracy in complex crowd scenarios. This step uses highly customized functions to optimize parameters in the latent space to adapt to various monitoring environments and changes in crowd density. The relevant formula is as follows:

[0052]

[0053] Where R is the total reward function of the latent space, used to evaluate and optimize the performance of the entire model; x j Let θ be the j-th feature in the latent space. These features are data extracted from the crowd density map in an abstract manner to capture key information and complexity of the image. θ represents the parameters of the crowd density recognition model, which control the behavior of the reward function f. These parameters are optimized through the training process. f is the reward function, which is customized according to the specific needs of crowd density recognition and is used to adjust the behavior of the crowd density recognition model in the latent space to maximize the accuracy and efficiency of the model recognition. N is the total number of features in the latent space.

[0054] Dynamic optimization of the latent space leverages latent feature learning theory from machine learning to extract deep features from images using deep learning networks such as autoencoders or generative adversarial networks. These features are transformed into a multi-dimensional latent space, where each dimension captures certain key aspects of the input data. In this latent space, the model does not operate directly on the raw image data, but rather on these higher-level abstract features, allowing for more flexible and efficient learning and optimization. The reward function f is designed to be adjustable, typically containing multiple levels and parameters to adapt to different environments and needs. For example, for crowd density recognition in scenes with large lighting variations, the reward function might particularly emphasize features insensitive to lighting changes, while in highly crowded scenes, it might focus more on the ability to distinguish different groups of people.

[0055] In practical applications, latent space dynamic optimization provides crowd density recognition systems with significant flexibility and adaptability. For example, in a large, multifunctional commercial center, crowd density and mobility vary considerably throughout the day and are also influenced by different activities and time periods. Through latent space dynamic optimization, the recognition system can dynamically adjust its parameters based on the current situation to ensure accurate crowd density estimation regardless of the environment. Furthermore, this technology is particularly effective in handling large-scale public events such as music festivals or sporting events. In these events, crowd density changes rapidly and is often accompanied by significant environmental noise and visual obstruction. Through dynamic optimization of the latent space, the system can adjust in real time and quickly adapt to these changes, improving recognition accuracy and response speed. In conclusion, latent space dynamic optimization not only enhances the adaptability of crowd density recognition technology but also significantly improves its performance in complex environments, making it an indispensable part of modern intelligent monitoring systems.

[0056] IV. Population Density Estimation

[0057] A dynamic optimization diffusion model under optimal parameter conditions is used to identify crowd density and obtain the final crowd density estimation result.

[0058] In this embodiment, a crowd density identification system based on a dynamic optimization diffusion model is also provided. The identification system can implement the methods described above, and includes...

[0059] (1) Layer-by-layer diffusion simulation module: By gradually increasing the noise in the image, the process of image data gradually decaying is simulated to obtain a noisy image;

[0060] (2) Enhanced learning correction module: Utilizes the denoising process and image quality preservation strategy to recover a high-quality crowd density map from noisy images;

[0061] (3) Latent space dynamic optimization module: Based on the crowd density map, the parameters of the crowd density recognition model are dynamically adjusted in the latent space to obtain the crowd density recognition model under the optimal parameter state.

[0062] This invention combines three core steps—layer-by-layer diffusion simulation, enhanced learning correction, and latent space dynamic optimization—to produce an overall effect superior to the sum of its parts. Layer-by-layer diffusion simulation provides a robust foundation, enabling the model to learn how to maintain stability in variable environments; enhanced learning correction further improves image processing quality, ensuring data accuracy and usability; finally, latent space dynamic optimization finely adjusts model behavior, ensuring a rapid and accurate response to various scene changes in practical applications.

[0063] The essential characteristic of this three-step combination lies in its high adaptability and accuracy, significantly improving the performance of crowd density identification in complex environments. In key areas such as public safety and urban surveillance, this invention offers significant advancements, not only handling complex and changing environmental conditions but also providing high-precision data support in real time, greatly enhancing the practicality and reliability of surveillance systems. Furthermore, this method reduces reliance on high-performance computing resources, enabling the deployment of efficient surveillance systems even in resource-constrained environments. By utilizing these three core steps, this invention not only overcomes the limitations of individual technologies but also, through their organic combination, significantly improves the performance and practical value of the entire method.

[0064] In summary, this invention improves the accuracy and reliability of crowd density recognition systems under complex environmental conditions, particularly addressing the degradation of image quality due to environmental factors such as changes in illumination and occlusion. By introducing a method based on dynamic optimization diffusion simulation, this invention effectively adapts to and handles these image quality variations, thereby ensuring accurate crowd density estimation even in dynamic or adverse visual environments. Simultaneously, this method reduces reliance on high-performance computing resources, enabling crowd density recognition systems to operate more efficiently in resource-constrained environments.

[0065] By adopting the above-disclosed technical solution of this invention, the following beneficial effects are obtained:

[0066] This invention provides a crowd density recognition method and system based on a dynamically optimized diffusion model. Through layer-by-layer diffusion simulation, this invention simulates various image quality degradation factors that may be encountered in real-world environments, such as changes in illumination and occlusion. This simulation allows the recognition system to better adapt to these changes in practical applications, thereby maintaining high accuracy. In experiments, this invention improved the accuracy of crowd density recognition by approximately 10% to 15% compared to existing technologies in environments with unstable illumination and occlusion. The enhanced learning correction and latent space dynamic optimization steps significantly reduce the demand for computing resources by optimizing the model processing flow and parameter adjustments. This makes this invention suitable for environments with limited computing power, such as achieving efficient operation on low-cost hardware. Compared to existing technologies, this invention reduces computing resource consumption by approximately 20% when processing the same amount of data. The latent space dynamic optimization enables the system to dynamically adjust its behavior according to different monitoring scenarios, a flexibility not found in existing technologies. This invention can automatically adjust parameters according to environmental changes, making it suitable for various monitoring scenarios from indoors to outdoors and from day to night, greatly expanding its application scope. In testing, this invention conducted a series of crowd density recognition experiments under different lighting and occlusion conditions to evaluate the stability and accuracy of the model. The test results show that the recognition accuracy of this invention remains above 85% under extreme lighting conditions, while the accuracy of existing technologies under the same conditions is only around 70%.

[0067] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for identifying crowd density based on a dynamic optimization diffusion model, characterized in that: Includes the following steps, S1. Layer-by-layer diffusion simulation: By gradually increasing the noise in the image, the process of image data gradually decaying is simulated to obtain a noisy image; Step S1 specifically involves gradually introducing noise into the image data to simulate the degradation of image quality in a real-world environment, in order to train a dynamically optimized diffusion model to adapt to the impact of interference factors that may be encountered in actual monitoring scenarios on the quality of image data. The calculation formula is as follows: Among them, I t and I t-1 These are the noisy images at time steps t and t-1, respectively; N(0,1) is a random variable following a standard normal distribution with a mean of 0 and a standard deviation of 1, used to simulate the random noise added in each step; k t dt is the diffusion coefficient at time step t; dt is the time increment. S2, Enhanced Learning Correction: Utilizes denoising processes and image quality preservation strategies to recover high-quality crowd density maps from noisy images; Step S2 specifically involves determining the total loss function based on the denoising function and the regularization function, and then using the total loss function to recover a high-quality, denoised crowd density map from the noisy image; the calculation formula is as follows: Where L is the total loss function; D(I) t ) is the denoising function; ||I t -D(I t )‖ 2 R(I) represents the mean square error between the crowd density map and the noisy image; λ is the regularization coefficient; R(I) t ) is the regularization function; T is the total number of time steps; S3. Dynamic optimization of latent space: Based on the crowd density map, the parameters of the crowd density recognition model are dynamically adjusted in the latent space to obtain the crowd density recognition model under the optimal parameter state. Step S3 specifically involves optimizing the parameters of the crowd density identification model in the latent space using the total reward function of the latent space, making it adaptable to various monitoring environments and changes in crowd density; the calculation formula is as follows: Where R is the total reward function of the potential space; x j Let θ be the j-th feature in the latent space, which is the data extracted from the crowd density map; θ is the parameter of the crowd density recognition model; f is the reward function, which is used to adjust the behavior of the crowd density recognition model in the latent space to maximize the accuracy and efficiency of the model recognition; N is the total number of features in the latent space. Dynamic optimization of the latent space utilizes feature learning theory in machine learning to extract deep features from crowd density maps through autoencoders or generative adversarial networks. These features are transformed into a multi-dimensional latent space, where each dimension captures certain key aspects of the crowd density map. In the latent space, the model does not operate directly on the crowd density map, but rather on higher-level abstract features, thereby enabling the model to learn and optimize more flexibly and effectively.

2. The crowd density identification method based on a dynamic optimization diffusion model according to claim 1, characterized in that: In the layer-by-layer diffusion simulation, starting from the initial clear image, random noise is gradually introduced. The noise at each step is calculated based on the diffusion coefficient, which is dynamically adjusted according to the environmental changes in the actual monitoring scenario.

3. The crowd density identification method based on a dynamic optimization diffusion model according to claim 1, characterized in that: The denoising function is obtained by training a deep learning model on the corresponding training dataset. Denoising functions can learn to identify and eliminate noise components in an image while preserving the original features and quality of the image.

4. The crowd density identification method based on a dynamic optimization diffusion model according to claim 1, characterized in that: The regularization function is obtained by training a deep learning model on the corresponding training dataset; the regularization function can automatically adjust based on the content complexity of the image, and it provides additional constraints to maintain important details and quality of noisy images.

5. The crowd density identification method based on a dynamic optimization diffusion model according to claim 1, characterized in that: Step S3 is followed by, S4. Crowd density estimation: Crowd density is identified using a dynamic optimization diffusion model under optimal parameter conditions, and the final crowd density estimation result is obtained.

6. A crowd density identification system based on a dynamic optimization diffusion model, characterized in that: The identification system is capable of implementing the method described in any one of claims 1 to 5, and the identification system includes, Layer-by-layer diffusion simulation module: By gradually increasing the noise in the image, it simulates the process of image data gradually decaying to obtain a noisy image; Enhanced learning correction module: Utilizes denoising processes and image quality preservation strategies to recover high-quality crowd density maps from noisy images; Latent Space Dynamic Optimization Module: Based on the crowd density map, the parameters of the crowd density recognition model are dynamically adjusted in the latent space to obtain the crowd density recognition model under the optimal parameter state.

Images