Rainfall monitoring method, system, storage medium and electronic device

Abstract

The application is suitable for the technical field of rainfall monitoring, and provides a rainfall monitoring method, system, storage medium and electronic equipment. The rainfall monitoring method provided by the application acquires a rainfall event stream collected by a dynamic visual sensor, reconstructs a rainfall image based on the rainfall event stream, identifies the rainfall intensity corresponding to the rainfall occurring in the rainfall image, and generates an alarm prompt information according to the rainfall intensity and the rainfall occurrence period when the rainfall intensity meets the alarm condition, so as to solve the technical problem that the existing rainfall monitoring method cannot accurately and timely monitor rainfall data to meet the demand of rainfall monitoring and early warning.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of rainfall monitoring, and more particularly to a rainfall monitoring method, system, storage medium and electronic device. BACKGROUND

[0002] In the existing rainfall monitoring technology, ground rain stations, radars and satellite remote sensing rain measurement and rainfall measurement technology based on traditional cameras are generally used.

[0003] However, the above existing methods have some problems, which cannot accurately and timely monitor rainfall data to meet the needs of rainfall monitoring and early warning, for example, the ground rain station observation points are unevenly distributed and limited in number, mostly offline operation, high construction and maintenance cost, and it is difficult to obtain real-time collaborative and high spatial resolution rainfall data. The hardware facilities and inversion rainfall intensity of radars and satellite remote sensing rain measurement have deficiencies, the rain measurement accuracy needs to be improved, and it is not sensitive to heavy rain, and real-time accurate data cannot be provided. The imaging effect of the rainfall measurement technology based on traditional cameras decreases when the background changes, it is difficult to capture raindrops in a strong rain white fog environment, and the data volume is large and difficult to process in real time online, resulting in delayed result transmission. SUMMARY

[0004] The purpose of the embodiments of the present application is to provide a rainfall monitoring method, system, storage medium and electronic device, which aims to solve the technical problem that the existing rainfall monitoring method cannot accurately and timely monitor rainfall data to meet the needs of rainfall monitoring and early warning.

[0005] To achieve the above-mentioned purpose, according to the first aspect of the present application, a rainfall monitoring method is provided, which comprises:

[0006] In response to obtaining the rainfall event stream collected by the dynamic vision sensor, a rainfall image is reconstructed based on the rainfall event stream;

[0007] Identify the rainfall intensity corresponding to the rainfall occurring in the rainfall image;

[0008] If the rainfall intensity meets the alarm condition, generate an alarm prompt information according to the rainfall intensity and the rainfall occurrence period.

[0009] Optionally, in a possible implementation manner of the first aspect, the rainfall image is reconstructed based on the rainfall event stream, comprising:

[0010] Obtain the event parameters of each rainfall event in the rainfall event stream, wherein the event parameters include: pixel coordinates, illumination change polarity and timestamp;

[0011] determine a change process of each of the rainfall events over time according to a mapping relationship between the timestamp and the pixel coordinates, the illumination change polarity;

[0012] reconstruct the rainfall image based on the change processes of all the rainfall events over time.

[0013] Optionally, in a possible implementation manner of the first aspect, the identifying the rainfall intensity corresponding to the rainfall occurring in the rainfall image comprises:

[0014] obtaining feature information contained in the rainfall image, wherein the feature information at least comprises at least one of the following: background feature, noise feature and raindrop layer feature;

[0015] inputting the feature information corresponding to the rainfall image into a pre-trained rainfall identification model to obtain the rainfall intensity output by the rainfall identification model, wherein the rainfall identification model is obtained by training an initial neural network model by using a plurality of sets of training data, and each set of the training data comprises sample feature information of a sample rainfall image and a sample rainfall intensity corresponding to the sample feature information.

[0016] Optionally, in a possible implementation manner of the first aspect, the rainfall identification model performs the following processing on the feature information of the rainfall image to obtain the rainfall intensity:

[0017] extracting a raindrop part in the rainfall image based on the feature information of the rainfall image;

[0018] calculating raindrop density, raindrop average particle size and raindrop terminal moving speed of the raindrop part in the rainfall image;

[0019] calculating the rainfall intensity according to the raindrop density, the raindrop average particle size and the raindrop terminal moving speed.

[0020] Optionally, in a possible implementation manner of the first aspect, the extracting the raindrop part in the rainfall image based on the feature information of the rainfall image comprises:

[0021] extracting an approximate raindrop part in the rainfall image based on the feature information of the rainfall image by using an edge detection algorithm;

[0022] calculating image gradient values of the approximate raindrop part in a horizontal direction and a vertical direction by using an image gradient calculation algorithm;

[0023] performing non-maximum suppression processing on the image gradient values to determine edge pixels of the approximate raindrop part;

[0024] According to edge pixels of the approximate raindrop part, a raindrop part in the rainfall image is determined.

[0025] Optionally, in a possible implementation manner of the first aspect, according to the edge pixels of the approximate raindrop part, the raindrop part in the rainfall image is determined, including:

[0026] In a case where the edge pixel is greater than a first edge threshold, it is determined to retain the edge pixel;

[0027] In a case where the edge pixel is less than a second edge threshold, it is determined to eliminate the edge pixel, wherein the first edge threshold is greater than the second edge threshold;

[0028] In a case where the edge pixel is less than the first edge threshold and greater than the second edge threshold, according to connection information between the edge pixel and a retained edge pixel, it is determined whether to eliminate the edge pixel;

[0029] According to the retained edge pixel, the raindrop part in the rainfall image is determined.

[0030] Optionally, in a possible implementation manner of the first aspect, the raindrop end moving speed of the raindrop part in the rainfall image is calculated, including:

[0031] A raindrop end position pixel point of the raindrop part in the rainfall image is determined.

[0032] According to the raindrop end position pixel point, a motion track, a motion time and a pixel displacement of the raindrop part are determined.

[0033] According to the motion track, the motion time and the pixel displacement of the raindrop part, the raindrop end moving speed of the raindrop part in the rainfall image is calculated.

[0034] Optionally, in a possible implementation manner of the first aspect, the method further includes:

[0035] Measurement information of a pluviometer arranged in a plurality of regions is acquired.

[0036] According to a change process of the measurement information, a change process of the rainfall intensity is determined.

[0037] Optionally, in a possible implementation manner of the first aspect, before the rainfall intensity of the rainfall in the rainfall image is identified, the method further includes:

[0038] The rainfall image is subjected to image preprocessing, wherein the image preprocessing includes at least one of the following: denoising processing and enhancement processing.

[0039] According to a second aspect of the present application, a rainfall monitoring system is provided, comprising:

[0040] a dynamic visual sensor arranged at an observation point, configured to collect rainfall events to obtain a rainfall event stream;

[0041] a processor connected to the plurality of dynamic visual sensors, configured to reconstruct a rainfall image based on the rainfall event stream, and identify a rainfall intensity corresponding to rainfall occurring in the rainfall image, and generate an alarm prompt information according to the rainfall intensity and a rainfall occurrence period when the rainfall intensity meets an alarm condition.

[0042] The second aspect and any one of the implementation manners of the second aspect correspond to the first aspect and any one of the implementation manners of the first aspect respectively. The technical effects corresponding to the second aspect and any one of the implementation manners of the second aspect can be referred to the technical effects corresponding to the first aspect and any one of the implementation manners of the first aspect, which will not be described herein again.

[0043] In a third aspect, an electronic device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the method according to any one of the aspects when executing the computer program.

[0044] In a fourth aspect, a computer readable storage medium is provided, and the computer readable storage medium stores a computer program, and the computer program is executable by a processor to implement the method according to any one of the aspects.

[0045] In a fifth aspect, a computer program product is provided, and when the computer program product is executed on an electronic device, the electronic device executes the method according to any one of the first aspect.

[0046] It can be understood that the beneficial effects of the second aspect to the fifth aspect can be referred to the related description of the first aspect, which will not be described herein again.

[0047] The rainfall monitoring method provided by the present application, by responding to the rainfall event stream collected by the dynamic visual sensor, reconstructing the rainfall image based on the rainfall event stream, identifying the rainfall intensity corresponding to the rainfall occurring in the rainfall image, and generating the alarm prompt information according to the rainfall intensity and the rainfall occurrence period when the rainfall intensity meets the alarm condition. To provide accurate rainfall information for relevant departments and personnel in a timely manner, to make up for the defects of offline operation data transmission of ground rain gauge and delay of result transmission caused by difficulty in real-time online processing of traditional camera data; solve the technical problem that the existing rainfall monitoring method cannot accurately and timely monitor rainfall data to meet the demand of rainfall monitoring and early warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the embodiments or descriptions of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0049] Figure 1 1 is a flow chart of a rainfall monitoring method provided in an embodiment of the present application;

[0050] Figure 2 1 is a flow chart of an optional rainfall monitoring method provided in an embodiment of the present application;

[0051] Figure 3 1 is a flow chart of an optional rainfall monitoring method provided in an embodiment of the present application;

[0052] Figure 4 1 is a flow chart of an optional rainfall monitoring method provided in an embodiment of the present application;

[0053] Figure 5 1 is a flow chart of an optional rainfall monitoring method provided in an embodiment of the present application;

[0054] Figure 6 Schematic diagram of the structure of the rainfall monitoring system provided in the embodiment of the present application;

[0055] Figure 7 This is a structural diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0056] In the following description, specific details such as specific system structures and techniques are provided for purposes of illustration rather than limitation to facilitate a thorough understanding of the embodiments of the present application. However, it will be apparent to those skilled in the art that the present application may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid obscuring the description of the present application with unnecessary detail.

[0057] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or collections thereof.

[0058] It should also be understood that the term “and / or” as used herein, means any one or more of the associated listed items, and includes all possible combinations of the associated listed items. It should also be understood that, in the description and claims of the application, relative terms such as “first”, “second”, “third”, etc., are used to distinguish one element from another, and are not necessarily intended to imply relative importance or order.

[0059] As used in the description and the appended claims of the application, the term “if’ can be interpreted as meaning “when”, or “once”, or “in response to a determination”, or “in response to a detection” depending on the context. Similarly, the phrase “if determined” or “if detected [the described condition or event]” can be interpreted as meaning “once determined” or “in response to a determination” or “once detected [the described condition or event]” or “in response to a detection [the described condition or event]” depending on the context.

[0060] In addition, in the description and claims of the application, the terms “first”, “second”, “third”, etc. are used merely to distinguish descriptions and are not intended to imply relative importance or order.

[0061] Reference in the specification to “one embodiment” or “some embodiments” means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the application. The appearances of the phrases “in one embodiment”, “in some embodiments”, “in other embodiments”, “in additional embodiments”, and so on, in various places in the specification are not necessarily all referring to the same embodiment, unless otherwise specifically stated. The terms “comprising”, “including”, “having” and their variants are meant to be construed as “including but not limited to”, unless otherwise specifically stated.

[0062] First, some terms in the embodiments of the application are explained to facilitate understanding by those skilled in the art.

[0063] Prewitt operator: an operator for image edge detection, which detects edges by calculating the gradient of pixels in an image, using a specific template to convolve the image to determine the gradient components in horizontal and vertical directions, and then determine the edge position.

[0064] Laplacian operator: an edge detection operator that detects edges by calculating the second derivative of an image, which is sensitive to changes in gray scale in the image and can detect abrupt changes and edges in the image.

[0065] Lucas-Kanade (LK) optical flow method: a method for calculating the motion of pixels in an image, which assumes that the motion of pixels is consistent within a small local area, and determines the motion vector of the pixels by solving the optical flow equation.

[0066] Horn-Schunck optical flow algorithm: a method for calculating optical flow by minimizing the smoothness constraint and brightness constancy constraint of optical flow, which can obtain more accurate optical flow estimation.

[0067] PCA-Flow algorithm: a principal component analysis optical flow algorithm that uses principal component analysis to model and estimate optical flow, which can improve the accuracy and efficiency of optical flow calculation to some extent.

[0068] LeNet-5 convolutional neural network model: an early convolutional neural network model mainly used for tasks such as handwritten digit recognition, with structures such as convolutional layers, pooling layers and fully connected layers.

[0069] Visual Geometry Group (VGG) series model: a group of deep convolutional neural network models known for their simple structure and good performance, usually with a deep network layer and small size convolution kernel.

[0070] Residual Network (ResNet) series model, which solves the gradient vanishing problem in deep neural network training by introducing residual connection, can train very deep network and achieve excellent performance.

[0071] The above is a brief introduction to the terms involved in the embodiments of the present application, which will not be repeated hereinafter.

[0072] In existing rainfall monitoring technology, ground rain stations, radars, satellite remote sensing rain measurement and traditional camera-based rainfall measurement technology are generally used.

[0073] First, ground rain stations are devices used to measure precipitation, usually placed on the ground to monitor and record the amount of rain. Rain stations are composed of various types of rain gauges, such as float-type, tipping-bucket or pressure-type rain gauges. They receive rain falling into the collector and record it, then transmit the data to data centers or monitoring centers for analysis and use. However, the observation points of ground rain stations are unevenly distributed and limited in number, and most of them are offline, and the cost of building and maintaining high-density rain gauges is high, so it is difficult to obtain real-time collaborative and high spatial resolution rainfall data.

[0074] Second, radar rain measurement is a method of measuring precipitation using radar technology. The radar system transmits radio waves, and when the radio waves encounter raindrops, part of the energy is reflected back to the radar receiver to form an echo. The intensity of the echo and the reflectivity are used to estimate the amount of precipitation. Satellite remote sensing rain measurement uses satellite-borne remote sensing sensors to estimate precipitation. The satellite sensor receives microwave or infrared radiation signals reflected by the ground and atmosphere, processes and analyzes the received signals to infer cloud characteristics and precipitation probability. However, radar and satellite remote sensing rain measurement has limitations in hardware facilities and inversion of rainfall intensity. The accuracy of rain measurement needs to be further improved, and the observation of heavy rain is not sensitive enough to provide real-time and accurate rainfall data.

[0075] Third, with the development of optical technology, rain measurement methods based on optical sensors have emerged and gradually developed. These methods mainly use traditional frame cameras and other optical detection technologies to observe precipitation. However, when the background is cluttered and changes, the imaging effect of raindrops based on traditional frame cameras decreases, leading to a decrease in algorithm effectiveness. In the white fog environment caused by heavy rain, it is difficult to capture raindrops. Moreover, the data volume of traditional frame cameras is large, and the data is difficult to process in real time, resulting in delayed transmission of results.

[0076] As can be seen, the above existing methods have some problems, which result in the inability to accurately and real-time monitor rainfall data to meet the demand for rainfall monitoring and early warning.

[0077] The present application provides an example of a rainfall monitoring method. Please refer to Figure 1 Figure 1 The present application provides a schematic flowchart of a rainfall monitoring method. As an example, but not limited to, this method can be applied to a rainfall monitoring system.

[0078] S101, in response to obtaining the rainfall event stream collected by the dynamic vision sensor, reconstructing the rainfall image based on the rainfall event stream.

[0079] S102, identifying the rainfall intensity corresponding to the rainfall occurring in the rainfall image.

[0080] S103, in the case where the rainfall intensity meets the alarm condition, generating an alarm prompt information according to the rainfall intensity and the rainfall occurrence period.

[0081] ​In the examples of the present application, the rainfall monitoring system is distributed at multiple observation points in a geographical area, for example, the rainfall monitoring system can be installed in a location prone to heavy rainfall or weak flood control to monitor the rainfall status of the city online. In specific use scenarios, the rainfall events at the observation point are collected by the dynamic visual sensor in the rainfall monitoring system, solving the problem of uneven distribution and limited number of observation points of ground rain stations, and enabling more comprehensive coverage of the monitoring area and richer rainfall data to improve the spatial resolution.

[0082] Optionally, the dynamic visual sensor can also be understood as an event camera, which can be a traditional camera or a hybrid camera combining a traditional camera and a frame camera.

[0083] Secondly, for example, a dynamic visual analysis algorithm can be used to reconstruct a rainfall image (rainfall process image) based on a rainfall event stream and identify the rainfall intensity in the rainfall image. Compared with radar and satellite remote sensing and traditional camera rain measurement methods, the accuracy of rainfall intensity is improved, avoiding the shortcomings of radar and satellite in hardware facilities and rainfall intensity inversion, and the problem of declining imaging effect of traditional cameras in a cluttered background and strong rainfall white fog environment.

[0084] Finally, when the rainfall intensity meets the alarm condition, an alarm prompt message is generated and sent to the client to provide accurate rainfall information to relevant departments and personnel in a timely manner, realize real-time early warning, and make up for the defects of ground rain station offline operation data transmission not timely and traditional camera data difficult to be processed online in real time, resulting in delayed result transmission.

[0085] In one possible implementation manner, please refer to Figure 2 as shown, Figure 2 a schematic flowchart of a rainfall monitoring method provided by the present application is shown, which reconstructs a rainfall image based on a rainfall event stream, including:

[0086] S201, obtaining event parameters of each rainfall event in the rainfall event stream, wherein the event parameters include: pixel coordinates, illumination change polarity and timestamp.

[0087] S202, determining the change process of each rainfall event over time according to the mapping relationship between the timestamp and the pixel coordinates and the illumination change polarity.

[0088] S203, reconstructing a rainfall image based on the change process of all rainfall events over time.

[0089] In an optional implementation manner, the pixel coordinates, illumination change polarity and timestamp of each rainfall event in the rainfall event stream are obtained to accurately obtain the position and time information of each rainfall event.

[0090] Specifically, the pixel coordinates explicitly indicate the location information of each rainfall event in the image, and obtaining the pixel coordinates can accurately track the distribution of raindrops at different observation points. The illumination change polarity reflects the influence of raindrops on light. Since the presence of raindrops changes the propagation path of light, causing changes in illumination, by recording the illumination change polarity, raindrops can be further distinguished from other objects in their different effects on light. The timestamp records the time point of each rainfall event, providing time dimension information for analyzing the dynamic changes of rainfall.

[0091] According to the mapping relationship between the timestamp and the pixel coordinates, and the timestamp and the illumination change polarity, the change process of each rainfall event over time can be determined. It should be understood that according to the mapping relationship between the timestamp and the pixel coordinates, and the timestamp and the illumination change polarity, different event parameters can be linked to construct the change process or trajectory of each raindrop over time. Specifically, the higher the rainfall intensity, the greater the amplitude of event change, and the number of rainfall events has a certain positive correlation with the amplitude of time change and rainfall intensity.

[0092] For example, through the correspondence between the timestamp and the pixel coordinates, the position change of raindrops at different time points can be determined; combined with the illumination change polarity, the degree to which the raindrops affect the light during the movement process can be understood.

[0093] Since this change process can detail the entire process of each raindrop from appearance to disappearance, it provides rich dynamic information for reconstructing the rainfall image. Further, by dynamically tracking the development trajectory of the rainfall event, the dynamic change process of the rainfall can be more accurately understood, overcoming the problem of decreased imaging effect of traditional cameras when the background changes chaotically, and the difficulty in capturing raindrops in a white fog environment caused by heavy rainfall.

[0094] Finally, the rainfall image is reconstructed based on the change process of all rainfall events, specifically by integrating the change processes of each raindrop to form a complete rainfall scene image. By comprehensively considering the position, illumination change, and time change of all raindrops, a more realistic and accurate rainfall image is restored, so that the rainfall information collected by the integrated dynamic vision sensor can achieve comprehensive rainfall monitoring of the observation point or monitoring point, making up for the uneven distribution and limited number of ground rain gauges, and also solving the problem of result transmission delay caused by the large amount of data of traditional frame cameras difficult to process in real time online, thereby improving the accuracy and real-time performance of rainfall monitoring.

[0095] In a possible implementation, before identifying the rainfall intensity of the rainfall in the rainfall image, the method further includes: performing image preprocessing on the rainfall image, where the image preprocessing includes at least one of the following: denoising processing and enhancement processing.

[0096] In actual acquisition, due to various factors, the rainfall image may contain some noise, such as sensor noise, environmental noise, and the like. These noises will affect the subsequent accurate identification of the rainfall intensity. Through denoising processing, the noise interference in the rainfall image can be removed, the image is made clearer, the raindrop part is highlighted, and the extraction accuracy of the rainfall feature is improved.

[0097] In some cases, the rainfall image may appear relatively dark or blurred due to poor lighting conditions or other reasons. Since the enhancement processing can enhance the contrast and brightness of the rainfall image and make the raindrops more obvious, subsequent analysis and identification are facilitated. Meanwhile, the enhancement processing can also improve the quality of the image, making the details more abundant, which is helpful for more accurately calculating the raindrop density, average particle size, terminal moving speed, and the like.

[0098] Compared with the prior art, for the problem of decreased imaging effect of a traditional camera when the background is cluttered and changes, the denoising processing and the enhancement processing can improve the image quality, so that the raindrops are more easily identified in a complex background. For the problem of low accuracy of radar and satellite remote sensing rain measurement, the preprocessed rainfall image can provide more accurate input for the rainfall identification model, thereby improving the identification accuracy of the rainfall intensity. For the problem that it is difficult for a ground rain gauge station to obtain high spatial resolution rainfall data, by preprocessing the rainfall image collected by the dynamic visual sensor, information of different observation points can be better integrated, and the spatial resolution and the comprehensiveness of monitoring can be improved.

[0099] In a possible implementation, identifying the rainfall intensity corresponding to the rainfall occurring in the rainfall image includes:

[0100] Obtaining feature information contained in the rainfall image, where the feature information at least includes at least one of the following: background features, noise features, and raindrop layer features;

[0101] Inputting the feature information corresponding to the rainfall image into a pre-trained rainfall identification model to obtain the rainfall intensity output by the rainfall identification model.

[0102] The rainfall identification model is obtained by training an initial neural network model using multiple sets of training data, and each set of training data includes: sample feature information of a sample rainfall image, and a sample rainfall intensity corresponding to the sample feature information.

[0103] In the above optional implementation, first, it is of great significance to obtain various feature information such as background features, noise features and raindrop layer features contained in the rainfall image. For example, the background features can help to distinguish different environmental scenes, such as cities, mountains or plains, etc., and different backgrounds may have an impact on the performance and monitoring of rainfall. The extraction of noise features helps to exclude other interference factors in the environment and improve the accuracy of rainfall monitoring. The raindrop layer features directly reflect the state of rainfall, including the size, density and distribution of raindrops, etc.

[0104] Then, the feature information is input into a pre-trained rainfall recognition model. The rainfall recognition model is pre-trained through a large number of training data. In the pre-training process, each set of training data contains sample feature information of a sample rainfall image and a sample rainfall intensity corresponding to the sample feature information. By continuously adjusting the parameters of the model, the model can learn the complex mapping relationship between different feature combinations and a specific rainfall intensity.

[0105] Analysis shows that compared with traditional radar rainfall measurement, this method is not limited by hardware facilities and can more flexibly adapt to different monitoring environments. Radar rainfall measurement may be disturbed by various factors when inverting rainfall intensity, while the method based on neural network model can consider various features and reduce the influence of interference factors to improve accuracy.

[0106] Compared with satellite remote sensing rainfall measurement, this method can provide higher spatial resolution. Although satellite remote sensing can cover a larger range, it may not be fine enough for local rainfall monitoring. By using dynamic visual sensors, more detailed local rainfall information can be obtained, and combined with neural network model analysis, the rainfall intensity of different regions can be accurately identified.

[0107] Compared with ground rain stations, this method overcomes the problem of uneven distribution and limited number of observation points. Ground rain stations have high construction and maintenance costs and are difficult to realize real-time collaborative monitoring. By using dynamic visual sensors and neural network models, a large amount of data can be quickly processed to realize real-time monitoring and accurate early warning, providing stronger support for flood control and disaster reduction work.

[0108] In one possible implementation, please refer to Figure 3 It is shown that Figure 3 A schematic flow chart of a rainfall monitoring method provided by the present application is shown, and the rainfall recognition model processes the feature information of the rainfall image as follows to obtain the rainfall intensity:

[0109] S301, based on the feature information of the rainfall image, extracting the raindrop part in the rainfall image;

[0110] S302, calculating the raindrop density, average raindrop particle size, and final moving speed of the raindrop portion in the rainfall image;

[0111] S303: Calculate the rainfall intensity based on the raindrop density, the average raindrop particle size, and the raindrop terminal moving speed.

[0112] Optionally, in the example of this application, the rainfall recognition model can be a LeNet-5 convolutional neural network model, or a VGG series model, a ResNet series model, etc.

[0113] In this implementation, the rainfall recognition model first extracts raindrops from rainfall images using feature information. This step overcomes the difficulty traditional cameras have in accurately capturing raindrops in environments with cluttered and changing backgrounds and heavy rainfall and white mist. By leveraging multiple feature information, raindrops can be more accurately separated, improving rainfall recognition capabilities.

[0114] Next, the rainfall recognition model calculates the raindrop density, average droplet size, and final droplet velocity within the rainfall image. It should be understood that these three parameters can reflect rainfall characteristics from different perspectives. For example, raindrop density can reflect the intensity and range of rainfall; a higher density generally indicates heavier rainfall; average droplet size reflects the size distribution of raindrops, which is important for determining rainfall type and intensity; and final droplet velocity reflects the dynamic changes in rainfall and the influence of factors such as wind.

[0115] Finally, the rainfall identification model calculates rainfall intensity based on the calculated raindrop density, average raindrop size, and final movement speed. This calculation method, which integrates multiple parameters, is more accurate and reliable than traditional single-parameter measurement methods. Compared with ground-based rain gauges, this method does not require a large number of observation points and high construction and maintenance costs. It can obtain a wider range of rainfall data through dynamic visual sensors and accurately calculate rainfall intensity through model calculations. Compared with radar and satellite remote sensing rainfall measurements, this method has significantly improved accuracy, avoiding deficiencies in hardware facilities and rainfall intensity inversion. It is also more sensitive to heavy rainfall observations and can provide real-time and accurate rainfall data.

[0116] For a possible implementation, please refer to Figure 4 As shown, Figure 4 A schematic flow chart of a rainfall monitoring method provided by the present application is shown, which extracts raindrops from a rainfall image based on feature information of the rainfall image, including:

[0117] S401 , using an edge detection algorithm to extract an approximate raindrop portion in the rainfall image based on feature information of the rainfall image.

[0118] S402, calculate image gradient values of the approximate raindrop part in the horizontal direction and the vertical direction by using an image gradient calculation algorithm.

[0119] S403, perform non-maximum suppression on the image gradient values to determine edge pixels of the approximate raindrop part.

[0120] S404, determine the raindrop part in the rain image according to the edge pixels of the approximate raindrop part.

[0121] Optionally, the edge detection algorithm in the examples of the present application can be a Canny edge detection algorithm for images, and can also be a Prewitt operator, a Laplacian operator, or other edge detection methods; and the image gradient calculation algorithm in the examples of the present application can be a Sobel operator.

[0122] In addition, the Lucas-Kanade (LK) optical flow method can also be used in the examples of the present application to track the pixel points of the end position of the raindrops in the picture, track the motion trajectory, motion time and pixel displacement of the raindrops, and calculate the moving speed of the raindrop end. The Lucas-Kanade (LK) optical flow method used in the examples of the present application can also be replaced by a Horn-Schunck optical flow algorithm, a PCA-Flow algorithm, etc.

[0123] In the above optional implementation, the raindrop part in the rain image can be effectively extracted by the raindrop recognition model through the following steps. Specifically, first, an edge detection algorithm is used to extract an approximate raindrop part based on feature information of the rain image. The edge detection algorithm can quickly locate the edge contour of an object according to the changes in the features of the image, such as color, brightness, etc. In the rain image, the possible raindrop area can be more accurately found by using information such as background features, noise features and raindrop layer features, thereby overcoming the problem that a traditional camera is difficult to capture raindrops in a complex environment.

[0124] Then, an image gradient calculation algorithm is used to calculate image gradient values of the approximate raindrop part in the horizontal direction and the vertical direction. Since the image gradient reflects the degree of change of the pixel value in the image, it is very important for determining the edge of an object. Furthermore, by calculating the gradient values, the edge information of the raindrops can be further refined, and the accuracy of raindrop extraction can be improved.

[0125] Then, non-maximum suppression is performed on the image gradient values to determine edge pixels of the approximate raindrop part. Non-maximum suppression can remove those pixel points that are not local maximum values, thereby retaining the most obvious edge pixels, and can remove some fuzzy edges and noise, so that the edge of the raindrop is more clear and explicit.

[0126] Finally, the raindrops in the rainfall image are determined based on the edge pixels that approximate the raindrops. Through the above series of processing, raindrops can be accurately extracted from the rainfall image, providing a reliable basis for the subsequent calculation of parameters such as raindrop density, average particle size, and final moving speed.

[0127] Compared with traditional rainfall monitoring technologies, ground rain gauges are limited by the distribution and number of observation points, making it difficult to monitor rainfall comprehensively and accurately. Radar and satellite remote sensing rainfall detection is insufficient in accuracy and sensitivity to heavy rainfall. This method, based on image feature information and a series of algorithm processing, can extract raindrop information in real time and accurately, providing strong support for providing real-time and accurate rainfall data and alarm prompt information.

[0128] For a possible implementation, please refer to Figure 5 As shown, Figure 5 A schematic flow chart of a rainfall monitoring method provided by the present application is shown, which determines the raindrop portion in a rainfall image based on edge pixels approximating the raindrop portion, including:

[0129] S501 : When the edge pixel is greater than a first edge threshold, determine to retain the edge pixel.

[0130] S502 : When the edge pixel is smaller than the second edge threshold, determine to remove the edge pixel, wherein the first edge threshold is greater than the second edge threshold.

[0131] S503 : When the edge pixel is smaller than the first edge threshold and larger than the second edge threshold, determine whether to remove the edge pixel according to connection information between the edge pixel and the retained edge pixels.

[0132] S504: Determine the raindrop portion in the rainfall image based on the retained edge pixels.

[0133] In this implementation, a pre-set first and second edge thresholds are first determined to classify edge pixels that approximate raindrops. If an edge pixel's value exceeds the first threshold, it is retained. This allows for rapid screening of distinct raindrop edges, ensuring accurate identification of larger, clearer raindrops.

[0134] Then, when the edge pixel is less than the second edge threshold, it is determined to be removed. Since the second edge threshold is relatively low, edge pixels less than this threshold are likely to be noise or unclear parts. Removing these pixels can reduce interference and improve the accuracy of raindrop extraction.

[0135] In addition, for the case where the edge pixel is less than the first edge threshold and greater than the second edge threshold, it is determined whether to reject according to the connection information between the edge pixel and the retained edge pixel. This processing mode fully considers the uncertainty of the edge pixel, and through connection analysis with the determined reliable edge pixel, it can further judge whether the edge pixel in the intermediate range belongs to the part of the raindrop. For example, if the edge pixel has a good connection relationship with the retained edge pixel, it indicates that the edge pixel is likely to be the edge of the raindrop, and thus the edge pixel is retained; if the edge pixel has no connection relationship with the retained edge pixel, the edge pixel is rejected.

[0136] Finally, the raindrop part in the rainfall image is determined according to the retained edge pixel. After the above processing process, the true raindrop part can be accurately extracted from the approximate raindrop part, so as to provide a data basis for subsequent calculation of rainfall intensity and other parameters.

[0137] Compared with the traditional rainfall monitoring technology, this method based on edge pixel classification and connection analysis is more detailed and accurate in processing the edge of the raindrop, can accurately extract the raindrop in a complex rainfall image, improves the accuracy and reliability of the rainfall intensity calculation, and provides strong support for real-time and accurate rainfall monitoring and early warning.

[0138] In one possible implementation manner, the raindrop end moving speed of the raindrop part in the rainfall image is calculated, including:

[0139] The raindrop end position pixel point of the raindrop part in the rainfall image is determined.

[0140] The motion trajectory, motion time and pixel displacement of the raindrop part are determined according to the raindrop end position pixel point.

[0141] The raindrop end moving speed of the raindrop part in the rainfall image is calculated according to the motion trajectory, motion time and pixel displacement of the raindrop part.

[0142] In this implementation manner, the raindrop end moving speed of the raindrop part in the rainfall image is accurately calculated through the following steps, effectively solving the problems in the prior art and achieving good technical effects.

[0143] Firstly, the raindrop end position pixel point of the raindrop part in the rainfall image is determined. This step can find the end position of the raindrop in the image by analyzing the edge pixel of the raindrop part. Since the raindrop forms a certain trajectory during falling, determining the end position pixel point provides key reference information for subsequent calculation of the motion trajectory, motion time and pixel displacement of the raindrop.

[0144] Then, the motion trajectory, motion time and pixel displacement of the raindrop part are determined according to the pixel point of the raindrop end position. Through the analysis of the rainfall images of multiple consecutive frames, the motion path of the raindrop from the starting position to the end position can be tracked to determine the motion trajectory. The pixel displacement is obtained by calculating the position change of the raindrop between different frames, reflecting the moving distance of the raindrop in the image.

[0145] Finally, the raindrop end moving speed of the raindrop part in the rainfall image is calculated according to the motion trajectory, motion time and pixel displacement of the raindrop part. The pixel displacement can be converted into actual physical displacement by using the relationship between time information and pixel displacement, and the raindrop end moving speed can be calculated by combining the time interval.

[0146] Compared with the traditional rainfall monitoring technology, this method can calculate the raindrop end moving speed in real time by analyzing the raindrop part in the rainfall image, which provides an important parameter for a more comprehensive understanding of the rainfall characteristics. This helps to improve the accuracy of rainfall intensity calculation and provides stronger support for timely and accurate rainfall monitoring and early warning.

[0147] In a possible implementation manner, measurement information of rain gauges distributed in multiple regions can also be acquired; and a change process of rainfall intensity can be determined according to a change process of the measurement information.

[0148] In this implementation manner, the rain gauges, as a kind of traditional rainfall measurement equipment, can directly measure the precipitation at a specific location. By setting rain gauges in multiple regions, actual rainfall data at different locations can be acquired, which makes up for the limitations of a single monitoring method.

[0149] Since the measurement information of the rain gauges usually includes data such as accumulated precipitation in a period of time, by analyzing the change of these data over time, the change trend of rainfall intensity can be inferred. For example, if the precipitation measured by the rain gauges increases rapidly in a short time, it indicates that the rainfall intensity is increasing; on the contrary, if the precipitation gradually decreases, it indicates that the rainfall intensity is weakening.

[0150] Compared with the prior art, for the problem of uneven distribution and limited number of ground rain gauges, by setting rain gauges in multiple regions, the density of monitoring points can be increased and the spatial resolution can be improved. At the same time, the measurement information of the rain gauges can be verified with other monitoring methods such as dynamic visual sensors, to improve the accuracy of rainfall intensity calculation. For the problems of low accuracy of radar and satellite remote sensing rain measurement and insensitivity to heavy rain, the field measurement data of the rain gauges can be used as a supplement to more accurately reflect the change of rainfall intensity in local areas. For the problem of large amount of data of traditional cameras that is difficult to process in real time, the measurement information of the rain gauges is relatively simple and easy to analyze in real time, which helps to determine the change process of rainfall intensity in a timely manner and provides more timely support for early warning and decision-making.

[0151] It should be understood that the size of the serial number of each step in the above embodiment does not mean the order of execution, and the execution order of each process should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0152] The rainfall monitoring method corresponding to the above embodiment, Figure 6 is a structural schematic diagram of a rainfall monitoring system provided by the embodiments of the present application, referring to Figure 6 , the rainfall monitoring system comprises:

[0153] The dynamic visual sensor 601 (the number of the dynamic visual sensor can be one or more, Figure 6 which is schematically shown in one), is arranged at an observation point, and is used to collect rainfall events to obtain a rainfall event stream.

[0154] The processor 602 is connected with the dynamic visual sensor, and is used to reconstruct a rainfall image based on the rainfall event stream, identify a rainfall intensity corresponding to rainfall occurring in the rainfall image, and generate an alarm prompt information according to the rainfall intensity and a rainfall occurring period when the rainfall intensity meets an alarm condition.

[0155] Optionally, the rainfall monitoring system is arranged at multiple different observation points in a geographical area, for example, the rainfall monitoring system can be installed at a position where heavy rainfall is prone to occur or flood control is weak, to perform online monitoring of the rainfall state of a city. In a specific use scenario, the dynamic visual sensor in the rainfall monitoring system collects rainfall events at the observation point, solves the problem of uneven distribution and limited number of ground rain gauge observation points, can more comprehensively cover the monitoring area, and obtain more abundant rainfall data, thereby improving the spatial resolution.

[0156] Secondly, the rainfall image is reconstructed based on the rainfall event stream and the rainfall intensity is identified, which has improved the accuracy of the rainfall intensity compared with radar and satellite remote sensing rain measurement and traditional camera rain measurement, avoids the deficiencies of radar and satellite in hardware facilities and rainfall intensity inversion, and the problem of decreased imaging effect of the traditional camera in a cluttered background and a strong rainfall white fog environment.

[0157] Finally, when the rainfall intensity meets the alarm condition, the alarm prompt information is generated and sent to the client, so as to provide accurate rainfall information for relevant departments and personnel in time, realize real-time early warning, and make up for the defects of the ground rain gauge offline operation data transmission not timely and the traditional camera data difficult to be processed in real time, resulting in delayed result transmission.

[0158] In summary, the rainfall monitoring system provided in the examples of the present application is realized based on dynamic vision, and through the cooperative work of the dynamic vision sensor and the processor, the common problems of the existing rainfall monitoring technology are effectively solved, and accurate, real-time and high-resolution rainfall monitoring can be realized.

[0159] In addition, in the examples of the present application, as shown in Figure 6 The rainfall monitoring system further includes a storage module 603, a power module 604 (such as a solar power module), a communication module 605 (such as a 4G communication module, a Beidou positioning and timing module), and a tipping bucket rain gauge 606. The rainfall monitoring system integrated with the modules is installed at a location prone to heavy rainfall or weak in flood control for online monitoring of the urban rainfall state.

[0160] By encapsulating the dynamic vision sensor 601, the processor 602 (the calculation module), the storage module 603, the power module 604 (such as a solar power module), and the communication module 605 (such as a 4G communication module, a Beidou positioning and timing module), a stable and reliable modular encapsulation model is built to realize the functions of data collection, calculation, analysis, and transmission. The dynamic vision sensor is responsible for sensing the change in light intensity, the calculation module is responsible for analyzing the event stream data to determine whether an event is rainfall and the rainfall grade, the storage module is responsible for storing the rainfall process event data, and the data is transmitted through the 4G communication module under normal monitoring conditions, and the key data information is transmitted through the Beidou positioning and timing module when a natural disaster such as heavy rainfall occurs. In addition, Figure 6 The tipping bucket rain gauge 606 in the examples of the present application is used to provide the actual rainfall intensity to provide the basis for error judgment of the monitoring results of the dynamic vision sensor.

[0161] An algorithm is written using event stream information to reconstruct the image. First, an optimization algorithm is established according to the image characteristics of the background, noise, and raindrop layer to preliminarily extract raindrop information, the input event information is classified and processed, and the classification of rainfall and non-rainfall is realized through methods such as optical flow estimation, feature extraction, and fusion. Secondly, the event data is denoised and enhanced to make the image restoration more real and clear, the low-contrast high-noise rainfall image is processed using the contrast enhancement technology and the denoising method, and the local contrast is enhanced to solve the problem of uneven illumination.

[0162] Then, the raindrop-related parameters are extracted using the restored rainfall image, the Canny edge detection algorithm is used to extract the raindrop part in the picture, the Sobel operator is used to calculate the gradient of the image, the non-maximum suppression is performed on the gradient image to thin the edges, the double-threshold strategy is used to adjust the threshold to detect and connect the edges, and the raindrop density and the average raindrop size are calculated according to the extracted raindrops; the Lucas-Kanade (LK) optical flow method is used to track the pixel points at the end of the raindrops in the picture, the motion trajectory, motion time, and pixel displacement of the raindrops are tracked, and the moving speed of the raindrops at the end is calculated.

[0163] Finally, the mapping relationship between the rainfall image information R(x, y, z) (x is the raindrop density, y is the average particle size of the raindrop, and z is the terminal velocity of the raindrop) and the rain intensity I is established using the LeNet-5 convolutional neural network model, so as to realize the calculation of the rain intensity by using the trained neural model. For the rainfall video, the recurrent neural network is used to extract the time sequence characteristics of the video, so as to realize the rainfall recognition based on the video information.

[0164] In a specific application process, the dynamic vision sensor receives light intensity change information input, senses dynamic light change and outputs an event stream, each event has information such as pixel coordinates, light change polarity and time stamp, indicating that the light intensity received by the pixel sensing point at the corresponding coordinate position changes at a certain time. The higher the rainfall intensity, the more pixel points in the picture that change in light intensity, the faster the light intensity changes, the more events are generated, and the number of events has a certain correlation with the rainfall intensity. Therefore, the correlation between the number of events and the rainfall intensity is explored. The event stream timestamp is taken as the independent variable, and the event number information in the rainfall process is superimposed by integrating the time scale from microseconds to milliseconds, and the change process of the event number with time is restored. In addition, according to the rainfall intensity change process obtained in the second step and the measurement results of the tipping bucket rain gauge, the correlation between the event number change process and the rainfall intensity change process is observed and analyzed, so as to realize the inversion of the rainfall intensity change by the change of the event number, which is helpful to directly use the event data to realize the rapid analysis of the rainfall process and the preliminary judgment of whether it is raining.

[0165] A plurality of event cameras are used to monitor the rainfall state of multiple points in the city in real time, data fusion and communication of the multiple event cameras are realized by means of the Beidou system, the Beidou command machine can realize the two-way communication between the Beidou system and the rainfall monitoring device, so as to fuse the data and algorithm processing results (whether it is raining, rainfall intensity and rainfall occurrence time) of the multiple event cameras, and when a natural disaster such as heavy rain or short-time heavy rain occurs, a warning prompt is sent out through the short message system.

[0166] In addition, the examples of the present application can also be combined with the ground rain gauges currently used by the meteorological station as a real-time monitoring, analysis and management urban rainfall monitoring system. The monitoring data is sent to the center station through various communication methods, and after the data is collected in the center station, the rainfall intensity is analyzed, and the multiple event cameras arranged at different positions are used to restore the rainfall state area distribution map.

[0167] Corresponding to the rainfall monitoring method of the above embodiment, an example of a rainfall monitoring device is provided, which can be realized by software, hardware or a combination of the two as part of or all of a computer device, which can be an electronic device as shown. Figure 7 The rainfall monitoring device comprises:

[0168] a monitoring unit configured to, in response to obtaining a rainfall event stream collected by the dynamic vision sensor, reconstruct a rainfall image based on the rainfall event stream;

[0169] a recognition unit configured to recognize a rainfall intensity corresponding to the rainfall occurring in the rainfall image;

[0170] an alarm unit configured to, in a case where the rainfall intensity satisfies an alarm condition, generate an alarm prompt information according to the rainfall intensity and a rainfall occurring time period.

[0171] It should be noted that the rainfall monitoring device provided in the above embodiment is only exemplified by the division of the above functional modules, and in actual application, the above functions can be completed by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above.

[0172] Each functional unit and module in the above embodiment can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit. In addition, the specific names of each functional unit and module are only for convenient distinction, and are not used to limit the protection scope of the embodiments of the present application.

[0173] It should be noted that the information interaction, execution process and the like between the above device / unit, since based on the same concept as the method embodiments of the present application, the specific functions and the technical effects brought by it can be referred to the method embodiment part, and will not be repeated here.

[0174] The embodiments of the present application also provide an electronic device, which includes one or more processors and a memory;

[0175] The memory is coupled with the one or more processors, and the memory is configured to store computer program codes including computer instructions. The one or more processors invoke the computer instructions to enable the electronic device to execute the above-mentioned rainfall monitoring method.

[0176] The electronic device can be a mobile phone, a smart screen, a tablet computer, a wearable electronic device, a vehicle-mounted electronic device, an augmented reality (AR) device, a virtual reality (VR) device, a notebook computer, an ultra-mobile personal computer (UMPC), a netbook, a personal digital assistant (PDA), a projector, or a communication device such as a server, a memory, a base station, or an intelligent automobile. The embodiments of the present application do not limit the specific type of the electronic device.

[0177] The embodiments of the present application also provide a computer readable storage medium, which stores computer instructions; when the computer readable storage medium is run on an electronic device, the electronic device executes the above-mentioned rainfall monitoring method.

[0178] The computer instructions can be stored in the computer readable storage medium or transmitted from one computer readable storage medium to another computer readable storage medium, for example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center through wired (for example, coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (for example, infrared, wireless, microwave, etc.) mode. The computer readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. integrated with one or more media. The available medium can be a magnetic medium (for example, floppy disk, hard disk, magnetic tape), optical medium or semiconductor medium (for example, solid state disk (SSD)) and the like.

[0179] The embodiments of the present application also provide a computer program product containing computer instructions, which, when run on an electronic device, enables the electronic device to execute the above-mentioned rainfall monitoring method.

[0180] The computer storage medium and the computer program product provided by the above embodiments of the present application are used to execute the above-mentioned method, so the beneficial effects that can be achieved by them can refer to the beneficial effects of the above-mentioned method, which will not be repeated here.

[0181] In the above embodiments, all or part of the embodiments can be implemented by software, hardware, firmware or any combination thereof. When implemented by software, all or part of the embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer instructions are loaded and executed on a computer, all or part of the processes or functions according to the embodiments of the present application are generated. The computer can be a general purpose computer, a special purpose computer, a computer network or other programmable apparatus. The computer instructions can be stored in a computer readable storage medium or transmitted from one computer readable storage medium to another computer readable storage medium, for example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center through a wired (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, microwave, etc.) manner. The computer readable storage medium can be any available medium accessible by a computer, or a data storage device such as a server, data center, etc. integrated with one or more available media. The available media can be magnetic media (such as floppy disk, hard disk, magnetic tape), optical media (such as digital versatile disc (DVD)) or semiconductor media (such as solid state disk (SSD)) and the like.

[0182] Figure 7 A structural schematic diagram of an electronic device provided by the embodiments of the present application is shown. The electronic device 700 can be a mobile phone, a computer, a digital broadcast terminal, a messaging device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, etc.

[0183] The memory 701 can be used to store computer software programs 702 and modules, and the processor 703 executes various function applications and data processing of the mobile phone by running the software programs and modules stored in the memory 701. The memory 701 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application program required by a function (such as a sound playing function, an image playing function, etc.), etc.; and the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory 701 can include a high-speed random access memory, and can also include a non-volatile memory, for example, at least one magnetic disk storage device, a flash memory device or other volatile solid state memory device.

[0184] The processor 703 can include one or more of a central processing unit, an application processor (AP), a baseband processor, or any other processor. The processor can be a neural processing unit and a command center of the wireless router. The processor 703 can generate operation control signals according to instruction codes and timing signals, and complete the control of fetching and executing instructions. The memory 701 can be used to store computer executable program codes, and the executable program codes include instructions. The processor 703 executes various functional applications and data processing of the network device by running the instructions stored in the memory. The memory 701 can include a program storage area and a data storage area, such as data of a sound signal to be played, etc. For example, the memory can be a double data rate synchronous dynamic random access memory (DDR) or a flash memory (Flash), etc.

[0185] In the above embodiments, the description of each embodiment has its own focus, and the parts not described or recorded in detail in a certain embodiment can be referred to the related description of other embodiments.

[0186] Those skilled in the art can appreciate that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be implemented by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are implemented in hardware or software depends on the specific application and design constraints of the technical solutions. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0187] In the embodiments provided in the present application, it should be understood that the disclosed apparatus / network device and method can be implemented in other ways. For example, the apparatus / network device embodiments described above are only schematic. The division of the modules or units is only a logical function division, and there can be another division manner in actual implementation. For example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the shown or discussed components can be indirect coupling or communication connection through some interfaces, devices or units, and can be electrical, mechanical or other forms.

[0188] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, i.e. they can be located in one place, or distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiments.

[0189] The above-described embodiments are only used to illustrate the technical solutions of the present application, but not limit them; although the present application is described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A method of rainfall monitoring, characterised in that, The method comprises: in response to obtaining a rainfall event stream collected by a dynamic vision sensor, reconstructing a rainfall image based on the rainfall event stream; identifying a rainfall intensity corresponding to rainfall occurring in the rainfall image; in a case where the rainfall intensity satisfies an alarm condition, generating alarm prompt information according to the rainfall intensity and a rainfall occurrence period; wherein the reconstructing the rainfall image based on the rainfall event stream comprises: obtaining event parameters of each rainfall event in the rainfall event stream, wherein the event parameters comprise: pixel coordinates, illumination change polarity, and a timestamp; determining a change process of each rainfall event over time according to a mapping relationship between the timestamp and the pixel coordinates and the illumination change polarity; reconstructing the rainfall image based on the change processes of all the rainfall events over time.

2. The method of claim 1, wherein, The identifying the rainfall intensity corresponding to rainfall occurring in the rainfall image comprises: obtaining feature information contained in the rainfall image, wherein the feature information at least contains at least one of the following: background features, noise features, and raindrop layer features; inputting the feature information corresponding to the rainfall image into a pre-trained rainfall identification model to obtain the rainfall intensity output by the rainfall identification model, wherein the rainfall identification model is obtained by training an initial neural network model using multiple sets of training data, and each set of training data includes sample feature information of a sample rainfall image and a sample rainfall intensity corresponding to the sample feature information.

3. The method of claim 2, wherein, The rainfall identification model processes the feature information of the rainfall image as follows to obtain the rainfall intensity: extracting a raindrop part in the rainfall image based on the feature information of the rainfall image; calculating a raindrop density, a raindrop average particle size, and a raindrop terminal moving speed of the raindrop part in the rainfall image; calculating the rainfall intensity according to the raindrop density, the raindrop average particle size, and the raindrop terminal moving speed.

4. The method of claim 3, wherein, The extracting the raindrop part in the rainfall image based on the feature information of the rainfall image comprises: using an edge detection algorithm to extract an approximate raindrop part in the rainfall image based on the feature information of the rainfall image; using an image gradient calculation algorithm to calculate image gradient values of the approximate raindrop part in a horizontal direction and a vertical direction; performing non-maximum suppression processing on the image gradient values to determine edge pixels of the approximate raindrop part; determining the raindrop part in the rainfall image according to the edge pixels of the approximate raindrop part.

5. The method of claim 4, wherein, The determining the raindrop part in the rainfall image according to the edge pixels of the approximate raindrop part comprises: in a case where the edge pixel is greater than a first edge threshold, determining to retain the edge pixel; in a case where the edge pixel is less than a second edge threshold, determining to eliminate the edge pixel, wherein the first edge threshold is greater than the second edge threshold; in a case where the edge pixel is less than the first edge threshold and greater than the second edge threshold, determining whether to eliminate the edge pixel according to connection information between the edge pixel and retained edge pixels. According to the reserved edge pixels, a raindrop part in the rain image is determined.

6. The method of claim 3, wherein, A raindrop end moving speed of the raindrop part in the rain image is calculated, including: A raindrop end position pixel point of the raindrop part in the rain image is determined; According to the raindrop end position pixel point, a motion track, a motion time and a pixel displacement of the raindrop part are determined; According to the motion track, the motion time and the pixel displacement of the raindrop part, the raindrop end moving speed of the raindrop part in the rain image is calculated.

7. A rainfall monitoring system characterised in that, The method is implemented as any one of claims 1 to 6, including: A dynamic visual sensor is arranged at an observation point to collect a rain event and obtain a rain event stream; A processor is connected to the dynamic visual sensor to reconstruct a rain image based on the rain event stream and identify a rain intensity corresponding to the rain in the rain image, and in a case that the rain intensity meets an alarm condition, generate an alarm prompt information according to the rain intensity and a rain occurrence period.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The processor executes the computer program to implement the method as any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, the computer program comprising instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 8. The computer program is executed by the processor to implement the method as any one of claims 1 to 6.

Images