Monitoring image quality diagnosis method and system based on weather perception

Abstract

The invention relates to the technical field of image processing, and particularly discloses a monitoring image quality diagnosis method and system based on weather perception, and the method comprises the steps: collecting monitoring image data from a target monitoring camera; calling a plurality of third-party weather information APIs to obtain real-time weather data of the monitored area, comparing weather conditions and visibility information provided by the APIs, and confirming the credibility of the weather information when an error is within a preset range; according to the verified weather information, an image classification model which is completely matched with the weather condition and has the closest visibility is retrieved in an image quality recognition model library; performing quality defect identification on the monitoring image data by adopting the matched image classification model, wherein the quality defects comprise one or more of lens blurring, stain pollution and foreign matter shielding; and when the quality is identified to be abnormal, obtaining the monitoring image data of the target monitoring camera again for secondary verification, and if the monitoring image data is still abnormal, outputting fault alarm information.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to a monitoring image quality diagnosis method and system based on weather perception. Background Art

[0002] With the rapid development of new urban management concepts such as digital twin cities, smart cities, smart transportation, and intelligent security, video surveillance systems have become crucial data collection and perception terminals in urban infrastructure. By capturing real-time dynamic images of the environment, surveillance cameras not only support public safety and traffic management but also provide multi-dimensional real-world data for the construction of digital twin cities. This includes real-time updates on building status, weather conditions, traffic flow, and pedestrian distribution. This data, integrated with the virtual and the real within the digital twin platform, supports the precise simulation, analysis, and optimization of urban operations, forming an indispensable technical foundation for smart city management and decision-making.

[0003] As a key front-end device in surveillance systems, camera image quality directly impacts the overall performance of the surveillance network and the accuracy of subsequent intelligent analysis (such as object detection and behavior recognition). Therefore, real-time, automated monitoring and diagnosis of surveillance camera image quality, and the timely detection of abnormalities such as lens blur, contamination, and occlusion, are crucial for ensuring the stable operation and data reliability of surveillance systems and digital twin platforms.

[0004] Traditional surveillance camera image quality testing relies primarily on manual inspections and reviews, a time-consuming and labor-intensive approach that struggles to meet the real-time, continuous monitoring requirements of large-scale video surveillance systems. The recent rise of deep learning, particularly convolutional neural networks (CNNs), has made quality assessment based on automated image classification models possible. Most existing methods achieve automated quality testing by training a single, general model and utilizing datasets containing both normal and various abnormal images to classify and assess surveillance images.

[0005] However, existing technologies have significant shortcomings, particularly in complex and changeable outdoor weather conditions, where single, general-purpose image quality assessment models often perform poorly. Outdoor weather conditions, such as sunny, cloudy, rainy, snowy, foggy, and dust storms, can have complex and significant effects on the visual characteristics of images, causing them to exhibit raindrop interference, natural blur, low contrast, graying, or color casts. For example, the image characteristics changes caused by natural environmental factors, such as raindrop interference and overall blurring caused by heavy rain, loss of distant details due to heavy fog, and yellow casts and granular noise caused by sandstorms, are highly similar to the abnormalities caused by camera hardware failures, such as lens defocus, stains, or occlusions.

[0006] Because most existing models fail to incorporate real-time weather information as a basis for judgment and lack the ability to perceive environmental changes, they are unable to effectively distinguish between weather-induced image changes and actual faults. For example, the natural blur caused by haze is often misinterpreted as lens defocus; rain blur is confused with lens stains; and the loss of image clarity caused by dusty weather is mistakenly identified as equipment failure. Such misjudgments lead to frequent false positives and missed negatives, which not only severely reduce the accuracy and reliability of automated quality diagnosis systems, but also increase the troubleshooting burden on maintenance personnel and affect the timeliness of fault resolution.

[0007] The fundamental reason for the aforementioned shortcomings of existing technologies lies in the fact that single, general-purpose image quality detection models fail to fully consider the dynamic impact of external environmental conditions, such as weather, on image features. They also fail to incorporate real-time weather information to adjust detection strategies and parameters for intelligent adaptation. In other words, these methods adopt a "one-size-fits-all" approach, attempting to address diverse and ever-changing environments with a fixed model. This inability to adjust weights and criteria for different weather scenarios results in significant accuracy degradation and insufficient robustness in practical applications.

[0008] In summary, given the high requirements of digital twin cities for the accuracy of multi-source real-time data and the intelligent development trend of monitoring system management, there is an urgent need for a technical solution that can intelligently perceive weather changes in the monitoring environment and dynamically select the most appropriate image classification model based on real-time weather information. Summary of the Invention

[0009] The purpose of this invention is to design a monitoring image quality diagnosis method and system based on weather perception, which solves the problem of insufficient diagnostic accuracy of traditional general methods under different weather conditions. By perceiving and utilizing real-time, verified weather information, it dynamically matches a dedicated image quality recognition model suitable for current weather conditions, thereby achieving more accurate and robust image quality defect identification and diagnosis, reducing false alarms and missed alarms, and improving the reliability and operation and maintenance efficiency of the monitoring system.

[0010] The present invention provides a monitoring image quality diagnosis method based on weather perception, comprising: S1: Collect surveillance image data from the target surveillance camera; S2: Call several third-party weather information APIs to obtain real-time weather data for the monitored area, compare the weather conditions and visibility information provided by each API, and confirm the credibility of the weather information when the error is within a preset range; S3: Based on the verified weather information, searching for an image classification model that fully matches the weather conditions and has the closest visibility in an image quality recognition model library, wherein the image quality recognition model library stores dedicated models for different weather conditions in a metadata management manner; S4: Using the matched image classification model to identify quality defects of the monitoring image data, the quality defects include one or more of lens blur, stain contamination, and foreign object occlusion; S5: When quality abnormality is identified, the monitoring image data of the target monitoring camera is obtained again for secondary verification. If it is still abnormal, a fault alarm message is output.

[0011] Preferably, in step S1, collecting monitoring image data from a target monitoring camera includes: Capture surveillance image data from the target surveillance camera at a preset period; Respond to the trigger command within the system and actively capture monitoring image data from the target monitoring camera in real time. Preferably, in step S2, comparing the weather conditions and visibility information provided by each API includes: Calculate the standard deviation of the visibility values provided by each API. When the standard deviation is less than the preset threshold, the visibility information is confirmed to be credible. The weather conditions provided by each API are voted and counted, and when the consistency reaches the preset ratio, the weather conditions are confirmed to be credible.

[0012] Preferably, in step S3, searching the image quality recognition model library for an image classification model that fully matches the weather conditions and has the closest visibility based on the verified weather information includes: Using the verified weather conditions as an exact matching condition, filtering the metadata records; and selecting, from the filtered results, a model whose applicable visibility range includes the current real-time visibility or is closest to the current real-time visibility value; When an image classification model matching the current weather information cannot be retrieved from the model library, the model training process is triggered.

[0013] Preferably, in step S3, the image quality recognition model library adopts metadata management including: A metadata record is created for each image classification model in the image quality recognition model library, wherein the metadata record contains at least a unique model identifier, a weather type corresponding to the model, an applicable visibility range, and a storage path of the model file.

[0014] Preferably, the model training process includes: The monitoring image data under the current weather conditions are cached as training samples, and quality defect category labels are added to the samples through manual annotation. A new image classification model is trained using the labeled samples, and finally the trained image classification model and its corresponding weather metadata are stored in the model library.

[0015] Preferably, the image classification model is constructed based on a pre-trained ResNet-18 network, including a hierarchical feature extraction layer and a feature fusion module; The hierarchical feature extraction layer includes the first three feature extraction layers layer1, layer2, and layer3 of ResNet-18, which output feature maps f1, f2, and f3 of different spatial resolutions respectively; The feature fusion module concatenates the feature maps f1, f2, and f3 in the channel dimension, performs dimensionality reduction and channel fusion through 1×1 convolution, and sequentially undergoes batch normalization and ReLU activation function processing.

[0016] Preferably, in step S4, the using the matched image classification model to identify quality defects of the monitoring image data includes: The feature fusion module of the image classification model splices feature maps of different levels in the channel dimension, and then uses the ReLU activation function and global average pooling to output a fixed-dimensional vector, which is further sent to the Dropout and fully connected layers for classifying quality defect categories.

[0017] Preferably, in step S5, the method of outputting the fault alarm information includes: An alarm notification containing camera identification, abnormality type and current weather information is sent to the preset operation and maintenance personnel or management system through the robot interface or message queue of the instant messaging tool.

[0018] Preferably, a monitoring image quality diagnosis system based on weather perception includes: An image information acquisition module is used to collect monitoring image data from a target monitoring camera; The weather information acquisition module calls several third-party weather information APIs to obtain real-time weather data for the monitored area, compares the weather conditions and visibility information provided by each API, and confirms the credibility of the weather information when the error is within a preset range; A model selection module is used to retrieve an image classification model with a completely matched weather condition and the closest visibility from the image quality recognition model library based on the verified weather information; A defect recognition module is used to identify quality defects of the monitoring image data using the matched image classification model, wherein the quality defects include one or more of lens blur, stain contamination, and foreign object occlusion; The secondary verification module is used to obtain the monitoring image data of the target monitoring camera again for secondary verification when quality abnormalities are identified. If the abnormality is still found, a fault alarm message is output.

[0019] Compared with the prior art, the present invention has the following beneficial effects: The present invention discloses a monitoring image quality diagnosis method and system based on weather perception. In combination with the real-time weather environment, a corresponding environmental model is selected to reduce misjudgment caused by weather differences and improve the accuracy and stability of image quality diagnosis; multi-source weather data verification and secondary image acquisition verification reduce the probability of false alarms and missed alarms, and ensure the credibility of system alarms; the model library is based on the classified storage and dynamic retrieval of weather and visibility, so that the system can adapt to the changing natural environment and improve the intelligent level of monitoring quality management; the secondary verification and alarm mechanism after abnormality identification ensures the timeliness of maintenance, so that maintenance personnel can quickly locate and handle problems, and reduce monitoring blind spots and safety hazards. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 This is a flow chart of a method for diagnosing monitoring image quality based on weather perception provided by an embodiment of the present invention; Figure 2 This is a schematic diagram of a monitoring image quality diagnosis system module based on weather perception provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0021] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0022] like Figure 1 As shown, the present application provides a monitoring image quality diagnosis method based on weather perception, comprising: S1: Collect surveillance image data from the target surveillance camera; S2: Call several third-party weather information APIs to obtain real-time weather data for the monitored area, compare the weather conditions and visibility information provided by each API, and confirm the credibility of the weather information when the error is within a preset range; S3: Based on the verified weather information, searching for an image classification model that fully matches the weather conditions and has the closest visibility in an image quality recognition model library, wherein the image quality recognition model library stores dedicated models for different weather conditions in a metadata management manner; S4: Using the matched image classification model to identify quality defects of the monitoring image data, the quality defects include one or more of lens blur, stain contamination, and foreign object occlusion; S5: When quality abnormality is identified, the monitoring image data of the target monitoring camera is obtained again for secondary verification. If it is still abnormal, a fault alarm message is output.

[0023] In this solution, real-time weather information is integrated with a dynamic model selection mechanism to achieve intelligent diagnosis of surveillance camera image quality. Specifically, it involves five key steps: First, image data is collected from designated surveillance cameras as the basis for diagnosis; second, the reliability of acquired weather data (such as weather conditions and visibility) is ensured by invoking multiple third-party weather APIs and performing cross-validation; third, based on the verified weather information, the image classification model that best matches the current weather conditions is retrieved from a pre-built library of image quality recognition models; fourth, this matching model is used to identify quality defects in the collected image data and classify common camera fault types (such as lens blur, contamination, or obstruction); fifth, if the diagnostic results indicate an anomaly, a secondary image acquisition and verification mechanism is implemented to confirm the authenticity of the fault and output an alarm message for timely resolution.

[0024] In step S1, real-time image data is collected from a specific surveillance camera via a network communication protocol (e.g., HTTP). This collection can be periodic or event-triggered, ensuring that the acquired image data represents the current visual state of the monitored area. This image data, serving as the basis for subsequent diagnostics, is typically stored in a digital format (e.g., JPEG or RAW) and can be managed via an object-based storage system. In step S2, multiple external weather service interfaces, such as Hefeng Weather and OpenWeatherMap, are called to obtain weather data for the current monitored area, including weather conditions (e.g., sunny, cloudy, rainy, snowy) and visibility (in meters or kilometers, quantifying air transparency). Subsequently, a data comparison is performed: the consistency of weather condition categories (e.g., string matching) and the difference in visibility values (e.g., absolute error or relative error) between the APIs are calculated. Only when the difference is within a preset threshold (the weather conditions are completely consistent and the visibility error is less than 10%) is the data considered credible; otherwise, an alert mechanism is triggered, such as notifying operations and maintenance personnel to change the API. In step S3, the weather data verified in S2 is used as the query condition to search the pre-built model library. The model library uses a relational database to store model metadata (including model ID, weather type, visibility, model path, etc.), and the model file itself is saved in the form of object storage. The retrieval logic prioritizes ensuring an exact match of weather conditions (such as exact string comparison), and then optimizes visibility matching (such as minimizing the absolute difference) to select the most suitable model. The image classification model is designed based on a deep learning framework and fine-tuned for specific weather conditions through transfer learning; In step S4, the image data collected in step S1 is processed using the image classification model retrieved in step S3 (a custom ResNet-based network). The model extracts features using a convolutional neural network and outputs a classification result, identifying the category of image quality defects (such as "no abnormality," "lens blur," "stain contamination," and "foreign object occlusion"). The output includes a predicted label and a confidence score, with the defect type defined based on a standard dataset. In step S5, if the diagnostic result from S4 indicates an anomaly, such as if the confidence level exceeds a threshold, a secondary verification is triggered: S1 is re-executed to acquire image data and the model from S4 is re-applied for diagnosis. If the secondary result is still an anomaly, an alarm is generated, including fault details, camera ID, and timestamp. The verification mechanism can be configured with a delay or multiple iterations to filter out transient interference.

[0025] Preferably, in step S1, collecting monitoring image data from a target monitoring camera includes: Capture surveillance image data from the target surveillance camera at a preset period; Respond to the trigger command within the system and actively capture monitoring image data from the target monitoring camera in real time.

[0026] In the above solution, it is supported to specify the cycle for different surveillance cameras Capture surveillance images and support other modules in the system to actively trigger the acquisition of surveillance images from designated cameras.

[0027] In terms of specific implementation, an IP camera that provides a snapshot interface through the http communication protocol can be used, and the scheduled snapshot and active snapshot functions can be further encapsulated through Python / Flask on the back end to provide a corresponding RESTful style interface. Preferably, in step S2, comparing the weather conditions and visibility information provided by each API includes: Calculate the standard deviation of the visibility values provided by each API. When the standard deviation is less than the preset threshold, the visibility information is confirmed to be credible. The weather conditions provided by each API are voted and counted, and when the consistency reaches the preset ratio, the weather conditions are confirmed to be credible.

[0028] In the above solution, multiple third-party weather information APIs can be called to compare whether the information is consistent, with a focus on weather conditions (such as sunny, overcast, cloudy, light rain, heavy rain, thunderstorms, snow, etc.) and visibility (a measure of air transparency, commonly seen in haze weather). The weather information will only be trusted if the error is within a certain range. Otherwise, the operation and maintenance personnel will be notified to replace the API.

[0029] Preferably, in step S3, searching the image quality recognition model library for an image classification model that fully matches the weather conditions and has the closest visibility based on the verified weather information includes: Using the verified weather conditions as an exact matching condition, filtering the metadata records; and selecting, from the filtered results, a model whose applicable visibility range includes the current real-time visibility or is closest to the current real-time visibility value; When an image classification model matching the current weather information cannot be retrieved from the model library, the model training process is triggered.

[0030] Preferably, in step S3, the image quality recognition model library adopts metadata management including: A metadata record is created for each image classification model in the image quality recognition model library, wherein the metadata record contains at least a unique model identifier, a weather type corresponding to the model, an applicable visibility range, and a storage path of the model file.

[0031] Preferably, the model training process includes: The monitoring image data under the current weather conditions are cached as training samples, and quality defect category labels are added to the samples through manual annotation. A new image classification model is trained using the labeled samples, and finally the trained image classification model and its corresponding weather metadata are stored in the model library.

[0032] In one embodiment provided in this application, the trained model file is stored in the form of an object through MinIO, and the model metadata is stored through a relational database (such as MySQL / PostgreSQL). Table 1 is an example of a table of model metadata: Table 1: Example of a table of model metadata Based on the acquired weather information, this module searches the existing model library for the most suitable image classification model (one that perfectly matches the weather conditions and has visibility as close as possible) (this can be fine-tuned based on an existing pre-trained model, such as MobileNet). Retrieval can be achieved using MySQL statements.

[0033] If a suitable model cannot be found, captured surveillance images are saved for manual annotation. These images can be categorized as normal, blurry, contaminated, or obstructed, forming a corresponding weather data set for supervised training. Finally, the trained new model is saved in the model library. The dataset format uses a PyTorch-compatible ImageFolder structure, and transfer learning is accomplished using PyTorch interfaces (such as defining loss functions using torch.nn.CrossEntropyLoss()).

[0034] Preferably, the image classification model is constructed based on a pre-trained ResNet-18 network, including a hierarchical feature extraction layer and a feature fusion module; The hierarchical feature extraction layer includes the first three feature extraction layers of ResNet-18, layer1, layer2, and layer3 (the layer naming is based on Pytorch), which output feature maps f1, f2, and f3 of different spatial resolutions respectively; The feature fusion module concatenates the feature maps f1, f2, and f3 in the channel dimension, performs dimensionality reduction and channel fusion through 1×1 convolution, and sequentially undergoes batch normalization and ReLU activation function processing.

[0035] In one embodiment provided in this application, the image classification model is a self-designed convolutional neural network based on the existing pre-trained model ResNet. A feature fusion module is added to improve the model complexity and nonlinear modeling capabilities. It is suitable for improving performance when fine-tuning on small sample data. The image classification model is implemented using Pytorch as follows: First, the first three feature extraction layers of ResNet-18, specifically layer1, layer2 and layer3 in torchvision.models.resnet18, are hierarchically intercepted to form the shared basic part of the network. Assume that the input image is , the network extracts features through three sub-modules ,in: Each level of output feature map has different spatial resolution and semantic expression capabilities, focusing on edges, textures and local structures respectively.

[0036] In order to fully exploit the complementary information of each layer, a simple but efficient cross-layer feature fusion module is designed. After splicing the outputs of multiple layers in the channel dimension, a 1×1 convolution is performed to achieve dimensionality reduction and channel fusion, namely: in[ ] indicates channel splicing, To fuse the convolution kernel, BN stands for Batch Normalization. is the ReLU activation function. This fusion mechanism not only improves feature utilization, but also effectively alleviates the deep overfitting problem.

[0037] The fused feature map is compressed into a fixed-length vector through global average pooling (GAP), and then fed into the Dropout and fully connected layers to output the final classification result. Let the final output be: The entire network structure takes into account both migration efficiency and training speed by freezing the bottom layer and training the top layer, and is particularly suitable for rapid modeling of small-scale image datasets.

[0038] Preferably, in step S4, the using the matched image classification model to identify quality defects of the monitoring image data includes: The feature fusion module of the image classification model splices feature maps of different levels in the channel dimension, and then uses the ReLU activation function and global average pooling to output a fixed-dimensional vector, which is further sent to the Dropout and fully connected layers for classifying quality defect categories.

[0039] In one embodiment provided in the present application, after selecting the image classification model, the module will call the model to classify the captured image (no abnormality, lens blur, stain contamination, occlusion, etc.). If an abnormality occurs, the monitoring image of the corresponding camera will be captured again and classified again. If an abnormality occurs again, the operation and maintenance personnel will be notified in the same way as the weather information acquisition module for timely processing.

[0040] In addition, a user-friendly graphical web management interface can be built to support manual review of random image classification results. For example, the classification results can be stored in a database, and its table structure can be defined as shown in Table 2: Table 2: Table structure of classification results Preferably, in step S5, the method of outputting the fault alarm information includes: An alarm notification containing camera identification, abnormality type and current weather information is sent to the preset operation and maintenance personnel or management system through the robot interface or message queue of the instant messaging tool.

[0041] Preferably, a monitoring image quality diagnosis system based on weather perception includes: An image information acquisition module is used to collect monitoring image data from a target monitoring camera; The weather information acquisition module calls several third-party weather information APIs to obtain real-time weather data for the monitored area, compares the weather conditions and visibility information provided by each API, and confirms the credibility of the weather information when the error is within a preset range; A model selection module is used to retrieve an image classification model with a completely matched weather condition and the closest visibility from the image quality recognition model library based on the verified weather information; A defect recognition module is used to identify quality defects of the monitoring image data using the matched image classification model, wherein the quality defects include one or more of lens blur, stain contamination, and foreign object occlusion; The secondary verification module is used to obtain the monitoring image data of the target monitoring camera again for secondary verification when quality abnormalities are identified. If the abnormality is still found, a fault alarm message is output.

[0042] In the above solution, through the implementation of the above technical solution, the following beneficial technical effects can be produced: Improving the reliability of weather information acquisition: By invoking and comparing several third-party weather information APIs within the weather information acquisition module, the accuracy of the weather information used as the basis for subsequent model selection is ensured. When data from different APIs differ significantly, the system can detect potential issues with the weather data source, preventing errors in the diagnostic process caused by a single, erroneous weather information input, and laying a solid foundation for subsequent accurate model matching.

[0043] Dynamically optimize and select diagnostic models to enhance adaptability: Based on reliable weather information provided by the weather information acquisition module, the model selection module searches and selects the model that "fully matches the weather conditions and has the closest visibility" from a library of weather-specific models. This means that the system no longer uses a single, universal model to handle all weather conditions. Instead, it selects the specialized model that performs best under the specific weather conditions (such as "heavy rain, visibility 500 meters" or "sunny, visibility 10 kilometers"). This enables the diagnostic method to dynamically adapt to environmental changes, significantly improving its ability to understand and process image features under different weather backgrounds.

[0044] Improved accuracy in identifying image quality defects, reducing false alarms and missed alarms: Because the defect recognition module utilizes an image classification model that closely matches current weather conditions, the model has already learned the characteristics of normal images under those conditions, as well as various fault characteristics, during training. Consequently, it can more effectively distinguish between normal image changes (such as blur, contrast loss, and color shift) caused by the prevailing weather (fog, rain, and dust) and those caused by real faults such as lens blur, stains, and occlusions. This directly improves the accuracy of identifying real faults and significantly reduces the misclassification of normal images as faults due to weather conditions (false alarms), or the overlooking of real faults in inclement weather (missed alarms).

[0045] Enhanced fault alarm confidence: The secondary verification module introduces a secondary verification mechanism that, after initially identifying an anomaly, recaptures images and re-identifies them to confirm the persistence of the fault. This helps filter out short-term image anomalies caused by transient interference (such as birds flying overhead, brief strong light reflections, and temporary vehicle occlusions), ensuring that only persistent conditions with a high probability of true hardware failure or stable occlusion trigger alarms. This further improves alarm accuracy and reliability and reduces unnecessary maintenance responses.

[0046] To summarize, this application overcomes the defect in existing technologies that a single model is difficult to adapt to complex weather changes by introducing mechanisms such as weather perception, multi-source weather information verification, dynamic model matching, and secondary verification. This enables high-precision, high-reliability intelligent diagnosis of monitoring image quality under various weather conditions, ultimately improving the operation and maintenance efficiency and overall effectiveness of the monitoring system.

[0047] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A monitoring image quality diagnosis method based on weather perception, characterized in that: include: S1: Collect surveillance image data from the target surveillance camera; S2: Call several third-party weather information APIs to obtain real-time weather data for the monitored area, compare the weather conditions and visibility information provided by each API, and confirm the credibility of the weather information when the error is within a preset range; S3: Based on the verified weather information, searching for an image classification model that fully matches the weather conditions and has the closest visibility in the image quality recognition model library, wherein the image quality recognition model library stores dedicated models for different weather conditions in a metadata management manner; S4: Using the matched image classification model to identify quality defects of the monitoring image data, the quality defects include one or more of lens blur, stain contamination, and foreign object occlusion; S5: When quality abnormality is identified, the monitoring image data of the target monitoring camera is obtained again for secondary verification. If it is still abnormal, a fault alarm message is output.

2. The method for diagnosing surveillance image quality based on weather perception according to claim 1, characterized in that: In step S1, collecting monitoring image data from a target monitoring camera includes: Capture surveillance image data from the target surveillance camera at a preset period; Respond to the trigger command within the system and actively capture monitoring image data from the target monitoring camera in real time.

3. The method for diagnosing surveillance image quality based on weather perception according to claim 1, characterized in that: In step S2, the comparison of weather conditions and visibility information provided by each API includes: Calculate the standard deviation of the visibility values provided by each API. When the standard deviation is less than the preset threshold, the visibility information is confirmed to be credible. The weather conditions provided by each API are voted and counted, and when the consistency reaches the preset ratio, the weather conditions are confirmed to be credible.

4. The method for diagnosing surveillance image quality based on weather perception according to claim 3, characterized in that: In step S3, searching the image quality recognition model library for an image classification model that fully matches the weather conditions and has the closest visibility based on the verified weather information includes: Using the verified weather conditions as an exact matching condition, filtering the metadata records; and selecting, from the filtered results, a model whose applicable visibility range includes the current real-time visibility or is closest to the current real-time visibility value; When an image classification model matching the current weather information cannot be retrieved from the model library, the model training process is triggered.

5. The method for diagnosing surveillance image quality based on weather perception according to claim 4, characterized in that: In step S3, the image quality recognition model library adopts metadata management including: A metadata record is created for each image classification model in the image quality recognition model library, wherein the metadata record contains at least a unique model identifier, a weather type corresponding to the model, an applicable visibility range, and a storage path of the model file.

6. The method for diagnosing surveillance image quality based on weather perception according to claim 4, characterized in that: The model training process includes: The monitoring image data under the current weather conditions are cached as training samples, and quality defect category labels are added to the samples through manual annotation. A new image classification model is trained using the labeled samples, and finally the trained image classification model and its corresponding weather metadata are stored in the model library.

7. The method for diagnosing surveillance image quality based on weather perception according to claim 6, characterized in that: The image classification model is built based on a pre-trained ResNet-18 network, including a hierarchical feature extraction layer and a feature fusion module; The hierarchical feature extraction layer includes the first three feature extraction layers layer1, layer2, and layer3 of ResNet-18, which output feature maps f1, f2, and f3 of different spatial resolutions respectively; The feature fusion module splices the feature maps f1, f2, and f3 in the channel dimension, performs dimensionality reduction and channel fusion through 1×1 convolution, and then sequentially undergoes batch normalization and ReLU activation function processing.

8. The method for diagnosing surveillance image quality based on weather perception according to claim 7, characterized in that: In step S4, the use of the matched image classification model to identify quality defects of the monitoring image data includes: The feature fusion module of the image classification model splices feature maps of different levels in the channel dimension, and then uses the ReLU activation function and global average pooling to output a fixed-dimensional vector, which is further sent to the Dropout and fully connected layers for classifying quality defect categories.

9. The method for diagnosing surveillance image quality based on weather perception according to claim 1, characterized in that: In step S5, the method of outputting fault alarm information includes: An alarm notification containing camera identification, abnormality type and current weather information is sent to the preset operation and maintenance personnel or management system through the robot interface or message queue of the instant messaging tool.

10. A monitoring image quality diagnosis system based on weather perception, characterized in that: include: An image information acquisition module is used to collect monitoring image data from a target monitoring camera; The weather information acquisition module calls several third-party weather information APIs to obtain real-time weather data for the monitored area, compares the weather conditions and visibility information provided by each API, and confirms the credibility of the weather information when the error is within a preset range; A model selection module is used to retrieve an image classification model with a completely matched weather condition and the closest visibility from the image quality recognition model library based on the verified weather information; A defect recognition module is used to identify quality defects of the monitoring image data using the matched image classification model, wherein the quality defects include one or more of lens blur, stain contamination, and foreign object occlusion; The secondary verification module is used to obtain the monitoring image data of the target monitoring camera again for secondary verification when quality abnormalities are identified. If the abnormality is still found, a fault alarm message is output.

Images