A tower-based omnidirectional sky camera system

Abstract

The present invention discloses an omnidirectional sky camera system based on a tower base, comprising: an omnidirectional sky camera fixed on the tower base, and an analysis terminal connected to the omnidirectional sky camera via a data transmission module, wherein the airspace of a meteorological site is photographed by the omnidirectional sky camera to obtain an airspace image, the airspace image is regularly transmitted to the analysis terminal via the data transmission module, the airspace image and atmospheric electric field data are received by the analysis terminal, and the airspace image and the atmospheric electric field data are jointly analyzed to generate an analysis report, wherein the analysis report includes meteorological feature analysis results and causes of atmospheric electric field anomalies.

Description

Technical Field

[0001] The invention belongs to the technical field of atmospheric electric field observation and image acquisition, and in particular relates to an omnidirectional sky camera system based on a tower base. Background Art

[0002] Importance of atmospheric electric field observation: In atmospheric electric field observation, accurate detection of meteorological factors such as wind speed, wind direction, precipitation, cloud cover, etc. is crucial. These data not only help analyze the environmental status of the atmospheric electric field, but also are the key basis for determining the causes of abnormal atmospheric electric field.

[0003] Limitations of traditional observation methods: Traditional atmospheric electric field observation stations mainly rely on a single atmospheric electric field probe for detection, while other key meteorological data such as wind speed, wind direction, precipitation, cloud cover, etc. usually need to be obtained from meteorological data centers far away from the observation station. This post-acquisition method is not only inefficient, but also because there is usually a large distance between the meteorological station and the atmospheric electric field instrument, the data obtained is difficult to accurately reflect the actual meteorological environment of the observation station. Summary of the invention

[0004] In order to solve the above technical problems, the present invention proposes an omnidirectional sky camera system based on a tower base, which is designed to observe the sky environment at the location of the atmospheric electric field instrument in real time, and perform real-time, high-precision image observation and data transmission on the omnidirectional sky conditions at the site where the atmospheric electric field instrument is located. By capturing high-definition image data, it provides key meteorological information including cloud cover, rainfall, snowfall, etc., thereby providing more accurate and timely meteorological data support for atmospheric electric field observation.

[0005] To achieve the above object, the present invention provides an omnidirectional sky camera system based on a tower base, comprising:

[0006] An omnidirectional sky camera fixed on a tower base, and an analysis terminal connected to the omnidirectional sky camera via a data transmission module, wherein the omnidirectional sky camera is used to photograph the airspace of the meteorological site to obtain an airspace image, and the airspace image is regularly transmitted to the analysis terminal via the data transmission module, and the airspace image and atmospheric electric field data are received by the analysis terminal, and the airspace image and the atmospheric electric field data are jointly analyzed to generate an analysis report, wherein the analysis report includes meteorological characteristic analysis results and causes of atmospheric electric field anomalies.

[0007] Optionally, the omnidirectional sky camera includes a lens, an image sensor and an image processing chip connected in sequence, and the image processing chip performs preliminary denoising and enhancement processing on the original image collected by the image sensor from the lens to generate a spatial image; a shooting control unit is connected to the image sensor, and the shooting parameters of the camera are controlled by the shooting control unit, wherein the shooting parameters include resolution, frame rate and shooting range.

[0008] Optionally, the data transmission module includes an FPGA chip and an FTP system, wherein the spatial domain image is compressed and packaged by the FPGA chip, and the packaged data is transmitted at fixed time intervals through a transmission channel established by the FTP system.

[0009] Optionally, in the analysis terminal, the process of jointly analyzing the airspace image and the atmospheric electric field data includes: performing image recognition on the airspace image to generate meteorological features, and jointly analyzing the meteorological features with the atmospheric electric field data to generate atmospheric electric field anomalies and corresponding causes.

[0010] Optionally, in the analysis terminal, the process of performing image recognition on the spatial domain image includes:

[0011] The spatial domain image is preprocessed and enhanced, and the feature extraction of the enhanced spatial domain image is performed through a deep learning model to obtain meteorological features, wherein the deep learning model is trained and generated by transfer learning and semi-supervised learning, and the feature extraction of the enhanced spatial domain image is performed by multimodal fusion in the deep learning model, wherein the multimodal fusion is to set the input data of the deep learning model as the image features of the enhanced spatial domain image of several consecutive frames, fit the correlation between the input data and the meteorological features through the deep learning model, and output the meteorological features through the deep learning model; the deep learning model is also updated through continuous learning and iterative improvement.

[0012] Optionally, in the analysis terminal, the process of jointly analyzing the meteorological characteristics and the atmospheric electric field data includes:

[0013] Obtain historical meteorological characteristics and historical atmospheric electric field data;

[0014] Performing data matching on the historical meteorological characteristics and historical atmospheric electric field data;

[0015] Conducting a correlation test on the historical meteorological characteristics and historical atmospheric electric field data on the trend of changes;

[0016] Based on the test results, the historical meteorological characteristics and historical atmospheric electric field data are pattern recognized through statistical analysis and machine learning algorithms to generate electric field change patterns and electric field anomaly data corresponding to the meteorological characteristics;

[0017] Based on the electric field variation pattern and the electric field anomaly data, combined with meteorological characteristics, anomaly detection is performed on the atmospheric electric field data to generate atmospheric electric field data anomalies and causes.

[0018] Optionally, in the analysis terminal, the analysis report is displayed through an interactive analysis interface.

[0019] Optionally, in the analysis terminal, the airspace image and atmospheric electric field data are backed up, and the functions in the analysis terminal are subsequently adjusted and optimized.

[0020] Compared with the prior art, the present invention has the following advantages and technical effects:

[0021] The all-round sky camera system based on the tower base can effectively realize real-time meteorological observation and data analysis at the atmospheric electric field observation site. Through all-round shooting and precise joint analysis, the system provides strong support for meteorological research and forecasting, helps to timely discover and predict abnormal meteorological events, and provides a scientific basis for relevant decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0023] Figure 1 It is an overall schematic diagram of the tower crane fixing device and the atmospheric electric field instrument of the omnidirectional sky camera according to an embodiment of the present invention after installation;

[0024] Figure 2 This is a partial enlarged view of the connection between the omnidirectional sky camera fixing bracket and the tower crane according to an embodiment of the present invention.

[0025] Figure 3 4 is a structural diagram of an omnidirectional sky camera system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0026] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0027] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0028] The present invention provides an omnidirectional sky camera system based on a tower base, which relates to the field of atmospheric electric field observation. The system includes an omnidirectional sky camera and data transmission software matched therewith. The camera is used to capture the meteorological conditions around the site, including but not limited to cloud cover, rainfall, snowfall, etc. The data transmission system transmits the captured image to a designated terminal through an FTP system once every 15 minutes.

[0029] The advantage of the present invention is that it makes the observation of the atmospheric electric field no longer rely on the data provided by the traditional meteorological station, thereby ensuring that the observed meteorological conditions are the actual meteorological conditions at the location and time of the station. This method not only reduces the cost of data acquisition, but also improves the accuracy of meteorological observation. In addition, since the system has the ability to take all-round shots, it can also provide more comprehensive and detailed meteorological observation data.

[0030] like Figure 3 As shown, a tower-based omnidirectional sky camera system comprises:

[0031] An all-round sky camera installed at the atmospheric electric field observation site is used to capture and photograph the meteorological conditions of the entire airspace where the site is located;

[0032] Data transmission module, which transmits the images captured by the camera to the designated terminal every 15 minutes through the FPGA chip and the FTP system on the chip;

[0033] The analysis terminal is used to receive and analyze the transmitted images, and conduct joint analysis in combination with the atmospheric electric field data to generate an analysis report.

[0034] like Figure 1-2 As shown, Figure 1 The tower crane fixing device of the omnidirectional sky camera and the overall schematic diagram of the atmospheric electric field instrument after installation are shown. The figure details the relative positions and assembly relationships of key components such as the fixing bracket, adjustment mechanism and locking device, and clearly shows how the omnidirectional sky camera is stably fixed on the tower crane structure through the device. Figure 2 The enlarged partial view of the connection between the omnidirectional sky camera fixing bracket and the tower crane is shown in detail, especially highlighting the fine-tuning function of the adjustment mechanism. Through this picture, we can clearly see how the adjustment mechanism can achieve precise adjustment of the installation angle and position of the omnidirectional sky camera to meet the needs of different tower crane structures and installation environments.

[0035] exist Figure 1-2 In the invention, the camera is fixed on a high tower, i.e., a tower base, through a fixing device, wherein the fixing device is fixed to the tower base through a steel hoop, and the fixing device is fixed to the steel hoop through a flat head screw and a square nut. The fixing device includes an outer frame, and a camera base bracket is arranged on the outer frame. The camera is fixed to a storage place in the center of the camera base bracket. A left-side punch-free telescopic rod, a right-side punch-free telescopic rod and a top punch-free telescopic rod are arranged on the outer frame to clamp and fix the camera, wherein the punched telescopic rod includes mutually fixed anti-skid pads, a base and a stainless steel pipe in sequence from the camera to the outer frame, a wing screw is welded on the outer frame, a nut is arranged on the screw, and the stainless steel pipe is sleeved on the screw, and the stainless steel pipe, the base and the anti-skid pad are pushed to tighten the camera by tightening the nut. At least one side of the outer frame is open, and the camera is used to shoot the airspace by adjusting the angle of the fixing device and the camera.

[0036] The omnidirectional sky camera is designed to capture the entire airspace where the station is located, thus providing comprehensive and continuous meteorological observation data;

[0037] The analysis terminal can receive and analyze images transmitted by the camera in real time, and conduct joint analysis with atmospheric electric field data to improve the accuracy and real-time performance of meteorological observations.

[0038] Among the above-mentioned related modules, the omnidirectional sky camera module includes camera hardware and a shooting control unit, wherein the camera hardware includes a lens, an image sensor, an FPGA chip, etc. connected in sequence. The shooting control unit is connected to the image sensor and is responsible for setting the camera's shooting parameters, such as resolution, frame rate, shooting range, etc. The FPGA chip performs image preprocessing on the original image, and the image preprocessing is used to perform preliminary processing on the captured original image, such as denoising, color correction, etc.

[0039] The data transmission module mainly uses the FTP system, which is responsible for establishing a stable transmission channel with the terminal. It also includes the FPGA chip, which compresses and packages the spatial image data after pre-processing the original image, and controls the FTP transmission. The FPGA chip controls the data transmission of the FTP system to ensure that the image data is transmitted on time every 15 minutes.

[0040] The analysis terminal mainly includes the following functions: Image reception: responsible for receiving and processing image data transmitted from the FTP system. Atmospheric electric field data reception: simultaneously receiving data collected by the atmospheric electric field instrument. Joint analysis of images and electric fields: comparing and analyzing image data with atmospheric electric field data to obtain joint analysis results, which include meteorological characteristics and atmospheric electric field abnormal data and causes.

[0041] The improved analysis method of joint analysis mainly uses the following techniques:

[0042] Image recognition technology: Using computer vision algorithms, the images captured by the camera are automatically identified for key meteorological features such as cloud cover, precipitation, and snow, reducing the workload of manual analysis.

[0043] Machine learning model: Establish a correlation model between the atmospheric electric field and meteorological characteristics based on machine learning, and improve the accuracy and efficiency of joint analysis by training the model.

[0044] Time series analysis: Combined with historical data, time series analysis is performed on the meteorological characteristics and atmospheric electric fields of continuous image data to discover trends and laws of meteorological changes.

[0045] Interactive Analysis Interface: Design a user-friendly interactive analysis interface to facilitate meteorologists and analysts to intuitively compare and analyze images and data.

[0046] Feedback mechanism: A feedback mechanism is established to feed back the results of the joint analysis to the camera module and the data transmission module so that parameters can be dynamically adjusted according to actual needs.

[0047] Through the combination of the above modules and improved analysis methods, efficient and accurate meteorological observation and data analysis of the tower-based all-round sky camera system can be achieved.

[0048] The above contents are explained through the relevant implementation steps of the tower-based all-round sky camera system:

[0049] 1. System deployment and preparation

[0050] At a meteorological observation station in a certain mountain area, the present invention selected a 10-meter-high observation tower as the tower base. This location has a wide field of view and no obstructions, and can capture the entire airspace where the station is located. On the tower base, the present invention installed a high-definition omnidirectional sky camera, which was precisely adjusted to ensure that the camera can capture the entire sky in the area where the station is located. Subsequently, a data transmission module was configured near the camera, including a high-performance FPGA chip and an FTP system. The FPGA chip is configured to automatically capture an image every 15 minutes, and control and transmit it to a remote terminal in real time through the FTP system.

[0051] 2. System Operation and Image Acquisition

[0052] After the camera is turned on, it starts to continuously capture the sky. Every 15 minutes, the FPGA chip preprocesses the original image, including removing noise and correcting color, to improve image quality. The preprocessed image data, i.e., the airspace image, is transmitted to the storage server of the remote terminal through the FTP system. At the same time, the atmospheric electric field instrument is also continuously collecting atmospheric electric field data.

[0053] 3. Terminal Analysis and Joint Processing

[0054] At the remote terminal, after receiving the image data, the dedicated analysis module immediately stores it and associates it with the atmospheric electric field data. Through computer vision algorithms, the analysis terminal automatically identifies meteorological features such as cloud cover, precipitation, and snow in the image. Subsequently, these meteorological features are jointly analyzed with the atmospheric electric field data to find potential correlations between them. After comparison and analysis, the analysis terminal found that when the atmospheric electric field intensity reaches a certain threshold, the changes in cloud cover and precipitation show obvious patterns. The above patterns are fitted by historical data to generate thresholds for atmospheric electric field intensity data to determine whether the atmospheric electric field is abnormal, abnormal data and causes. Based on these analysis results, the terminal generates a detailed analysis report and presents it to meteorologists through an interactive analysis interface.

[0055] 4. System maintenance and management

[0056] In order to ensure the continuous and stable operation of the system, the present invention has formulated a regular inspection and maintenance plan. Every month, a professional team is sent to the observatory to conduct a comprehensive inspection and maintenance of the camera, data transmission module and analysis terminal. At the same time, in order to ensure the security of the data, a data backup and recovery mechanism has been established. Every week, a full backup of the data is performed to prevent accidental loss. In addition, with the continuous deepening of meteorological research and technological advancement, the system is upgraded and optimized according to the latest research results and technological developments to improve its performance and accuracy.

[0057] Through this implementation case, we can see that the tower-based all-round sky camera system has shown powerful functions and potential in practical applications. It can not only observe and record sky conditions in real time, but also conduct joint analysis with atmospheric electric field data to provide strong support for meteorological research and forecasting. At the same time, through regular inspection and maintenance, as well as data backup and recovery, the stability of the system and the security of data are ensured. With the continuous advancement of technology and continuous optimization of the system, this tower-based all-round sky camera system will play a more important role in future meteorological observations and research.

[0058] Based on the tower-based all-round sky camera system, the above-mentioned related implementation steps are explained in detail:

[0059] 1. System deployment and preparation

[0060] 1. Choose a suitable tower base location: Choose a suitable tower base location at the atmospheric electric field observation site to ensure that the omnidirectional sky camera can capture the entire airspace where the site is located.

[0061] 2. Install an all-round sky camera: Figure 1-2As shown, install an omnidirectional sky camera on the tower base, and adjust the angle and height of the camera to ensure that the shooting range covers the entire airspace.

[0062] 3. Configure the data transmission module: Install the data transmission module near the camera, including the FPGA chip and the FTP system. Configure the FPGA chip so that it can automatically capture an image every 15 minutes and transmit it to the designated terminal through the FTP system.

[0063] 4. Configure relevant functions in the analysis terminal: Install dedicated analysis software on the terminal device to receive and jointly analyze the image data transmitted by the camera.

[0064] 2. System Operation and Image Acquisition

[0065] 1. Start the omnidirectional sky camera: Start the camera and start shooting the sky conditions at the site. The camera will shoot continuously according to the preset shooting parameters.

[0066] 2. Image preprocessing: The original image taken by the camera is preprocessed through the FPGA chip, including denoising, color correction, etc., to improve image quality.

[0067] 3. Image data transmission: Every 15 minutes, the FPGA chip packages the pre-processed image data, i.e., the spatial image data, and transmits it to the designated terminal through the FTP system, ensuring the stability and real-time performance of data transmission.

[0068] 3. Terminal Analysis and Joint Processing

[0069] 1. Receiving image data: The analysis terminal receives the airspace image data transmitted from the FTP system and stores it in the local database. At the same time, the analysis terminal also receives the atmospheric electric field data collected by the atmospheric electric field instrument.

[0070] 2. Image recognition and analysis: Use computer vision algorithms to automatically identify and analyze received images to extract key meteorological features such as cloud cover, precipitation, and snow amount.

[0071] Image recognition and analysis specifically include the following:

[0072] 1) Preprocessing enhancement: Before the spatial domain image enters the recognition and analysis stage, a more sophisticated preprocessing step is performed, including more advanced denoising algorithms (such as non-local mean denoising or deep learning denoising models), as well as more precise color correction and contrast enhancement techniques, to further improve the quality of the spatial domain image.

[0073] 2) Deep learning models: Use deep convolutional neural networks (CNN) or more advanced models (such as ResNet, EfficientNet, etc.) to automatically extract image features. These models are trained with a large amount of image data and can more accurately identify complex meteorological features such as cloud types, precipitation types, and snow distribution.

[0074] 3) Transfer learning: If the existing dataset is not sufficient to train a completely new model, you can consider using transfer learning. By leveraging a model pre-trained on a large image dataset (such as ImageNet) and fine-tuning it to adapt to a specific meteorological image recognition task, you can improve the generalization ability of the model, accelerate the convergence of the model, and improve performance.

[0075] 4) Multimodal fusion: When the information features of a single-frame image are insufficient, multimodal fusion extraction can be selected to change the input of the deep learning model from a single-frame image to multiple-frame images in a time series, extract meteorological features from the perspective of the time series through the deep learning model, or extract image features such as color, texture or shape of the airspace image before input, combine with the above time series, extract the image features of the enhanced airspace image of several consecutive frames, use it as the input data of the deep learning model, and output the corresponding meteorological features. Combine multiple image features (such as color, texture, shape, etc.) and time series data to improve the recognition accuracy of meteorological features. For example, by analyzing the changes in cloud types in multiple consecutive frames of images, cloud cover and precipitation trends can be estimated more accurately.

[0076] 5) Semi-supervised learning: training with a small amount of labeled data and a large amount of unlabeled data. This method can pre-train the model by applying self-supervised learning on unlabeled data (such as predicting image rotation angle, color transformation, etc.), and then fine-tune it on limited labeled data.

[0077] 6) Post-processing optimization: Before the recognition results are output, post-processing algorithms are applied to optimize the recognition results. This may include removing isolated small areas, smoothing boundaries, or making corrections based on context information.

[0078] 7) Continuous learning and updating: As new meteorological image data is continuously collected, the deep learning model should be able to continuously learn and update to adapt to changes in meteorological conditions and improvements in recognition tasks. This can be achieved through incremental learning or online learning methods.

[0079] 8) User feedback integration: Combine user feedback and expert knowledge to iteratively improve the deep learning model. This can be achieved by collecting user opinions, error reports or suggestions on recognition results and integrating them into the model training process.

[0080] 9) Model compression and acceleration: Apply deep learning model compression technologies, such as pruning, quantization, knowledge distillation, etc., to reduce the computational complexity and storage requirements of the model and improve the real-time performance of the system; use model acceleration technologies, such as using GPUs or dedicated accelerators for inference calculations, to further improve the processing speed of the system.

[0081] 3. Joint analysis: Jointly analyze the meteorological features obtained by image recognition with the atmospheric electric field data. By comparing the image data and the atmospheric electric field data, the correlation and regularity between them can be found.

[0082] The joint analysis specifically includes the following:

[0083] 1) Data matching:

[0084] Match the image recognition data (such as cloud cover, precipitation, snow amount, etc.) at the same time point with the atmospheric electric field data to ensure the temporal consistency of the two.

[0085] 2) Feature comparison:

[0086] On the basis of historical data, statistical analysis or regression fitting methods are used to compare the changing trends of meteorological features extracted by image recognition and atmospheric electric field data to find the correlation between them; for example, if it is found that the atmospheric electric field intensity also increases accordingly when the cloud cover increases, then there may be some correlation between the two. When there is some correlation between the meteorological features and the atmospheric electric field data, subsequent pattern recognition is performed.

[0087] 3) Pattern recognition:

[0088] Based on historical data, statistical analysis and machine learning algorithms are used to identify the potential patterns or regularities between meteorological characteristics and atmospheric electric field data; statistical analysis is performed to determine whether there is a fitting relationship between meteorological characteristics and atmospheric electric field data. If there is a fitting relationship, it is saved as a sample and the machine learning algorithm is trained to identify different abnormal change patterns of electric field data according to different meteorological characteristics, or to identify meteorological characteristics under abnormal electric field, so as to determine the potential patterns or regularities between meteorological characteristics and atmospheric electric field data. This includes the electric field change pattern under different meteorological conditions, or the relationship between electric field anomalies and specific weather phenomena.

[0089] 4) Anomaly Detection:

[0090] Based on historical data and the potential patterns or laws in the above pattern recognition, a threshold is preset, and abnormal values ​​in the atmospheric electric field data are detected according to the preset threshold; combined with the meteorological features in the image data, it is determined whether the abnormal values ​​are caused by meteorological features, or abnormal values ​​may be caused by other circumstances other than meteorological features, and the meteorological conditions or phenomena that the abnormal values ​​may correspond to are analyzed, such as lightning activities, storms, etc. The joint analysis ultimately results in the analysis results of meteorological features, the association or pattern between meteorological features and atmospheric electric fields, and the abnormal data and causes of atmospheric electric fields.

[0091] Regarding the anomaly detection content, the present invention can introduce more advanced anomaly detection algorithms, such as autoencoders based on deep learning, first-class classifiers, etc., to detect anomalies in image data and atmospheric electric field data.

[0092] In the present invention, joint analysis is an innovative point of joint analysis of image data and atmospheric electric field data, which is closely integrated with other steps. The joint analysis is performed after receiving the image data and atmospheric electric field data. How to collect and transmit these data is the basis and input of the joint analysis; the result of the joint analysis will depend on the accuracy and completeness of these data. Image recognition and analysis provide key meteorological feature information for joint analysis; the joint analysis compares and analyzes these features with the atmospheric electric field data to find the relationship between them. The result of the joint analysis is an important basis for generating an analysis report; the meteorological feature analysis, abnormal cause judgment, etc. in the analysis report are all based on the results of the joint analysis. The results of the joint analysis are displayed to the user through an interactive analysis interface to help the user understand the relationship between meteorological conditions and the atmospheric electric field; user feedback can guide the further optimization and improvement of the joint analysis.

[0093] 4. Generate analysis report: Generate a detailed analysis report based on the results of the joint analysis. The report includes the analysis results of meteorological characteristics, the cause of the atmospheric electric field anomaly, and the corresponding recommended measures.

[0094] 5. Results display and feedback: The analysis report is displayed to the user through an interactive analysis interface. The user can provide feedback based on the report results, and the system will make corresponding adjustments and optimizations based on the feedback.

[0095] 4. System maintenance and management include the following:

[0096] (1) Regular inspection and maintenance: Regularly inspect and maintain the functions of the omnidirectional sky camera, data transmission module, and analysis terminal to ensure their normal operation.

[0097] (2) Data backup and recovery: The received image data and atmospheric electric field data are backed up regularly to prevent data loss. At the same time, a data recovery mechanism is established to ensure timely recovery when data is lost or damaged.

[0098] (3) System upgrade and optimization: The system is upgraded and optimized based on the latest meteorological data and image data, including improving the image recognition algorithm, i.e. the deep learning model, optimizing the data transmission module, etc., to improve the performance and accuracy of the system.

[0099] At the same time, in order to better improve the service performance of the system, the following technical solutions are proposed:

[0100] 1. Incremental learning and online updating:

[0101] Using incremental learning technology, the analysis terminal can update and improve its internal deep learning model in real time when it receives new meteorological data and image data, without the need to retrain the entire system.

[0102] 2. Data enhancement and expansion:

[0103] Apply data enhancement technology to increase the diversity of training data sets through operations such as rotation, translation, and scaling, thereby improving the generalization ability of deep learning models; use data expansion technology, combined with simulation and synthetic data, to increase the size of the data set and further improve the performance of deep learning models.

[0104] 3. Modular design and component replacement:

[0105] The system is decomposed into multiple independent modules, which are divided and encapsulated into different modules according to the above-mentioned models or data processing functions. Each module is responsible for different functions, such as data preprocessing, feature extraction, model training, etc.; when the performance of a module no longer meets the requirements, the module can be replaced individually without the need to reconstruct the entire system.

[0106] Through the above specific implementation methods, the tower-based all-round sky camera system can effectively realize real-time meteorological observation and data analysis at the atmospheric electric field observation site. The system provides strong support for meteorological research and forecasting through all-round shooting and precise joint analysis, helps to timely discover and predict abnormal meteorological events, and provides a scientific basis for relevant decision-making.

[0107] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. An omnidirectional sky camera system based on a tower base, characterized in that: include: An omnidirectional sky camera fixed on a tower base, and an analysis terminal connected to the omnidirectional sky camera via a data transmission module, wherein the omnidirectional sky camera is used to photograph the airspace of the meteorological station to obtain an airspace image, and the airspace image is regularly transmitted to the analysis terminal via the data transmission module, and the airspace image and atmospheric electric field data are received by the analysis terminal, and the airspace image and the atmospheric electric field data are jointly analyzed to generate a joint analysis result, wherein the joint analysis result includes a meteorological characteristic analysis result and atmospheric electric field abnormality data and causes; The omnidirectional sky camera includes a lens and an image sensor connected in sequence, wherein the image sensor is connected to an FPGA chip, and the FPGA chip performs preliminary denoising and enhancement processing on the original image collected by the image sensor from the lens to generate a spatial image; a shooting control unit is connected to the image sensor, and the shooting parameters of the camera are controlled by the shooting control unit, wherein the shooting parameters include resolution, frame rate and shooting range; The camera is fixed on a high tower, i.e., a tower base, through a fixing device, wherein the fixing device is fixed on the tower base through a steel hoop, and the fixing device is fixed on the steel hoop through a flat head screw and a square nut. The fixing device includes an outer frame, and a camera base bracket is arranged on the outer frame. The camera is fixed to a storage place in the center of the camera base bracket. The outer frame is provided with a left punch-free telescopic rod, a right punch-free telescopic rod, and a top punch-free telescopic rod to clamp and fix the camera, wherein the punched telescopic rod includes mutually fixed anti-skid pads, a base, and a stainless steel pipe in sequence from the camera to the outer frame, a wing screw is welded on the outer frame, a nut is arranged on the screw, and the stainless steel pipe is sleeved on the screw, and the stainless steel pipe, the base, and the anti-skid pad are pushed to tighten the camera by tightening the nut. At least one side of the outer frame is open, and the camera is used to shoot airspace by adjusting the angle of the fixing device and the camera; In the analysis terminal, the process of jointly analyzing the meteorological characteristics and the atmospheric electric field data includes: Obtain historical meteorological characteristics and historical atmospheric electric field data; Performing data matching on the historical meteorological characteristics and historical atmospheric electric field data; Conducting a correlation test on the historical meteorological characteristics and historical atmospheric electric field data on the trend of changes; Based on the test results, the historical meteorological characteristics and historical atmospheric electric field data are pattern recognized through statistical analysis and machine learning models to generate electric field change patterns and electric field anomaly data corresponding to the meteorological characteristics; Based on the electric field change pattern and the electric field abnormality data, combined with meteorological characteristics, anomaly detection is performed on the atmospheric electric field data to generate atmospheric electric field abnormality data and causes; The joint analysis specifically includes the following: 1) Data matching: Matching the image recognition data at the same time point with the atmospheric electric field data to ensure the temporal consistency of the two, wherein the image recognition data is cloud cover, precipitation or snow amount; 2) Feature comparison: Based on historical data, the meteorological features extracted by image recognition are compared with the changing trends of atmospheric electric field data through statistical analysis or regression fitting methods to find the correlation between them. It is found that when the cloud cover increases, the atmospheric electric field intensity also increases accordingly, so there may be some correlation between the two. When there is some correlation between the meteorological features and the atmospheric electric field data, subsequent pattern recognition is carried out. 3) Pattern recognition: Based on historical data, use statistical analysis and machine learning algorithms to identify potential patterns or laws between meteorological characteristics and atmospheric electric field data; statistically analyze whether there is a fitting relationship between meteorological characteristics and atmospheric electric field data. If there is a fitting relationship, save it as a sample, train the machine learning algorithm, and identify different abnormal change patterns of electric field data according to different meteorological characteristics, or identify meteorological characteristics under electric field anomalies, so as to determine the potential patterns or laws between meteorological characteristics and atmospheric electric field data; potential patterns or laws include electric field change patterns under different meteorological conditions, or the relationship between electric field anomalies and specific weather phenomena; 4) Anomaly Detection: Based on historical data and the potential patterns or laws existing in the above-mentioned pattern recognition, a threshold is preset, and based on the preset threshold, anomalies in the atmospheric electric field data are detected; combined with the meteorological characteristics in the image data, it is determined whether the anomalies are caused by meteorological characteristics, or whether the anomalies may be caused by other circumstances other than meteorological characteristics, and the meteorological conditions or phenomena corresponding to the anomalies are analyzed, where the phenomena are lightning activities or storms; the joint analysis ultimately results in the meteorological characteristic analysis results, the correlation or pattern between the meteorological characteristics and the atmospheric electric field, and the abnormal data and causes of the atmospheric electric field.

2. The system according to claim 1, characterized in that The data transmission module includes an FPGA chip and an FTP system, wherein the spatial domain image is compressed and packaged by the FPGA chip, and the packaged data is transmitted at fixed time intervals through a transmission channel established by the FTP system.

3. The system according to claim 1, characterized in that In the analysis terminal, the process of jointly analyzing the airspace image and the atmospheric electric field data includes: performing image recognition on the airspace image to generate meteorological features, and jointly analyzing the meteorological features with the atmospheric electric field data to generate atmospheric electric field anomalies and corresponding causes.

4. The system according to claim 1, characterized in that In the analysis terminal, the process of performing image recognition on the spatial domain image includes: The spatial domain image is preprocessed and enhanced, and the feature extraction of the enhanced spatial domain image is performed through a deep learning model to obtain meteorological features, wherein the deep learning model is trained and generated by transfer learning and semi-supervised learning, and the feature extraction of the enhanced spatial domain image is performed by multimodal fusion in the deep learning model, wherein the multimodal fusion is to set the input data of the deep learning model as the image features of the enhanced spatial domain image of several consecutive frames, fit the correlation between the input data and the meteorological features through the deep learning model, and output the meteorological features through the deep learning model; the deep learning model is also updated through continuous learning and iterative improvement.

5. The system according to claim 1, characterized in that In the analysis terminal, an analysis report is generated according to the joint analysis result, and the analysis report is displayed through an interactive analysis interface.

6. The system according to claim 1, characterized in that In the analysis terminal, the airspace image and atmospheric electric field data are backed up, and the functions in the analysis terminal are subsequently adjusted and optimized.

Images