Debris flow intelligent monitoring method and system based on visual computing video image acquisition and analysis

Abstract

The invention relates to the field of geological disaster monitoring, in particular to an intelligent debris flow monitoring method based on visual calculation and video image acquisition and analysis, which comprises the following steps: controlling the angle and focal length of a camera mounted in a debris flow susceptible area, and acquiring video image data of topographic full view, topographic texture and topographic color in a plurality of current areas; transmitting the plurality of video image data to an edge computing device; the edge computing device performs real-time image preprocessing, target detection, feature extraction and state analysis on the video image acquired by the camera, judges whether debris flow occurs or not and the danger degree of the debris flow, and correspondingly generates a monitoring image and data; after data compression is carried out on the analyzed and processed monitoring image and data, the data is transmitted to a remote monitoring center through a wireless network; and further data analysis and processing are carried out on the analyzed and processed monitoring image and data, so that the processing reaction time is shortened through data earlier stage processing.

Description

Technical Field

[0001] The present invention relates to the field of geological disaster monitoring, and in particular to a method and system for intelligent monitoring of debris flows by visual computing video image acquisition and analysis. Background Art

[0002] Debris flow is a common natural disaster, characterized by suddenness, fast flow rate, large flow, large material capacity and strong destructive power. Timely monitoring and early warning of debris flow are crucial to protecting people's lives and property. Traditional debris flow monitoring methods mainly include rainfall monitoring, surface displacement monitoring, etc., but these methods have limitations. For example, rainfall monitoring can only indirectly reflect the possibility of debris flow, while surface displacement monitoring requires the installation of sensors at specific locations, which is costly and has limited coverage. With the development of technology, video image monitoring technology has gradually been applied to the field of debris flow monitoring.

[0003] The prior art CN108241182B includes a satellite monitoring unit for acquiring satellite image data through remote sensing technology; a remote sensing identification unit for identifying debris flow areas on satellite image data; a meteorological monitoring unit for collecting relevant meteorological data; a mud and water level monitoring unit for monitoring mud and water level data; and a comprehensive analysis unit for comprehensively analyzing the debris flow monitoring situation based on the above remote sensing identification results, meteorological monitoring results, mud and water level monitoring results, and in combination with local historical disasters and geological and geomorphological conditions.

[0004] However, in the existing technology, a large amount of video image data obtained by satellite is transmitted to the central server for manual processing and judgment. The huge amount of data requires high network bandwidth, which is prone to delays and data loss problems. In addition, the amount of unprocessed data is large, which affects the response time of debris flow disaster handling. Summary of the invention

[0005] The purpose of the present invention is to provide a method and system for intelligent monitoring of debris flow based on visual computing video image acquisition and analysis, which solves the problem in the prior art that a large amount of video image data acquired through satellites is transmitted to a central server for manual processing and judgment. The huge amount of data requires high network bandwidth, is prone to delays and data loss, and the amount of unprocessed data is large, which affects the response time for debris flow disaster processing.

[0006] To achieve the above object, the present invention provides a method for intelligent monitoring of debris flow based on visual computing video image acquisition and analysis, comprising the following steps:

[0007] Control the angle and focal length of the camera installed in the debris flow prone area, obtain the video image data of the terrain, terrain texture and terrain color in multiple current areas, and transmit the multiple video image data to the edge computing device;

[0008] The edge computing device performs real-time image preprocessing, target detection, feature extraction, and state analysis on the video images collected by the camera to determine whether a debris flow has occurred and the degree of danger of the debris flow, and generates corresponding monitoring images and data;

[0009] After the analyzed and processed monitoring images and data are compressed, they are transmitted to the remote monitoring center via a wireless network;

[0010] The analyzed and processed monitoring images and data are further analyzed and processed to generate video images of the monitoring area and debris flow monitoring data for display.

[0011] Among them, in "the edge computing device performs real-time image preprocessing, target detection, feature extraction and state analysis on the video images collected by the camera, determines whether a debris flow has occurred and the degree of danger of the debris flow, and generates monitoring images and data accordingly", the following steps are included:

[0012] Perform image preprocessing on the collected video images, crop and splice the image data to obtain a complete monitoring video image;

[0013] The acquired video images are firstly subjected to preliminary noise removal using filters, and then subjected to fine noise removal using machine learning methods;

[0014] The denoised video image data is used to detect debris flow targets in the video image using a deep learning algorithm to identify the location, size and motion state of the debris flow;

[0015] Extract shape features, texture features and color features of the detected debris flow targets;

[0016] Based on the extracted feature information, the state of the debris flow is analyzed and judged to determine whether a debris flow has occurred and the degree of danger of the debris flow, and corresponding monitoring images and data are generated.

[0017] Among them, in “preliminarily removing noise from the acquired video image using a filter, and then performing fine denoising using a machine learning method”, the method also includes the following steps:

[0018] Analyze the noise in the image data to determine the type and distribution of the noise;

[0019] Select mean filter, median filter and Gaussian filter to perform preliminary denoising on noise of certain type and distribution;

[0020] Select denoising autoencoder and convolutional neural network for model training, use multiple sets of paired data sets containing clear images and their corresponding noisy images as training data sets to train the model and learn the mapping relationship from noisy images to clear images;

[0021] The trained model is applied to the image data processed by the filter method, the image after preliminary denoising is imported into the model for fine denoising, and the processed image data is output.

[0022] Among them, in “selecting a denoising autoencoder and a convolutional neural network for model training, using multiple sets of paired data sets containing clear images and their corresponding noisy images as training data sets to train the model, and learning the mapping relationship from noisy images to clear images”, the method includes:

[0023] Obtain existing clear and noisy images of debris flow monitoring through big data;

[0024] The clear image and its corresponding noisy image are imported into the denoising autoencoder and convolutional neural network for pairing to form a training data pair;

[0025] Performing preprocessing of normalization, cropping, and adjusting resolution on the training data to obtain processed training data;

[0026] The processed training data is rotated, flipped, and scaled to output diverse training data, and the training model learns the mapping relationship from noisy images to clear images.

[0027] Among them, in “using a deep learning algorithm to detect debris flow targets in the video image after denoising, and identifying the location, size and motion state of the debris flow”, the method includes the following steps:

[0028] Splitting the clear image data into individual image frames;

[0029] Redistribute the pixel values ​​in the image, calculate the histogram of the image through the histogram equalization function, apply the equalization algorithm to evenly distribute the image pixel values ​​and increase the image contrast;

[0030] Scan the video frames of multiple denoised and enhanced image information frame by frame, and identify the boundary box and width and height of the debris flow target in each image;

[0031] The graphics acquisition model Premiere decomposes and arranges the graphics frames in time sequence to form a series of continuous images corresponding to different times;

[0032] The flow velocity, flow direction and morphological change characteristics of debris flow in different frames are analyzed, and the movement trend and energy state information of debris flow are calculated.

[0033] Among them, in “analyzing the flow velocity, flow direction and morphological change characteristics of the debris flow in different frames, and calculating the movement trend and energy state information of the debris flow”, the method includes the following steps:

[0034] The optical flow method is used to calculate the displacement vector of debris flow pixel points and obtain the flow velocity field of debris flow.

[0035] Perform edge detection on the front line of the debris flow image frame, extract the graphic contour, calculate the direction vector of each point on the front line contour, count and analyze the distribution of the direction vector, and determine the flow direction of the debris flow;

[0036] The debris flow was separated from the background using image segmentation algorithms, and the shape parameters of the debris flow area, perimeter, aspect ratio, and volume parameters of volume change rate were calculated to quantify the morphological changes;

[0037] According to the extracted characteristic parameters, the numerical analysis method is used to solve and obtain the movement trend of the speed change and direction change of the debris flow and the energy state information of kinetic energy and potential energy.

[0038] A debris flow intelligent monitoring system for visual computing video image acquisition and analysis, comprising a camera acquisition module, an edge computing and analysis module, a remote monitoring module and a data display module, wherein the edge computing and analysis module receives a wireless signal from the camera acquisition module, the remote monitoring module receives a wireless signal from the edge computing and analysis module, and the data display module receives a signal from the remote monitoring module;

[0039] The camera acquisition module is used to control the angle and focal length of the camera installed in the debris flow prone area, obtain the video image data of the terrain overall appearance, terrain texture and terrain color in multiple current areas, and transmit the multiple video image data to the edge computing device;

[0040] The edge computing analysis module is used to perform real-time image preprocessing, target detection, feature extraction and state analysis on the video images collected by the camera, determine whether a debris flow has occurred and the degree of danger of the debris flow, and generate monitoring images and data accordingly;

[0041] The remote monitoring module is used to compress the analyzed monitoring images and data and transmit them to the remote monitoring center via a wireless network;

[0042] The data display module is used to perform further data analysis and processing on the analyzed and processed monitoring images and data, and generate video images of the monitoring area and debris flow monitoring data for display.

[0043] The present invention provides a method and system for intelligent monitoring of debris flow by visual computing video image acquisition and analysis, comprising the following steps: controlling the angle and focal length of a camera installed in an area prone to debris flow, obtaining video image data of the overall terrain, terrain texture and terrain color in multiple current areas, and transmitting the multiple video image data to an edge computing device; the edge computing device performs real-time image preprocessing, target detection, feature extraction and state analysis on the video images acquired by the camera, determines whether a debris flow has occurred and the degree of danger of the debris flow, and generates monitoring images and data accordingly; compresses the analyzed and processed monitoring images and data, and transmits them to a remote monitoring center via a wireless network; further analyzes and processes the analyzed and processed monitoring images and data, generates video images of the monitoring area and debris flow monitoring data for display, and reduces the processing response time through data pre-processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art are briefly introduced below.

[0045] Figure 1 It is a flow chart of the intelligent debris flow monitoring method of visual computing video image acquisition and analysis of the present invention.

[0046] Figure 2 The edge computing device of the present invention performs real-time image preprocessing, target detection, feature extraction and state analysis on the video images collected by the camera to determine whether a mudslide has occurred and the degree of danger of the mudslide, and generates a flowchart of monitoring images and data accordingly.

[0047] Figure 3 The present invention is a flowchart of first using a filter to preliminarily remove noise from the acquired video image, and then using a machine learning method to perform fine denoising.

[0048] Figure 4 The present invention selects a denoising autoencoder and a convolutional neural network for model training, uses multiple sets of paired data sets containing clear images and their corresponding noisy images as training data sets to train the model, and learns the mapping relationship from noisy images to clear images.

[0049] Figure 5 It is a flowchart of the present invention that uses a deep learning algorithm to detect debris flow targets in video images after denoising, and identifies information about the location, size and motion state of the debris flow.

[0050] Figure 6 It is a flow chart of the present invention for analyzing the flow velocity, flow direction and morphological change characteristics of a debris flow in different frames and calculating the movement trend and energy state information of the debris flow.

[0051] Figure 7 It is a system diagram of the debris flow intelligent monitoring system for visual computing video image acquisition and analysis of the present invention.

[0052] In the figure: camera acquisition module 101, edge computing and analysis module 102, remote monitoring module 103 and data display module 104. DETAILED DESCRIPTION

[0053] Embodiments of the present invention are described in detail below. Examples of the embodiments are shown in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, but should not be construed as limiting the present invention.

[0054] See also Figures 1 to 6 , a debris flow intelligent monitoring method based on visual computing video image acquisition and analysis, comprising the following steps:

[0055] S1: Control the angle and focal length of the camera installed in the debris flow prone area, obtain the video image data of the terrain, terrain texture and terrain color in multiple current areas, and transmit the multiple video image data to the edge computing device;

[0056] In this embodiment, a suitable area prone to debris flow is selected for system installation. When installing the camera, ensure that its position can cover the entire monitoring area, and adjust the angle and focal length of the camera. When installing the edge computing device, ensure that its connection with the camera is stable and reliable, and perform corresponding parameter settings and debugging; debug the system as a whole, including the image acquisition of the camera, the analysis and processing of the edge computing device, data transmission, and the display and early warning functions of the remote monitoring center to ensure the normal operation of the system. After the system is started, the camera collects video images of the monitoring area in real time and transmits the image data to the edge computing device. The edge computing device analyzes and processes the video image in real time to determine whether a debris flow has occurred and the degree of danger of the debris flow.

[0057] S2: The edge computing device performs real-time image preprocessing, target detection, feature extraction, and state analysis on the video images collected by the camera to determine whether a debris flow has occurred and the degree of danger of the debris flow, and generates corresponding monitoring images and data;

[0058] S21: performing image preprocessing on the collected video images, cropping and splicing the image data, and obtaining a complete monitoring video image;

[0059] S22: using a filter to initially remove noise from the acquired video image, and then using a machine learning method to perform fine denoising;

[0060] S221: Analyze the noise in the image data to determine the type and distribution of the noise;

[0061] S222: Select a mean filter, a median filter, and a Gaussian filter to perform preliminary denoising on noise of a determined type and distribution;

[0062] S223: selecting a denoising autoencoder and a convolutional neural network for model training, using multiple sets of paired data sets including clear images and their corresponding noisy images as training data sets to train the model, and learning a mapping relationship from noisy images to clear images;

[0063] S2231: Obtain existing clear images and noise images of debris flow monitoring through big data;

[0064] S2232: importing the clear image and its corresponding noisy image into the denoising autoencoder and the convolutional neural network for pairing to form a training data pair;

[0065] S2233: performing pre-processing of normalizing, cropping, and adjusting resolution on the training data to obtain processed training data;

[0066] S2234: Rotate, flip, and scale the processed training data to output diverse training data, and train the model to learn the mapping relationship from the noisy image to the clear image.

[0067] S224: Apply the trained model to the image data processed by the filter method, import the image after preliminary denoising into the model for fine denoising, and output the processed image data.

[0068] In this embodiment, the edge computing device processes the video images collected by the camera in real time to monitor debris flow, which integrates a variety of image processing and machine learning technologies. First, in the S21 stage, the device pre-processes the video images, which mainly includes image cropping and splicing, with the aim of constructing a complete monitoring video image from possibly scattered or overlapping image data. This process usually relies on image registration and splicing algorithms to ensure seamless connection between images, thereby providing a comprehensive and accurate monitoring field of view.

[0069] Entering the S22 stage, the device first analyzes the noise in the image. This is because in the wild environment, the images collected by the camera are often interfered by various noises, such as Gaussian noise, salt and pepper noise, etc. In order to preliminarily remove these noises, the device will select a suitable filter according to the type and distribution of the noise, such as the mean filter is used to remove Gaussian noise, the median filter is used to remove salt and pepper noise, and the Gaussian filter can smooth the image to a certain extent. However, the image after preliminary denoising may still contain complex or unpredictable noise, so it is necessary to use machine learning methods for fine denoising.

[0070] In the S223 stage, the device uses a denoising autoencoder and a convolutional neural network for model training to achieve mapping from noisy images to clear images. In order to train these models, the device first collects clear images and corresponding noisy images of debris flow monitoring from big data, and then imports these images into the network for pairing to form training data pairs. Before training, the device also performs preprocessing operations such as normalization, cropping, and adjusting the resolution of the image data to obtain more standardized and consistent training data. In order to improve the generalization ability of the model, the device also performs data enhancement on the processed training data, such as rotation, flipping, scaling, etc., to output diverse training data. Through these steps, the device can learn the effective mapping relationship of recovering clear images from noisy images, thereby providing high-quality image input for subsequent debris flow detection and hazard level judgment.

[0071] S23: Using the denoised video image data, a deep learning algorithm is used to detect debris flow targets in the video image, and information about the location, size, and motion state of the debris flow is identified;

[0072] S231: splitting the clear image data into separate image frames;

[0073] S232: reallocate pixel values ​​in the image, calculate the histogram of the image through a histogram equalization function, apply an equalization algorithm to evenly distribute the image pixel values, and increase the image contrast;

[0074] S233: Scanning the video frames of the denoised and enhanced image information frame by frame, identifying the boundary box and width and height of the debris flow target in each image;

[0075] S234: Decomposing and arranging the graphic frames in time sequence through the graphic acquisition model Premiere to form a series of continuous images corresponding to different times;

[0076] S235: Analyze the flow velocity, flow direction and morphological change characteristics of the debris flow in different frames, and calculate the movement trend and energy state information of the debris flow.

[0077] S2351: Apply the optical flow method to calculate the displacement vector of the debris flow pixel point and obtain the flow velocity field of the debris flow;

[0078] S2352: performing edge detection on the front line of the debris flow image frame, extracting the graphic contour, calculating the direction vector of each point on the front line contour, counting and analyzing the distribution of the direction vector, and determining the flow direction of the debris flow;

[0079] S2353: Use image segmentation algorithm to separate debris flow from background, calculate shape parameters of debris flow area, perimeter, aspect ratio and volume parameters of volume change rate, and quantify morphological changes;

[0080] S2354: Based on the extracted characteristic parameters, numerical analysis methods are used to solve and obtain the movement trend of the speed change and direction change of the debris flow and the energy state information of the kinetic energy and potential energy.

[0081] In this embodiment, in the S23 stage, the edge computing device uses a deep learning algorithm to detect debris flow targets on the denoised video image data, and identifies the location, size, and motion state information of the debris flow in detail. The following is a description of the specific implementation method of this stage:

[0082] First, in step S231, the device splits the clear and denoised video image data into separate image frames. This step is the basis of video processing, because video is essentially a collection of continuous image frames, and frame-by-frame processing can more accurately capture and analyze dynamic changes in the image.

[0083] Next, in step S232, the device redistributes the pixel values ​​in the image through a histogram equalization function. Histogram equalization is a technique for enhancing image contrast. It calculates the histogram of the image and adjusts the pixel values ​​according to the distribution of the histogram so that the pixel values ​​in the image are more evenly distributed. This can increase the contrast between the debris flow target and the background in the image and improve the accuracy of target detection.

[0084] Entering step S233, the device scans the denoised and enhanced image frames frame by frame, and uses deep learning algorithms (such as object detection network YOLO, Faster R-CNN, etc.) to identify the bounding box and width and height of the debris flow target in each image. These bounding boxes can accurately mark the location and size of the debris flow in the image, providing basic data for subsequent analysis.

[0085] In step S234, the device uses a graphics acquisition model (such as Premiere or similar software) to decompose and arrange the image frames in chronological order to form a series of continuous images. These continuous images can show the state of the debris flow at different time points and provide intuitive visual information for analyzing the movement trend of the debris flow.

[0086] Finally, in step S235, the device conducts an in-depth analysis of the flow velocity, flow direction, and morphological change characteristics of the debris flow in different frames. Specifically:

[0087] In sub-step S2351, the device calculates the displacement vector of the debris flow pixel point using the optical flow method, thereby obtaining the flow velocity field of the debris flow. The optical flow method is a method for estimating the pixel motion in an image sequence, which can accurately capture the dynamic changes of the debris flow in the image.

[0088] In sub-step S2352, the device performs edge detection on the front line of the debris flow image frame, extracts the graphic contour, and calculates the direction vector of each point on the front line contour. By statistically analyzing the distribution of these direction vectors, the device can determine the flow direction of the debris flow.

[0089] In substep S2353, the device uses image segmentation algorithms (such as GrabCut, U-Net, etc.) to separate the debris flow from the background and calculates the shape parameters of the debris flow, such as area, perimeter, aspect ratio, and volume parameters such as volume change rate. These parameters can quantify the morphological changes of the debris flow and provide an important basis for assessing the danger of the debris flow.

[0090] In sub-step S2354, the equipment uses numerical analysis methods (such as finite difference method, finite element method, etc.) based on the extracted characteristic parameters to solve the movement trend of the speed change and direction change of the debris flow, as well as the energy state information such as kinetic energy and potential energy. This information can fully reflect the movement state and energy characteristics of the debris flow, and provide a scientific basis for subsequent early warning and emergency response.

[0091] S24: extracting shape features, texture features and color features of the detected debris flow target;

[0092] In this embodiment:

[0093] Shape feature extraction:

[0094] The shape characteristics of debris flow targets mainly include their boundary shape, aspect ratio, area, etc. Through image processing technology, such as edge detection and contour extraction, the precise boundary of debris flow targets can be obtained. Furthermore, the shape parameters such as area, perimeter, aspect ratio, etc. of debris flow targets can be calculated to reflect their overall shape characteristics.

[0095] Texture feature extraction:

[0096] The texture features of debris flow targets reflect their surface roughness, particle distribution and other characteristics. Texture analysis methods such as gray level co-occurrence matrix (GLCM) and local binary pattern (LBP) can be used to extract the texture features of debris flow targets. These texture features help distinguish debris flow from other natural landforms and improve the accuracy of recognition.

[0097] Color feature extraction:

[0098] The color characteristics of debris flow targets are mainly related to their material composition, water content, etc. The color characteristics of debris flow targets can be extracted through color space conversion (such as RGB to HSV), color histogram statistics, etc. These color characteristics can be used to assist in determining the type and danger level of debris flow.

[0099] S25: Analyze and judge the state of the debris flow based on the extracted characteristic information to determine whether a debris flow has occurred and the degree of danger of the debris flow, and generate monitoring images and data accordingly.

[0100] In this embodiment:

[0101] Analysis and judgment of debris flow status:

[0102] Based on the extracted shape, texture and color features, a debris flow recognition and classification model can be constructed. The model can be trained and optimized using machine learning algorithms (such as support vector machines, random forests, etc.) or deep learning algorithms (such as convolutional neural networks, etc.). By matching the debris flow targets detected in real time with the model, it is possible to determine whether a debris flow has occurred and preliminarily assess its degree of danger.

[0103] Generation of monitoring images and data:

[0104] Once a debris flow target is detected, the system should immediately generate a monitoring image containing information such as the debris flow's location, size, shape, texture, and color. At the same time, the system should also record the dynamic changes of the debris flow target, such as flow speed and direction, and generate corresponding monitoring data. These monitoring images and data can be used for subsequent debris flow warning, emergency response, and disaster assessment.

[0105] S3: compress the analyzed monitoring images and data and transmit them to the remote monitoring center via a wireless network;

[0106] In this embodiment, the edge computing device will perform efficient data compression on the monitored images and data after analysis and processing. This process uses advanced compression algorithms, such as JPEG, H.264 or more advanced encoding technologies, to significantly reduce the amount of data, thereby reducing the bandwidth requirements and transmission time of wireless transmission. The compressed data is securely and stably transmitted to the remote monitoring center via wireless networks, such as 4G / 5G, Wi-Fi or dedicated wireless communication networks. This step ensures real-time and data integrity, allowing the remote monitoring center to quickly receive monitoring information from the front end, providing a valuable time window for subsequent disaster response and decision-making.

[0107] S4: Further data analysis and processing are performed on the analyzed and processed monitoring images and data to generate video images of the monitoring area and debris flow monitoring data for display.

[0108] In this embodiment: after the remote monitoring center receives the compressed monitoring images and data, it first performs a decompression operation to restore the quality of the original images and data. Subsequently, advanced data analysis and processing technologies, such as deep learning models, statistical analysis methods, and geographic information systems (GIS), are used to deeply analyze the monitoring data and extract key information, such as the velocity, flow rate, and accumulation range of the debris flow. After this information is integrated, high-quality video images of the monitoring area and debris flow monitoring reports are generated. Ultimately, these monitoring results are displayed to monitoring personnel in an intuitive and easy-to-understand manner through a large-screen display system, a monitoring software interface, or a mobile application platform, allowing them to quickly grasp the dynamic situation of the monitoring area and make accurate judgments and decisions. This process not only improves monitoring efficiency, but also enhances the timeliness and effectiveness of disaster response.

[0109] See also Figure 7 , a debris flow intelligent monitoring system for visual computing video image acquisition and analysis, including a module for executing the debris flow intelligent monitoring method for visual computing video image acquisition and analysis, and also including a camera acquisition module 101, an edge computing analysis module 102, a remote monitoring module 103 and a data display module 104, the edge computing analysis module receives the wireless signal of the camera acquisition module, the remote monitoring module receives the wireless signal of the edge computing analysis module, and the data display module receives the signal of the remote monitoring module;

[0110] The camera acquisition module 101 is used to control the angle and focal length of the camera installed in the debris flow prone area, obtain the video image data of the terrain overall appearance, terrain texture and terrain color in multiple current areas, and transmit the multiple video image data to the edge computing device;

[0111] The edge computing analysis module 102 is used to perform real-time image preprocessing, target detection, feature extraction and state analysis on the video images collected by the camera, determine whether a debris flow has occurred and the degree of danger of the debris flow, and generate monitoring images and data accordingly;

[0112] The remote monitoring module 103 is used to compress the analyzed monitoring images and data and transmit them to the remote monitoring center via a wireless network;

[0113] The data display module 104 is used to perform further data analysis and processing on the analyzed monitoring images and data, and generate video images of the monitoring area and debris flow monitoring data for display.

[0114] In this embodiment, the debris flow intelligent monitoring system of the present invention mainly consists of a camera, an edge computing device, and a remote monitoring center. The camera is installed in the debris flow prone area to collect video images. The edge computing device is connected to the camera and is responsible for real-time analysis and processing of the video images. The remote monitoring center receives the monitoring results sent by the edge computing device and conducts further data analysis and early warning release.

[0115] Camera acquisition:

[0116] The camera uses a high-definition digital camera and can collect video images of the debris flow prone area in real time. The installation location of the camera should be selected at a key position that can cover the entire monitoring area to ensure comprehensive video information can be obtained.

[0117] Edge computing analysis:

[0118] The edge computing device uses a high-performance embedded processor with powerful computing and image processing capabilities. The edge computing device conducts real-time analysis on the video images collected by the camera, mainly including the following steps:

[0119] Image preprocessing: Perform preprocessing operations such as denoising and enhancement on the collected video images to improve the image quality.

[0120] Object detection: Use deep learning algorithms to detect debris flow objects in the video images and identify information such as the location, size, and motion state of the debris flow.

[0121] Feature extraction: Extract features from the detected debris flow objects, including shape features, texture features, color features, etc.

[0122] Status analysis: Analyze and judge the status of the debris flow based on the extracted feature information to determine whether a debris flow has occurred and the degree of danger of the debris flow.

[0123] Data transmission:

[0124] The edge computing device transmits the analyzed and processed monitoring results to the remote monitoring center through a wireless network. In order to reduce the data transmission volume and improve the transmission efficiency, data compression and encryption technologies can be used.

[0125] Remote monitoring center:

[0126] The remote monitoring center receives the monitoring results sent by the edge computing device and conducts further data analysis and processing. The remote monitoring center can display the video images and debris flow monitoring data of the monitoring area in real time, and at the same time can also issue early warning information according to the monitoring results to notify relevant departments and personnel to take corresponding emergency measures.

[0127] Software composition: The software of the debris flow intelligent monitoring system of the present invention mainly includes a camera driver, edge computing software, a wireless network driver, a remote monitoring software, etc.

[0128] Camera driver: used to drive the camera to collect video and transmit the collected video data to the edge computing device. The camera driver should have the characteristics of high stability and good compatibility, and be able to support multiple models of cameras.

[0129] Edge computing software: used to analyze and process the collected video images in real time and transmit the monitoring results to the remote monitoring center. Edge computing software should have the characteristics of high efficiency, high accuracy, strong adaptability, and be able to adapt to different environmental conditions and monitoring needs.

[0130] Wireless network driver: used to drive wireless network devices for data transmission and ensure the stability and reliability of data transmission. Wireless network driver should have the characteristics of good compatibility, high transmission rate, high security, etc., and be able to support multiple models of wireless network devices.

[0131] Remote monitoring software: used to display, store and analyze monitoring results in real time and issue early warning information. Remote monitoring software should have the characteristics of user-friendly interface, simple operation and powerful functions to meet the needs of different users.

[0132] What is disclosed above is only one or more preferred embodiments of the present application, and cannot be used to limit the scope of rights of the present application. Ordinary technicians in this field can understand that all or part of the processes of implementing the above embodiments and equivalent changes made according to the claims of the present application are still within the scope covered by the present application.

Claims

1. A method for intelligent monitoring of debris flow based on visual computing video image acquisition and analysis, characterized in that: The steps include: Control the angle and focal length of the camera installed in the debris flow prone area, obtain the video image data of the terrain, terrain texture and terrain color in multiple current areas, and transmit the multiple video image data to the edge computing device; The edge computing device performs real-time image preprocessing, target detection, feature extraction, and state analysis on the video images collected by the camera to determine whether a debris flow has occurred and the degree of danger of the debris flow, and generates corresponding monitoring images and data; After the analyzed and processed monitoring images and data are compressed, they are transmitted to the remote monitoring center via a wireless network; The analyzed and processed monitoring images and data are further analyzed and processed to generate video images of the monitoring area and debris flow monitoring data for display.

2. The method for intelligent monitoring of debris flow based on visual computing video image acquisition and analysis as claimed in claim 1, characterized in that: In "the edge computing device performs real-time image preprocessing, target detection, feature extraction, and state analysis on the video images collected by the camera, determines whether a debris flow has occurred and the degree of danger of the debris flow, and generates monitoring images and data accordingly", the following steps are included: Perform image preprocessing on the collected video images, crop and splice the image data to obtain a complete monitoring video image; The acquired video images are firstly subjected to preliminary noise removal using filters, and then subjected to fine noise removal using machine learning methods; The denoised video image data is used to detect debris flow targets in the video image using a deep learning algorithm to identify the location, size and motion state of the debris flow; Extract shape features, texture features and color features of the detected debris flow targets; Based on the extracted feature information, the state of the debris flow is analyzed and judged to determine whether a debris flow has occurred and the degree of danger of the debris flow, and corresponding monitoring images and data are generated.

3. The intelligent monitoring method for debris flow based on visual computing video image acquisition and analysis as claimed in claim 2, characterized in that: In "preliminarily removing noise from the acquired video image using a filter, and then performing fine denoising using a machine learning method", the method further includes the following steps: Analyze the noise in the image data to determine the type and distribution of the noise; Select mean filter, median filter and Gaussian filter to perform preliminary denoising on noise of certain type and distribution; Select denoising autoencoder and convolutional neural network for model training, use multiple sets of paired data sets containing clear images and their corresponding noisy images as training data sets to train the model and learn the mapping relationship from noisy images to clear images; The trained model is applied to the image data processed by the filter method, the image after preliminary denoising is imported into the model for fine denoising, and the processed image data is output.

4. The method for intelligent monitoring of debris flow based on visual computing video image acquisition and analysis as claimed in claim 3, characterized in that: In "selecting a denoising autoencoder and a convolutional neural network for model training, using multiple sets of paired data sets including clear images and their corresponding noisy images as training data sets to train the model, and learning the mapping relationship from noisy images to clear images", the method includes: Obtain existing clear and noisy images of debris flow monitoring through big data; The clear image and its corresponding noisy image are imported into the denoising autoencoder and convolutional neural network for pairing to form a training data pair; Performing preprocessing of normalization, cropping, and adjusting resolution on the training data to obtain processed training data; The processed training data is rotated, flipped, and scaled to output diverse training data, and the training model learns the mapping relationship from noisy images to clear images.

5. The method for intelligent monitoring of debris flow based on visual computing video image acquisition and analysis as claimed in claim 2, characterized in that: In "Using a deep learning algorithm to detect debris flow targets in the video image after denoising, and identifying information about the location, size and motion state of the debris flow", the method includes the following steps: Splitting the clear image data into individual image frames; Redistribute the pixel values ​​in the image, calculate the histogram of the image through the histogram equalization function, apply the equalization algorithm to evenly distribute the image pixel values ​​and increase the image contrast; Scan the video frames of multiple denoised and enhanced image information frame by frame, and identify the boundary box and width and height of the debris flow target in each image; The graphics acquisition model Premiere decomposes and arranges the graphics frames in chronological order to form a series of continuous images corresponding to different times; The flow velocity, flow direction and morphological change characteristics of debris flow in different frames are analyzed, and the movement trend and energy state information of debris flow are calculated.

6. The method for intelligent monitoring of debris flow based on visual computing video image acquisition and analysis as claimed in claim 5, characterized in that: In "analyzing the flow velocity, flow direction and morphological change characteristics of the debris flow in different frames, and calculating the movement trend and energy state information of the debris flow", the method includes the following steps: The optical flow method is used to calculate the displacement vector of debris flow pixel points and obtain the flow velocity field of debris flow. Perform edge detection on the front line of the debris flow image frame, extract the graphic contour, calculate the direction vector of each point on the front line contour, count and analyze the distribution of the direction vector, and determine the flow direction of the debris flow; The debris flow was separated from the background using image segmentation algorithms, and the shape parameters of the debris flow area, perimeter, aspect ratio, and volume parameters of volume change rate were calculated to quantify the morphological changes; According to the extracted characteristic parameters, the numerical analysis method is used to solve and obtain the movement trend of the speed change and direction change of the debris flow and the energy state information of kinetic energy and potential energy.

7. A debris flow intelligent monitoring system for visual computing video image acquisition and analysis, comprising a module for executing the debris flow intelligent monitoring method for visual computing video image acquisition and analysis as claimed in any one of claims 1 to 6, characterized in that: It includes a camera acquisition module, an edge computing and analysis module, a remote monitoring module and a data display module, wherein the edge computing and analysis module receives the wireless signal of the camera acquisition module, the remote monitoring module receives the wireless signal of the edge computing and analysis module, and the data display module receives the signal of the remote monitoring module; The camera acquisition module is used to control the angle and focal length of the camera installed in the debris flow prone area, obtain the video image data of the terrain overall appearance, terrain texture and terrain color in multiple current areas, and transmit the multiple video image data to the edge computing device; The edge computing analysis module is used to perform real-time image preprocessing, target detection, feature extraction and state analysis on the video images collected by the camera, determine whether a debris flow has occurred and the degree of danger of the debris flow, and generate monitoring images and data accordingly; The remote monitoring module is used to compress the analyzed monitoring images and data and transmit them to the remote monitoring center via a wireless network; The data display module is used to perform further data analysis and processing on the analyzed and processed monitoring images and data, and generate video images of the monitoring area and debris flow monitoring data for display.

Images