CN114119642A

CN114119642A - Method, device, equipment and storage medium for extracting water body in flood disaster

Info

Publication number: CN114119642A
Application number: CN202111210954.1A
Authority: CN
Inventors: 王福涛; 王世新; 周艺; 王敬明; 王振庆; 熊义兵; 王丽涛; 刘文亮; 朱金峰
Original assignee: Aerospace Information Research Institute of CAS
Current assignee: Aerospace Information Research Institute of CAS
Priority date: 2021-10-18
Filing date: 2021-10-18
Publication date: 2022-03-01

Abstract

The invention provides a flood disaster water body extraction method, a device, equipment and a storage medium, wherein the method comprises the following steps: acquiring an SAR image to be extracted from a water body, wherein the SAR image to be extracted from the water body comprises an image consisting of a dual polarization wave band and a spectral characteristic wave band of an SAR image of a flood disaster; inputting the SAR image to be extracted of the water body into a preset flood extraction convolutional neural network to obtain a flood disaster water body information distribution map; the preset flood extraction convolution neural network is used for performing residual convolution and attention mechanism correction on the SAR image to be extracted from the water body, and performing cavity convolution on the corrected SAR image to be extracted from the water body to obtain the flood disaster water body information distribution map. The method can realize the purpose of extracting the water body information based on the SAR image of the flood disaster with high precision and high accuracy, and effectively improves the extraction precision and the extraction accuracy of the water body of the flood disaster.

Description

Method, device, equipment and storage medium for extracting water body in flood disaster

Technical Field

The invention relates to the technical field of natural disaster remote sensing, in particular to a method, a device, equipment and a storage medium for extracting a water body in a flood disaster.

Background

Flood disasters are one of the most serious natural disasters in the world, and cause a great amount of casualties and economic losses to the world. Especially in China, flood disasters frequently occur, and the development of national economy and the life and property safety of people in China are seriously threatened. Therefore, timely and accurate acquisition of the space-time distribution range of flood flooding has extremely important significance for relevant departments to make disaster-resistant and relief schemes and post-disaster loss assessment.

In the conventional technology, a Synthetic Aperture Radar (SAR) and a deep learning model are usually combined to extract a water body from an optical image of a flood disaster, and a convolutional neural network is adopted to extract the water body at first; then extracting urban surface water bodies in the high-resolution remote sensing images based on the convolutional neural network model, and obtaining higher water body extraction precision; and then, a multi-scale lake water body extraction network model can be used for extracting the small lake water body.

However, the research of extracting water by the conventional technology mainly focuses on the optical image of the flood disaster, and the water and the shadow are difficult to distinguish, the small water rivers are difficult to extract completely, and the extraction precision is not high, so that the extraction precision and the extraction accuracy of the flood disaster water are not high.

Disclosure of Invention

The invention provides a flood disaster water body extraction method, a flood disaster water body extraction device, equipment and a storage medium, which are used for solving the defects of low extraction precision and extraction accuracy of a flood disaster water body in the prior art and achieving the purpose of extracting water body information based on a flood disaster SAR image with high precision and high accuracy.

The invention provides a flood disaster water body extraction method, which comprises the following steps:

acquiring an SAR image to be extracted from a water body, wherein the SAR image to be extracted from the water body comprises an image consisting of a dual polarization wave band and a spectral characteristic wave band of an SAR image of a flood disaster;

inputting the SAR image to be extracted of the water body into a preset flood extraction convolutional neural network to obtain a flood disaster water body information distribution map;

the preset flood extraction convolution neural network is used for performing residual convolution and attention mechanism correction on the SAR image to be extracted from the water body, and performing cavity convolution on the corrected SAR image to be extracted from the water body to obtain the flood disaster water body information distribution map.

According to the flood disaster water body extraction method provided by the invention, the preset flood extraction convolutional neural network comprises a preset residual error neural network and a preset cavity convolutional network, and the SAR image to be extracted of the water body is input into the preset flood extraction convolutional neural network to obtain a flood disaster water body information distribution map, which comprises the following steps:

inputting the SAR image to be extracted of the water body into a preset residual error neural network for residual error convolution and attention mechanism correction to obtain a first water body information distribution map after the flood disaster water body edge is subjected to fine correction;

and inputting the first water body information distribution map into a preset cavity convolution network to carry out change processing of a preset expansion scale, so as to obtain a flood disaster water body information distribution map with complete flood disaster water body edge extraction.

According to the method for extracting the flood disaster water body, provided by the invention, the preset residual error neural network comprises P residual error sub-networks, and when each residual error sub-network is respectively connected with an attention mechanism correction sub-network, the SAR image to be extracted from the water body is input into the preset residual error neural network for residual error convolution and attention mechanism correction, so as to obtain a first water body information distribution map after the flood disaster water body edge is subjected to fine correction, and the method comprises the following steps:

inputting the SAR image to be extracted of the water body into a q-th residual error sub-network for convolution, batch normalization and linear arrangement processing to obtain a q-th residual error convolution image;

inputting the q-th residual convolution image into a q-th attention mechanism correction sub-network for channel correction and space correction processing to obtain a q-th corrected image;

inputting the q-th corrected image into a q + 1-th residual error sub-network for convolution, batch normalization and linear arrangement processing to obtain a q + 1-th residual error convolution image;

inputting the (q + 1) th residual convolution image into a (q + 1) th attention mechanism correction sub-network for channel correction and spatial correction processing to obtain a (q + 1) th corrected image;

adding 1 to the value q, and repeatedly executing the step of inputting the (q + 1) th residual convolution image into a (q + 1) th attention mechanism correction sub-network for channel correction and space correction processing until a P-th corrected image is obtained, wherein the P-th corrected image is a first water body information distribution map after flood disaster water body edge fine correction; q belongs to [1, …, P ], q +1 belongs to [2, …, P ].

According to the method for extracting the flood disaster water body, provided by the invention, when the preset cavity convolution network comprises Q different expansion rates, the first water body information distribution map is input into the preset cavity convolution network to carry out change processing of the preset expansion scale, and a complete flood disaster water body information distribution map extracted from the edge of the flood disaster water body is obtained, and the method comprises the following steps:

inputting the first water body information distribution map into a preset cavity convolution network for different expansion scale processing to obtain Q fine water body characteristic information maps with different expansion scales;

performing fusion processing on the fine water body characteristic information graph to obtain a fusion image;

and decoding the fused image corresponding to the residual convolution and attention mechanism correction to obtain a flood disaster water body information distribution map with complete flood disaster water body edge extraction.

According to the flood disaster water body extraction method provided by the invention, the training process of the preset flood extraction convolutional neural network comprises the following steps:

acquiring a training sample image set; the training sample image set comprises training sample images, and the training sample images are SAR images obtained by carrying out sample labeling, regular grid cutting and data enhancement processing on the SAR images to be extracted from the water body;

and training a preset initial flood extraction convolutional neural network according to the training sample image set to obtain the preset flood extraction convolutional neural network.

According to the flood disaster water body extraction method provided by the invention, the training of the preset initial flood extraction convolutional neural network according to the training sample image set to obtain the preset flood extraction convolutional neural network comprises the following steps:

performing iterative training on a preset initial flood extraction convolutional neural network according to the training sample image set to obtain an intermediate flood extraction convolutional neural network obtained after the iterative training;

judging whether the cumulative training round number corresponding to the current round of iterative training reaches a preset round number threshold value or not;

if the accumulated training round number reaches the preset round number threshold value, taking the intermediate flood extraction convolutional neural network obtained after the iterative training as the preset flood extraction convolutional neural network;

and if the accumulated training round number does not reach the preset round number threshold value, training the intermediate flood extraction convolutional neural network to obtain the preset flood extraction convolutional neural network.

According to the flood disaster water body extraction method provided by the invention, the process of determining the preset turn number threshold value comprises the following steps:

dividing the training sample image set into training samples and verification samples according to a preset proportion;

training a preset initial flood extraction convolutional neural network according to the training samples to obtain an intermediate flood extraction convolutional neural network obtained after training in preset rounds;

verifying the intermediate flood extraction convolutional neural network according to the verification sample to obtain a value of an evaluation index of the intermediate flood extraction convolutional neural network;

judging whether the value of the evaluation index reaches a preset standard value or not;

if the value of the evaluation index reaches the preset standard value, taking the accumulated training round number corresponding to the intermediate flood extraction convolutional neural network as the preset round number threshold value;

and if the value of the evaluation index does not reach the preset standard value, training the intermediate flood extraction convolutional neural network to obtain the preset round number threshold value.

The invention also provides a water body extraction device for flood disasters, which comprises:

the acquisition module is used for acquiring an SAR image to be extracted from a water body, wherein the SAR image to be extracted from the water body comprises an image consisting of a dual polarization wave band and a spectral characteristic wave band of the SAR image of a flood disaster;

the determining module is used for inputting the SAR image to be extracted of the water body into a preset flood extraction convolutional neural network to obtain a flood disaster water body information distribution map;

The invention also provides electronic equipment which comprises a memory, a processor and a computer program which is stored on the memory and can run on the processor, wherein the processor executes the program to realize the steps of any one of the flood disaster water body extraction methods.

The present invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, performs the steps of the flood disaster water extraction method as described in any of the above.

According to the flood disaster water body extraction method, the device, the equipment and the storage medium, firstly, an image formed by a dual polarization wave band and a spectral characteristic wave band of a flood disaster SAR image is used as a water body SAR image to be extracted, and then the water body SAR image to be extracted is input into a preset flood extraction convolutional neural network to be subjected to a water body extraction process to obtain a flood disaster water body information distribution map. The preset flood extraction convolution neural network is used for carrying out residual convolution and attention mechanism correction on the SAR image to be extracted from the water body, and carrying out cavity convolution on the corrected SAR image to be extracted from the water body to obtain the flood disaster water body information distribution map, so that the purpose of further capturing more detailed and feature-unobvious characteristic information of the water body edge on the basis of carrying out accurate correction and reliable extraction on the water body edge with obvious features of the SAR image to be extracted from the water body can be realized, and therefore, the water body information in the obtained flood disaster water body information distribution map is richer and more complete, and the extraction accuracy of the flood disaster water body are effectively improved.

Drawings

In order to more clearly illustrate the technical solutions of the present invention or the prior art, the drawings needed for the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and those skilled in the art can also obtain other drawings according to the drawings without creative efforts.

FIG. 1 is a schematic flow chart of a water body extraction method for flood disasters, provided by the invention;

FIG. 2 is a schematic diagram of a flood disaster water body information distribution diagram obtained based on the method of the present invention;

FIG. 3 is a schematic flow chart of convolution, batch normalization and linear sorting processes performed by each residual sub-network according to the present invention;

FIG. 4 is a schematic flow diagram of a channel correction and spatial correction process performed by each attention mechanism correction subnetwork provided in the present invention;

FIG. 5 is a schematic diagram of a preset hole convolution network according to the present invention executing hole convolutions with different expansion rates;

FIG. 6 is a schematic diagram of a precision test area extracted using the method of the present invention;

FIG. 7 is a schematic structural diagram of a flood disaster water body extraction device provided by the invention;

fig. 8 is a schematic structural diagram of an electronic device provided in the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is obvious that the described embodiments are some, but not all embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

Flood disasters are one of the most serious natural disasters in the world, and cause a great amount of casualties and economic losses to the world. According to the Chinese water and drought disaster bulletin issued by the ministry of water conservancy, the flood disaster causes direct economic loss of 1682.88 million yuan in China all the year after 1998 to 2019, and 1288 people die; therefore, flood disasters bring huge losses to the economic development of China and the life safety of people. Therefore, timely and accurate acquisition of the space-time distribution range of flood flooding has extremely important significance for relevant departments to make disaster-resistant and relief schemes and post-disaster loss assessment.

At present, flood monitoring mainly comprises a hydrological site monitoring method and a remote sensing monitoring method, wherein the hydrological site monitoring method has high accuracy, but has the defects of few sites, incapability of monitoring a spatial distribution range and the like; the remote sensing technology has the advantages of short revisit time, capability of monitoring the spatial distribution range and the like, and gradually becomes a main method for flood monitoring. However, since flood water is often accompanied by bad weather, the conventional optical sensor is affected by cloud rain, and it is difficult to provide a cloudless high-quality optical image. On the contrary, the SAR has all-weather monitoring capability and is not easily affected by bad weather such as cloud and rain, and thus plays an increasingly important role in flood monitoring and extraction.

A flood monitoring and extracting method based on SAR data mainly comprises a threshold value method, an object-oriented method, an active contour method and a data fusion method. However, due to the defects of speckle noise, uneven gray scale distribution and the like of the SAR image, the conventional SAR data flood monitoring method cannot well meet the requirements in large-area flood monitoring under the conditions of large data volume and complex flood flooding scene. Under the background, combining the SAR data and the intelligent water body extraction algorithm gradually becomes a research hotspot.

In addition, the deep learning algorithm is mainly applied to tasks such as semantic segmentation, target detection and image classification in the remote sensing field, and a good effect is achieved. In 2015, Long et al proposed a semantic segmentation method based on a full convolution network FCN, which solves the problem of image segmentation at semantic level and performs pixel-level classification on images; the FCN is the first convolutional neural network used for semantic segmentation, however, the network has the disadvantages of over-simple upsampling operation, loss of detail information, and the like. And then, Ronneberger and the like improve the FCN network, and provide a U-Net network with a coding and decoding structure, wherein the network can integrate the characteristics of low resolution and high resolution, so that the image segmentation precision is greatly improved. Since then, in order to improve the precision and performance of semantic segmentation, researchers have proposed many classical semantic segmentation models, such as deep v3, UNet + +, HRNet, and so on.

In recent years, the deep learning model is gradually applied to remote sensing image water body extraction, for example, before old, a convolutional neural network is adopted to extract a water body, and compared with the traditional extraction methods such as NDWI (non-uniform distribution of the parameters) and the like, the effectiveness of the deep learning method in extracting the water body is proved; chen and the like design a new convolutional neural network model to extract urban surface water in high-resolution remote sensing images, and obtain higher water extraction precision; wang et al propose a multi-scale lake water body extraction network model, which can well extract small lake water bodies.

However, the research of extracting water by the existing deep learning model mainly focuses on optical images, and the research on SAR images is relatively few, and the defects that the water and shadows are difficult to distinguish, small water rivers are difficult to extract completely, the extraction accuracy is not high, and the like exist. The key of the SAR image flood monitoring is the recognition and extraction of water body information as well as the optical remote sensing image.

Based on the above problems, the present invention provides a method for extracting a water body from a flood disaster, wherein an execution main body of the method for extracting a water body from a flood disaster may be a device for extracting a water body from a flood disaster, and the device for extracting a water body from a flood disaster may be implemented as a part or all of a terminal device by software, hardware, or a combination of software and hardware. Alternatively, the terminal device may be a Personal Computer (PC), a portable device, a notebook Computer, a smart phone, a tablet Computer, a portable wearable device, and other electronic devices, such as a tablet Computer, a mobile phone, and the like. The present invention does not limit the specific form of the terminal device.

It should be noted that the execution subject of the method embodiments described below may be part or all of the terminal device described above. The following method embodiments take the execution subject as an example of the terminal device.

Fig. 1 is a schematic flow chart of the flood disaster water body extraction method provided by the present invention, and as shown in fig. 1, the flood disaster water body extraction method includes the following steps:

step S110, acquiring an SAR image to be extracted from the water body, wherein the SAR image to be extracted from the water body comprises an image consisting of a dual polarization wave band and a spectral characteristic wave band of the SAR image from a flood disaster.

Specifically, when the terminal device obtains the SAR image to be extracted from the water body, the SAR image for flood disaster can be obtained first, preprocessing operations such as track correction, thermal noise removal, radiometric calibration, speckle filtering, terrain correction, decibelization, clipping and embedding are performed on the SAR image by using SNAP software, a backscattering distribution map of the SAR image is obtained, a dual polarization band of the backscattering distribution map is further obtained, the dual polarization band can include a VH band and a VV band, spectral characteristics SDWI bands derived from polarization of the VH band and the VV band are obtained by using Python programming, and finally a three-band image composed of the VH band, the VV band and the SDWI band is used as the SAR image to be extracted from the water body.

It should be noted that the electric field vector of the energy pulse emitted by the radar may be polarized in the vertical or horizontal plane. And, regardless of the wavelength, the radar signal may transmit a horizontal (H) or vertical (V) electric field vector, receive a return signal of either the horizontal (H) or vertical (V) or both. Therefore, the radar remote sensing system for SAR radar generally employs four polarization modes of HH, VV, HV and VH, where HH and VV are co-polarized and HV and VH are counter (or cross) polarized.

And S120, inputting the SAR image to be extracted of the water body into a preset flood extraction convolutional neural network to obtain a flood disaster water body information distribution map.

Specifically, when acquiring the SAR image of the water body to be extracted, the terminal device may extract the water body information with obvious features in the SAR image of the water body to be extracted, such as a wide lake surface, a small pond and the like, according to a preset feature extraction algorithm and a feature expansion algorithm, and then further extract more water body information with unobvious features, such as more detailed small rivers and the like, unidentifiable micro ditches and the like, on the basis of ensuring the extraction accuracy and the extraction precision, so as to obtain a complete flood disaster water body information distribution map extracted from the water body. When the SAR image to be extracted from the water body is determined according to the 26-day Sentinel-1SAR image in 7/2020 during the occurrence period of the flood disaster, the flood disaster water body information distribution map shown in fig. 2 can be obtained.

It should be noted that, in order to improve the extraction accuracy and the extraction precision, the feature extraction algorithm used in the present invention includes an algorithm combining a deep residual error network and an attention mechanism, because the deep residual error network can reduce the complexity of a network model and improve the processing speed of the network model, and the attention mechanism can not only extract more semantic features, but also enable the terminal device to have the function of ignoring irrelevant information and paying attention to key information like a human brain. Therefore, the method can accurately and accurately extract the water body information with obvious water body characteristics in the SAR image to be extracted, such as water body information with obvious characteristics on a wide lake surface, a small water body, a pond and other characteristics which are easily submerged during flood disasters. Alternatively, the depth residual network may be a depth residual network (ResNet) 18, and the attention mechanism may be a simultaneous Spatial compression and Channel Excitation (scan) attention mechanism.

The method for extracting the flood disaster water body comprises the steps of firstly taking an image composed of a dual polarization wave band and a spectral characteristic wave band of a flood disaster SAR image as a water body SAR image to be extracted, and then inputting the water body SAR image to be extracted into a preset flood extraction convolutional neural network for water body extraction to obtain a flood disaster water body information distribution map. The preset flood extraction convolution neural network is used for carrying out residual convolution and attention mechanism correction on the SAR image to be extracted from the water body, and carrying out cavity convolution on the corrected SAR image to be extracted from the water body to obtain the flood disaster water body information distribution map, so that the purpose of further capturing more detailed and feature-unobvious characteristic information of the water body edge on the basis of carrying out accurate correction and reliable extraction on the water body edge with obvious features of the SAR image to be extracted from the water body can be realized, and therefore, the water body information in the obtained flood disaster water body information distribution map is richer and more complete, and the extraction accuracy of the flood disaster water body are effectively improved.

Based on the defects of low Extraction precision and low Extraction accuracy of a Flood disaster water body in the prior art, the invention provides a SAR data-based water body information Extraction method for a Flood disaster water body, wherein the water body information Extraction is carried out by a Flood Extraction convolutional neural Network (FENet), the FENet comprises a coding correction Network and a cavity convolutional Network, the coding correction Network adopts a Resnet18 Network, and a residual error sub-Network in a Resnet18 Network is connected with an attention mechanism correction sub-Network, so that the coded and corrected characteristic images are input into the cavity convolutional networks with different expansion rates, deep semantic features are extracted by different scales, and the images subjected to convolutional cavities are decoded to be in the original size through a decoding process corresponding to the coding correction process. In addition, when the attention mechanism correction sub-network adopted by the FENet network is the Scse attention mechanism, the accuracy of model prediction can be further improved.

Therefore, when the preset flood extraction convolutional neural network includes a preset residual neural network and a preset void convolutional network, step S120 may be implemented by:

firstly, inputting the SAR image to be extracted of the water body into a preset residual error neural network for residual error convolution and attention mechanism correction to obtain a first water body information distribution map after the flood disaster water body edge is subjected to fine correction; and then inputting the first water body information distribution map into a preset cavity convolution network for change processing of a preset expansion scale, so as to obtain a flood disaster water body information distribution map with complete flood disaster water body edge extraction.

Specifically, when the terminal device obtains the to-be-extracted SAR image of the water body, the to-be-extracted SAR image of the water body may be input into a preset residual neural network for residual convolution and attention correction, that is, the preset residual neural network performs residual convolution on the to-be-extracted SAR image of the water body first, and then performs attention correction on the SAR image after the residual convolution, so as to obtain a first water body information distribution map after the flood disaster water body edge is subjected to fine correction, where the first water body information distribution map includes water body information with obvious characteristics and water body information after the water body edge is subjected to fine correction.

Further, when the terminal device determines that the preset residual error neural network outputs the first water body information distribution map, the first water body information distribution map can be further input into the preset cavity convolution neural network to be expanded at a plurality of different expansion rates, so that the characteristic information of the edge of the water body with more details and unobvious characteristics in the SAR image to be extracted from the water body can be captured, and the water body information is richer and comprehensive.

According to the flood disaster water body extraction method provided by the invention, the water body information with obvious water body characteristics is extracted through the preset residual error neural network with the residual error convolution and attention mechanism correction functions, and more small water body information which is not obvious in water body characteristics and is easy to ignore is extracted through the preset cavity convolution neural network with different expansion processing functions, so that the accuracy and the reliability of the flood disaster water body information extraction are improved.

Optionally, the specific process of inputting the SAR image to be extracted from the water body into the preset residual error neural network for residual error convolution and attention mechanism correction when the preset residual error neural network includes P residual error subnetworks and each residual error subnetwork is connected with one attention mechanism correction subnetwork, so as to obtain the first water body information distribution map after flood disaster water body edge fine correction includes:

Specifically, when the preset residual neural network is a Resnet18 network including 4 residual subnetworks and each attention mechanism correction subnetwork represents a se attention mechanism, the terminal device may, for the obtained SAR image to be extracted from the water body, input the SAR image to be extracted from the water body into a 1 st residual subnetwork to perform convolution, batch normalization and linear arrangement to obtain a 1 st residual convolution image, and then input the 1 st residual convolution image into the 1 st attention mechanism correction subnetwork to perform channel correction and spatial correction to obtain a 1 st correction image; inputting the 1 st corrected image into a 2 nd residual error sub-network for convolution, batch normalization and linear arrangement to obtain a 2 nd residual error convolution image, and inputting the 2 nd residual error convolution image into a 2 nd attention mechanism correction sub-network for channel correction and space correction to obtain a 2 nd corrected image; inputting the 2 nd corrected image into the 3 rd residual error sub-network for convolution, batch normalization and linear arrangement to obtain a 3 rd residual error convolution image, and inputting the 3 rd residual error convolution image into the 3 rd attention mechanism correction sub-network for channel correction and space correction to obtain a 3 rd corrected image; and inputting the 3 rd corrected image into a 4 th residual error sub-network for convolution, batch normalization and linear arrangement to obtain a 4 th residual error convolution image, inputting the 4 th residual error convolution image into a 4 th attention mechanism correction sub-network for channel correction and spatial correction to obtain a 4 th corrected image, wherein the obtained 4 th corrected image is the first water body information distribution map after flood disaster water body edge fine correction, and the first water body information distribution map is also a characteristic image obtained after P steps (such as 4 steps) coding correction.

In the actual processing process, each residual sub-network may perform convolution, batch normalization, and linear arrangement as shown in fig. 3, and may perform convolution twice, batch normalization twice, and linear arrangement twice as shown in (a) of fig. 3, and then directly add the result to the image x to obtain a corrected image, or may perform convolution twice, batch normalization twice, and linear arrangement twice as shown in (b) of fig. 3, and then add the result to the image x after convolution once to obtain a corrected image, where x is an SAR image to be extracted from the water body or any one of the 1 st corrected image to the P-1 st corrected image, y is any one of the 1 st corrected image to the P-th corrected image, and Relu represents linear arrangement.

Furthermore, each attention mechanism correction sub-network may perform the channel correction and the spatial correction processing of each residual convolution image as shown in fig. 4 and the size of each residual convolution image is (C, H, W), respectively, C denotes the color channel of each residual convolution image, H denotes the height of each residual convolution image, and W denotes the width of each residual convolution image; the channel correction processing is to perform global average pooling processing on the residual convolution image, then perform activation processing on a first feature image obtained after two convolutions by using a relu function and a sigmoid function respectively, and finally perform channel correction in a channel multiplication mode to obtain a first image after channel correction; the spatial correction is to perform convolution operation on the residual convolution image, then use a sigmoid function to perform activation processing, and finally perform spatial information correction in a spatial multiplication mode to obtain a second image after spatial correction; furthermore, the first image after channel correction and the second image after spatial correction are combined in a channel addition mode to obtain a corrected image after channel correction and spatial correction, and the size of the corrected image is the same as that of the corresponding residual convolution image. In the invention, after each attention mechanism correction sub-network is added to each residual error sub-network, the dimension of each residual error convolution image cannot be changed, so that the aim of accurately correcting the water body edge in each residual error convolution image is fulfilled.

It should be noted that, although the commonly used deep residual error network includes Resnet18, Resnet34, Resnet50, Resnet101 and Resnet152, in order to reduce the complexity of the network model and increase the processing speed of the network model, the present invention adopts the Resnet18 network and selects 4 residual error subnetworks in the Resnet18 network to achieve the purpose of extracting water body information with obvious water body characteristics, which greatly improves the utilization efficiency of the Resnet18 network, and when each residual error subnetwork is connected with a Scse attention machine mechanism, the accuracy of model prediction can be further improved.

According to the flood disaster water body extraction method, the purpose of quickly and accurately extracting water body information with obvious characteristics is achieved through the mode that P residual error sub-networks are connected with the attention mechanism correction sub-network respectively, and therefore the purpose of greatly improving the prediction accuracy of a network model on the basis of accurately correcting the water body edge can be achieved.

Optionally, when the preset cavity convolution network includes Q different expansion rates, the first water information distribution map is input into the preset cavity convolution network to perform change processing of a preset expansion scale, and a specific process of extracting a complete flood disaster water information distribution map from the edge of the flood disaster water is obtained, which may include:

inputting the first water body information distribution map into a preset cavity convolution network for different expansion scale processing to obtain Q fine water body characteristic information maps with different expansion scales; performing fusion processing on the fine water body characteristic information graph to obtain a fusion image; and decoding the fused image corresponding to the residual convolution and attention mechanism correction to obtain a flood disaster water body information distribution map with complete flood disaster water body edge extraction.

Specifically, when the preset cavity convolution network includes 4 different expansion rates, the terminal device may input the first water information distribution map into the preset cavity convolution network for performing cavity convolution processing at different expansion rates on the first water information distribution map after the flood disaster water edge is subjected to fine correction, so as to obtain 4 feature images with different expansion scales, further perform fusion processing (such as channel addition processing) on the 4 feature images with different expansion scales, obtain a fusion image, and finally perform P-step decoding operation corresponding to the P-step coding correction on the fusion image, so as to obtain a flood disaster water information distribution map with the same size as the original water to-be-extracted SAR image, where the flood disaster water information distribution map includes more obvious water information and more detailed and less obvious water information, fig. 5 shows schematic diagrams of convolution of holes with different expansion rates, where in fig. 5, (a) is a schematic diagram of holes and holes at an expansion rate of 1, (b) is a schematic diagram of convolution of holes at an expansion rate of 2, and (c) is a schematic diagram of convolution of holes at an expansion rate of 4, where a convolution kernel with an expansion rate of 1 can be regarded as a standard convolution kernel. Because the cavity convolutions with different expansion rates have different receptive fields, the different receptive fields are very important for distinguishing mountain shadows and water bodies with similar spectral characteristics, but if the expansion rate is set to be too large, the characteristic information of the small water bodies is lost too much. Therefore, the invention combines the hole convolutions with the expansion rates of 1, 2, 4 and 8, extracts the semantic features with different scales, and realizes the purpose of more completely extracting the water body edge by capturing the feature information of more tiny water body rivers when the expansion rate is properly reduced.

According to the flood disaster water body extraction method provided by the invention, through the modes of carrying out cavity convolution processing of different expansion scales on the first water body information distribution diagram and carrying out fusion processing and decoding processing on the characteristic images corresponding to the expansion scales, the purpose that more tiny water body information with unobvious water body characteristics can be captured on the premise that more obvious water body information with obvious characteristics in the SAR images of the flood disaster area can be accurately and accurately extracted is realized, and the extraction precision and the extraction efficiency of the flood disaster water body are greatly improved.

Optionally, in order to improve flexibility of training the initial flood extraction convolutional neural network and improve reliability of the preset flood extraction convolutional neural network, a training process of the preset flood extraction convolutional neural network may include:

acquiring a training sample image set; the training sample image set comprises training sample images, and the training sample images are SAR images obtained by carrying out sample labeling, regular grid cutting and data enhancement processing on the SAR images to be extracted from the water body; and training a preset initial flood extraction convolutional neural network according to the training sample image set to obtain the preset flood extraction convolutional neural network.

Specifically, the terminal device trains a preset initial flood extraction convolutional neural network, which can also be regarded as a process of training a preset initial FENet network, and the initial FENet network is composed of an initial Resnet18 network and an initial hole convolutional network.

In addition, before training the initial FENet network, a training sample image set may be obtained from the SAR image to be extracted from the water body, where the obtaining process of the training image set includes:

firstly, cutting out two typical areas from an SAR image to be extracted of a water body as a training image and a test image respectively, wherein the two typical areas do not have an intersection and the conditions met by each typical area comprise: the water body containing different types, such as calm open water body, water body submerging ground objects, small river water body and the like, the more the water body types, the better; the larger the size of each representative region, the better.

Then carry out the sample mark processing that becomes more meticulous respectively to training image and test image, obtain the training label image that the training image corresponds and the test label image that the test image corresponds, its sample mark processing procedure that becomes more meticulous includes: and (3) combining the existing high-resolution optical remote sensing image data as assistance, and respectively carrying out boundary marking on the suspected water bodies in the training image and the test image based on the experience of researchers so as to obtain a training label image and a test label image.

Then, regular grid cutting and data enhancement are carried out on the training images and the training label images to obtain a training sample image set comprising M training sample images; and performing regular grid cutting and data enhancement on the test images and the test label images to obtain a test sample image set comprising N test sample images. The data enhancement comprises horizontal turning, vertical turning and diagonal mirror image operation on the image and the corresponding label image, and the data enhancement is to obtain sufficient samples and avoid the overfitting phenomenon in the network training process.

Taking a test image as an example to perform regular grid cutting description, the process comprises the following steps: each sample is cut from left to right and from top to bottom in the test image according to a preset sample size, overlapping regions exist between adjacent samples in the cut samples, and the size of the overlapping regions is preset (for example, 20). And (3) clipping samples adjacent to the boundary of the row or the column, wherein the last row is clipped from top to bottom (or from bottom to top) to obtain samples with a preset sample size, and the last column is also clipped from left to right (or from right to left) to obtain samples with a preset sample size, and for such samples, the overlapping area between the samples and the adjacent samples is smaller than a preset size (for example, smaller than 20).

And finally, training the initial FENet network by using the training sample image set until the trained network reaches the preset network precision, thus obtaining the preset flood extraction convolutional neural network.

According to the flood disaster water body extraction method provided by the invention, the training sample image set for training the initial flood extraction convolutional neural network is acquired from the SAR image to be extracted of the water body in a mode of training the initial flood extraction convolutional neural network from the training sample image set acquired from the SAR image to be extracted of the water body, and the SAR image to be extracted of the water body comprises the dual polarization wave band and the spectral characteristic wave band, so that the network precision and the network reliability can be improved by training the initial flood extraction convolutional neural network by using the training sample image set, and a foundation is laid for quickly and accurately extracting water body information subsequently.

Optionally, training a preset initial flood extraction convolutional neural network according to the training sample image set, to obtain a specific process of the preset flood extraction convolutional neural network, which may include:

performing iterative training on a preset initial flood extraction convolutional neural network according to the training sample image set to obtain an intermediate flood extraction convolutional neural network obtained after the iterative training; judging whether the cumulative training round number corresponding to the current round of iterative training reaches a preset round number threshold value or not; if the accumulated training round number reaches the preset round number threshold value, taking the intermediate flood extraction convolutional neural network obtained after the iterative training as the preset flood extraction convolutional neural network; and if the accumulated training round number does not reach the preset round number threshold value, training the intermediate flood extraction convolutional neural network to obtain the preset flood extraction convolutional neural network.

Specifically, the terminal device trains an initial flood extraction convolutional neural network by using M training sample images, each training sample image is called to be trained once after the FENet network is trained once, and each training sample image is called to be trained once after the FENet network is trained once. After the terminal equipment trains the preset number of rounds by using M training sample images, whether the accumulated number of training rounds after the training round reaches the preset number of rounds threshold value or not can be judged, the training is stopped when the accumulated number of training rounds reaches the preset number of rounds threshold value, and the corresponding network when the test sample image set comprising N test sample images is used for testing and training is stopped is used for determining that the values of 5 indexes, namely Precision (Precision), Recall (Recall), F1 index (F1), water body cross-over-parallel ratio (IoU) and cross-over-parallel ratio (mIoU), are all the highest, at the moment, the corresponding network is determined to be converged when the training is stopped, namely the preset flood extraction convolutional neural network is obtained; otherwise, when the cumulative training round number does not reach the preset round number threshold, continuing to execute the preset round number training of the network obtained after the training round until the cumulative training round number reaches the preset round number threshold, and obtaining the preset flood extraction convolutional neural network. The calculation formulas of the 5 indexes are respectively as follows:

wherein n represents the number of network-predicted water body categories, TP represents the number of actual water body pixels of the water body predicted by the network, FP represents the number of actual non-water body pixels of the water body predicted by the network, FN represents the number of actual water body pixels of the non-water body predicted by the network, Precision represents Precision, Recall represents Recall rate, F1 represents an F1 index, mIoU represents an average cross-over-parallel ratio, and IoU represents a water body cross-over-parallel ratio; the F1 index comprehensively considers the precision rate and the recall rate, and the higher the value is, the better the model extraction effect is; the average cross ratio mIoU comprehensively considers the cross ratio of the model extraction water body and the cross ratio of the non-water body, and the higher the value is, the better the model extraction effect is; both the F1 index and the mIoU index are comprehensive evaluation indexes for measuring the network model.

It should be noted that, since training the fent network once for each training sample image is referred to as performing training once, and training the fent network once for all the M training sample images is referred to as performing training once, it is known that when a training process is performed for each training sample image, a training process is performed for all the M training sample images for each training sample image.

The process of performing one training for each training sample image comprises the following steps: the method comprises the steps that a training sample image enters an initial flood extraction convolutional neural network to be subjected to P-step coding correction, when the value of P is 4 and the initial flood extraction convolutional neural network comprises 4 initial residual error sub-networks and each initial residual error sub-network is connected with an initial Scse attention mechanism correction sub-network, the training sample image is subjected to first-step residual error convolution and attention correction processing to obtain a first correction coding image; carrying out second-step residual convolution and attention correction processing on the first correction coded image to obtain a second correction coded image; carrying out residual convolution and attention correction processing in the third step on the second correction coded image to obtain a third correction coded image; and performing residual convolution and attention correction processing in the fourth step on the third corrected and encoded image to obtain a fourth corrected and encoded image, wherein the fourth encoded image can be called a feature image obtained after encoding and correction.

And then performing hole convolution processing on the characteristic image map, wherein Q different expansion rates (preferably 4 expansion rates, such as 1, 2, 4 and 8) can be selected, performing hole convolution processing on the characteristic image map to obtain Q characteristic images with different expansion scales, performing fusion processing (such as channel addition) on each characteristic image with different expansion scales, and performing P-step decoding corresponding to the P-step coding correction on the image map subjected to fusion processing to obtain a decoded image which has the same size as the original training sample image and is obtained after the initial flood extraction convolutional neural network is trained once.

According to the flood disaster water body extraction method provided by the invention, the aim of quickly and efficiently obtaining the preset flood extraction convolutional neural network is realized by training whether the accumulated round number of the preset initial flood extraction convolutional neural network reaches the preset round number threshold value or not through the training sample image set, so that the network precision is ensured, the training complexity is greatly reduced, and the network training speed is accelerated.

Optionally, the process of determining the preset round number threshold may include:

dividing the training sample image set into training samples and verification samples according to a preset proportion; training a preset initial flood extraction convolutional neural network according to the training samples to obtain an intermediate flood extraction convolutional neural network obtained after training in preset rounds; verifying the intermediate flood extraction convolutional neural network according to the verification sample to obtain a value of an evaluation index of the intermediate flood extraction convolutional neural network; judging whether the value of the evaluation index reaches a preset standard value or not; if the value of the evaluation index reaches the preset standard value, taking the accumulated training round number corresponding to the intermediate flood extraction convolutional neural network as the preset round number threshold value; and if the value of the evaluation index does not reach the preset standard value, training the intermediate flood extraction convolutional neural network to obtain the preset round number threshold value.

The preset standard value can be used for representing the loss value, the precision and other parameters of the network, and the values of the parameters are enough to indicate that the network meets the convergence requirement.

Specifically, the basis for the terminal device to determine whether the network after the current round of training converges is to compare the accumulated number of training rounds after the current round of training with a preset round number threshold. Therefore, the specific value of the preset round number threshold is important.

For the determination process of the preset round number threshold, the M training sample images may be divided into training samples and verification samples according to a preset ratio, where the preset ratio may be 6:4 or 7:3, or other ratios as long as the number of training samples is ensured to be greater than the number of verification samples.

After the terminal device trains a preset initial flood extraction convolutional neural network preset turn number by using a training sample, obtaining an intermediate flood extraction convolutional neural network obtained after the preset turn number training, and evaluating the intermediate flood extraction convolutional neural network by using a verification sample to obtain an evaluation index value of the intermediate flood extraction convolutional neural network, wherein the evaluation index value can comprise a loss function value, network precision and the like; when the value of the evaluation index of the intermediate flood extraction convolutional neural network is determined to reach a preset standard value, taking the accumulated training round number corresponding to the intermediate flood extraction convolutional neural network as a preset round number threshold value; when it is determined that the value of the evaluation index of the intermediate flood extraction convolutional neural network does not reach the preset standard value, the intermediate flood extraction convolutional neural network can be trained again by using the training sample until the value of the evaluation index verified by the trained network using the verification sample reaches the preset standard value.

According to the flood disaster water body extraction method provided by the invention, the preset round threshold value is accurately and reliably determined by the way of training the network for a certain round by using the training samples in the training sample image set and verifying whether the network meets the precision requirement by using the verification samples in the training sample image set, so that a powerful basis is provided for quickly and accurately obtaining the preset flood extraction convolutional neural network in the follow-up process.

In the actual processing process, in order to verify the effectiveness of the method, the method is compared and analyzed with a traditional flood monitoring method based on SAR images, and two typical areas as shown in fig. 6 can be selected, wherein the water content of the first area is 33.4%, and the water content of the second area is 5.8%. The method provided by the invention is tested by obtaining real water body information through visual interpretation under the assistance of a high-resolution optical remote sensing image, comparing the artificially marked real water body information with water body information obtained through an algorithm, and quantitatively evaluating the accuracy of water body information extraction by adopting various evaluation indexes, so that the method provided by the invention is particularly characterized by comprising the steps of recall ratio, precision ratio, false alarm rate and false alarm rate

Evaluation indexes four indexes are used for testing the effectiveness and reliability of the method provided by the invention, and the calculation formulas of the four indexes are as follows:

F_a＝1-P (8)

wherein, P (T)_W) Real water body picture element, P (A) representing artificial visual interpretation_W) Representing the water body pixels extracted by using a method; r represents the recall ratio of the extraction method, the recall ratio represents the degree that the range of the extracted water body is close to the real water body, and the higher the value is, the higher the value isGood; p represents the precision ratio of the extraction method, the precision ratio represents the accuracy of the extracted water body, and the higher the value is, the better the value is; f_aRepresenting the false alarm rate of the extraction method;

the comprehensive evaluation index of the extraction method is expressed, the recall ratio and the precision ratio are comprehensively considered, and the higher the value is, the better the effect of the method is expressed; the three methods include the method of the present invention, the Otsu method (i.e., Otsu global threshold method), and the object-oriented method, and the extraction accuracy pairs of the three methods are shown in table 1.

TABLE 1

Based on fig. 6 and table 1, the method of the present invention is examined to find the advantages and disadvantages of the conventional method with respect to different water body ratios in selected comparison areas. The result shows that the Otsu global threshold value method has high recall ratio and precision ratio and good water body extraction effect under the condition of high water body occupation ratio; under the condition of low water occupation ratio, the Otsu global threshold method is the worst in water extraction precision and the worst in water extraction accuracy; the object-oriented method has certain advantages in the aspect of small water body extraction compared with an Otsu global threshold method, but the extraction precision is not as good as that of the preset flood extraction convolutional neural network in the method; for the water body information extraction condition of large-range flood disasters, the object-oriented method has the worst extraction precision. Among the three methods, the preset flood extraction convolutional neural network in the method has the best effect and the highest precision. For the extraction of water body information in a large-scale macroscopic view, an Otsu global threshold method is difficult to find a proper threshold value for extracting the water body information, and a large amount of prior knowledge is needed for a segmentation rule and a classification rule of an object-oriented method. Compared with the traditional flood monitoring method, the accuracy of the preset flood extraction convolutional neural network in the method is higher and the effect is better.

The following describes the flood disaster water body extraction device provided by the present invention, and the flood disaster water body extraction device described below and the flood disaster water body extraction method described above can be referred to in correspondence with each other.

Fig. 7 illustrates a flood disaster water body extracting apparatus, as shown in fig. 7, the flood disaster water body extracting apparatus 700, including: the acquiring module 710 is configured to acquire an SAR image to be extracted from a water body, where the SAR image to be extracted from the water body includes an image composed of a dual polarization band and a spectral characteristic band of an SAR image of a flood disaster; the determining module 720 is configured to input the to-be-extracted SAR image of the water body into a preset flood extraction convolutional neural network, so as to obtain a flood disaster water body information distribution map; the preset flood extraction convolution neural network is used for performing residual convolution and attention mechanism correction on the SAR image to be extracted from the water body, and performing cavity convolution on the corrected SAR image to be extracted from the water body to obtain the flood disaster water body information distribution map.

Optionally, the determining module 720 may be specifically configured to input the to-be-extracted SAR image of the water body into a preset residual neural network for residual convolution and attention mechanism correction, so as to obtain a first water body information distribution map after flood disaster water body edge fine correction; and inputting the first water body information distribution map into a preset cavity convolution network to carry out change processing of a preset expansion scale, so as to obtain a flood disaster water body information distribution map with complete flood disaster water body edge extraction.

Optionally, the determining module 720 may be further configured to input the to-be-extracted SAR image of the water body into a q-th residual sub-network for convolution, batch normalization, and linear arrangement processing, so as to obtain a q-th residual convolution image; inputting the q-th residual convolution image into a q-th attention mechanism correction sub-network for channel correction and space correction processing to obtain a q-th corrected image; inputting the q-th corrected image into a q + 1-th residual error sub-network for convolution, batch normalization and linear arrangement processing to obtain a q + 1-th residual error convolution image; inputting the (q + 1) th residual convolution image into a (q + 1) th attention mechanism correction sub-network for channel correction and spatial correction processing to obtain a (q + 1) th corrected image; adding 1 to the value q, and repeatedly executing the step of inputting the (q + 1) th residual convolution image into a (q + 1) th attention mechanism correction sub-network for channel correction and space correction processing until a P-th corrected image is obtained, wherein the P-th corrected image is a first water body information distribution map after flood disaster water body edge fine correction; q belongs to [1, …, P ], q +1 belongs to [2, …, P ].

Optionally, the determining module 720 may be further configured to input the first water information distribution map into a preset cavity convolution network for different expansion scale processing, so as to obtain Q fine water characteristic information maps with different expansion scales; performing fusion processing on the fine water body characteristic information graph to obtain a fusion image; and decoding the fused image corresponding to the residual convolution and attention mechanism correction to obtain a flood disaster water body information distribution map with complete flood disaster water body edge extraction.

Optionally, the determining module 720 may be further configured to obtain a training sample image set; the training sample image set comprises training sample images, and the training sample images are SAR images obtained by carrying out sample labeling, regular grid cutting and data enhancement processing on the SAR images to be extracted from the water body; and training a preset initial flood extraction convolutional neural network according to the training sample image set to obtain the preset flood extraction convolutional neural network.

Optionally, the determining module 720 may be further configured to perform iterative training on a preset initial flood extraction convolutional neural network according to the training sample image set, and obtain an intermediate flood extraction convolutional neural network obtained after the iterative training of the current round; judging whether the cumulative training round number corresponding to the current round of iterative training reaches a preset round number threshold value or not; if the accumulated training round number reaches the preset round number threshold value, taking the intermediate flood extraction convolutional neural network obtained after the iterative training as the preset flood extraction convolutional neural network; and if the accumulated training round number does not reach the preset round number threshold value, training the intermediate flood extraction convolutional neural network to obtain the preset flood extraction convolutional neural network.

Optionally, the determining module 720 may be further configured to divide the training sample image set into a training sample and a verification sample according to a preset ratio; training a preset initial flood extraction convolutional neural network according to the training samples to obtain an intermediate flood extraction convolutional neural network obtained after training in preset rounds; verifying the intermediate flood extraction convolutional neural network according to the verification sample to obtain a value of an evaluation index of the intermediate flood extraction convolutional neural network; judging whether the value of the evaluation index reaches a preset standard value or not; if the value of the evaluation index reaches the preset standard value, taking the accumulated training round number corresponding to the intermediate flood extraction convolutional neural network as the preset round number threshold value; and if the value of the evaluation index does not reach the preset standard value, training the intermediate flood extraction convolutional neural network to obtain the preset round number threshold value.

Fig. 8 illustrates a physical structure diagram of an electronic device, and as shown in fig. 8, the electronic device 800 may include: a processor (processor)810, a communication Interface 820, a memory 830 and a communication bus 840, wherein the processor 810, the communication Interface 820 and the memory 830 communicate with each other via the communication bus 840. The processor 810 may invoke logic instructions in the memory 830 to perform a flood disaster water body extraction method comprising: acquiring an SAR image to be extracted from a water body, wherein the SAR image to be extracted from the water body comprises an image consisting of a dual polarization wave band and a spectral characteristic wave band of an SAR image of a flood disaster; inputting the SAR image to be extracted of the water body into a preset flood extraction convolutional neural network to obtain a flood disaster water body information distribution map; the preset flood extraction convolution neural network is used for performing residual convolution and attention mechanism correction on the SAR image to be extracted from the water body, and performing cavity convolution on the corrected SAR image to be extracted from the water body to obtain the flood disaster water body information distribution map.

In addition, the logic instructions in the memory 830 may be implemented in software functional units and stored in a computer readable storage medium when the logic instructions are sold or used as independent products. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.

In another aspect, the present invention also provides a computer program product comprising a computer program stored on a non-transitory computer readable storage medium, the computer program comprising program instructions, which when executed by a computer, enable the computer to perform the flood disaster water body extraction method provided by the above methods, the method comprising: acquiring an SAR image to be extracted from a water body, wherein the SAR image to be extracted from the water body comprises an image consisting of a dual polarization wave band and a spectral characteristic wave band of an SAR image of a flood disaster; inputting the SAR image to be extracted of the water body into a preset flood extraction convolutional neural network to obtain a flood disaster water body information distribution map; the preset flood extraction convolution neural network is used for performing residual convolution and attention mechanism correction on the SAR image to be extracted from the water body, and performing cavity convolution on the corrected SAR image to be extracted from the water body to obtain the flood disaster water body information distribution map.

In yet another aspect, the present invention also provides a non-transitory computer readable storage medium having stored thereon a computer program, which when executed by a processor, is implemented to perform the flood disaster water body extraction methods provided above, the method comprising: acquiring an SAR image to be extracted from a water body, wherein the SAR image to be extracted from the water body comprises an image consisting of a dual polarization wave band and a spectral characteristic wave band of an SAR image of a flood disaster; inputting the SAR image to be extracted of the water body into a preset flood extraction convolutional neural network to obtain a flood disaster water body information distribution map; the preset flood extraction convolution neural network is used for performing residual convolution and attention mechanism correction on the SAR image to be extracted from the water body, and performing cavity convolution on the corrected SAR image to be extracted from the water body to obtain the flood disaster water body information distribution map.

The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.

Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware. With this understanding in mind, the above-described technical solutions may be embodied in the form of a software product, which can be stored in a computer-readable storage medium such as ROM/RAM, magnetic disk, optical disk, etc., and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in the embodiments or some parts of the embodiments.

Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.

Claims

1. A flood disaster water body extraction method is characterized by comprising the following steps:

2. The method according to claim 1, wherein the preset flood extraction convolutional neural network comprises a preset residual error neural network and a preset cavity convolutional network, and the inputting the SAR image to be extracted from the water body into the preset flood extraction convolutional neural network to obtain the flood disaster water body information distribution map comprises:

3. The method for extracting a flood disaster water body according to claim 2, wherein the preset residual error neural network comprises P residual error sub-networks, and when each residual error sub-network is connected to one attention mechanism correction sub-network, the SAR image to be extracted from the water body is input into the preset residual error neural network for residual error convolution and attention mechanism correction, so as to obtain the first water body information distribution map after the flood disaster water body edge is corrected, including:

4. The method according to claim 2, wherein when the preset cavity convolution network includes Q different expansion rates, the first water information distribution map is input into the preset cavity convolution network to perform change processing of a preset expansion scale, so as to obtain a complete flood disaster water information distribution map extracted from the edge of the flood disaster water, and the method includes:

5. The flood disaster water body extraction method according to any one of claims 1 to 4, wherein the training process of the preset flood extraction convolutional neural network comprises:

6. The method of claim 5, wherein the training a preset initial flood extraction convolutional neural network according to the training sample image set to obtain the preset flood extraction convolutional neural network comprises:

7. The flood disaster water body extraction method according to claim 6, wherein the determination process of the preset turn number threshold value comprises:

8. The utility model provides a flood disaster water extraction element which characterized in that includes:

9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor when executing the program performs the steps of the flood disaster water body extraction method according to any one of claims 1 to 7.

10. A non-transitory computer readable storage medium having stored thereon a computer program, wherein the computer program when executed by a processor implements the steps of the flood disaster water extraction method according to any one of claims 1 to 7.