Image identification detection system based on AI

Abstract

The invention, which relates to the technical field of image identification processing, discloses an AI-based image identification detection system comprising the following components: an optical physical model processing module, an environmental factor perception and analysis module, a color deviation correction module, a micro-feature enhancement module, an AI image identification module and a data storage and dynamic update module. By combining the optical physical model and the micro-feature enhancement algorithm based on the physical principle, the accuracy and reliability of object recognition under complex weather and illumination conditions are remarkably improved, and by introducing the optical physical model, the interaction between light and the object surface under different environments is accurately simulated, so that the object recognition accuracy and reliability are improved. And meanwhile, the micro-feature enhancement algorithm is optimally designed for specific environments such as low illumination and strong wind, so that the signal strength of the micro-feature region is effectively enhanced, and micro-feature deformation caused by environmental factors is recovered.

Description

Technical Field

[0001] The present invention relates to the field of image recognition processing technology, and in particular to an AI-based image recognition detection system. Background Art

[0002] As a core component in the field of computer vision, image recognition technology has shown broad application prospects in many fields such as intelligent transportation, security monitoring, and industrial automation in recent years. With the rapid development of artificial intelligence technology, image recognition technology is not only limited to simple object classification and recognition, but has also gradually expanded to higher-level tasks such as posture estimation and scene understanding. In actual environments, the image acquisition process may be affected by various weather conditions such as sunny, cloudy, rainy, and foggy days, as well as different light intensities and angles, resulting in color deviation, contrast reduction, blurred details, and other problems in the image, which in turn affects the accuracy and reliability of object recognition.

[0003] Traditional image recognition and detection systems are often designed without fully considering the combined impact of complex weather and lighting conditions on the micro-features and colors of objects. These systems usually rely on fixed image preprocessing algorithms and feature extraction methods, and are difficult to adapt to significant changes in image quality in different environments. For example, in rainy weather or strong light conditions, image parameters such as contrast, brightness, and color saturation will change significantly, causing the micro-features of objects to become blurred or difficult to identify. In addition, traditional systems often lack the ability to monitor and analyze environmental factors in real time and are unable to dynamically adjust recognition strategies according to changes in environmental conditions. Therefore, in practical applications, traditional image recognition and detection systems are difficult to meet the needs of high-precision recognition, especially under complex and changeable weather and lighting conditions.

[0004] In response to the above problems, it is necessary to optimize the existing image recognition and detection system. By integrating optical physics models and micro-feature enhancement algorithms based on physical principles, accurate recognition and detection of objects under complex weather and lighting conditions can be achieved. Therefore, it is of great significance to develop an AI-based image recognition and detection system that can comprehensively realize the above characteristics. Summary of the invention

[0005] The purpose of the present invention is to make up for the shortcomings of the prior art and provide an AI-based image recognition and detection system, which can simulate the color changes of objects under different weather and lighting conditions by utilizing optical physics models, thereby providing a basic basis for subsequent color deviation correction. At the same time, environmental data is acquired in real time through the environmental factor perception and analysis module, and a relationship model between environmental factors and object feature changes is established. On this basis, a color deviation correction algorithm and a micro-feature enhancement algorithm are used to pre-process the image to improve image quality and visibility of object features. Finally, an AI image recognition model is constructed using a deep learning algorithm to perform feature extraction and classification on the pre-processed image. In addition, the system also has a data storage and dynamic update mechanism, which can continuously optimize recognition performance and adaptability as new data accumulates.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: an AI-based image recognition and detection system, which includes the following components: an optical physical model processing module, an environmental factor perception and analysis module, a color deviation correction module, a micro-feature enhancement module, an AI image recognition module, and a data storage and dynamic update module; The optical physics model processing module selects the optical physics model according to the characteristics of the color change of the object under different weather and lighting conditions, and calculates the color and light intensity change of the object in combination with the environmental parameters during image acquisition; The environmental factor perception and analysis module uses the built-in sensor of the image acquisition device to synchronously obtain light intensity, angle, temperature, humidity, wind speed and wind direction data when acquiring images, and interacts with the meteorological data interface to obtain weather information, and uses data analysis algorithms to perform in-depth analysis of environmental data; The color deviation correction module designs a color deviation correction algorithm based on the color change results simulated by the optical physics model and the influence rules obtained by the environmental factor analysis, and corrects the color deviation by adjusting the color space parameters of the image pixels; The micro-feature enhancement module designs a micro-feature enhancement algorithm for low-light environments, enhances the signal strength of the micro-feature area by analyzing the photon distribution, and constructs a micro-feature deformation prediction and compensation model in combination with aerodynamic principles for strong wind environments, predicts micro-feature deformation based on wind speed, wind direction, object shape and material parameters, and performs reverse compensation; The AI ​​image recognition module performs normalization preprocessing on the image that has undergone color deviation correction and micro-feature enhancement processing, and annotates the image data to construct a training data set, uses a deep learning algorithm to construct an AI image recognition model and performs training, adjusts parameters by optimizing the algorithm to improve the model recognition performance, inputs the preprocessed image to be recognized into the model, analyzes the image features and outputs the recognition result; The data storage and dynamic update module constructs a database system to store and manage data related to system operation, and regularly updates various parameters and models as new environmental data, object samples and physical research results emerge.

[0007] Furthermore, the optical physics model processing module calculates the color intensity change of the object in combination with the environmental parameters during image acquisition, and the calculation formula is: ,in, represents the wavelength of light, Indicates the reflected light intensity, is the wavelength of the incident light The light intensity at is the wavelength of the object surface The reflectivity at is the wavelength of the atmosphere The transmittance at is the wavelength of the atmosphere The scattering coefficient at .

[0008] Furthermore, the environmental factor perception and analysis module uses a data analysis algorithm to perform in-depth analysis on environmental data, and the algorithm formula is: ,in, represents the comprehensive impact coefficient of environmental factors, represent Real-time measurements of environmental factors of type, yes Reference mean of type environmental factors, for The fluctuation range of type environmental factors, yes The weight coefficient of the type of environmental factors, Corresponding light intensity , Lighting Angle ,temperature ,humidity and wind speed , through this formula, a quantitative relationship model between environmental factors and object color and micro-feature changes is established.

[0009] Furthermore, the color deviation correction module designs a color deviation correction algorithm to correct the color deviation by adjusting the color space parameters of the image pixels. Specifically, the RGB value of the original pixel is set to , the corrected RGB value is , the correction formula is: ,in, Represents the red, green, and blue component values ​​of the original pixel, Represents the red, green, and blue component values ​​of the corrected pixel, are the correction coefficients for the red, green and blue channels respectively, It is the comprehensive influence coefficient of environmental factors.

[0010] Furthermore, in the micro-feature enhancement module, a micro-feature enhancement algorithm is designed for low-light environments, and the signal strength of the micro-feature area is enhanced by analyzing the photon distribution. The enhancement formula is: ,in, is the original grayscale value of the pixel in the low-light image, is the gray value of the pixel after enhancement, β is the enhancement coefficient, It is the mean photon count based on Poisson distribution, which means the mean photon count of image pixels under low light conditions. By introducing the enhancement coefficient β, the enhanced gray value is adjusted. , thereby achieving image brightness adjustment. Furthermore, in the micro-feature enhancement module, for strong wind environments, a micro-feature deformation prediction and compensation model is constructed in combination with aerodynamic principles to perform reverse compensation. Specifically, in strong wind environments, , assuming that the original position of a micro-feature point on the surface of the object is , the position after compensation is According to the principles of aerodynamics, the displacement of micro-feature points is related to wind speed V, wind direction φ, and elastic modulus of the object surface. The model formula is: ,in, is the original position coordinate of the micro feature point, is the position coordinate of the micro-feature point after deformation under the action of strong wind, is the actual wind speed, is the wind direction, expressed in degrees, is the elastic modulus of the surface, They are and The compensation coefficient of the direction is determined by experiments and simulations. Furthermore, in the AI ​​image recognition module, a deep learning algorithm is used to construct an AI image recognition model. Specifically, the image feature vector extracted by the convolutional neural network is , the environmental factor vector is , the image feature vector and the environmental factor vector are transformed through two fully connected layers to obtain the transformed vector and , let the weight matrix of the fully connected layer be and , the biases are and ,but , , by introducing an adaptive weight adjustment mechanism, the fusion weights of image features and environmental factor features are dynamically adjusted according to the current environmental factors, and the adaptive weight α is defined. , where σ is the Sigmoid function, and is a learnable parameter, then the fusion feature vector By formula It can be obtained that through this formula, the model can use environmental information to assist image recognition.

[0011] Furthermore, the AI ​​image recognition module improves model recognition by adjusting parameters through optimization algorithm, and its algorithm formula is: ,in, is the loss function, is the total number of categories, Yes Category The weight factor of is the true category label of the sample. When the sample belongs to the category hour, ,otherwise , The model predicts that the sample belongs to the category The probability of the model being lost is determined by the labeled data set. During the training process, the weight parameters of the model are continuously adjusted to minimize the loss function S.

[0012] Furthermore, the Yes Category The weight factor is given by the formula Calculate, where is the total number of samples, For Category Compared with the existing technology, the AI-based image recognition detection system has the following beneficial effects: 1. The present invention significantly improves the accuracy and reliability of object recognition under complex weather and lighting conditions by combining optical physics models and micro-feature enhancement algorithms based on physical principles. It also introduces optical physics models to accurately simulate the interaction between light and object surfaces in different environments, providing a scientific basis for subsequent color deviation correction. At the same time, the micro-feature enhancement algorithm is optimized for specific environments such as low light and strong wind, effectively enhancing the signal strength of the micro-feature area and restoring the micro-feature deformation caused by environmental factors, thereby greatly improving the accuracy and reliability of recognition.

[0013] 2. The present invention ensures that the system can be continuously optimized with the accumulation of new data and adapt to various complex practical application scenarios through data storage and dynamic update mechanism. By constructing a database system, optical physics model parameters, color correction data, micro-feature enhancement algorithm parameters, AI image recognition model training data and system operation-related data are efficiently stored and quickly queried, thereby realizing centralized management and effective utilization of data. With the emergence of new environmental data, object samples and physical research results, the system can regularly update the above-mentioned parameters and models, thereby continuously improving recognition performance and adaptability, so that the present invention can maintain a high degree of flexibility and accuracy when facing a constantly changing external environment.

[0014] Other advantages, objectives and features of the present invention will be set forth in part in the following description and, in part, will be apparent to those skilled in the art based on an examination of the following or may be taught from the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0016] Figure 1 It is a structural diagram of the AI-based image recognition detection system; Figure 2 Flowchart of the AI-based image recognition detection system. DETAILED DESCRIPTION

[0017] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation mode, structure, characteristics and effects of the present invention are described in detail below in combination with the accompanying drawings and preferred embodiments.

[0018] Embodiment 1 In a modern industrial production workshop, there are many types of parts on the production line and the production rhythm is tight. In order to ensure product quality, it is necessary to perform real-time, high-precision detection of parts in the production process, accurately identify the model of the parts, and detect whether there are defects such as scratches, cracks, and wear on the surface to ensure that only qualified parts enter the next production link. However, the environment in the workshop is relatively complex and there are differences in lighting conditions. For example, strong lighting in local areas may cause excessive reflection on the surface of parts, while some corners or equipment shadows are insufficiently illuminated. In addition, in some special process links, such as welding and grinding, smoke, dust, etc. will be generated. These factors bring challenges to the accurate identification of parts.

[0019] In view of the complex optical environment in the workshop, the system needs to comprehensively consider multiple factors to select the appropriate optical physics model. For the interaction between conventional illumination and component surface, the spectral reflection model is used to describe the reflection characteristics of light. In special scenes with smoke and dust, the optical model considering particle scattering is combined to more accurately simulate the light propagation process. The formula is used. Carefully calculate the interaction between light and the surface of the component. In different production areas, the parameters in the formula are determined according to the actual situation. For example, in the area near the welding station, the smoke concentration is high and the atmospheric scattering coefficient is will increase accordingly, and the atmospheric transmittance Reduce, by accurately calculating the reflected light intensity, predicting the changes in surface color and gloss of parts under different lighting and environmental conditions, providing a solid theoretical basis for subsequent color deviation correction, so that the system can accurately perceive the true appearance of parts in complex environments.

[0020] In order to comprehensively monitor the environmental conditions in the workshop, light sensors, temperature and humidity sensors, dust concentration sensors, smoke alarms and other equipment are installed at key locations. These sensors collect detailed environmental data such as light intensity, light angle, temperature, humidity, wind speed, etc. in real time. At the same time, the system uses formula To calculate the comprehensive impact coefficient of environmental factors ,in, represents the comprehensive impact coefficient of environmental factors, yes Reference mean of type environmental factors, for The fluctuation range of type environmental factors, yes The weight coefficient of the type of environmental factors, Corresponding light intensity , Lighting Angle ,temperature ,humidity and wind speed , The weight coefficient is used to deeply analyze the impact of various environmental factors on the surface micro-features of components (such as potential scratches, cracks and other defects) and colors, providing quantitative environmental impact information for subsequent processing modules, so that the system can be adjusted in a targeted manner.

[0021] Based on the comprehensive impact coefficient of environmental factors As well as the color characteristics of different parts, using the formula Accurately correct the color deviation of the collected component images. Since components of different materials and uses have unique color characteristics, a large number of sample tests and analyses are performed to determine the corresponding color characteristics for each type of component. , , For example, for a silver component made of aluminum alloy, its color change pattern under different lighting and environments is different from that of other materials. After careful experiments, the appropriate , , The RGB channels of the image can be precisely adjusted to effectively correct color deviations caused by lighting and environmental factors, ensuring that the surface color of components can truly reflect their actual status, providing an accurate image basis for subsequent defect detection and model identification.

[0022] In the insufficiently illuminated areas of the workshop, micro-defects on the surface of parts, such as fine scratches and small pits, are difficult to be clearly identified. At this time, the formula These micro features are enhanced. In practical applications, the optimal enhancement factor is determined through multiple experiments and comparisons based on different component materials and surface characteristics. For example, for a smooth metal part, under certain low light conditions, find a value that can maximize the contrast between micro defects and the normal surface. Value, based on the original grayscale value of the micro-feature area and the mean photon count based on Poisson distribution , highlighting the details of micro defects, making the originally imperceptible micro defects clearly presented in the image, greatly improving the detection accuracy of micro defects.

[0023] The images of parts that have been corrected for color deviation and enhanced for micro-features are preprocessed in a comprehensive and detailed manner. The image size is adjusted to a standard size that meets the input requirements of the AI ​​model, and standardized to make the image data numerically consistent and comparable. In order to build a high-quality training data set, a large number of images of parts with different models and defects are annotated in detail. The annotation content includes not only the model and defect type (such as scratches, cracks, wear, etc.) of the parts, but also the location, size, shape, etc. of the defects. The deep learning algorithm is used to carefully design the network structure according to the actual needs and data characteristics of industrial inspection, and the appropriate number of network layers, convolution kernel size, step size and other hyperparameters are selected to build the AI ​​image recognition model. During the training process, the formula is used The loss function is calculated, and the model parameters are continuously adjusted through optimization algorithms such as stochastic gradient descent and Adam optimizer to minimize the loss function, thereby improving the model's recognition accuracy for different types of parts and various defects. In order to enhance the generalization ability of the model, data enhancement technology is used, such as rotating, flipping, scaling, adding noise and other operations on the part images to simulate various image changes that may occur in the actual production process. The processed part images are input into the trained model. The model can quickly and accurately identify the model of the parts and detect whether there are defects on the surface. The recognition results are promptly fed back to the production control system. The production control system makes real-time adjustments to the production process based on these results to ensure that the produced parts meet the quality standards.

[0024] The system uses a high-performance database system to store various important information such as optical physical model parameters, environmental factor data, color correction data, micro-feature enhancement algorithm parameters, and AI image recognition model training data. With the continuous improvement of industrial production processes, the continuous introduction of new types of parts, and changes in the workshop environment, the stored data is regularly updated. For example, when a new production process is adopted, which causes the smoke and dust characteristics in the workshop to change, it is necessary to re-collect environmental data, analyze its impact on light propagation and component characteristics, and update the weight coefficients in the formula of optical physical model parameters and comprehensive influence coefficients of environmental factors. When a new model of parts is introduced, a large amount of image data of the model of parts is collected, including normal samples and samples with various defects, and the AI ​​image recognition model is retrained and optimized to ensure that the system can always accurately identify and detect new parts to meet the development needs of industrial production. Embodiment 2 In modern urban traffic networks, traffic intersections are key nodes with a large number of vehicles shuttling back and forth every day. To achieve efficient traffic management, it is necessary to accurately identify vehicles under different weather and lighting conditions, including key information such as vehicle brand, color, license plate number, etc., so as to provide strong support for functions such as traffic flow monitoring, violation capture, and intelligent traffic scheduling.

[0025] Weather conditions in cities are complex and changeable, and lighting conditions vary with time and location. For common local weather conditions, such as strong direct sunlight on sunny days, slippery roads on rainy days causing complex light reflections, light scattering on foggy days seriously affecting vehicle visibility, as well as low light in the morning and evening and strong direct light at noon, the system will intelligently select an adaptive optical physics model. For example, on foggy days, in order to accurately simulate the propagation of light in foggy particles and its interaction with the vehicle surface, the system uses the formula To calculate the reflected light intensity of the vehicle surface in detail, in this formula, the values ​​of various parameters are determined according to the actual weather and lighting conditions. For example, on a sunny day, the atmospheric transmittance Higher, and in haze days, the atmospheric scattering coefficient Increase, through precise calculation, predict the subtle changes in vehicle color in different environments, provide accurate basis for subsequent color correction, and ensure that the vehicle color can be accurately perceived in various environments.

[0026] In order to obtain comprehensive environmental information, various devices such as light sensors, temperature sensors, humidity sensors, and wind speed sensors are reasonably deployed around traffic cameras. At the same time, a real-time data interface is established with the meteorological department to ensure that environmental data such as light intensity, light angle, temperature, humidity, wind speed, and detailed weather information can be obtained in real time and accurately. In-depth analysis of the impact of various environmental factors on micro-features such as vehicle color and license plate. For example, in heavy rain, raindrops will strongly refract and scatter light. At the same time, strong winds may cause raindrops on the surface of the vehicle to flow, affecting the clarity of the license plate. At this time, the weight coefficients of humidity and wind speed are and It will be relatively large. Through comprehensive consideration of various factors, the impact of the quantitative environment on vehicle characteristics can be accurately determined to provide a scientific basis for subsequent processing.

[0027] Based on the comprehensive impact coefficient of environmental factors As well as the characteristic differences of different vehicle colors, using the formula Accurately correct the color of the vehicle images captured by the camera, and determine the color of vehicles of different colors through a large number of experiments and data distribution. For example, for a blue vehicle, the color of blue changes under different lighting and weather conditions in a unique way. value, and make targeted adjustments to the blue channel so that the vehicle color can be accurately restored in various environments, providing a guarantee for subsequent accurate identification of the vehicle color.

[0028] In the low-light period in the morning and evening, the license plate number is a key micro-feature, and its clarity is crucial for vehicle recognition. Enhance the micro features of the license plate. In practical applications, through multiple experiments and field tests, the most appropriate enhancement coefficient β is determined according to different lighting conditions and license plate materials. For example, for a metal license plate, in a specific low-light environment, the β value that can effectively improve the contrast between the license plate feature and the background is determined. and the mean photon count based on Poisson distribution , highlighting the details of the license plate number, making the license plate still clearly visible in low light, greatly improving the accuracy of license plate recognition.

[0029] In strong winds, vehicles may experience slight vibrations due to wind force, which may cause deformation of micro features such as license plates. Compensate for micro features that may be deformed, and determine the elastic modulus of vehicle materials through actual measurement and simulation analysis of different vehicle models , combined with the real-time wind speed and wind direction Data, accurately calculate the compensation position of micro feature points , thereby effectively restoring the original form of the micro-features. For example, when the wind speed reaches a certain level, the system can calculate and adjust the position and shape of the license plate characters according to the formula to ensure that the license plate number can be accurately identified even in strong winds.

[0030] For the vehicle images after color deviation correction and micro-feature enhancement, we first perform comprehensive preprocessing to uniformly adjust the image size to a standard size suitable for AI model input, and perform standardization to make the image data meet the numerical range requirements of the model. In order to build a high-quality training data set, a large number of vehicle images of different brands, colors and license plates are annotated in detail. The annotated information includes not only the vehicle brand, color, and license plate number, but also the type of vehicle (such as cars, trucks, buses, etc.), body posture and other information. A deep learning algorithm is used to build an AI image recognition model. During the training process, the formula is used Calculate the loss function and continuously adjust the model parameters through optimization algorithms such as stochastic gradient descent and Adam optimizer to minimize the loss function, thereby improving the recognition accuracy of the model. In order to enhance the generalization ability of the model, use a variety of data enhancement techniques, such as random rotation, flipping, scaling, adding noise and other operations on vehicle images to simulate vehicle image changes in various actual scenarios, so that the model can learn more comprehensive vehicle features. The processed real-time vehicle images are input into the trained model. The model can quickly and accurately identify the vehicle's brand, color, license plate number and other information, and transmit the recognition results to the traffic management system in a timely and accurate manner, providing key data support for traffic management decisions.

[0031] The system is equipped with a high-performance data storage system for storing various key information such as optical physics model parameters, environmental factor data, color correction data, micro-feature enhancement algorithm parameters, and AI image recognition model training data. With the continuous changes in the urban traffic environment, such as road facility updates, changes in the obstruction of surrounding buildings, the emergence of new vehicle models and license plate styles, the stored data is regularly updated. For example, when a city introduces new energy vehicle models, their body materials and color characteristics are different from those of traditional models. At this time, the spectral reflection model parameters need to be updated to accurately simulate the interaction between light and the surface of the new model. When new high-rise buildings are built around traffic intersections, causing significant changes in lighting conditions, a large amount of environmental data is re-collected, the new impact of environmental factors on vehicle recognition is analyzed, and the weight coefficients in the comprehensive impact coefficient formula of environmental factors are adjusted to ensure that the system can always adapt to the changing traffic environment and maintain high-precision vehicle recognition capabilities.

[0032] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technical personnel in this field can make some changes or modify the technical contents disclosed above into equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any brief modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. AI-based image recognition detection system, characterized by: The system includes the following components: optical physical model processing module, environmental factor perception and analysis module, color deviation correction module, micro-feature enhancement module, AI image recognition module and data storage and dynamic update module; The optical physics model processing module selects the optical physics model according to the characteristics of the color change of the object under different weather and lighting conditions, and calculates the color and light intensity change of the object in combination with the environmental parameters during image acquisition; The environmental factor perception and analysis module uses the built-in sensor of the image acquisition device to synchronously obtain light intensity, angle, temperature, humidity, wind speed and wind direction data when acquiring images, and interacts with the meteorological data interface to obtain weather information, and uses data analysis algorithms to perform in-depth analysis of environmental data; The color deviation correction module designs a color deviation correction algorithm based on the color change results simulated by the optical physics model and the influence rules obtained by the environmental factor analysis, and corrects the color deviation by adjusting the color space parameters of the image pixels; The micro-feature enhancement module designs a micro-feature enhancement algorithm for low-light environments, enhances the signal strength of the micro-feature area by analyzing the photon distribution, and constructs a micro-feature deformation prediction and compensation model in combination with aerodynamic principles for strong wind environments, predicts micro-feature deformation based on wind speed, wind direction, object shape and material parameters, and performs reverse compensation; The AI ​​image recognition module performs normalization preprocessing on the image that has undergone color deviation correction and micro-feature enhancement processing, and annotates the image data to construct a training data set, uses a deep learning algorithm to construct an AI image recognition model and performs training, adjusts parameters by optimizing the algorithm to improve the model recognition performance, inputs the preprocessed image to be recognized into the model, analyzes the image features and outputs the recognition result; The data storage and dynamic update module constructs a database system to store and manage data related to system operation, and regularly updates various parameters and models as new environmental data, object samples and physical research results emerge.

2. The AI-based image recognition detection system according to claim 1, characterized in that: The optical physics model processing module calculates the color intensity change of the object in combination with the environmental parameters during image acquisition, and the calculation formula is: in, represents the wavelength of light, Indicates the reflected light intensity, is the wavelength of the incident light The light intensity at is the wavelength of the object surface The reflectivity at is the wavelength of the atmosphere The transmittance at is the wavelength of the atmosphere The scattering coefficient at .

3. The AI-based image recognition detection system according to claim 1, characterized in that: The environmental factor perception and analysis module uses a data analysis algorithm to perform in-depth analysis on environmental data. The algorithm formula is: ,in, represents the comprehensive impact coefficient of environmental factors, represent Real-time measurements of environmental factors of type, yes Reference mean of type environmental factors, for The fluctuation range of type environmental factors, yes The weight coefficient of the type of environmental factors, Corresponding light intensity , Lighting Angle ,temperature ,humidity and wind speed , through this formula, a quantitative relationship model between environmental factors and object color and micro-feature changes is established.

4. The AI-based image recognition detection system according to claim 3, characterized in that: The color deviation correction module designs a color deviation correction algorithm to correct the color deviation by adjusting the color space parameters of the image pixels. Specifically, the RGB value of the original pixel is set to , the corrected RGB value is , the correction formula is: ,in, Represents the red, green, and blue component values ​​of the original pixel, Represents the red, green, and blue component values ​​of the corrected pixel, are the correction coefficients for the red, green and blue channels respectively, It is the comprehensive influence coefficient of environmental factors.

5. The AI-based image recognition detection system according to claim 1, characterized in that: In the micro-feature enhancement module, a micro-feature enhancement algorithm is designed for low-light environments, and the signal strength of the micro-feature area is enhanced by analyzing the photon distribution. The enhancement formula is: ,in, is the original grayscale value of the pixel in the low-light image, is the gray value of the pixel after enhancement, β is the enhancement coefficient, It is the mean photon count based on Poisson distribution, which means the mean photon count of image pixels under low light conditions. By introducing the enhancement coefficient β, the enhanced gray value is adjusted. , thereby achieving image brightness adjustment.

6. The AI-based image recognition detection system according to claim 1, characterized in that: In the micro-feature enhancement module, for strong wind environment, a micro-feature deformation prediction and compensation model is constructed in combination with aerodynamic principles to perform reverse compensation. Specifically, in strong wind environment , assuming that the original position of a micro-feature point on the surface of the object is , the position after compensation is According to the principles of aerodynamics, the displacement of micro-feature points is related to wind speed V, wind direction φ, and elastic modulus of the object surface. The model formula is: ,in, is the original position coordinate of the micro feature point, is the position coordinate of the micro-feature point after deformation under the action of strong wind, is the actual wind speed, is the wind direction, expressed in degrees, is the elastic modulus of the surface, They are and The compensation coefficient of the direction is determined through experiments and simulations.

7. The AI-based image recognition detection system according to claim 1, characterized in that: In the AI ​​image recognition module, a deep learning algorithm is used to construct an AI image recognition model. Specifically, the image feature vector extracted by the convolutional neural network is , the environmental factor vector is , the image feature vector and the environmental factor vector are transformed through two fully connected layers to obtain the transformed vector and , let the weight matrix of the fully connected layer be and , the biases are and ,but , , by introducing an adaptive weight adjustment mechanism, the fusion weights of image features and environmental factor features are dynamically adjusted according to the current environmental factors, and the adaptive weight α is defined. , where σ is the Sigmoid function, and is a learnable parameter, then the fusion feature vector By formula It can be obtained that through this formula, the model can use environmental information to assist image recognition.

8. The AI-based image recognition detection system according to claim 1, characterized in that: The AI ​​image recognition module improves model recognition by adjusting parameters through optimization algorithm, and its algorithm formula is: ,in, is the loss function, is the total number of categories, Yes Category The weight factor of is the true category label of the sample. When the sample belongs to the category hour, ,otherwise , The model predicts that the sample belongs to the category The probability of the model being lost is determined by the labeled data set. During the training process, the weight parameters of the model are continuously adjusted to minimize the loss function S.

9. The AI-based image recognition detection system according to claim 8, characterized in that: Said Yes Category The weight factor is given by the formula Calculate, where is the total number of samples, For Category The number of samples.

Images