An image capture and recognition method based on dynamic attenuation network and algorithm

Abstract

The present invention discloses an image capture and recognition method based on a dynamic attenuation network and algorithm, and relates to the field of image recognition technology, including acquiring image data and environmental data respectively through acquisition equipment, constructing a multi-level dynamic attenuation network model; setting an initial attenuation coefficient based on historical data, and dynamically adjusting the attenuation coefficient based on environmental data. The method of the present invention dynamically adjusts the attenuation coefficient by adaptively adjusting the attenuation coefficient based on the current environmental data, so that the attenuation coefficient changes more smoothly and naturally in different environments. By adopting the Gaussian weighting and time modulation method, the output image data can capture and express dynamic information that changes with time, and is suitable for complex environmental changes that occur with time. The processing results of each layer are integrated by weighted summation, so that the final output image data can integrate the characteristics of each layer, which can significantly improve the overall quality and detail performance of the image data.

Description

Technical Field

[0001] The invention relates to the technical field of image recognition, in particular to an image capture and recognition method based on a dynamic attenuation network and an algorithm. Background Art

[0002] In the field of modern image processing and recognition, with the rapid development of computer vision technology, image capture and recognition technology has been widely used. From intelligent monitoring, autonomous driving to medical image analysis and industrial automation, various application scenarios rely on efficient and accurate image processing technology.

[0003] Some methods attempt to introduce environmental data to compensate for image capture and recognition, but most of them are simple linear adjustments, which fail to fully utilize the complex characteristics of environmental data and time changes, and are inconvenient to smoothly adjust and process images under changes in different environments and time, thereby reducing the refinement and quality of image processing. In particular, they often show great limitations in response to changes in the natural environment, resulting in a decrease in recognition rate and reduced image data processing quality. Summary of the invention

[0004] In view of the problems existing in the above-mentioned existing image capture and recognition methods based on dynamic attenuation networks and algorithms, the present invention is proposed.

[0005] Therefore, the problem to be solved by the present invention is that some methods attempt to introduce environmental data to compensate for the capture and recognition of images, but most of them are simple linear adjustments, which fail to fully utilize the complex characteristics of environmental data and time changes, and are inconvenient to smoothly adjust and process images under changes in different environments and time, thereby reducing the refinement and quality of image processing. In particular, they often show great limitations in response to changes in the natural environment, resulting in a decrease in recognition rate and a reduction in the quality of image data processing.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: an image capture and recognition method based on a dynamic attenuation network and an algorithm, comprising:

[0007] Image data and environmental data are acquired through acquisition equipment to build a multi-level dynamic attenuation network model;

[0008] Set the initial attenuation coefficient based on historical data and dynamically adjust the attenuation coefficient based on environmental data;

[0009] Optimizing the multi-level dynamic attenuation network model according to the dynamically adjusted attenuation coefficient and generating the final processed image data;

[0010] The network model is deployed through the cloud server and stored in the local server and cloud server according to the data type.

[0011] As a preferred solution of the image capture and recognition method based on the dynamic attenuation network and algorithm of the present invention, wherein: the image data and environmental data are obtained by the acquisition device respectively, including:

[0012] The environmental data includes light illuminance and temperature. The light illuminance L is obtained by a light sensor, and the temperature T is obtained by a temperature sensor.

[0013] The image data is collected by a high-resolution camera, and the acquisition frequency of the environmental data is set to be synchronized with the acquisition frequency of the image data;

[0014] The environmental data changes over time periods.

[0015] As a preferred solution of the image capture and recognition method based on dynamic attenuation network and algorithm described in the present invention, wherein: the construction of a multi-level dynamic attenuation network model includes:

[0016] Construct a multi-level dynamic attenuation network model, including input layer, hidden layer and output layer;

[0017] In the input layer, the input image data is in a two-dimensional matrix format;

[0018] In the hidden layer, each hidden layer receives the output of the previous layer and processes the image data, which is expressed as;

[0019] I i =α i I i-1 ;

[0020] where α i represents the attenuation coefficient of the i-th layer, I i Indicates the image data output processed by the i-th layer, I i-1 Indicates the image data output processed by the i-1th layer;

[0021] The output layer outputs the processing results of the hidden layer.

[0022] As a preferred solution of the image capture and recognition method based on the dynamic attenuation network and algorithm of the present invention, wherein: the initial attenuation coefficient is set based on historical data, including:

[0023] Collect historical light intensity L' i And calculate the historical light illuminance L' i The mean and standard deviation of the historical light intensity L' i The mean of is taken as the reference brightness L0, and the initial attenuation coefficient is set by the Sigmoid function, which is expressed as:

[0024]

[0025] Where α0 represents the initial attenuation coefficient, L i represents the light brightness of the hidden layer in the i-th time period, L0 represents the reference brightness, σ L Indicates the standard deviation of light illuminance;

[0026] The initial attenuation coefficient is input into the hidden layer of the multi-level dynamic attenuation network model.

[0027] As a preferred solution of the image capture and recognition method based on the dynamic attenuation network and algorithm of the present invention, the method of dynamically adjusting the attenuation coefficient based on environmental data includes:

[0028] Based on the environmental data of the hidden layer in the current time period, the attenuation coefficient is dynamically adjusted, which is expressed as:

[0029]

[0030] where ΔE i represents the brightness change of the hidden layer in the i-th time period, E ref represents the reference brightness change, k represents the adjustment parameter, T0 represents the reference temperature, σ T represents the standard deviation of temperature, T i represents the temperature of the hidden layer in the i-th time period, α i represents the attenuation coefficient of the hidden layer in the i-th time period;

[0031] Collect historical temperature T' i And calculate the historical temperature T' i The mean and standard deviation of the historical temperature T' i The mean value of is taken as the reference temperature T0;

[0032] Use the training set to train and minimize the loss by combining the cross entropy loss function with the gradient descent method, update and determine the parameter E ref and k;

[0033] Use the current time period environment data to dynamically adjust the attenuation coefficient to obtain α i .

[0034] As a preferred solution of the image capture and recognition method based on the dynamic attenuation network and algorithm of the present invention, wherein: the multi-level dynamic attenuation network model is optimized according to the dynamic adjustment of the attenuation coefficient, and the final processed image data is generated, including:

[0035] The dynamically adjusted attenuation coefficient is applied to the multi-level dynamic attenuation network model. The hidden layer of the multi-level dynamic attenuation network model is optimized by comprehensively considering the light brightness and time changes of the image data, which is expressed as:

[0036]

[0037] Among them I o represents the final processed image data, f represents the ReLU nonlinear transformation function, R i represents the time period of the i-th layer;

[0038] The output layer is optimized and the hidden layer optimization results of all time periods are integrated, which is expressed as:

[0039]

[0040] Among them I ob represents the image data of the comprehensive output of the output layer, and n represents the total number of hidden layers;

[0041] The environmental data and image data collected by the acquisition device in real time are input into the optimized multi-level dynamic attenuation network model, and the final processed image data is output.

[0042] As a preferred solution of the image capture and recognition method based on dynamic attenuation network and algorithm described in the present invention, wherein: the network model is deployed through a cloud server, including:

[0043] Select the system scenario that requires image recognition and deploy a multi-level dynamic attenuation network model;

[0044] The image data and environmental data collected by the acquisition device are transmitted to the cloud server via a wireless network;

[0045] Based on the cloud server, the image data and environmental data are received using the HTTPS encrypted transmission protocol;

[0046] After receiving the image data, the cloud server calls and deploys a multi-level dynamic attenuation network model in real time for image processing and analysis.

[0047] As a preferred solution of the image capture and recognition method based on dynamic attenuation network and algorithm described in the present invention, wherein: the data is stored in the local server and the cloud server respectively according to the data type, including:

[0048] The real-time processed image data and analysis results are stored in a local server;

[0049] Store real-time processed image data on local servers and use a distributed storage system to store data on multiple nodes;

[0050] Historical data and non-real-time processed data are stored in the cloud server via wireless network, and local data transmission is carried out using high-bandwidth and low-latency LAN.

[0051] A computer device comprises: a memory and a processor; the memory stores a computer program, and the processor implements the steps of the above-mentioned image capture and recognition method based on a dynamic attenuation network and algorithm when executing the computer program.

[0052] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the above-mentioned image capture and recognition method based on a dynamic attenuation network and algorithm.

[0053] The beneficial effects of the present invention are as follows: by adaptively adjusting the attenuation coefficient dynamically based on the current environmental data, the change of the attenuation coefficient in different environments is made smoother and more natural; by adopting the Gaussian weighting and time modulation method, the output image data can capture and express dynamic information that changes with time, and is suitable for complex environmental changes that occur over time; by integrating the processing results of each layer in a weighted summation manner, the final output image data can integrate the characteristics of each layer, which can significantly improve the overall quality and detail expression of the image data, and make the output image data clearer and richer. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0055] Figure 1 Schematic diagram of the process of image capture recognition method based on dynamic attenuation network and algorithm.

[0056] Figure 2 Schematic diagram of the data processing flow of the image capture recognition method based on dynamic attenuation network and algorithm. DETAILED DESCRIPTION

[0057] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below in conjunction with the accompanying drawings.

[0058] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0059] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor is it a separate or selective embodiment that is mutually exclusive with other embodiments.

[0060] Example 1

[0061] Reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, and which provides an image capture and recognition method based on a dynamic attenuation network and an algorithm. The image capture and recognition method based on a dynamic attenuation network and an algorithm includes:

[0062] S1, obtain image data and environmental data through acquisition equipment respectively, and build a multi-level dynamic attenuation network model;

[0063] Preferably, the image data and the environmental data are acquired respectively by a collection device, including:

[0064] Environmental data includes light brightness and temperature. Light brightness L is obtained through a light sensor, and temperature T is obtained through a temperature sensor.

[0065] Image data is collected through a high-resolution camera, and the acquisition frequency of environmental data is set to be synchronized with the acquisition frequency of image data;

[0066] Environmental data changes with time periods. The time period is defined according to the periodic changes of the environmental data. Setting the time period includes setting it as day and night changes and seasonal changes.

[0067] By collecting high-quality image data, a reliable foundation is provided for subsequent processing and analysis, and the accuracy of image recognition is improved. By acquiring environmental data, a basis is provided for the use of the model and the dynamic adjustment of parameters according to current environmental conditions, which improves the adaptability and robustness of image processing. The adaptive ability of the model can be used in various complex environments and is suitable for a variety of application scenarios.

[0068] Furthermore, a multi-level dynamic attenuation network model is constructed, including:

[0069] Construct a multi-level dynamic attenuation network model, including input layer, hidden layer and output layer;

[0070] In the input layer, the input image data is in a two-dimensional matrix format;

[0071] In the hidden layer, each hidden layer receives the output of the previous layer and processes the image data, which is expressed as;

[0072] I i =αi I i-1 ;

[0073] where α i represents the attenuation coefficient of the i-th layer, I i Indicates the image data output processed by the i-th layer, I i-1 Indicates the image data output processed by the i-1th layer;

[0074] The output layer outputs the processing results of the hidden layer.

[0075] By using image data in a two-dimensional matrix format, the data format is unified. This format is widely used in the fields of computer vision and image processing, and can efficiently perform data operations and transmission. Through the output of the output layer, it ensures that the characteristics of the collected data can be comprehensively reflected in the final output, thereby improving the overall quality and detail expression of the image data, and further enhancing the feature representation and recognition capabilities of the image data.

[0076] S2, setting the initial attenuation coefficient based on historical data and dynamically adjusting the attenuation coefficient based on environmental data;

[0077] Preferably, the initial attenuation coefficient is set based on historical data, including:

[0078] Collect historical light intensity L' i And calculate the historical light illuminance L' i The mean and standard deviation of the historical light intensity L' i The mean of is taken as the reference brightness L0, and the initial attenuation coefficient is set by the Sigmoid function, which is expressed as:

[0079]

[0080] Where α0 represents the initial attenuation coefficient, L i represents the light brightness of the hidden layer in the i-th time period, L0 represents the reference brightness, σ L Indicates the standard deviation of light illuminance;

[0081] The initial attenuation coefficient is input into the hidden layer of the multi-level dynamic attenuation network model.

[0082] The initial attenuation coefficient setting and application provide a basis for the calculation of the network model. The Sigmoid function is used to standardize the difference between the brightness data and the reference brightness data between 0 and 1. The purpose is to smoothly map the change of light brightness to a controllable range to avoid the instability caused by mutations. At the same time, the Sigmoid function makes the attenuation coefficient change less when the light brightness is close to the reference value, and the attenuation coefficient changes more when the difference is large, thereby realizing adaptive adjustment of the light brightness, making the initial attenuation coefficient setting more accurate and reducing the initial error.

[0083] Furthermore, the attenuation coefficient is dynamically adjusted based on environmental data, including:

[0084] Based on the environmental data of the hidden layer in the current time period, the attenuation coefficient is dynamically adjusted, which is expressed as:

[0085]

[0086] where ΔE i represents the brightness change of the hidden layer in the i-th time period (ΔE = L i -L0), E ref represents the reference brightness change, k represents the adjustment parameter, T0 represents the reference temperature, σ T represents the standard deviation of temperature, T i represents the temperature of the hidden layer in the i-th time period, α i represents the attenuation coefficient of the hidden layer in the i-th time period;

[0087] Collect historical temperature T' i And calculate the historical temperature T' i The mean and standard deviation of the historical temperature T' i The mean value of is taken as the reference temperature T0;

[0088] Use the training set to train and minimize the loss by combining the cross entropy loss function with the gradient descent method, update and determine the parameter E ref and k;

[0089] Use the current time period environment data to dynamically adjust the attenuation coefficient to obtain α i ;

[0090] formula Through the current illumination change, refer to the illumination value E ref This ensures that when the illumination changes significantly, the attenuation coefficient can be dynamically adjusted according to the change amount, which increases the flexibility and adaptability of the formula. Considering the temperature T iThe use of Gaussian function can smoothly reduce the impact of temperature change and avoid extreme impact of temperature on the attenuation coefficient. T The influence range of temperature change on the attenuation coefficient is determined, ensuring that the attenuation coefficient can be adaptively adjusted under different temperature conditions.

[0091] By dynamically adjusting the attenuation coefficient, each hidden layer can be adaptively adjusted according to the current environmental data, and can be adaptively adjusted when facing different light brightness and temperature changes, so as to effectively cope with complex and changeable actual scenes and improve the robustness of image data processing. By dynamically adjusting the attenuation coefficient and combining the Sigmoid function, Gaussian function and linear adjustment, the attenuation coefficient changes more smoothly and naturally in different environments. Gaussian weighting can smoothly process the changes in light brightness and avoid the loss of image data details, making the processed image data more consistent in time series. Even under extreme light or temperature conditions, it can still maintain a high image data processing quality through adaptive adjustment, thereby improving the stability and reliability of the network model.

[0092] S3, optimizing the multi-level dynamic attenuation network model according to the dynamically adjusted attenuation coefficient, and generating the final processed image data;

[0093] Preferably, the multi-level dynamic attenuation network model is optimized according to the dynamically adjusted attenuation coefficient, and the final processed image data is generated, including:

[0094] The dynamically adjusted attenuation coefficient is applied to the multi-level dynamic attenuation network model. The hidden layer of the multi-level dynamic attenuation network model is optimized by comprehensively considering the light brightness and time changes of the image data, which is expressed as:

[0095]

[0096] Among them I o represents the final processed image data, f represents the ReLU nonlinear transformation function (which is a nonlinear transformation of the final processed image data), R i represents the time period of the i-th layer;

[0097] The output layer is optimized and the hidden layer optimization results of all time periods are integrated, which is expressed as:

[0098]

[0099] Among them I ob represents the image data of the comprehensive output of the output layer, and n represents the total number of hidden layers;

[0100] Input the environmental data and image data collected by the acquisition device in real time into the optimized multi-level dynamic attenuation network model, and output the final processed image data;

[0101] in the formula The image data is weighted using a Gaussian function, taking into account the current layer light intensity L i Compared with the reference light illumination L0, Gaussian weighting makes the effect of light illumination on the image smoother and more natural, which helps to maintain the details and overall quality of the image. The time factor is introduced, and the image data is modulated by a sinusoidal function to simulate the time-varying lighting or other environmental changes. This can help the network model learn and adapt to dynamically changing scenes, making image data processing more timely and real-time.

[0102] By adopting the method of Gaussian weighting and time modulation, the factors of light brightness and time can be considered in the integration process to ensure the spatiotemporal consistency of image data, so that the output image data can capture and express dynamic information that changes over time, enhancing the real-time and dynamic expressiveness of image data. By introducing a multi-level structure, the network model has higher robustness when processing complex scenes, can effectively cope with a variety of scene changes, and improves the stability and reliability of the network model. The processing results of the hidden layer can provide rich features for the final output and provide better input for subsequent layers, which significantly improves the performance of the entire network model. The processing results of each layer are integrated by weighted summation of the output layer, so that the final output image data can integrate the features of each layer, which can significantly improve the overall quality and detail expression of the image data, make the output image data clearer and richer, avoid the limitations of single-level processing, ensure the balance and fusion of features at different levels in the final output, and improve the comprehensive expressiveness of the image data.

[0103] S4, deploy the network model through the cloud server and store it in the local server and cloud server according to the data type;

[0104] Preferably, the network model is deployed via a cloud server, including:

[0105] Select the system scenario that requires image recognition and deploy a multi-level dynamic attenuation network model;

[0106] The image data and environmental data collected by the acquisition device are transmitted to the cloud server via a wireless network;

[0107] Based on the cloud server, the image data and environmental data are received using the HTTPS encrypted transmission protocol;

[0108] After receiving the image data, the cloud server calls and deploys a multi-level dynamic attenuation network model in real time for image processing and analysis.

[0109] Data transmission via wireless networks greatly improves the deployment flexibility of image acquisition equipment and is suitable for a variety of different scenarios and application requirements, including urban monitoring, intelligent transportation, medical imaging, industrial detection, etc. The use of HTTPS encrypted transmission protocol ensures that image data is not stolen or tampered with during transmission, ensuring data security and privacy protection. The high scalability of cloud servers enables the system to expand computing and storage resources as business needs grow, ensuring the long-term stable operation of the system.

[0110] Further, according to the data type, it is stored in the local server and the cloud server respectively, including:

[0111] The real-time processed image data and analysis results are stored in a local server;

[0112] Store real-time processed image data on local servers and use a distributed storage system to store data on multiple nodes;

[0113] Historical data and non-real-time processed data are stored in the cloud server via wireless network, and local data transmission is carried out using high-bandwidth and low-latency LAN.

[0114] By storing real-time processed image data and analysis results on local servers, fast data access and processing speed are ensured. Local storage reduces data transmission delays, improves the system's real-time response capabilities and user experience. A distributed storage system is used to store data on multiple nodes, which improves data access speed and system reliability. The high scalability and high availability of cloud storage support long-term preservation and management of large-scale data.

[0115] Example 2

[0116] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0117] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0118] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0119] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used to implement: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0120] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. An image capture and recognition method based on a dynamic attenuation network and algorithm, characterized in that: include, Image data and environmental data are acquired through acquisition equipment to build a multi-level dynamic attenuation network model; Set the initial attenuation coefficient based on historical data and dynamically adjust the attenuation coefficient based on environmental data; Optimizing the multi-level dynamic attenuation network model according to the dynamically adjusted attenuation coefficient and generating the final processed image data; Deploy the network model through the cloud server and store it on the local server and cloud server according to the data type; The image data and the environmental data are obtained by the acquisition device respectively, including: The environmental data includes light illuminance and temperature. The light illuminance L is obtained by a light sensor, and the temperature T is obtained by a temperature sensor. The image data is collected by a high-resolution camera, and the acquisition frequency of the environmental data is set to be synchronized with the acquisition frequency of the image data; The environmental data changes over time periods; The multi-level dynamic attenuation network model is constructed, including: Construct a multi-level dynamic attenuation network model, including input layer, hidden layer and output layer; In the input layer, the input image data is in a two-dimensional matrix format; In the hidden layer, each hidden layer receives the output of the previous layer and processes the image data, which is expressed as; I i =α i ·I i-1 ; where α i represents the attenuation coefficient of the i-th layer, I i Indicates the image data output processed by the i-th layer, I i-1 Indicates the image data output processed by the i-1th layer; The output layer outputs the processing results of the hidden layer; The step of setting the initial attenuation coefficient based on historical data includes: Collect historical light intensity L' i And calculate the historical light illuminance L' i The mean and standard deviation of the historical light intensity L' i The mean of is taken as the reference brightness L0, and the initial attenuation coefficient is set by the Sigmoid function, which is expressed as: Where α0 represents the initial attenuation coefficient, L i represents the light brightness of the hidden layer in the i-th time period, L0 represents the reference brightness, σ L Indicates the standard deviation of light illuminance; Inputting the initial attenuation coefficient into the hidden layer of the multi-level dynamic attenuation network model; The dynamically adjusting the attenuation coefficient based on environmental data includes: Based on the environmental data of the hidden layer in the current time period, the attenuation coefficient is dynamically adjusted, which is expressed as: Where △E i represents the brightness change of the hidden layer in the i-th time period, E ref represents the reference brightness change, k represents the adjustment parameter, T0 represents the reference temperature, σ T represents the standard deviation of temperature, T i represents the temperature of the hidden layer in the i-th time period, α i represents the attenuation coefficient of the hidden layer in the i-th time period; Collect historical temperature T' i And calculate the historical temperature T' i The mean and standard deviation of the historical temperature T' i The mean value of is taken as the reference temperature T0; Use the training set to train and minimize the loss by combining the cross entropy loss function with the gradient descent method, update and determine the parameter E ref and k; Use the current time period environment data to dynamically adjust the attenuation coefficient to obtain α i ; The method of optimizing the multi-level dynamic attenuation network model according to the dynamic adjustment of the attenuation coefficient and generating the final processed image data includes: The dynamically adjusted attenuation coefficient is applied to the multi-level dynamic attenuation network model. The hidden layer of the multi-level dynamic attenuation network model is optimized by comprehensively considering the light brightness and time changes of the image data, which is expressed as: Among them I o represents the final processed image data, f represents the ReLU nonlinear transformation function, R i represents the time period of the i-th layer; The output layer is optimized and the hidden layer optimization results of all time periods are integrated, which is expressed as: Among them I ob represents the image data of the comprehensive output of the output layer, and n represents the total number of hidden layers; The environmental data and image data collected by the acquisition device in real time are input into the optimized multi-level dynamic attenuation network model, and the final processed image data is output.

2. The image capture and recognition method based on dynamic attenuation network and algorithm as claimed in claim 1, characterized in that: The network model is deployed through the cloud server, including: Select the system scenario that requires image recognition and deploy a multi-level dynamic attenuation network model; The image data and environmental data collected by the acquisition device are transmitted to the cloud server via a wireless network; Based on the cloud server, the image data and environmental data are received using the HTTPS encrypted transmission protocol; After receiving the image data, the cloud server calls and deploys a multi-level dynamic attenuation network model in real time for image processing and analysis.

3. The image capture and recognition method based on dynamic attenuation network and algorithm as claimed in claim 2, characterized in that: The data types are stored in the local server and the cloud server respectively, including: The real-time processed image data and analysis results are stored in a local server; Store real-time processed image data on local servers and use a distributed storage system to store data on multiple nodes; Historical data and non-real-time processed data are stored in the cloud server via wireless network, and local data transmission is carried out using high-bandwidth and low-latency LAN.

4. A computer device comprising: Memory and processor; The memory stores a computer program, characterized in that when the processor executes the computer program, the steps of the image capture and recognition method based on a dynamic attenuation network and algorithm described in any one of claims 1 to 3 are implemented.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the image capture and recognition method based on a dynamic attenuation network and algorithm described in any one of claims 1 to 3 are implemented.

Images