Low-bandwidth remote monitoring method for monitoring meteorological service system and related device

Abstract

The invention relates to the field of data processing, and particularly provides a low-bandwidth remote monitoring method for a meteorological service supervision system and a related device, and the method comprises the steps: judging whether a change region of a meteorological image to be detected is larger than a change threshold value or not compared with a change region of a previous meteorological image after the meteorological image to be detected is obtained; or detecting whether the meteorological image to be detected contains an abnormal detection result; when the change area of the meteorological image to be detected is greater than a change threshold compared with the change area of the previous meteorological image, coding the change area, and sending a coding result of the change area to a receiving end; or, when the meteorological image to be detected contains an abnormal detection result, encoding the abnormal detection result, and sending an abnormal encoding result to the receiving end, so that the receiving end synthesizes with the previous meteorological image based on the change region encoding result or the abnormal encoding result to obtain a synthesized meteorological image. According to the invention, low-bandwidth, low-overhead and efficient remote monitoring and protection of the meteorological service system can be completed.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a low-bandwidth remote monitoring method for a regulatory meteorological business system and a related device. Background Art

[0002] In the operation and management of meteorological business systems, remote monitoring plays a vital role. It can obtain meteorological information in real time, detect meteorological anomalies in time, and provide key support for meteorological disaster warnings and meteorological resource management. However, there are many shortcomings in the existing remote monitoring methods of meteorological business systems. On the one hand, in terms of data transmission, due to the large amount of meteorological image data, the traditional transmission method requires a large amount of bandwidth resources, which will lead to slow data transmission, high latency and even data loss in low-bandwidth environments, seriously affecting the real-time and reliability of remote monitoring. On the other hand, in terms of abnormal meteorological detection, the existing detection methods often lack pertinence and efficiency, and it is difficult to accurately focus on specific meteorological anomaly types for detection. The accuracy and reliability of the detection results need to be improved, and they cannot meet the needs of meteorological business systems for timely and accurate detection of abnormal meteorology. In addition, some monitoring systems lack effective means for resource management and traffic control, and cannot achieve efficient and stable remote monitoring under limited bandwidth conditions, resulting in slow system response speed and poor stability, which in turn affects the normal operation of meteorological business systems and the timeliness and effectiveness of meteorological disaster warnings. Therefore, there is an urgent need for a remote monitoring method for meteorological business systems that can achieve efficient remote monitoring in a low-bandwidth environment, accurately detect abnormal meteorological conditions, and effectively manage network resources. Summary of the invention

[0003] In view of this, the present invention provides a low-bandwidth remote monitoring method and related device for a regulatory meteorological business system. The technical solution of the present invention is implemented as follows:

[0004] On the one hand, the present invention provides a low-bandwidth remote monitoring method for a regulatory meteorological business system, which is applied to a remote monitoring device, and the remote monitoring device is communicatively connected to a receiving end, and the method includes: acquiring a meteorological image to be detected; judging whether a change area of ​​the meteorological image to be detected compared with a previous meteorological image is greater than a change threshold; or, detecting whether the meteorological image to be detected contains an abnormal detection result; when a change area of ​​the meteorological image to be detected compared with a previous meteorological image is greater than a change threshold, encoding the change area to obtain a change area encoding result, and sending the change area encoding result to a receiving end, so that the receiving end synthesizes the change area encoding result and the previous meteorological image to obtain a synthetic meteorological image; or, when the meteorological image to be detected contains an abnormal detection result, encoding the abnormal detection result to obtain an abnormal encoding result, and sending the abnormal encoding result to the receiving end, so that the receiving end synthesizes the abnormal encoding result and the previous meteorological image to obtain a synthetic meteorological image.

[0005] On the other hand, the present invention provides a remote monitoring device, including a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and the processor implements the steps in the above method when executing the program.

[0006] The present invention can complete a low-bandwidth, low-overhead, and efficient remote monitoring and protection meteorological service system, and only transmits the changed area or abnormal meteorological data, which significantly reduces the bandwidth occupation. At the same time, the data transmission rate is effectively controlled by combining the token bucket and the leaky bucket algorithm, ensuring efficient and stable remote monitoring under limited bandwidth, and ensuring the integrity and reliability of the data. Efficient remote monitoring under low bandwidth is achieved, which is suitable for remote operation and management of various meteorological service systems. In the training process of the abnormal meteorological detection network, when detecting abnormal meteorological weather, the target detection branch can guide the abnormal detection in the meteorological training image, and the abnormal meteorological detection of the meteorological training image can focus on the desired detection direction, so as to promote the detection of abnormal information in the meteorological training image. Then, the target priori mark is determined in at least one priori mark matched by the meteorological training image, and the detection branch corresponding to the target priori mark is the same as the target detection branch. Based on the abnormal meteorological detection result of the meteorological training image under the target detection branch and the abnormal meteorological detection result under the target detection branch, the meteorological image detection network can be trained so that the meteorological image detection network has the ability to detect abnormal weather under the target detection branch. The debugged meteorological image detection network is used to perform abnormal meteorological detection of meteorological images under the corresponding detection branch. Compared with the network without target detection branches, the anomaly detection effect of the debugged meteorological image detection network is better. In addition, based on at least one prior marker matched by the meteorological training image, the meteorological image detection network can be debugged based on multiple target detection branches, so that the meteorological image detection network can have the ability to detect multiple abnormal meteorological conditions in one debugging, saving the computing power dependence of debugging. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Figure 1 A schematic diagram of the implementation flow of a low-bandwidth remote monitoring method for a regulatory meteorological business system provided in an embodiment of the present invention.

[0008] Figure 2 A hardware entity schematic diagram of a remote monitoring device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0009] The embodiment of the present invention provides a low-bandwidth remote monitoring method for a regulatory meteorological business system, which can be executed by a processor of a remote monitoring device. The remote monitoring device is communicatively connected to at least one receiving end. The remote monitoring device can refer to a device with data processing capabilities such as a server, a laptop, a tablet computer, a desktop computer, a mobile device, etc. The receiving end can also refer to a device with data processing capabilities such as a server, a laptop, a tablet computer, a desktop computer, a mobile device, etc.

[0010] Figure 1A schematic diagram of the implementation flow of a low-bandwidth remote monitoring method for a regulatory meteorological business system provided by an embodiment of the present invention is shown in FIG. Figure 1 As shown, the method includes: Step S100: Acquire a meteorological image to be detected.

[0011] Step S200: determining whether a change area of ​​a meteorological image to be detected compared with a previous meteorological image is greater than a change threshold; or, detecting whether the meteorological image to be detected contains an abnormal detection result.

[0012] In step S100, the remote monitoring device obtains the meteorological image to be detected. The meteorological image to be detected refers to the meteorological related image that needs to be detected and analyzed at present. These images contain various meteorological information, such as cloud distribution, precipitation area, temperature change, etc., which are very important for the monitoring of the meteorological business system. The remote monitoring device can obtain the meteorological image to be detected in a variety of ways, such as directly obtaining real-time captured or monitored image data from meteorological monitoring equipment such as meteorological satellites and meteorological radars.

[0013] In step S200, the remote monitoring device determines whether the change area of ​​the meteorological image to be detected compared with the previous meteorological image is greater than the change threshold, or detects whether the meteorological image to be detected contains abnormal detection results. The previous meteorological image here refers to the meteorological image acquired at the adjacent time point before the meteorological image to be detected is acquired. By comparing the difference between the two images, it can be determined whether the meteorological conditions have changed significantly. The change threshold is a pre-set standard value used to measure the degree of image change. For example, when the change area exceeds a certain proportion of the total area of ​​the image (such as 10%), it is considered that the change area is greater than the change threshold.

[0014] In order to determine whether the change area of ​​the meteorological image to be detected compared with the previous meteorological image is greater than the change threshold, the remote monitoring device can use image matching and difference analysis methods. Specifically, the previous meteorological image and the meteorological image to be detected can be compared at the pixel level, the difference value of each pixel point can be calculated, and then the number of pixels whose difference values ​​exceed a certain threshold can be counted to determine the size of the change area. For example, the mean square error (MSE) method can be used to calculate the difference between two images, and the formula is: ; where m and n are the number of rows and columns of the image, respectively. 1(i,j) and I 2(i,j) are the pixel values ​​of the previous meteorological image and the meteorological image to be detected at the (i, j)th pixel point. The larger the MSE value calculated, the greater the difference between the two images. When the MSE value exceeds the preset change threshold, it can be judged that the change area of ​​the meteorological image to be detected compared with the previous meteorological image is greater than the change threshold.

[0015] In addition to determining the changed area of ​​the image, the remote monitoring device also needs to detect whether the meteorological image to be detected contains abnormal detection results. Abnormal detection results refer to situations that do not conform to normal meteorological laws or characteristics in meteorological images, such as sudden heavy rainfall, abnormal high temperature areas, etc. In order to detect whether the meteorological image to be detected contains abnormal detection results, the remote monitoring device can use machine learning or deep learning methods to analyze and identify the image using a trained meteorological image detection network. The meteorological image detection network can be debugged and optimized through a large number of meteorological training images to improve its detection ability for abnormal meteorological conditions. During the detection process, the remote monitoring device first obtains the meteorological image to be detected and branch prompt information, and the branch prompt information is used to indicate the detection branch of the meteorological image to be detected, and the number of branches corresponding to the detection branch. For example, the detection branch can include different meteorological anomaly types such as heavy rainfall detection, high temperature detection, and typhoon detection, and each detection branch corresponds to a specific detection task. The remote monitoring device obtains a branch meteorological image that matches the detection branch of the corresponding number of branches from at least one candidate meteorological image based on the branch prompt information, and generates a target meteorological image based on the branch meteorological image and the meteorological image to be detected. Then, through the debugged meteorological image detection network, according to the detection branch indicated by the branch prompt information, the target meteorological image is subjected to abnormal meteorological detection, and the abnormal detection result of the target meteorological image under the detection branch is obtained. Taking heavy precipitation detection as an example, the remote monitoring device obtains the prompt information of the meteorological image to be detected and the heavy precipitation detection branch, obtains the branch meteorological image related to heavy precipitation detection from the candidate meteorological image, and synthesizes it with the meteorological image to be detected into the target meteorological image. Then, the target meteorological image is analyzed using the trained meteorological image detection network to determine whether there is a heavy precipitation area in the image. If a heavy precipitation area is detected, it is considered that the meteorological image to be detected contains an abnormal detection result.

[0016] Step S300: When the change area of ​​the meteorological image to be detected compared with the previous meteorological image is greater than the change threshold, the change area is encoded to obtain the change area encoding result, and the change area encoding result is sent to the receiving end so that the receiving end can synthesize it based on the change area encoding result and the previous meteorological image to obtain a synthesized meteorological image.

[0017] In the remote monitoring method of the low-bandwidth regulatory meteorological business system, step S300 is a key operation performed by the remote monitoring device when the change area of ​​the meteorological image to be detected compared with the previous meteorological image is greater than the change threshold, which is intended to efficiently transmit the change area information to reduce bandwidth occupancy and ensure that the receiving end can synthesize an accurate meteorological image. When the remote monitoring device determines that the change area of ​​the meteorological image to be detected compared with the previous meteorological image is greater than the change threshold, it means that the meteorological conditions have changed significantly, and at this time, these change information needs to be communicated to the receiving end in a timely and accurate manner.

[0018] The change area refers to the part of the meteorological image to be detected that is different from the previous meteorological image. These differences may reflect the dynamic changes of meteorological elements, such as the movement of clouds, the expansion of precipitation areas, etc. Coding is the process of converting the information of the change area to make it more suitable for transmission in a low-bandwidth environment. The remote monitoring device encodes the change area to obtain the encoding result of the change area. This encoding process can adopt a variety of technical means, such as lossless compression coding and lossy compression coding. Lossless compression coding can reduce the amount of data without losing image information. The feasible lossless compression algorithms include Huffman coding, run-length coding, etc.

[0019] After obtaining the coding result of the changed area, the remote monitoring device sends it to the receiving end. After receiving the coding result of the changed area, the receiving end needs to synthesize it based on the result and the previous meteorological image to obtain a synthetic meteorological image. The synthesis process is to accurately integrate the change information represented by the coding result of the changed area into the previous meteorological image to reflect the current meteorological conditions. For example, suppose that a certain area in the previous meteorological image is displayed as sunny, and clouds appear in the area in the meteorological image to be detected. After the remote monitoring device detects this changed area, it encodes it and sends it to the receiving end. The receiving end adds the cloud information represented by the coding result to the corresponding position of the previous meteorological image, thereby obtaining a synthetic meteorological image containing the latest meteorological information.

[0020] When transmitting the coding results of the change area, in order to ensure stable and efficient data transmission in a low-bandwidth environment, the remote monitoring device uses a token bucket algorithm or a leaky bucket algorithm to control the bandwidth. The token bucket algorithm allows a certain amount of burst traffic during communication, but it will achieve the purpose of controlling the average bandwidth by controlling the long-term average rate. When the traffic exceeds the limit, there are no tokens in the bucket, and the data packet will be discarded or delayed. For example, the token bucket adds tokens to the bucket at a fixed rate, such as 100 tokens per second, and each data packet consumes 1 token. When there is data to be sent, the remote monitoring device checks whether there are enough tokens in the token bucket. If so, the data can be sent immediately and consume the corresponding tokens; if the bucket is full, the extra tokens will be discarded; if the bucket is empty, the data will need to wait until a token is available. In this way, the average bandwidth can be controlled while ensuring a certain burst traffic processing capacity. The leaky bucket algorithm strictly controls the bandwidth according to the specified traffic, and any traffic exceeding the bucket capacity will be immediately discarded. Data flows into the bucket at an irregular rate, and data flows out of the bucket at a constant rate, such as 1000 bytes per second. When the data inflow speed exceeds the output rate of the leaky bucket, the data will accumulate in the bucket. The bucket capacity is limited. If the maximum capacity is reached, new data will be discarded to prevent network congestion, thereby achieving the goal of accurately controlling bandwidth.

[0021] In order to ensure the accuracy and completeness of the changed area coding results, the remote monitoring device can also use error detection and correction technology during the coding and transmission process. For example, a parity bit or a cyclic redundancy check (CRC) code is added to the coding result. After receiving the coding result, the receiving end can verify the accuracy of the data by calculating the check bit or CRC code. If an error is found in the data, the remote monitoring device can be required to resend the data. In addition, the remote monitoring device can also use a retransmission mechanism to resend the data packets lost or damaged during the transmission process to ensure that the receiving end can receive the changed area coding results completely.

[0022] Step S400: When the meteorological image to be detected contains an abnormal detection result, the abnormal detection result is encoded to obtain an abnormal encoding result, and the abnormal encoding result is sent to the receiving end so that the receiving end synthesizes it based on the abnormal encoding result and the previous meteorological image to obtain a synthetic meteorological image.

[0023] The remote monitoring device encodes the abnormal detection result to obtain an abnormal coding result. The encoding process is to convert the abnormal detection result into a data form that is more suitable for transmission in a low-bandwidth network environment. For example, a variety of coding techniques can be used, such as run-length coding, arithmetic coding, etc. Run-length coding is a simple and effective lossless coding method that represents continuously repeated data with a count value and the value of the data.

[0024] After obtaining the abnormal coding result, the remote monitoring device sends it to the receiving end. After receiving the abnormal coding result, the receiving end will perform a synthesis operation in combination with the previous meteorological image to obtain a synthetic meteorological image. The synthesis process is to accurately add the abnormal information represented by the abnormal coding result to the corresponding position of the previous meteorological image to reflect the actual current meteorological conditions. For example, the previous meteorological image shows that the weather in a certain area is normal, and the meteorological image to be detected detects the abnormal situation of rainstorm clouds in the area. The remote monitoring device encodes the abnormal information and sends it to the receiving end. The receiving end adds the rainstorm cloud information represented by the code to the corresponding area of ​​the previous meteorological image to obtain a synthetic meteorological image containing the latest abnormal information.

[0025] As an implementation manner, in step S200, determining whether the change area of ​​the meteorological image to be detected compared with the previous meteorological image is greater than the change threshold includes: Step S201: performing image segmentation on the previous meteorological image and the meteorological image to be detected respectively, and obtaining image segmentation sets corresponding to the previous meteorological image and the meteorological image to be detected respectively, wherein the image segmentation set of the previous meteorological image and the image segments in the image segmentation set of the meteorological image to be detected correspond to each other one by one; Step S202: respectively calculating the differences between two one-to-one corresponding image blocks (such as calculating the MSE value) to obtain a plurality of block difference results; Step S203: averaging the difference results of the multiple blocks to obtain a mean difference result. If the mean difference result is greater than the change threshold, it is determined that the change area of ​​the meteorological image to be detected compared with the previous meteorological image is greater than the change threshold; Step S204: when sending the changed region encoding result to the receiving end, or sending the abnormal encoding result to the receiving end, a token bucket algorithm or a leaky bucket algorithm is used to control the bandwidth.

[0026] In step S201, the remote monitoring device performs image segmentation on the previous meteorological image and the meteorological image to be detected, respectively, to obtain the image segmentation sets corresponding to the previous meteorological image and the meteorological image to be detected, and the image segments in the image segmentation set of the previous meteorological image and the image segmentation set of the meteorological image to be detected correspond one to one. Image segmentation is a technical means of dividing an image into multiple small areas, and its purpose is to simplify the image processing and analysis process. In the processing of meteorological images, since meteorological images are usually large and contain rich information, directly comparing the entire image will consume a lot of computing resources and time. After the image is segmented, each small block can be processed and compared independently, thereby improving the processing efficiency. For example, for a meteorological image with a resolution of 1024×1024 pixels, the remote monitoring device can divide it into small blocks of 64×64 pixels, so that the entire image is divided into (1024÷64)×(1024÷64)=256 small blocks. During the segmentation process, the previous meteorological image and the meteorological image to be detected will be segmented according to the same rules to ensure that the image blocks in their image block sets correspond one to one, which is convenient for subsequent difference calculation.

[0027] Step S202 requires the remote monitoring device to calculate the difference between the two corresponding image blocks respectively, and obtain multiple block difference results. The difference calculation here can be performed in a variety of ways, and a feasible method is to calculate the mean square error (MSE) value. To illustrate with a simple example, suppose there are two 3×3 pixel image blocks, and the image block pixel values ​​of the previous meteorological image are [10, 20, 30; 40, 50, 60; 70, 80, 90], and the corresponding image block pixel values ​​of the meteorological image to be detected are [12, 22, 32; 42, 52, 62; 72, 82, 92]. According to the mean square error formula, first calculate the square of the difference of each pixel point, such as the first pixel point (12-10) 2=4, then add the squares of the differences of all pixels and divide by the total number of pixels 3×3=9 to get the mean square error value. The larger the mean square error value, the greater the difference between the two image blocks. In addition to the mean square error, other difference calculation methods can also be used, such as peak signal-to-noise ratio (PSNR). The calculation formula of PSNR is: ; Among them, MAX I is the maximum possible value of the image pixel value. For an 8-bit image, MAX I = 255. The larger the PSNR value, the more similar the two image blocks are and the smaller the difference is.

[0028] Step S203 is that the remote monitoring device averages the difference results of multiple blocks to obtain a mean difference result, and determines whether the mean difference result is greater than the change threshold. The change threshold is a pre-set standard value used to measure the degree of overall change of the meteorological image. By averaging the difference results of all blocks, an indicator that can reflect the change of the entire image can be obtained. For example, assuming that 100 block difference results are obtained after a meteorological image is divided into blocks, these 100 results are added and divided by 100 to obtain the mean difference result. If this mean difference result is greater than the pre-set change threshold, such as the change threshold is set to 10, and the mean difference result is 15, then the remote monitoring device determines that the change area of ​​the meteorological image to be detected compared to the previous meteorological image is greater than the change threshold. This means that the meteorological image has undergone a more significant change as a whole, which may indicate a change in meteorological conditions, such as the movement of clouds, the expansion of precipitation areas, etc.

[0029] In step S204, when sending the changed area encoding result to the receiving end, or sending the abnormal encoding result to the receiving end, the remote monitoring device uses a token bucket algorithm or a leaky bucket algorithm to control the bandwidth.

[0030] As an implementation manner, in step S200, detecting whether the meteorological image to be detected contains an abnormal detection result includes: Step S210: obtaining a meteorological image to be detected and branch prompt information, where the branch prompt information is used to indicate a detection branch of the meteorological image to be detected and the number of branches corresponding to the detection branch; Step S220: acquiring a branch meteorological image matching a detection branch of a corresponding number of branches from at least one candidate meteorological image according to the branch prompt information, and generating a target meteorological image according to the branch meteorological image and the meteorological image to be detected; Step S230: Through the debugged meteorological image detection network, according to the detection branch indicated by the branch prompt information, abnormal meteorological detection is performed on the target meteorological image to obtain the abnormal detection result of the target meteorological image under the detection branch.

[0031] In step S210, the remote monitoring device obtains the meteorological image to be detected and branch prompt information, and the branch prompt information is used to indicate the detection branch of the meteorological image to be detected, and the number of branches corresponding to the detection branch. Detection branches refer to different categories or directions for abnormality detection of meteorological images, and each detection branch corresponds to a specific abnormal meteorological condition or feature. For example, detection branches may include heavy precipitation detection, high temperature detection, typhoon detection, etc. The number of branches indicates the number of branches that currently need to be detected. Assuming that the remote monitoring device is to detect the meteorological image of a certain area, the branch prompt information indicates that the detection branches are heavy precipitation detection and high temperature detection, and the number of branches is 2. The remote monitoring device can obtain the meteorological image to be detected and the branch prompt information by establishing a connection with a meteorological database or other data source, or it can read relevant data from local storage.

[0032] In step S220, the remote monitoring device obtains a branch meteorological image that matches the detection branch of the corresponding number of branches from at least one candidate meteorological image according to the branch prompt information, and generates a target meteorological image according to the branch meteorological image and the meteorological image to be detected. The candidate meteorological image is a series of meteorological images stored in advance, and these images contain feature information under various meteorological conditions. The remote monitoring device selects branch meteorological images related to these detection branches from the candidate meteorological images according to the detection branches in the branch prompt information. For example, in the above-mentioned example of heavy precipitation detection and high temperature detection, the remote monitoring device finds images related to heavy precipitation and high temperature from the candidate meteorological images as branch meteorological images. Then, these branch meteorological images are combined with the meteorological image to be detected to generate a target meteorological image. There are many ways to combine, such as superimposing the branch meteorological image as reference information on the meteorological image to be detected, or performing feature fusion on them to enhance the feature information related to the detection branch in the target meteorological image.

[0033] In step S230, the remote monitoring device performs abnormal meteorological detection on the target meteorological image according to the detection branch indicated by the branch prompt information through the debugged meteorological image detection network, and obtains the abnormal detection result of the target meteorological image under the detection branch. The meteorological image detection network is debugged after a large number of meteorological training images, and it can accurately identify abnormal conditions in meteorological images. During the detection process, the detection branch plays a guiding role, allowing the meteorological image detection network to focus on specific abnormal meteorological features for detection. For example, for the heavy precipitation detection branch, the meteorological image detection network will focus on precipitation-related features in the target meteorological image, such as cloud density, color, etc., to determine whether there is an abnormal situation of heavy precipitation.

[0034] In order to implement steps S210-S230, the remote monitoring device adopts a series of technical means. When obtaining the meteorological image to be detected and the branch prompt information, the data interface technology can be used to connect with the meteorological monitoring equipment or database to ensure the accurate transmission of data. For screening branch meteorological images from candidate meteorological images, an image classification algorithm can be used to classify the candidate meteorological images according to the characteristics of the detection branch to find the matching branch meteorological image. When generating the target meteorological image, an image fusion algorithm can be used to effectively fuse the branch meteorological image and the meteorological image to be detected. In terms of abnormal meteorological detection, the meteorological image detection network can use a deep learning algorithm, such as a convolutional neural network (CNN), which can automatically extract feature information from the image and classify and judge abnormal situations.

[0035] As an implementation mode, step S230, through the debugged meteorological image detection network, according to the detection branch indicated by the branch prompt information, performs abnormal meteorological detection on the target meteorological image, and obtains the abnormal detection result of the target meteorological image under the detection branch, including: Step S231: the meteorological image detection network that has been debugged performs abnormal meteorological detection on the target meteorological image according to the detection branch indicated by the branch prompt information to obtain a first abnormal detection result; wherein the first abnormal detection result includes the element classification corresponding to the maximum support coefficient of each image element in the target meteorological image, and any support coefficient is used to indicate the probability that the element classification of any image element belongs to the corresponding element classification; Step S232: Based on the element classification corresponding to the maximum support coefficient of each image element in the target meteorological image, the abnormality detection result of the to-be-detected meteorological image in the target meteorological image under the detection branch is detected.

[0036] In step S231, the remote monitoring device performs abnormal meteorological detection on the target meteorological image according to the detection branch indicated by the branch prompt information through the debugged meteorological image detection network, and obtains the first abnormal detection result, which includes the element classification corresponding to the maximum support coefficient of each image element in the target meteorological image, wherein any support coefficient is used to indicate the probability that the element classification of any image element belongs to the corresponding element classification. The image element here can be a pixel point, a pixel block or an area with specific meteorological characteristics in the target meteorological image. The element classification is a category predefined according to the meteorological characteristics. For example, under the heavy precipitation detection branch, the element classification can include heavy precipitation area, weak precipitation area, no precipitation area, etc. The support coefficient is a value between 0 and 1, which reflects the possibility of a certain image element belonging to a specific element classification. For example, for a pixel point in the target meteorological image, its support coefficient for belonging to the heavy precipitation area is 0.8, the support coefficient for belonging to the weak precipitation area is 0.1, and the support coefficient for belonging to the no precipitation area is 0.1, then the maximum support coefficient of the pixel point is 0.8, and the corresponding element classification is the heavy precipitation area.

[0037] In order to achieve this detection process, the meteorological image detection network extracts and analyzes the features of the target meteorological image. In actual operation, the meteorological image detection network can use deep learning models such as convolutional neural networks (CNN). CNN can automatically learn the feature patterns in the image. For the target meteorological image, it will extract the feature information of the image through structures such as convolutional layers and pooling layers. Then, in the fully connected layer, the network will classify and predict the extracted features according to the detection branch indicated by the branch prompt information, and calculate the support coefficient of each image element belonging to different element classifications. For example, in a trained CNN model, after the target meteorological image is input, the output layer of the model will output the support coefficient of each image element for each element classification.

[0038] In step S232, the remote monitoring device detects the abnormality detection result of the target meteorological image under the detection branch based on the element classification corresponding to the maximum support coefficient of each image element in the target meteorological image. This step is to comprehensively analyze and judge the first abnormality detection result obtained in step S231. The remote monitoring device determines whether there is an area that meets the abnormal situation in the target meteorological image based on the element classification corresponding to the maximum support coefficient of each image element. For example, under the heavy precipitation detection branch, if there are a large number of image elements in the target meteorological image whose element classification corresponding to the maximum support coefficient is a heavy precipitation area, and the distribution and scale of these areas meet the pre-set abnormality standard, then the remote monitoring device will determine that the target meteorological image has an abnormality under the heavy precipitation detection branch.

[0039] Taking the typhoon detection branch as an example, in step S231, the meteorological image detection network detects the target meteorological image and calculates the support coefficient of each image element for the element classifications such as the typhoon eye, typhoon cloud system, and peripheral airflow. Assuming that the support coefficient of a certain image element for the typhoon eye is 0.9, and the support coefficients for other element classifications are relatively small, then the element classification corresponding to the maximum support coefficient of the image element is the typhoon eye. In step S232, the remote monitoring device counts the distribution and number of image elements belonging to the element classifications such as the typhoon eye and typhoon cloud system in the target meteorological image. If it is found that there is an obvious typhoon eye area, and there are a large number of image elements belonging to the typhoon cloud system around it, forming a typical typhoon structure, and the scale and intensity of this structure meet the definition criteria of the typhoon, then the remote monitoring device will determine that the meteorological image to be detected is abnormal under the typhoon detection branch, that is, signs of a typhoon are detected.

[0040] Steps S231-S232 perform abnormal meteorological detection on the target meteorological image through the meteorological image detection network, obtain the element classification corresponding to the maximum support coefficient of each image element, and then perform comprehensive analysis and judgment based on these classifications. The remote monitoring device can accurately detect the abnormal situation of the meteorological image to be detected under the detection branch.

[0041] As another implementation, step S230, through the debugged meteorological image detection network, according to the detection branch indicated by the branch prompt information, abnormal meteorological detection is performed on the target meteorological image to obtain the abnormal detection result of the target meteorological image under the detection branch, including: Step S2301: The meteorological image detection network that has been debugged performs abnormal meteorological detection on the target meteorological image according to the target detection branch to obtain a second abnormal detection result; the second abnormal detection result includes the element classification included in the element classification trajectory corresponding to the maximum trajectory support coefficient; wherein any element classification trajectory includes the support coefficient of the element classification of each image element of the target meteorological image under the corresponding element classification trajectory, and the conversion support coefficient of the element classification corresponding to the adjacent image elements; the trajectory support coefficient of any element classification trajectory is determined based on the support coefficient and the conversion support coefficient corresponding to any element classification trajectory; Step S2302: Based on the element classification included in the element classification trajectory corresponding to the maximum trajectory support coefficient, an abnormality detection result of the to-be-detected meteorological image in the target meteorological image under the detection branch is detected.

[0042] In step S2301, the remote monitoring device performs abnormal meteorological detection on the target meteorological image according to the target detection branch through the debugged meteorological image detection network, and obtains a second abnormal detection result, which includes the element classification included in the element classification trajectory corresponding to the maximum trajectory support coefficient. The element classification trajectory here refers to a sequence composed of the support coefficient of the element classification of each image element of the meteorological training image under the corresponding element classification trajectory, and the conversion support coefficient of the element classification corresponding to the adjacent image elements. The trajectory support coefficient of any element classification trajectory is determined based on the support coefficient and conversion support coefficient corresponding to the element classification trajectory.

[0043] Specifically, element classification is the division of different characteristic areas in meteorological images. For example, in the heavy precipitation detection branch, element classification can include heavy precipitation area, weak precipitation area, no precipitation area, etc. The support coefficient reflects the possibility that an image element belongs to a specific element classification. The value range is between 0 and 1. The larger the value, the higher the possibility that the image element belongs to the classification. The conversion support coefficient describes the probability of adjacent image elements converting from one element classification to another element classification.

[0044] For a more intuitive understanding, assume a simple meteorological image consisting of three adjacent image elements. Under the heavy precipitation detection branch, there may be multiple element classification trajectories. For example, one trajectory is: the first image element belongs to the heavy precipitation area, the second image element belongs to the weak precipitation area, and the third image element belongs to the no precipitation area. For this trajectory, each image element has a corresponding support coefficient belonging to its own classification, and there is also a conversion support coefficient between adjacent image elements. For example, the support coefficient of the first image element belonging to the heavy precipitation area is 0.8, the conversion support coefficient from the heavy precipitation area to the weak precipitation area to which the second image element belongs is 0.6, the support coefficient of the second image element belonging to the weak precipitation area is 0.7, the conversion support coefficient from the weak precipitation area to the no precipitation area to which the third image element belongs is 0.5, and the support coefficient of the third image element belonging to the no precipitation area is 0.9. The trajectory support coefficient of the element classification trajectory can be obtained by a certain calculation method (such as multiplying each support coefficient and the conversion support coefficient).

[0045] In actual operation, the meteorological image detection network conducts a comprehensive analysis of the target meteorological image and calculates all possible element classification trajectories and their corresponding trajectory support coefficients. This can be achieved with the help of deep learning models such as recurrent neural networks (RNN) or their variants, long short-term memory networks (LSTM). These models are able to process sequence data and are suitable for analyzing the relationship between adjacent image elements. The model will classify and predict each image element based on the input target meteorological image and the information of the target detection branch, and calculate the support coefficient and conversion support coefficient, thereby obtaining the trajectory support coefficients of different element classification trajectories.

[0046] Step S2302 requires the remote monitoring device to detect the abnormality detection result of the target meteorological image under the detection branch based on the element classification included in the element classification trajectory corresponding to the maximum trajectory support coefficient. That is to say, among all the calculated element classification trajectories, find the trajectory with the largest trajectory support coefficient, and the element classification contained in the trajectory is an important basis for judging the abnormality.

[0047] For example, in the above-mentioned example of heavy rainfall detection, multiple element classification trajectories and their trajectory support coefficients were obtained through calculation. The element classification trajectory corresponding to the maximum trajectory support coefficient shows that there is a large area of ​​heavy rainfall in the image and the distribution conforms to the characteristic pattern of heavy rainfall anomalies. Then the remote monitoring device can determine that there is an anomaly in the meteorological image to be detected under the heavy rainfall detection branch.

[0048] In order to accurately judge anomalies, the remote monitoring device can combine meteorological expertise and historical data to pre-set some judgment rules. For example, for heavy precipitation anomalies, it is stipulated that when the area of ​​the heavy precipitation area reaches a certain proportion (such as 15%) and the distribution of the heavy precipitation area shows a specific concentration, it is judged as an anomaly. At the same time, other meteorological factors such as air pressure and humidity can also be considered to further verify the reliability of the anomaly detection results.

[0049] In practical applications, in order to improve the accuracy and efficiency of anomaly detection, remote monitoring devices can preprocess meteorological images, such as removing noise and enhancing contrast, to improve image quality and facilitate the meteorological image detection network to better extract features.

[0050] As an implementation method, the meteorological image detection network is debugged by the following steps: Step S10: obtaining a meteorological training image and a target detection branch of the meteorological training image; the meteorological training image is matched with at least one priori mark, and any priori mark is used to indicate an abnormal meteorological detection result of the meteorological training image under a corresponding detection branch; Step S20: performing abnormal meteorological detection on the meteorological training image according to the target detection branch to obtain an abnormal meteorological detection result of the meteorological training image under the target detection branch; Step S30: determining a target priori marker from at least one priori marker matched by the meteorological training image, and determining an abnormal meteorological detection result of the meteorological training image under the target detection branch from the target priori marker, wherein the target priori marker is a priori marker that corresponds to the detection branch and the target detection branch. Step S40: Based on the abnormal meteorological detection results of the meteorological training image under the target detection branch and the abnormal meteorological detection results under the target detection branch, the meteorological image detection network is debugged; the debugged meteorological image detection network is used to perform abnormal meteorological detection of meteorological images under the corresponding detection branch.

[0051] In step S10, the remote monitoring device obtains a meteorological training image and a target detection branch of the meteorological training image. The meteorological training image is matched with at least one priori mark, and any priori mark is used to indicate the abnormal meteorological detection result of the meteorological training image under the corresponding detection branch. The meteorological training image is the basic data for training the meteorological image detection network, which can come from a large number of meteorological images collected by equipment such as meteorological satellites and meteorological radars. The target detection branch refers to a specific type of meteorological anomaly that the meteorological image detection network is expected to focus on and detect, such as heavy precipitation, high temperature, typhoon, etc. The priori mark is the abnormal situation of the meteorological training image under different detection branches that is pre-annotated or determined by other reliable methods. It provides a correct reference standard for the training of the meteorological image detection network. For example, for a meteorological training image, its target detection branch is heavy precipitation detection, and the priori mark clearly indicates which areas in the image have heavy precipitation anomalies. The remote monitoring device can obtain meteorological training images and target detection branch information from data sources such as meteorological data centers and historical meteorological databases, and these data sources will also provide corresponding priori marks.

[0052] In step S20, the remote monitoring device performs abnormal meteorological detection on the meteorological training image according to the target detection branch, and obtains the abnormal meteorological detection result of the meteorological training image under the target detection branch. This step is to let the meteorological image detection network perform preliminary analysis and judgment on the meteorological training image. The remote monitoring device inputs the meteorological training image into the meteorological image detection network. The network extracts and analyzes the image according to the requirements of the target detection branch, and tries to find out the abnormal conditions that may exist in the image. For example, under the heavy precipitation detection branch, the meteorological image detection network will pay attention to the thickness, color, texture and other features of the cloud layer in the image, and judge whether there is a heavy precipitation area through these features. In order to realize this detection process, the meteorological image detection network can adopt a deep learning model such as a convolutional neural network (CNN). CNN can automatically extract features in the image, process the image through structures such as convolutional layers and pooling layers, and then classify and predict the extracted features in the fully connected layer, so as to obtain the abnormal meteorological detection result of the meteorological training image under the target detection branch.

[0053] In step S30, the remote monitoring device determines the target priori marker from at least one prior marker matched by the meteorological training image, and determines the abnormal meteorological detection result of the meteorological training image under the target detection branch from the target prior marker, wherein the target prior marker is the prior marker corresponding to the detection branch and the target detection branch. Since the meteorological training image may match multiple prior markers, each prior marker corresponds to a different detection branch, it is necessary to find the target prior marker that is consistent with the current target detection branch. For example, the meteorological training image simultaneously matches the prior markers of the three detection branches of heavy precipitation, high temperature, and typhoon, and the current target detection branch is heavy precipitation detection, then the remote monitoring device will select the prior marker corresponding to the heavy precipitation detection branch as the target prior marker, and obtain the real abnormal meteorological detection result of the meteorological training image under the heavy precipitation detection branch. The purpose of this step is to provide an accurate reference for subsequent network debugging.

[0054] In step S40, the remote monitoring device performs a network debugging operation on the meteorological image detection network based on the abnormal meteorological detection results of the meteorological training image under the target detection branch and the abnormal meteorological detection results under the target detection branch. The debugged meteorological image detection network is used to perform abnormal meteorological detection of meteorological images under the corresponding detection branch. Specifically, the remote monitoring device compares the difference between the abnormal meteorological detection results obtained by the meteorological image detection network and the actual abnormal meteorological detection results indicated by the target priori mark, and adjusts the parameters of the meteorological image detection network according to the difference, so that the detection results of the network gradually approach the actual results. For example, a cross entropy loss function is used to measure the difference between the two.

[0055] In practical applications, in order to improve the training effect of the meteorological image detection network, the remote monitoring device can adopt a variety of strategies. In terms of data acquisition, the diversity of meteorological training images can be increased, including images of different seasons, different regions, and different meteorological conditions, to improve the network's adaptability to various meteorological conditions. During the training process, a batch training method can be used to divide the meteorological training images into multiple small batches for training, which can reduce memory usage and improve training efficiency. Regularization techniques, such as L1 and L2 regularization, can also be used to prevent network overfitting and improve the generalization ability of the network.

[0056] Taking heavy precipitation detection as an example, the remote monitoring device first obtains a large number of meteorological training images containing heavy precipitation conditions from the meteorological database, and determines heavy precipitation detection as the target detection branch. Then, these meteorological training images are input into the meteorological image detection network, and the network analyzes the images to obtain abnormal meteorological detection results under the heavy precipitation detection branch. Next, the remote monitoring device finds the target priori mark corresponding to the heavy precipitation detection branch from the priori marks matched by the meteorological training images, and obtains the real heavy precipitation anomaly. Finally, the difference between the network detection results and the real results is compared, and the parameters of the meteorological image detection network are adjusted using the cross entropy loss function and the gradient descent algorithm. After multiple iterative training, the meteorological image detection network's ability to detect heavy precipitation anomalies will continue to improve, and it will be able to more accurately identify heavy precipitation areas in meteorological images.

[0057] As an implementation method, the meteorological image detection network includes a representation information extraction component and an abnormal reasoning network; the representation information extraction component is used to extract meteorological image representation vectors; the abnormal reasoning network is used to perform abnormal meteorological detection on meteorological training images to obtain corresponding abnormal meteorological detection results.

[0058] The main function of the representation information extraction component is to extract the meteorological image representation vector, which is a vector that can represent the characteristic information of the meteorological image and the relationship between different image elements. In meteorological images, there is a wealth of information such as cloud morphology, precipitation area distribution, temperature change characteristics, etc., and the role of the representation information extraction component is to convert these complex image information into a vector form that can be processed and analyzed by computers. Taking heavy precipitation meteorological images as an example, the representation information extraction component will focus on the characteristics of the cloud thickness, color, texture, etc. in the image, as well as the size, location and boundary of the precipitation area, and quantify and integrate these characteristics and information to form a meteorological image representation vector. Each dimension in this vector corresponds to a certain feature or feature combination of the image, and this vector can be used to fully describe the characteristics of the meteorological image.

[0059] The representation information extraction component can adopt a variety of technical means. One method is to use a convolutional neural network (CNN). CNN has a powerful feature extraction capability. It performs convolution operations on meteorological images through convolution layers to extract local features in the image, such as edges, textures, etc. The pooling layer is used to reduce the dimension of the extracted features, reduce the amount of calculation and enhance the robustness of the features. Through multiple layers of convolution and pooling operations, CNN can gradually extract high-level semantic features of meteorological images, and finally combine these features into meteorological image representation vectors. For example, a typical CNN structure may contain multiple convolution layers and pooling layers. Each convolution layer uses different convolution kernels to extract different types of features. After multiple convolutions and pooling, a fixed-length vector is obtained as the meteorological image representation vector.

[0060] The abnormal reasoning network uses the meteorological image representation vector extracted by the representation information extraction component to perform abnormal meteorological detection on the meteorological training image and obtain the corresponding abnormal meteorological detection results. The task of the abnormal reasoning network is to judge whether there are abnormal meteorological conditions in the meteorological image, as well as the type and degree of the abnormality, based on the meteorological image representation vector. Continuing with the example of a heavy precipitation meteorological image, the abnormal reasoning network will judge whether there is a heavy precipitation area in the image, as well as the range and intensity of the heavy precipitation area, based on the input meteorological image representation vector.

[0061] The abnormal reasoning network can be implemented using structures such as fully connected neural networks or recurrent neural networks. The fully connected neural network takes the meteorological image representation vector as input, performs nonlinear transformation through multiple fully connected layers, and finally outputs the abnormal meteorological detection results. For example, the output layer of the network can use the softmax function to output the probability of each abnormal category (such as heavy precipitation, high temperature, typhoon, etc.), and the category with the highest probability is the predicted abnormal type. The formula of the softmax function is: ; Among them, z j is the input of the jth neuron in the network output layer, and K is the number of abnormal categories.

[0062] Recurrent neural networks are suitable for processing meteorological data with sequence relationships, such as sequences of meteorological images at different time points. It can capture the changing trend of meteorological images over time, thereby more accurately detecting anomalies. For example, in the process of monitoring the development of a typhoon, a recurrent neural network can predict the movement direction and intensity changes of a typhoon based on the representation vectors of meteorological images at multiple time points.

[0063] In practical applications, the representation information extraction component and the anomaly reasoning network work together. The representation information extraction component provides the anomaly reasoning network with accurate meteorological image representation vectors, enabling the anomaly reasoning network to perform effective anomaly detection based on these vectors. The detection results of the anomaly reasoning network can be fed back to the representation information extraction component to help it further optimize the feature extraction method and parameters and improve the quality of the meteorological image representation vector. Through this collaborative work, the meteorological image detection network can continuously improve its ability to detect abnormal meteorological conditions in meteorological training images, providing reliable support for remote monitoring of meteorological business systems and meteorological disaster warnings.

[0064] As an implementation mode, the number of target detection branches is at least one, and the number of abnormal meteorological detection results obtained by performing abnormal meteorological detection on meteorological training images based on the target detection branches is the same as the number of target detection branches; the number of detection branches corresponding to the prior labels matched by the meteorological training images is not less than the number corresponding to the target detection branches; wherein, when the number of target detection branches is multiple, the number of abnormal meteorological detection results obtained is also multiple, and the number of detection branches corresponding to the target prior labels determined based on the prior labels matched by the meteorological training images is equal to the number of target detection branches.

[0065] In the remote monitoring of low-bandwidth regulatory meteorological business systems, when the meteorological image detection network performs abnormal meteorological detection on meteorological training images, there is a specific correspondence between the number of target detection branches and the abnormal meteorological detection results, prior labels, etc. related thereto. The target detection branch refers to the specific meteorological anomaly type that is concerned when performing anomaly detection on meteorological training images. The number of target detection branches can be at least one, which means that the remote monitoring device can detect only one type of meteorological anomaly or multiple types at the same time. The number of abnormal meteorological detection results obtained after abnormal meteorological detection of meteorological training images based on the target detection branch is the same as the number of target detection branches, which reflects a one-to-one correspondence, that is, each target detection branch will correspond to a corresponding abnormal meteorological detection result.

[0066] The number of detection branches corresponding to the priori marks matched by the meteorological training images is not less than the number corresponding to the target detection branches. The priori marks are the real abnormal conditions of the meteorological training images under different detection branches that are predetermined, ensuring that there are sufficient references to evaluate the accuracy of the detection results when performing abnormal meteorological detection. When the number of target detection branches is multiple, the number of abnormal meteorological detection results obtained is also multiple, and the number of detection branches corresponding to the target priori marks determined based on the priori marks matched by the meteorological training images is equal to the number of target detection branches, which ensures that in the case of multi-branch detection, each target detection branch has a corresponding accurate reference standard.

[0067] The number of detection branches corresponding to the priori marks matched by this meteorological training image is at least three, and may also include marks of some other detection branches, but at least the three target detection branches of heavy precipitation, high temperature and typhoon must be covered. The remote monitoring device determines the target priori marks from these priori marks. The number of detection branches corresponding to these target priori marks is the same as the number of target detection branches, that is, the priori marks of the three detection branches of heavy precipitation, high temperature and typhoon. By comparing the abnormal meteorological detection results and the actual situation indicated by the target priori marks, the remote monitoring device can evaluate the detection accuracy of the meteorological image detection network under each target detection branch, and debug and optimize the meteorological image detection network according to the evaluation results.

[0068] In actual operation, the remote monitoring device can use multi-task learning to realize abnormal weather detection of multiple target detection branches. Multi-task learning allows the meteorological image detection network to handle multiple related tasks at the same time. In this example, it is to detect heavy precipitation, high temperature and typhoon at the same time. The network can share a part of the underlying feature extraction layer, and then set an independent output layer for each target detection branch to output the abnormal weather detection results of each branch respectively. This can improve the detection efficiency and also improve the detection accuracy by utilizing the correlation between different detection branches.

[0069] In addition, in order to ensure the accuracy and completeness of the target prior labeling, the remote monitoring device can use a combination of manual labeling and automatic labeling. Manual labeling can be done by meteorological experts who carry out detailed labeling of meteorological training images based on their professional knowledge to ensure the accuracy of the labeling. Automatic labeling can use existing meteorological data and algorithms to perform preliminary labeling on a large number of meteorological training images, which can then be manually reviewed and corrected to improve the efficiency of labeling.

[0070] By reasonably setting the number of target detection branches and ensuring the correspondence between abnormal meteorological detection results, prior labels, etc., the remote monitoring device can perform abnormal meteorological detection on meteorological training images more comprehensively and accurately. This multi-branch detection method can simultaneously meet the monitoring needs of multiple types of meteorological anomalies, and provide richer and more accurate information for remote monitoring of meteorological business systems and meteorological disaster warnings. In future meteorological monitoring and warning work, with the continuous enrichment of meteorological data and the continuous advancement of technology, the abnormal meteorological detection method of multi-target detection branches is expected to be further optimized and improved, providing stronger support for the development of the meteorological field.

[0071] As an implementation method, the meteorological training image acquisition method includes: Step S11: obtaining at least one candidate meteorological image and branch prompt information; the branch prompt information is used to prompt the number of branches of the target detection branch; Step S12: determining a candidate meteorological image that matches a target detection branch with a corresponding number of branches in at least one candidate meteorological image according to the branch prompt information, and using the determined candidate meteorological image as a branch meteorological image of the meteorological training image; Step S13: Acquire a sample meteorological image, and combine the branch meteorological image with the sample meteorological image to obtain a meteorological training image.

[0072] In step S11, the remote monitoring device obtains at least one candidate meteorological image and branch prompt information, and the branch prompt information is used to prompt the number of branches of the target detection branch. The candidate meteorological image is a series of meteorological images collected in advance, which cover various features under different meteorological conditions, such as images of different meteorological scenes such as sunny, cloudy, rainy, typhoon, etc. The branch prompt information provides key guidance for subsequent processing, which informs the remote monitoring device of the number of target detection branches that are concerned. For example, the branch prompt information may indicate that the target detection branch is heavy precipitation detection and high temperature detection, and the number of branches is 2. The remote monitoring device can obtain candidate meteorological images and branch prompt information by establishing a connection with a data source such as a meteorological data center or meteorological monitoring equipment, or it can read relevant data from local storage.

[0073] In step S12, the remote monitoring device determines a candidate meteorological image that matches the target detection branch of the corresponding number of branches in at least one candidate meteorological image according to the branch prompt information, and uses the determined candidate meteorological image as a branch meteorological image of the meteorological training image. This step is to filter out images related to the target detection branch from a large number of candidate meteorological images according to the branch prompt information. For example, in the above-mentioned example of heavy precipitation detection and high temperature detection, the remote monitoring device finds images related to heavy precipitation and high temperature from the candidate meteorological images as branch meteorological images. An image classification algorithm can be used to implement this screening process, by extracting and analyzing the features of the candidate meteorological images and classifying them into different target detection branch categories. For example, a convolutional neural network (CNN) is used to extract features from the candidate meteorological images, and then the images are classified according to the feature model of the target detection branch to find images that match the heavy precipitation and high temperature detection branches.

[0074] In step S13, the remote monitoring device obtains a sample meteorological image, combines the branch meteorological image and the sample meteorological image, and obtains a meteorological training image. The sample meteorological image refers to a basic image that contains various meteorological features. It can be a representative meteorological image or a processed comprehensive meteorological image. There are many ways to combine, such as superimposing the branch meteorological image as supplementary information on the sample meteorological image, or fusing their features. For example, for the branch meteorological image of the heavy precipitation detection branch, the part representing the heavy precipitation feature can be superimposed on the corresponding position of the sample meteorological image to enhance the feature information related to heavy precipitation in the sample meteorological image. In terms of feature fusion, methods such as feature splicing or weighted averaging can be used to fuse the features of the branch meteorological image and the sample meteorological image to generate a new meteorological training image.

[0075] Taking typhoon detection as an example, the remote monitoring device first obtains a large number of candidate meteorological images and branch prompt information. The prompt information indicates that the target detection branch is typhoon detection and the number of branches is 1. Then, the remote monitoring device uses an image classification algorithm to filter out typhoon-related images from the candidate meteorological images as branch meteorological images. These images may contain typical features such as the eye of the typhoon and the typhoon cloud system. Next, the remote monitoring device obtains a sample meteorological image and combines the typhoon-related branch meteorological image with the sample meteorological image. The features of the typhoon eye and the typhoon cloud system in the branch meteorological image can be superimposed on the appropriate position of the sample meteorological image, or the features of the two can be weighted averaged and fused to finally obtain a meteorological training image for typhoon detection training.

[0076] In practical applications, in order to improve the quality and diversity of meteorological training images, the remote monitoring device can pre-process the candidate meteorological images, such as removing noise, adjusting brightness and contrast, etc., to improve the clarity and quality of the image. The sample meteorological images can also be enhanced, such as rotating, flipping, scaling, etc., to increase the diversity of samples and improve the generalization ability of the meteorological image detection network. In addition, in order to ensure that the combination of branch meteorological images and sample meteorological images is reasonable and effective, some evaluation indicators can be used to measure the quality of the combined meteorological training images, such as information entropy, correlation, etc. Through these methods, high-quality and diverse meteorological training images can be obtained, which provides strong support for the training of meteorological image detection networks, thereby improving the accuracy and reliability of meteorological anomaly detection and better serving the remote monitoring and meteorological disaster warning work of meteorological business systems.

[0077] As an implementation mode, step S20, performing abnormal meteorological detection on the meteorological training image according to the target detection branch to obtain the abnormal meteorological detection result of the meteorological training image under the target detection branch, includes: Step S21: extracting representation information of the meteorological training image according to the target detection branch to obtain a meteorological image representation vector corresponding to the meteorological training image; the meteorological image representation vector is used to represent feature information of image elements in the meteorological training image and the relationship between different image elements; Step S22: performing abnormal meteorological detection on the meteorological training image based on the meteorological image representation vector to obtain an abnormal meteorological detection result of the meteorological training image under the target detection branch.

[0078] In step S21, the remote monitoring device extracts the representation information of the meteorological training image according to the target detection branch, and obtains the meteorological image representation vector corresponding to the meteorological training image, and the meteorological image representation vector is used to represent the characteristic information of the image elements in the meteorological training image and the relationship between different image elements. The meteorological training image contains rich meteorological information, such as the shape of clouds, the distribution of precipitation areas, the change of temperature, etc., and the meteorological image representation vector is a vector representation obtained by quantifying and integrating these complex information.

[0079] In order to extract meteorological image representation vectors, remote monitoring devices can use convolutional neural networks (CNNs) in deep learning. Through the alternating operation of multiple convolutional layers and pooling layers, CNN can gradually extract different levels of features of meteorological training images. For example, in the first convolutional layer, some simple edge and texture features may be extracted; as the number of network layers increases, the subsequent convolutional layers will extract more advanced semantic features, such as cloud types, boundaries of precipitation areas, etc.

[0080] In step S22, the remote monitoring device performs abnormal weather detection on the meteorological training image based on the meteorological image representation vector to obtain the abnormal weather detection result of the meteorological training image under the target detection branch. This step uses the meteorological image representation vector extracted above to determine whether the meteorological training image has abnormal weather conditions corresponding to the target detection branch.

[0081] Remote monitoring devices can use fully connected neural networks to detect abnormal weather. The fully connected neural network takes the meteorological image representation vector as input and performs nonlinear transformation through multiple fully connected layers. Assume that the input meteorological image representation vector is x, and the weight matrix of the lth layer is , the bias vector is , then the input of the l+1th layer It can be expressed as: ; where x l is the output of the lth layer. After being processed by a nonlinear activation function (such as the ReLU function, ReLU(z)=max(0,z)), the output x of the l+1th layer is obtained. l+1 .

[0082] In abnormal weather detection, the output layer of the fully connected neural network outputs the probability that the weather training image belongs to different abnormal categories under the target detection branch. For example, under the heavy precipitation detection branch, the output layer may have two neurons, representing the probability of the presence and absence of heavy precipitation anomalies. By comparing the sizes of these two probabilities, the remote monitoring device can determine whether the weather training image is abnormal under the heavy precipitation detection branch.

[0083] Continuing with the heavy precipitation detection as an example, the remote monitoring device inputs the meteorological image representation vector obtained in step S21 into the fully connected neural network. The network processes and analyzes the vector, and calculates the probability of the presence of heavy precipitation anomaly in the meteorological training image based on the characteristic information and relationship contained in the vector. If this probability exceeds a preset threshold (such as 0.5), the remote monitoring device determines that the meteorological training image has an anomaly under the heavy precipitation detection branch; otherwise, it determines that there is no anomaly.

[0084] In order to improve the accuracy of abnormal weather detection, the remote monitoring device can adopt the cross entropy loss function when training the fully connected neural network.

[0085] As an implementation manner, in step S22, before performing abnormal meteorological detection on the meteorological training image based on the meteorological image representation vector and obtaining the abnormal meteorological detection result of the meteorological training image under the target detection branch, the method further includes: Step S2201: obtaining the meteorological image type corresponding to the meteorological image to which each image element in the meteorological training image belongs; Step S2202: generating a shielding matrix based on the meteorological image type; the shielding matrix includes a plurality of shielding elements, and the image elements included in the meteorological training image correspond to the shielding elements in the shielding matrix; Step S2203: Based on the shielding matrix, the meteorological image representation vector of the meteorological training image is shielded to obtain the shielded meteorological image representation vector; the shielded meteorological image representation vector includes the element representation vectors corresponding to each image element in the sample meteorological image, and the shielded meteorological image representation vector is used to perform abnormal meteorological detection on the meteorological training image.

[0086] In step S2201, the remote monitoring device obtains the meteorological image type corresponding to the meteorological image to which each image element in the meteorological training image belongs. The meteorological image type is a way to classify meteorological images, and different meteorological image types have different meteorological features and patterns. For example, meteorological image types can be divided into sunny images, cloudy images, precipitation images, typhoon images, etc. In the meteorological training image, each image element may belong to a different meteorological image type, and understanding these types is helpful for subsequent analysis and processing. The remote monitoring device can obtain meteorological image type information in a variety of ways. One method is to use pre-labeled meteorological image data sets, which will label each meteorological image with its corresponding type. Another method is to use an image classification algorithm to extract and analyze the features of the meteorological training image and classify it into different meteorological image types. For example, a convolutional neural network (CNN) is used to extract features of the meteorological training image, and then the meteorological image type to which each image element belongs is determined based on a pre-trained classification model.

[0087] In step S2202, the remote monitoring device generates a shielding matrix based on the meteorological image type, the shielding matrix includes a plurality of shielding elements, and the image elements included in the meteorological training image correspond to the shielding elements in the shielding matrix. The function of the shielding matrix is ​​to shield some information irrelevant to the current abnormal meteorological detection according to the meteorological image type. For example, when performing heavy precipitation anomaly detection, if the meteorological image type to which a certain image element belongs is a sunny image, then this image element is likely to be irrelevant to the heavy precipitation anomaly, and its corresponding information can be shielded by the shielding matrix. The specific method of generating the shielding matrix can be designed according to the different meteorological image types. A simple method is to assign a shielding coefficient to each meteorological image type, and when a certain image element belongs to a specific meteorological image type, the value of the corresponding shielding element is set to the shielding coefficient of the type. For example, for the sunny image type irrelevant to the heavy precipitation detection, the shielding coefficient can be set to 0, and for the precipitation image type related to the heavy precipitation, the shielding coefficient can be set to 1. In this way, the elements in the shielding matrix can be assigned according to the meteorological image type of each image element in the meteorological training image.

[0088] In step S2203, the remote monitoring device performs a shielding operation on the meteorological image representation vector of the meteorological training image based on the shielding matrix to obtain the shielded meteorological image representation vector, which includes the element representation vector corresponding to each image element in the sample meteorological image, and the vector is used to perform abnormal meteorological detection on the meteorological training image. The shielding operation is to multiply the shielding matrix by the meteorological image representation vector element by element, so as to filter out unnecessary information. Through this shielding operation, the element value corresponding to the information irrelevant to the current abnormal meteorological detection becomes 0, while the information related to the detection is retained. For example, in the heavy precipitation detection, after the shielding operation, the value of the image element corresponding to the sunny image in the meteorological image representation vector becomes 0, while the value of the image element corresponding to the precipitation image remains unchanged, so that the characteristic information related to heavy precipitation can be highlighted and the accuracy of abnormal detection can be improved.

[0089] Taking heavy precipitation detection as an example, the remote monitoring device first obtains the meteorological image type corresponding to the meteorological image to which each image element in the meteorological training image belongs. Assuming that the meteorological training image contains image elements of sunny areas, cloudy areas and precipitation areas, the remote monitoring device determines the meteorological image type to which the image elements in different areas belong through the image classification algorithm. Then, a shielding matrix is ​​generated based on these meteorological image types. For the image elements in the sunny area, the corresponding shielding element value is set to 0; for the image elements in the precipitation area, the corresponding shielding element value is set to 1; for the image elements in the cloudy area, according to their correlation with heavy precipitation, the shielding element value is set to a value between 0 and 1, such as 0.2. Then, the remote monitoring device multiplies the generated shielding matrix by the meteorological image representation vector of the meteorological training image element by element, and obtains the shielded meteorological image representation vector. In this shielded vector, the value of the element representation vector corresponding to the image elements in the sunny area becomes 0, and is thus shielded, while the value of the element representation vector corresponding to the image elements in the precipitation area and some cloudy areas that have a certain correlation with heavy precipitation is retained. Finally, this shielded meteorological image representation vector is used to detect heavy precipitation anomalies in meteorological training images. Since irrelevant information is shielded, the detection process can focus more on features related to heavy precipitation, thereby improving the accuracy and efficiency of detection.

[0090] In practical applications, in order to improve the effect of steps S2201-S2203, the remote monitoring device can adopt some optimization strategies. When obtaining the type of meteorological image, a multimodal classification algorithm such as one based on deep learning can be used to combine the visual features of the image and the features of the meteorological data to improve the accuracy of classification. When generating a shielding matrix, the shielding coefficient can be dynamically adjusted according to different target detection branches to adapt to different abnormal meteorological detection needs. For example, when performing typhoon detection, for some cloudy image types that are somewhat related to the outer cloud system of the typhoon, the shielding coefficient can be appropriately increased to retain more information that may be related to the typhoon. In addition, in order to ensure the rationality and effectiveness of the shielding matrix, the remote monitoring device can be evaluated by a cross-validation method. By dividing the meteorological training image into a training set and a validation set, a shielding matrix is ​​generated on the training set and a shielding operation is performed, and then the effect of abnormal meteorological detection is evaluated on the validation set. According to the evaluation results, the generation method and shielding coefficient of the shielding matrix are adjusted until the best detection effect is achieved.

[0091] As an implementation manner, step S22, performing abnormal meteorological detection on the meteorological training image based on the meteorological image representation vector to obtain the abnormal meteorological detection result of the meteorological training image under the target detection branch, includes: Step S221: based on the meteorological image representation vector, inferring the abnormal classification prediction information of each image element in the meteorological training image; the abnormal classification prediction information of any image element represents the probability that the element classification of the corresponding image element belongs to the element classification of abnormal meteorology under the target detection branch; Step S222: Based on the abnormal classification prediction information corresponding to each image element in the meteorological training image, an abnormal meteorological detection result of the meteorological training image under the target detection branch is generated.

[0092] In step S221, the remote monitoring device infers the abnormal classification prediction information of each image element in the meteorological training image based on the meteorological image representation vector. The abnormal classification prediction information of any image element indicates the probability that the element classification of the corresponding image element belongs to the element classification of abnormal meteorology under the target detection branch. The meteorological image representation vector is obtained after feature extraction of the meteorological training image in the previous step. It contains the feature information of the image elements in the meteorological training image and the involved relationship between different image elements. The abnormal classification prediction information is the result of quantifying the possibility of each image element belonging to the abnormal meteorological element classification based on these feature information.

[0093] In order to realize this reasoning process, the remote monitoring device can adopt a neural network model, such as a fully connected neural network. The fully connected neural network takes the meteorological image representation vector as input and performs nonlinear transformation through multiple fully connected layers.

[0094] Taking the target detection branch of heavy precipitation detection as an example, it is assumed that the elements of abnormal weather are classified into three types: heavy precipitation area, weak precipitation area and no precipitation area. The remote monitoring device inputs the meteorological image representation vector of the meteorological training image into the fully connected neural network. After calculation and processing by the network, the three neurons in the output layer output the probability that the image element belongs to the heavy precipitation area, weak precipitation area and no precipitation area respectively. For example, for a specific pixel in the image, the network outputs that the probability of it belonging to the heavy precipitation area is 0.8, the probability of it belonging to the weak precipitation area is 0.1, and the probability of it belonging to the no precipitation area is 0.1. Then the abnormal classification prediction information of this pixel is such a set of probability values.

[0095] In step S222, the remote monitoring device generates an abnormal meteorological detection result of the meteorological training image under the target detection branch based on the abnormal classification prediction information corresponding to each image element in the meteorological training image. This step is to comprehensively analyze and judge the abnormal classification prediction information of each image element to determine the abnormal situation of the entire meteorological training image under the target detection branch.

[0096] The remote monitoring device can use a variety of methods to generate abnormal weather detection results. One method is to count the number or proportion of image elements belonging to each abnormal weather element classification, and then determine whether there is an abnormality based on a pre-set threshold.

[0097] More complex methods can also be used to combine the spatial relationship and feature correlation between different image elements to make judgments. For example, for adjacent image elements, if they all have a high probability of belonging to a heavy rainfall area, and these elements form a certain continuous area in space, then it can be more confidently judged that this is a heavy rainfall anomaly area.

[0098] Continuing with the example of heavy precipitation detection, the remote monitoring device counts the probability that all image elements in the meteorological training image belong to the heavy precipitation area. Assume that there are 1000 pixels in the image, and the probability that 200 of them belong to the heavy precipitation area is greater than 0.8. If the pre-set heavy precipitation anomaly threshold is that the number of pixels accounts for 15%, then since 200 / 1000=20% > 15%, the remote monitoring device determines that the meteorological training image has an anomaly under the heavy precipitation detection branch.

[0099] In order to improve the accuracy of abnormal weather detection results, the remote monitoring device can adopt the cross entropy loss function when training the neural network.

[0100] In practical applications, the effective execution of steps S221-S222 is crucial to the performance of the meteorological image detection network. Accurate abnormal classification prediction information can provide a basis for generating reliable abnormal meteorological detection results, and a reasonable result generation method can make accurate judgments based on these prediction information. The remote monitoring device can improve the accuracy of abnormal classification prediction information and the reliability of abnormal meteorological detection results by continuously optimizing the structure and parameters of the neural network and increasing the diversity and quantity of training data. At the same time, it can also be combined with other meteorological data, such as meteorological observation data, numerical forecast data, etc., to further improve the practicality and effectiveness of abnormal meteorological detection.

[0101] As an implementation method, the element classification of abnormal weather under the target detection branch includes at least one, and the abnormal classification prediction information of any image element includes: the support coefficient of the element classification of the corresponding image element belonging to different element classifications, and any support coefficient is used to indicate the probability that the element classification of the corresponding image element belongs to the corresponding element classification. Based on this, step S222, based on the abnormal classification prediction information corresponding to each image element in the meteorological training image, generates an abnormal meteorological detection result of the meteorological training image under the target detection branch, including:

[0102] Step S2221: determining a maximum support coefficient among the support coefficients included in the abnormal classification prediction information of any image element; Step S2222: classifying the maximum support coefficient and the element corresponding to the maximum support coefficient as target prediction information corresponding to any image element; Step S2223: generating an abnormal meteorological detection result of the meteorological training image under the target detection branch according to the target prediction information of each image element in the meteorological training image.

[0103] In step S2221, the remote monitoring device determines the maximum support coefficient among the support coefficients included in the abnormal classification prediction information of any image element. The support coefficient refers to the probability that the element classification of the corresponding image element belongs to different element classifications, which is obtained by predicting the abnormal classification of the meteorological training image in the previous step. For example, under the target detection branch of heavy precipitation detection, the element classification of abnormal meteorology may include heavy precipitation areas, weak precipitation areas, and no precipitation areas. For a specific image element in the meteorological training image, its abnormal classification prediction information may show that the support coefficient of the element belonging to the heavy precipitation area is 0.8, the support coefficient of the element belonging to the weak precipitation area is 0.1, and the support coefficient of the element belonging to the no precipitation area is 0.1. In this case, the maximum support coefficient is 0.8.

[0104] In step S2222, the remote monitoring device uses the maximum support coefficient and the element classification corresponding to the maximum support coefficient as the target prediction information corresponding to any image element. Continuing with the above example of heavy precipitation detection, the maximum support coefficient is 0.8, and its corresponding element classification is heavy precipitation area, then the target prediction information of the image element is (0.8, heavy precipitation area). This step simplifies the abnormal classification prediction information of each image element, focusing on the most likely element classification and its corresponding support coefficient, providing a clearer basis for the subsequent generation of the overall abnormal detection results of the meteorological training image.

[0105] In step S2223, the remote monitoring device generates an abnormal meteorological detection result of the meteorological training image under the target detection branch based on the target prediction information of each image element in the meteorological training image. This step requires comprehensive analysis and judgment of the target prediction information of all image elements. One feasible method is to count the number or proportion of image elements belonging to different element classifications, and judge whether the meteorological training image is abnormal based on a pre-set threshold.

[0106] For example, in heavy precipitation detection, the remote monitoring device counts the number of image elements in the meteorological training image whose target prediction information is (high support coefficient, heavy precipitation area). Assume that the meteorological training image has a total of 10,000 image elements, of which 1,500 image elements have target prediction information showing that they belong to the heavy precipitation area and the support coefficient is greater than 0.8. If the pre-set heavy precipitation anomaly threshold is that the proportion of image elements in the heavy precipitation area reaches 10%, since 1500 / 10000 = 15% > 10%, then the remote monitoring device can determine that the meteorological training image has an anomaly under the heavy precipitation detection branch.

[0107] In addition, the remote monitoring device can also consider the spatial distribution of image elements. If the image elements belonging to the heavy precipitation area form a continuous and large area in space, it is more likely to indicate the presence of heavy precipitation anomalies. On the contrary, if these elements are scattered, further analysis or combination with other meteorological data may be required to make a judgment. In practical applications, in order to improve the accuracy of anomaly detection results, the remote monitoring device can adopt some optimization strategies. When determining the maximum support coefficient, a confidence threshold can be set. Only when the maximum support coefficient exceeds the threshold, the corresponding element classification is used as part of the target prediction information. For example, the confidence threshold is set to 0.7. If the maximum support coefficient of an image element is 0.6, then the classification of the element may need to be further analyzed or the information of the element may need to be ignored.

[0108] When generating anomaly detection results, the correlation between different image elements can be combined. For example, adjacent image elements often have similar meteorological characteristics. If their target prediction information all shows that they belong to the heavy precipitation area, then the credibility of the judgment of heavy precipitation anomaly can be increased. This correlation can be evaluated by calculating the consistency of the target prediction information of adjacent image elements. In addition, the remote monitoring device can also dynamically adjust the threshold according to different meteorological scenes and target detection branches. For example, in some areas or seasons, the standard of heavy precipitation may be different. At this time, the threshold of the proportion of image elements in the heavy precipitation area can be adjusted according to the actual situation. Taking typhoon detection as an example, the element classification of abnormal meteorology may include typhoon eye, typhoon cloud system and peripheral airflow. For each image element in the meteorological training image, the remote monitoring device first determines the maximum support coefficient in its abnormal classification prediction information. Assuming that the support coefficient of an image element belonging to the typhoon eye is 0.9, the support coefficient of the typhoon cloud system is 0.05, and the support coefficient of the peripheral airflow is 0.05, then the maximum support coefficient is 0.9, the corresponding element classification is the typhoon eye, and the target prediction information of the image element is (0.9, typhoon eye).

[0109] Then, the remote monitoring device counts the target prediction information of all image elements in the meteorological training image. If it is found that there are multiple image elements with target prediction information of (high support coefficient, typhoon eye), and these elements form a relatively concentrated area in space, and there are also a large number of image elements with target prediction information of (high support coefficient, typhoon cloud system) around them, then it can be judged that the meteorological training image has an abnormality under the typhoon detection branch, that is, signs of a typhoon may be detected.

[0110] As another embodiment, the element classification of abnormal weather under the target detection branch includes at least one, and the abnormal classification prediction information is also used to indicate: when the element classification of the corresponding image element belongs to different element classifications, the probability that the element classification of adjacent image elements belongs to different element classifications, and the probability is expressed based on the conversion support coefficient of the adjacent image elements; the abnormal classification prediction information includes: the conversion support coefficient and the support coefficient that the element classification of the corresponding image element belongs to different element classifications; any support coefficient is used to indicate the probability that the element classification of the corresponding image element belongs to the corresponding element classification. Based on this, step S222, based on the abnormal classification prediction information corresponding to each image element in the meteorological training image, generates the abnormal weather detection result of the meteorological training image under the target detection branch, including:

[0111] Step S222A: generating at least one element classification trajectory corresponding to the meteorological training image according to the abnormal classification prediction information; any element classification trajectory includes the element classification of each image element in the meteorological training image; any element classification trajectory includes the support coefficient of the element classification of each image element of the meteorological training image under the corresponding element classification trajectory, and the conversion support coefficient of the adjacent image elements; Step S222B: determining a trajectory support coefficient of any element classification trajectory based on the support coefficient and the conversion support coefficient corresponding to any element classification trajectory; Step S222C: Determine the element classification trajectory corresponding to the maximum trajectory support coefficient in at least one element classification trajectory, and generate an abnormal meteorological detection result of the meteorological training image under the target detection branch based on the element classification included in the element classification trajectory corresponding to the maximum trajectory support coefficient.

[0112] In step S222A, the remote monitoring device generates at least one element classification trajectory corresponding to the meteorological training image based on the abnormal classification prediction information. Any element classification trajectory includes the element classification of each image element in the meteorological training image, and also includes the support coefficient of the element classification of each image element in the meteorological training image under the corresponding element classification trajectory, and the conversion support coefficient of the adjacent image elements. The abnormal classification prediction information not only indicates the probability that each image element belongs to a different element classification (i.e., the support coefficient), but also reflects the probability that the adjacent image elements are converted from one element classification to another element classification (i.e., the conversion support coefficient). The element classification trajectory is a sequence formed by arranging the element classifications of all image elements in the meteorological training image in a certain order and combining the corresponding support coefficients and conversion support coefficients.

[0113] Taking the target detection branch of heavy precipitation detection as an example, it is assumed that the element classification of abnormal meteorology includes three types: heavy precipitation area, weak precipitation area and no precipitation area. For the three adjacent image elements A, B and C in the meteorological training image, the abnormal classification prediction information shows that the support coefficient of image element A belonging to the heavy precipitation area is 0.8, the support coefficient of belonging to the weak precipitation area is 0.1, and the support coefficient of belonging to the no precipitation area is 0.1; the support coefficient of image element B belonging to the heavy precipitation area is 0.7, the support coefficient of belonging to the weak precipitation area is 0.2, and the support coefficient of belonging to the no precipitation area is 0.1; the support coefficient of image element C belonging to the heavy precipitation area is 0.6, the support coefficient of belonging to the weak precipitation area is 0.3, and the support coefficient of belonging to the no precipitation area is 0.1. At the same time, from image element A to image element B, the conversion support coefficient from heavy precipitation area to heavy precipitation area is 0.8, the conversion support coefficient from heavy precipitation area to weak precipitation area is 0.1, and the conversion support coefficient from heavy precipitation area to no precipitation area is 0.1; there is also a corresponding conversion support coefficient from image element B to image element C. Based on this information, an element classification trajectory can be generated: (A: heavy precipitation area, 0.8; B: heavy precipitation area, 0.7; C: heavy precipitation area, 0.6), where each element in the brackets represents the image element, element classification and support coefficient in turn, and the conversion support coefficient between adjacent image elements is implied in this trajectory. In order to generate the element classification trajectory, the remote monitoring device can use an exhaustive method or a search algorithm. The exhaustive method will traverse all possible element classification combinations, calculate the support coefficient and conversion support coefficient corresponding to each combination, and thus generate all possible element classification trajectories. The search algorithm can find a better element classification trajectory within a certain search space to reduce the amount of calculation. For example, using a dynamic programming algorithm, by recording intermediate results, repeated calculations can be avoided and the efficiency of generating element classification trajectories can be improved.

[0114] In step S222B, the remote monitoring device determines the trajectory support coefficient of any element classification trajectory based on the support coefficient and conversion support coefficient corresponding to the element classification trajectory. The trajectory support coefficient is an indicator to measure the rationality and possibility of the element classification trajectory. It comprehensively considers the support coefficient of each image element and the conversion support coefficient between adjacent image elements. A feasible method for calculating the trajectory support coefficient is to multiply the support coefficient of each image element in the element classification trajectory by the conversion support coefficient of the adjacent image element.

[0115] Assume that the element classification trajectory is , where I i represents the i-th image element, C i Indicates its element classification, s i Represents its support coefficient, adjacent image element Ii To I i+1 The conversion support coefficient is t i, i+1 , then the trajectory support coefficient S of the element classification trajectory can be expressed as: Continuing with the above example of heavy precipitation detection, for the element classification trajectory (A: heavy precipitation area, 0.8; B: heavy precipitation area, 0.7; C: heavy precipitation area, 0.6), assuming that the conversion support coefficient from A to B is t A, B =0.8, the conversion support coefficient from B to C is t B, C =0.7, then the trajectory support coefficient of this trajectory .

[0116] In step S222C, the remote monitoring device determines the element classification trajectory corresponding to the maximum trajectory support coefficient in at least one element classification trajectory, and generates an abnormal meteorological detection result of the meteorological training image under the target detection branch based on the element classification included in the element classification trajectory. The element classification trajectory corresponding to the maximum trajectory support coefficient represents the most likely element classification combination. By analyzing the distribution of element classifications in the trajectory, it can be determined whether the meteorological training image is abnormal under the target detection branch. For example, in heavy precipitation detection, if the element classification trajectory corresponding to the maximum trajectory support coefficient shows that a large area of ​​image elements in the meteorological training image is classified as a heavy precipitation area, and these areas have a certain continuity in space, then the remote monitoring device can determine that the meteorological training image is abnormal under the heavy precipitation detection branch. The specific judgment criteria can be pre-set according to the actual situation, such as the area proportion of the heavy precipitation area reaches a certain threshold (such as 15%), or the heavy precipitation area forms a specific shape or pattern, etc.

[0117] As an implementation mode, step S40, based on the abnormal meteorological detection result of the meteorological training image under the target detection branch and the abnormal meteorological detection result under the target detection branch, performs a network debugging operation on the meteorological image detection network, including: Step S41: generating a target training cost under the target detection branch according to the abnormal meteorological detection result of the meteorological training image under the target detection branch and the error between the abnormal meteorological detection results under the target detection branch; Step S42: According to the goal of reducing the target training cost, the parameters of the meteorological image detection network are iterated to perform network debugging operations on the meteorological image detection network.

[0118] In step S41, the remote monitoring device generates a target training cost under the target detection branch based on the abnormal meteorological detection results of the meteorological training image under the target detection branch and the error between the abnormal meteorological detection results under the target detection branch. Two groups of abnormal meteorological detection results are involved here, one group is the result obtained by the meteorological image detection network detecting the meteorological training image, and the other group is the real abnormal meteorological detection result of the meteorological training image indicated by the target priori mark under the target detection branch. The error is an indicator to measure the degree of difference between the two groups of results, and the target training cost is a value used to evaluate the current performance of the network, which reflects the degree of deviation between the network prediction result and the real result.

[0119] In order to generate the target training cost, the remote monitoring device can use a loss function, such as a cross entropy loss function.

[0120] In step S42, the remote monitoring device iterates the parameters of the meteorological image detection network according to the goal of reducing the target training cost, so as to perform network debugging operations on the meteorological image detection network. The parameters of the meteorological image detection network mainly include weights and biases in the network, which determine the calculation method and output results of the network. Iteration refers to continuously adjusting the parameters, and recalculating the target training cost after each adjustment until the target training cost reaches a satisfactory value or meets certain stop conditions.

[0121] In order to achieve iterative adjustment of parameter variables, remote monitoring devices usually use optimization algorithms, such as the stochastic gradient descent (SGD) algorithm. The basic idea of ​​the SGD algorithm is to update the parameter variables according to the gradient of the target training cost relative to the parameter variables. Suppose the parameter variables are , the target training cost is , then at the tth iteration, the update formula of the parameter variable is: ;in is the learning rate, which controls the step size of parameter update at each iteration. is the target training cost exist The gradient at .

[0122] In practical applications, in order to improve the effect of network debugging, the remote monitoring device can adopt some optimization strategies. The learning rate decay method can be used to gradually reduce the learning rate as the number of iterations increases, so that it can converge quickly in the early stage of training and adjust the parameters more finely in the later stage. Batch normalization technology can also be used to normalize the input of each layer in the network to accelerate the convergence speed of the network and improve stability. In addition, in order to prevent the network from overfitting, regularization technology, such as L2 regularization, can be used.

[0123] Steps S41-S42 continuously optimize the performance of the network by calculating the target training cost and iteratively adjusting the parameters of the meteorological image detection network. In the remote monitoring of the low-bandwidth regulatory meteorological business system, the debugged meteorological image detection network can more accurately perform abnormal meteorological detection of meteorological images under the target detection branch, providing strong support for the stable operation of the meteorological business system and meteorological disaster warning.

[0124] Figure 2 A hardware entity diagram of a remote monitoring device provided by an embodiment of the present invention, such as Figure 2 As shown, the hardware entity of the remote monitoring device 1000 includes: a processor 1001 and a memory 1002, wherein the memory 1002 stores a computer program that can be run on the processor 1001, and the processor 1001 implements the steps in the method of any of the above embodiments when executing the program.

Claims

1. A low-bandwidth remote monitoring method for a regulatory meteorological service system, characterized in that: Applied to a remote monitoring device, the remote monitoring device is communicatively connected to a receiving end, and the method comprises: Acquire the meteorological image to be detected; Determine whether the change area of ​​the meteorological image to be detected compared with the previous meteorological image is greater than the change threshold; or, detect whether the meteorological image to be detected contains an abnormal detection result; When a change area of ​​the meteorological image to be detected compared with the previous meteorological image is greater than a change threshold, the change area is encoded to obtain a change area encoding result, and the change area encoding result is sent to a receiving end, so that the receiving end synthesizes the change area encoding result and the previous meteorological image based on the change area encoding result to obtain a synthesized meteorological image; Alternatively, when the meteorological image to be detected contains an abnormal detection result, the abnormal detection result is encoded to obtain an abnormal encoding result, and the abnormal encoding result is sent to the receiving end so that the receiving end synthesizes it based on the abnormal encoding result and the previous meteorological image to obtain a synthetic meteorological image.

2. The method according to claim 1, characterized in that The determining whether a change area of ​​the to-be-detected meteorological image compared with a previous meteorological image is greater than a change threshold comprises: Performing image segmentation on the previous meteorological image and the meteorological image to be detected respectively, to obtain image segmentation sets corresponding to the previous meteorological image and the meteorological image to be detected respectively, wherein the image segmentation set of the previous meteorological image and the image segments in the image segmentation set of the meteorological image to be detected correspond to each other one by one; Calculate the differences between two image blocks that correspond to each other one by one, and obtain multiple block difference results; Averaging the plurality of block difference results to obtain a mean difference result, and if the mean difference result is greater than the change threshold, determining that the change area of ​​the meteorological image to be detected compared with the previous meteorological image is greater than the change threshold; When sending the changed region encoding result to the receiving end, or sending the abnormal encoding result to the receiving end, a token bucket algorithm or a leaky bucket algorithm is used to control the bandwidth; The detecting whether the meteorological image to be detected contains an abnormal detection result includes: Acquire a meteorological image to be detected and branch prompt information, wherein the branch prompt information is used to indicate a detection branch of the meteorological image to be detected and the number of branches corresponding to the detection branch; Acquire a branch meteorological image matching a detection branch of a corresponding number of branches from at least one candidate meteorological image according to the branch prompt information, and generate a target meteorological image according to the branch meteorological image and the meteorological image to be detected; Through the debugged meteorological image detection network, abnormal meteorological detection is performed on the target meteorological image according to the detection branch indicated by the branch prompt information, and the abnormal detection result of the target meteorological image under the detection branch is obtained.

3. The method according to claim 2, characterized in that The meteorological image detection network completed through debugging performs abnormal meteorological detection on the target meteorological image according to the detection branch indicated by the branch prompt information, and obtains the abnormal detection result of the target meteorological image under the detection branch, including: The meteorological image detection network that has been debugged performs abnormal meteorological detection on the target meteorological image according to the detection branch indicated by the branch prompt information to obtain a first abnormal detection result; wherein the first abnormal detection result includes the element classification corresponding to the maximum support coefficient of each image element in the target meteorological image, and any support coefficient is used to indicate the probability that the element classification of any image element belongs to the corresponding element classification; Based on the element classification corresponding to the maximum support coefficient of each image element in the target meteorological image, detecting and obtaining the abnormality detection result of the meteorological image to be detected in the target meteorological image under the detection branch; Alternatively, the meteorological image detection network completed through debugging performs abnormal meteorological detection on the target meteorological image according to the detection branch indicated by the branch prompt information, and obtains the abnormal detection result of the target meteorological image under the detection branch, including: The meteorological image detection network completed through debugging performs abnormal meteorological detection on the target meteorological image according to the target detection branch to obtain a second abnormal detection result; the second abnormal detection result includes the element classification included in the element classification trajectory corresponding to the maximum trajectory support coefficient; wherein any element classification trajectory includes the support coefficient of the element classification of each image element of the target meteorological image under the corresponding element classification trajectory, and the conversion support coefficient of the element classification corresponding to the adjacent image elements; the trajectory support coefficient of any element classification trajectory is determined based on the support coefficient and the conversion support coefficient corresponding to the any element classification trajectory; Based on the element classification included in the element classification trajectory corresponding to the maximum trajectory support coefficient, an abnormality detection result of the to-be-detected meteorological image in the target meteorological image under the detection branch is detected.

4. The method according to claim 3, characterized in that The meteorological image detection network is debugged by the following steps: Acquire a meteorological training image and a target detection branch of the meteorological training image; the meteorological training image is matched with at least one priori mark, and any priori mark is used to indicate an abnormal meteorological detection result of the meteorological training image under a corresponding detection branch; Performing abnormal meteorological detection on the meteorological training image according to the target detection branch to obtain an abnormal meteorological detection result of the meteorological training image under the target detection branch; Determine a target priori marker from at least one priori marker matched by the meteorological training image, and determine an abnormal meteorological detection result of the meteorological training image under the target detection branch from the target priori marker, wherein the target priori marker is a priori marker that corresponds to a detection branch and is the same as the target detection branch; According to the abnormal meteorological detection result of the meteorological training image under the target detection branch and the abnormal meteorological detection result under the target detection branch, a network debugging operation is performed on the meteorological image detection network; the debugged meteorological image detection network is used to perform abnormal meteorological detection of meteorological images under the corresponding detection branch; The number of the target detection branches is at least one, and the number of abnormal meteorological detection results obtained by performing abnormal meteorological detection on the meteorological training image according to the target detection branches is the same as the number of the target detection branches; the number of detection branches corresponding to the priori marks matched by the meteorological training image is not less than the number corresponding to the target detection branches; wherein, when the number of the target detection branches is multiple, the number of abnormal meteorological detection results obtained is also multiple, and the number of detection branches corresponding to the target priori marks determined according to the priori marks matched by the meteorological training image is equal to the number of the target detection branches; the meteorological image detection network includes a representation information extraction component and an abnormal reasoning network; the representation information extraction component is used to extract meteorological image representation vectors; the abnormal reasoning network is used to perform abnormal meteorological detection on the meteorological training image to obtain corresponding abnormal meteorological detection results; The method for acquiring the meteorological training image includes: Acquire at least one candidate meteorological image and branch prompt information; the branch prompt information is used to prompt the number of branches of the target detection branch; Determine, according to the branch prompt information, a candidate meteorological image that matches the target detection branch of the corresponding number of branches in the at least one candidate meteorological image, and use the determined candidate meteorological image as a branch meteorological image of the meteorological training image; A sample meteorological image is obtained, and the branch meteorological image and the sample meteorological image are combined to obtain a meteorological training image.

5. The method according to claim 4, characterized in that The step of performing abnormal meteorological detection on the meteorological training image according to the target detection branch to obtain an abnormal meteorological detection result of the meteorological training image under the target detection branch includes: Extracting representation information of the meteorological training image according to the target detection branch to obtain a meteorological image representation vector corresponding to the meteorological training image; the meteorological image representation vector is used to represent feature information of image elements in the meteorological training image and the relationship between different image elements; Abnormal meteorological detection is performed on the meteorological training image based on the meteorological image characterization vector to obtain an abnormal meteorological detection result of the meteorological training image under the target detection branch.

6. The method according to claim 5, characterized in that Before performing abnormal meteorological detection on the meteorological training image based on the meteorological image representation vector and obtaining the abnormal meteorological detection result of the meteorological training image under the target detection branch, the method further includes: Obtaining the meteorological image type corresponding to the meteorological image to which each image element in the meteorological training image belongs; Generate a shielding matrix based on the meteorological image type; the shielding matrix includes a plurality of shielding elements, and the image elements included in the meteorological training image correspond to the shielding elements in the shielding matrix; Based on the shielding matrix, a shielding operation is performed on the meteorological image representation vector of the meteorological training image to obtain a shielded meteorological image representation vector; the shielded meteorological image representation vector includes an element representation vector corresponding to each image element in the sample meteorological image, and the shielded meteorological image representation vector is used to perform abnormal meteorological detection on the meteorological training image; The performing abnormal meteorological detection on the meteorological training image based on the meteorological image representation vector to obtain the abnormal meteorological detection result of the meteorological training image under the target detection branch includes: Based on the meteorological image representation vector, the abnormal classification prediction information of each image element in the meteorological training image is inferred; the abnormal classification prediction information of any image element represents the probability that the element classification of the corresponding image element belongs to the element classification of abnormal meteorology under the target detection branch; Based on the abnormal classification prediction information corresponding to each image element in the meteorological training image, an abnormal meteorological detection result of the meteorological training image under the target detection branch is generated.

7. The method according to claim 6, characterized in that The element classification of abnormal weather under the target detection branch includes at least one, and the abnormal classification prediction information of any image element includes: the support coefficient of the element classification of the corresponding image element belonging to different element classifications, and any support coefficient is used to indicate the probability that the element classification of the corresponding image element belongs to the corresponding element classification; the abnormal weather detection result of the meteorological training image under the target detection branch is generated based on the abnormal classification prediction information corresponding to each image element in the meteorological training image, including: Determine a maximum support coefficient among the support coefficients included in the abnormal classification prediction information of any image element; classifying the maximum support coefficient and the elements corresponding to the maximum support coefficient as target prediction information corresponding to any one of the image elements; According to the target prediction information of each image element in the meteorological training image, an abnormal meteorological detection result of the meteorological training image under the target detection branch is generated.

8. The method according to claim 6, characterized in that The element classification of abnormal weather under the target detection branch includes at least one, and the abnormal classification prediction information is also used to indicate the probability that the element classification of the corresponding image element belongs to a different element classification, and the probability is expressed based on the conversion support coefficient of the adjacent image element; The abnormal classification prediction information includes: the conversion support coefficient and the support coefficient of the element classification of the corresponding image element belonging to different element classifications; any support coefficient is used to indicate the probability that the element classification of the corresponding image element belongs to the corresponding element classification; the abnormal classification prediction information corresponding to each image element in the meteorological training image is used to generate the abnormal meteorological detection result of the meteorological training image under the target detection branch, including: At least one element classification trajectory corresponding to the meteorological training image is generated according to the abnormal classification prediction information; any element classification trajectory includes the element classification of each image element in the meteorological training image; any element classification trajectory includes the support coefficient of the element classification of each image element in the meteorological training image under the corresponding element classification trajectory, and the conversion support coefficient of the adjacent image elements; Determining a trajectory support coefficient of any element classification trajectory based on a support coefficient and a conversion support coefficient corresponding to any element classification trajectory; An element classification trajectory corresponding to a maximum trajectory support coefficient is determined in the at least one element classification trajectory, and an abnormal meteorological detection result of the meteorological training image under the target detection branch is generated based on the element classification included in the element classification trajectory corresponding to the maximum trajectory support coefficient.

9. The method according to claim 4, characterized in that The network debugging operation is performed on the meteorological image detection network according to the abnormal meteorological detection result of the meteorological training image under the target detection branch and the abnormal meteorological detection result under the target detection branch, including: Generate a target training cost under the target detection branch according to an abnormal meteorological detection result of the meteorological training image under the target detection branch and an error between the abnormal meteorological detection results under the target detection branch; According to the goal of reducing the target training cost, the parameter variables of the meteorological image detection network are iterated to perform a network debugging operation on the meteorological image detection network.

10. A remote monitoring device, comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, characterized in that: When the processor executes the program, the steps in the method according to any one of claims 1 to 9 are implemented.

Images