Disaster condition assessment method and device based on multi-temporal images
By training models with pre- and post-disaster satellite images, the differences in disaster situation can be assessed, solving the problem of inaccurate disaster assessment in existing technologies and achieving more efficient disaster assessment.
Patent Information
- Application Number
- CN202210098620.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-01-27
- Publication Date
- 2025-11-28
- Estimated Expiration
- 2042-01-27
AI Technical Summary
Existing technologies cannot monitor complex disaster situations in real time, cannot effectively assess disaster situations, and cannot accurately evaluate satellite imagery.
By acquiring historical disaster prediction methods and devices, and employing multi-temporal image evaluation methods and devices, the accuracy of disaster prediction can be improved.
By acquiring pre-disaster and post-disaster images, models are trained to extract differences in disaster conditions, improve the accuracy and real-time nature of assessments, and provide effective references for emergency personnel.
Smart Images
Figure CN114494907B_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of disaster assessment, and in particular to a disaster assessment method and device based on multi-temporal images. BACKGROUND
[0002] Natural disasters such as floods, forest fires, and earthquakes always threaten people's life and property safety. In real life, one natural disaster often causes multiple disaster conditions. For example, floods are the most economically damaging natural disaster due to their widespread geographical distribution and frequent occurrence. Roads, railways, bridges, power lines, and natural gas lines are all damaged to varying degrees, and further mudslides or landslides, building or dam collapses may occur.
[0003] For natural disasters, it is crucial for on-site emergency teams to reduce response time, respond quickly, and take action to reduce losses and save lives. In addition, in order to better deploy resources in disaster areas, emergency personnel must understand the exact location and severity of the loss, which is also important. Currently, emergency personnel usually assess the extent of disaster damage by manually observing satellite images, but the assessment process can take several hours, which is extremely disadvantageous for rescue work.
[0004] The current ground monitoring system is limited by coverage and convenience, and cannot meet the more refined monitoring needs, and there is a problem of insufficient coverage and convenience. Today, unmanned aerial vehicles and satellite remote sensing are used in disaster relief and post-disaster reconstruction work. Among them, satellite remote sensing technology has become a necessary supplement, and its high repetition frequency and large-scale synchronous information collection capability have filled the gaps in previous monitoring to a greater extent, providing more extensive and accurate real-time data for indicators such as the affected area and extent of the target area.
[0005] However, the existing disaster assessment method based on neural networks only relies on post-disaster image memory training and cannot obtain disaster condition difference information. For satellite images, the range is too wide, and it is difficult to accurately assess and predict complex disaster conditions, making it difficult to provide effective recommendations for staff. SUMMARY
[0006] In view of the problems in the prior art, the main purpose of the embodiments of the present application is to provide a disaster assessment method and device based on multi-temporal images to improve the accuracy of disaster prediction.
[0007] In order to achieve the above purpose, the embodiments of the present application provide a disaster assessment method based on multi-temporal images, which comprises:
[0008] obtain historical pre-disaster images and historical post-disaster images, input the historical pre-disaster images and the historical post-disaster images into a preset initial evaluation model for processing to obtain a disaster condition prediction result;
[0009] update the initial evaluation model according to the disaster condition prediction result to obtain a disaster condition evaluation model;
[0010] input obtained actual pre-disaster images and actual post-disaster images into the disaster condition evaluation model for processing to obtain a disaster condition evaluation result.
[0011] Optionally, in an embodiment of the present application, the initial evaluation model comprises an initial feature model and an initial semantic model.
[0012] Optionally, in an embodiment of the present application, the inputting the historical pre-disaster images and the historical post-disaster images into the preset initial evaluation model for processing to obtain the disaster condition prediction result comprises:
[0013] inputting the historical pre-disaster images and the historical post-disaster images into a preset initial feature model for processing to obtain a first feature map corresponding to the historical pre-disaster images and a second feature map corresponding to the historical post-disaster images;
[0014] splicing the first feature map and the second feature map by using a channel splicing manner to obtain a feature splicing map;
[0015] inputting the feature splicing map into a preset initial semantic model for processing to obtain the disaster condition prediction result.
[0016] Optionally, in an embodiment of the present application, the disaster condition prediction result comprises a disaster condition prediction level corresponding to each pixel coordinate in the feature splicing map.
[0017] Optionally, in an embodiment of the present application, the updating the initial evaluation model according to the disaster condition prediction result to obtain the disaster condition evaluation model comprises:
[0018] updating the initial feature model and the initial semantic model according to the disaster condition evaluation result and a preset loss function to obtain a feature extraction model and a semantic segmentation model, and taking the feature extraction model and the semantic segmentation model as the disaster condition evaluation model.
[0019] Optionally, in an embodiment of the present application, the inputting the obtained actual pre-disaster images and actual post-disaster images into the disaster condition evaluation model for processing to obtain the disaster condition evaluation result comprises:
[0020] inputting the obtained actual pre-disaster images and actual post-disaster images into the feature extraction model for processing to obtain a pre-disaster feature map and a post-disaster feature map;
[0021] perform feature map splicing processing on the pre-disaster feature map and the post-disaster feature map to obtain a disaster condition feature map;
[0022] input the disaster condition feature map into the semantic segmentation model for processing to obtain the disaster condition assessment result; wherein the disaster condition assessment result includes a disaster condition prediction level corresponding to each pixel coordinate in the disaster condition feature map.
[0023] The embodiment of the present application also provides a disaster condition assessment device based on multi-temporal images, which comprises:
[0024] a historical image module configured to acquire historical pre-disaster images and historical post-disaster images, input the historical pre-disaster images and the historical post-disaster images into a preset initial assessment model for processing to obtain a disaster condition prediction result;
[0025] an assessment model module configured to update the initial assessment model according to the disaster condition prediction result to obtain a disaster condition assessment model;
[0026] an assessment result module configured to input acquired actual pre-disaster images and actual post-disaster images into the disaster condition assessment model for processing to obtain a disaster condition assessment result.
[0027] Optionally, in an embodiment of the present application, the initial assessment model comprises an initial feature model and an initial semantic model.
[0028] Optionally, in an embodiment of the present application, the historical image module comprises:
[0029] a feature map unit configured to input the historical pre-disaster images and the historical post-disaster images into a preset initial feature model for processing to obtain a first feature map corresponding to the historical pre-disaster images and a second feature map corresponding to the historical post-disaster images;
[0030] an image splicing unit configured to splice the first feature map and the second feature map by using a channel splicing method to obtain a feature splicing map;
[0031] a prediction result unit configured to input the feature splicing map into a preset initial semantic model for processing to obtain the disaster condition prediction result; wherein the disaster condition prediction result includes a disaster condition prediction level corresponding to each pixel coordinate in the feature splicing map.
[0032] Optionally, in an embodiment of the present application, the assessment model module is further configured to update the initial feature model and the initial semantic model according to the disaster condition assessment result and a preset loss function to obtain a feature extraction model and a semantic segmentation model, and use the feature extraction model and the semantic segmentation model as the disaster condition assessment model.
[0033] Optionally, in an embodiment of the present application, the evaluation result module comprises:
[0034] a feature extraction unit, configured to input the acquired actual pre-disaster image and actual post-disaster image into the feature extraction model for processing to obtain a pre-disaster feature map and a post-disaster feature map;
[0035] a disaster condition feature map unit, configured to perform feature map splicing processing on the pre-disaster feature map and the post-disaster feature map to obtain a disaster condition feature map;
[0036] an evaluation result unit, configured to input the disaster condition feature map into the semantic segmentation model for processing to obtain the disaster condition evaluation result; wherein the disaster condition evaluation result comprises a disaster condition prediction level corresponding to each pixel coordinate in the disaster condition feature map.
[0037] The present application also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above method when executing the program.
[0038] The present application also provides a computer readable storage medium, which stores a computer program for executing the above method.
[0039] The present application trains the model by simultaneously using pre-disaster satellite images and post-disaster satellite images, so that the disaster condition evaluation model can effectively extract the differences before and after the disaster, improve the prediction accuracy of complex disaster conditions, and provide effective reference for staff. BRIEF DESCRIPTION OF DRAWINGS
[0040] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.
[0041] Figure 1 A flowchart of a disaster condition evaluation method based on multi-temporal images according to an embodiment of the present application;
[0042] Figure 2 A flowchart of obtaining a disaster condition prediction result according to an embodiment of the present application;
[0043] Figure 3 A flowchart of obtaining a disaster condition evaluation result according to an embodiment of the present application;
[0044] Figure 4 A schematic diagram of a model training process according to an embodiment of the present application;
[0045] Figure 5 FIG. 1 is a schematic diagram of a model application process in an embodiment of the present application;
[0046] Figure 6 FIG. 2 is a structural schematic diagram of a disaster condition assessment device based on multi-temporal images in an embodiment of the present application;
[0047] Figure 7 FIG. 3 is a structural schematic diagram of a historical image module in an embodiment of the present application;
[0048] Figure 8 FIG. 4 is a structural schematic diagram of an assessment result module in an embodiment of the present application;
[0049] Figure 9 FIG. 5 is a structural schematic diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0050] The embodiments of the present application provide a disaster condition assessment method and device based on multi-temporal images, which can be applied to the financial field and other fields. It should be noted that the disaster condition assessment method and device based on multi-temporal images of the present application can be applied to the financial field, and can also be applied to any field other than the financial field. The application field of the disaster condition assessment method and device based on multi-temporal images of the present application is not limited.
[0051] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.
[0052] With the rapid development of satellite remote sensing technology, the data sources for disaster ecological impact assessment and monitoring are also becoming more and more abundant. Due to different types of disasters, different satellite data are applicable. For meteorological disasters, which mostly occur in a large spatial range, satellite data with a resolution of kilometers, such as MODIS and Fengyun, can be used. For fires and other disasters occurring in natural ecological systems such as forests and grasslands, satellite data with a spatial resolution of 15-30 m, such as HJ-1A, Landsat, and Gaofen-1 wide coverage satellites, can be used. For geological disasters, which occur in a small spatial range, satellite data with a spatial resolution of 10 m or less, such as QuickBird and Worldview, are needed. At the same time, due to the flexibility of unmanned aerial vehicles, they are not limited by the running track and can quickly obtain real-time high-resolution remote sensing data of the disaster area, which can clearly obtain the current situation information of the disaster area. At the same time, the flight height is low and is not affected by clouds, which has become an important data source in disaster emergency monitoring.
[0053] Since the existing disaster assessment method can only output one prediction value for each image, it is too wide for satellite images, and it is difficult to provide effective suggestions for staff, and it lacks more accurate location disaster assessment. The existing post-disaster assessment method mostly assesses the severity of a single disaster, lacks assessment of complex disasters, and has low practical value. In fact, multiple disasters often occur simultaneously, for example, when a rainstorm disaster occurs, it will not only cause the water level to rise, but also further cause landslides or landslides, building or dam collapses, etc.
[0054] Using only post-disaster satellite images to train the model cannot effectively extract disaster difference information, and there is a lot of room for improvement in accuracy. In addition, the actual range of satellite images is relatively large, and the existing method often outputs one label for one photo, which is a coarse-grained prediction, and cannot accurately identify the disaster location.
[0055] To solve the above two problems, the present application proposes to use pre-disaster satellite images and post-disaster satellite images to train the model, so that the neural network can effectively extract the difference between before and after, and improve the prediction accuracy. At the same time, the present application also adopts a semantic segmentation model to output a label (i.e. disaster degree) for each pixel point of the satellite image, and the output prediction result is more accurate. As Figure 1 As shown in the flowchart of the disaster assessment method based on multi-temporal images according to an embodiment of the present application, the execution subject of the disaster assessment method based on multi-temporal images provided by the present application includes but is not limited to a computer. The present application trains the model by using pre-disaster satellite images and post-disaster satellite images, so that the disaster assessment model can effectively extract the difference between before and after, and improve the prediction accuracy of complex disaster assessment, and provide effective suggestions for staff. The method shown in the figure includes:
[0056] Step S1, obtain historical pre-disaster images and historical post-disaster images, input the historical pre-disaster images and historical post-disaster images into a preset initial assessment model for processing to obtain a disaster prediction result.
[0057] Among them, the multi-temporal images in the present application refer to the use of pre-disaster images and post-disaster images at the same time. Specifically, remote sensing technology can be used to adopt multi-source satellite remote sensing images, and the satellite remote sensing images collected before and after the disaster can be used as historical pre-disaster images and historical post-disaster images.
[0058] Further, the preset initial assessment model includes an initial feature model and an initial semantic model. The initial feature model is used to extract the feature map of the satellite remote sensing image, and the initial semantic model is a semantic segmentation model, which is used to output a label (i.e. disaster degree) for each pixel point of the satellite remote sensing image, so as to output an accurate disaster assessment result.
[0059] Further, the labeled historical pre-disaster images and historical post-disaster images are input into the initial evaluation model for model training to obtain a disaster condition prediction result corresponding to the historical satellite remote sensing image.
[0060] In step S2, the initial evaluation model is updated according to the disaster condition prediction result to obtain a disaster condition evaluation model.
[0061] The disaster condition prediction result includes a disaster condition prediction level corresponding to each pixel coordinate of the image, and the initial evaluation model is updated using the disaster condition prediction result. Specifically, cross-entropy is used as a loss function, and the weights of the initial evaluation model are updated by minimizing the loss function in the model training process, thereby obtaining the trained disaster condition evaluation model.
[0062] In step S3, the obtained actual pre-disaster image and actual post-disaster image are input into the disaster condition evaluation model for processing to obtain a disaster condition evaluation result.
[0063] When an actual disaster occurs, images before and after the disaster, i.e., actual pre-disaster images and actual post-disaster images, are obtained by remote sensing technology. The actual pre-disaster images and actual post-disaster images are input into the trained disaster condition evaluation model to obtain a disaster condition evaluation result.
[0064] Further, the disaster condition evaluation result includes a disaster condition prediction level corresponding to each pixel coordinate of the actual disaster satellite remote sensing image, thereby improving the accuracy, real-time performance and accuracy of disaster condition evaluation and providing an effective reference for staff.
[0065] As an embodiment of the present application, the initial evaluation model includes an initial feature model and an initial semantic model.
[0066] In this embodiment, as shown in the figure, the historical pre-disaster images and historical post-disaster images are input into the preset initial evaluation model for processing to obtain a disaster condition prediction result, which includes: Figure 2
[0067] In step S21, the historical pre-disaster images and historical post-disaster images are input into the preset initial feature model for processing to obtain a first feature map corresponding to the historical pre-disaster images and a second feature map corresponding to the historical post-disaster images.
[0068] In step S22, the first feature map and the second feature map are spliced by a channel splicing method to obtain a feature splicing map.
[0069] In step S23, the feature splicing map is input into the preset initial semantic model for processing to obtain a disaster condition prediction result. The disaster condition prediction result includes a disaster condition prediction level corresponding to each pixel coordinate in the feature splicing map.
[0070] In the method, multi-source images are acquired by satellites with different resolutions using remote sensing technology, and a disaster condition evaluation model with practical application value is trained by combining a machine learning algorithm, so that the position and degree of complex disaster conditions can be grasped in time and rescue resources can be optimized. Different time-phase remote sensing images, i.e., images before and after a disaster, are used as inputs of the model to train the model, a classic CNN feature extraction network U-Net is used as an initial feature model, and a semantic segmentation network is used as an initial semantic model, and a disaster level is output at each pixel level.
[0071] Further, historical pre-disaster images and historical post-disaster images are simultaneously input during the model training process, and a disaster level is predicted for a pixel-level position.
[0072] The specific model training process is as shown in Figure 4 The historical pre-disaster satellite images and the post-disaster satellite images are input into the same feature extraction network, i.e., the initial feature model, and the image size can be 512*512*3.
[0073] The feature extraction network can adopt a ResNet-50 network, and the last pooling layer and the full connection layer are removed, so as to effectively extract multi-time-phase image features, and the output feature map size is 7*7*256.
[0074] The feature map splicing adopts feature map channel splicing instead of direct addition. The size of the spliced feature map is 7*7*512.
[0075] The semantic segmentation network, i.e., the initial semantic model, can adopt a classic U-net network. First, down-sampling is performed, then up-sampling is performed, and finally a 512*512 matrix is output. Each coordinate in the matrix has a K-dimensional vector, i.e., a one-hot vector representation of a disaster condition level prediction probability. Here, K is the disaster condition level in the scene, which is set to 5, and the specific meaning is as shown in Table 1.
[0076] Table 1
[0077] Severity of disaster 0 (none) 1 (mild) 2 (moderate) 3 (severe) 4 (crisis)
[0078] In the embodiment, the initial evaluation model is updated according to the disaster condition prediction result to obtain the disaster condition evaluation model. The disaster condition evaluation model is obtained by updating the initial feature model and the initial semantic model according to the disaster condition evaluation result and a preset loss function, and the feature extraction model and the semantic segmentation model are used as the disaster condition evaluation model.
[0079] The loss function is used to update the weight of the model, specifically, cross-entropy is used as the loss function, and the weight of the initial feature model and the initial semantic model is updated by minimizing the loss function in the training process. The feature extraction model and the semantic segmentation model are obtained, and the feature extraction model and the semantic segmentation model are used as the disaster assessment model.
[0080] In the embodiment, as shown in Figure 3 the actual pre-disaster image and the actual post-disaster image are input into the disaster assessment model for processing to obtain a disaster assessment result, which includes:
[0081] In step S31, the obtained actual pre-disaster image and actual post-disaster image are input into the feature extraction model for processing to obtain a pre-disaster feature map and a post-disaster feature map.
[0082] In step S32, the pre-disaster feature map and the post-disaster feature map are processed by feature map splicing to obtain a disaster feature map.
[0083] In step S33, the disaster feature map is input into the semantic segmentation model for processing to obtain a disaster assessment result.
[0084] In the embodiment, the disaster assessment result includes a disaster prediction level corresponding to each pixel coordinate in the disaster feature map.
[0085] The actual application process of the model is as shown in Figure 5 For example, after a city suffers from a rainstorm disaster, only the pre-disaster image and the post-disaster image need to be input into the trained model to obtain the disaster level of each pixel. Specifically, at the pixel coordinate (19, 29), the disaster level is 3-severe, and at this time, the staff needs to prioritize sending more disaster relief personnel to this place.
[0086] The present application improves the reliability of post-disaster assessment of the city, uses the pre-disaster remote sensing satellite image and the post-disaster remote sensing satellite image as input at the same time, uses the feature extraction network and the semantic segmentation network, predicts the disaster degree for each pixel, improves the accuracy, real-time and accuracy. Reducing the image analysis time, winning the race with time, predicting the disaster level at the accurate position.
[0087] The present application simultaneously uses the pre-disaster satellite image and the post-disaster satellite image, uses the image training set of multiple types of disaster to train the neural network, and the output result is the disaster severity and the disaster category. Specifically, the feature extraction network and the semantic segmentation network are used to predict the disaster degree at the pixel level, that is, the disaster degree is predicted for each pixel of a satellite image, which is more accurate than the existing model. The on-site emergency team can shorten the response time and take quick action, which is crucial for reducing losses and saving lives.
[0088] As shown in Figure 6As shown is a structural schematic diagram of a disaster situation assessment device based on multi-temporal images according to an embodiment of the present application. The device shown in the figure comprises:
[0089] A historical image module 10 is configured to acquire historical pre-disaster images and historical post-disaster images, input the historical pre-disaster images and the historical post-disaster images into a preset initial assessment model for processing, and obtain a disaster situation prediction result.
[0090] In the present application, the multi-temporal images refer to the use of both pre-disaster images and post-disaster images. Specifically, remote sensing technology can be used to adopt multi-source satellite remote sensing images, and the satellite remote sensing images collected before and after the disaster situation occurs can be used as the historical pre-disaster images and the historical post-disaster images.
[0091] Further, the preset initial assessment model comprises an initial feature model and an initial semantic model. The initial feature model is configured to extract a feature map of the satellite remote sensing images, and the initial semantic model is a semantic segmentation model configured to output a label (i.e., a disaster degree) for each pixel point of the satellite remote sensing images, so as to output an accurate disaster situation assessment result.
[0092] Further, the labeled historical pre-disaster images and the historical post-disaster images are input into the initial assessment model for model training, so as to obtain a disaster situation prediction result corresponding to the historical satellite remote sensing images.
[0093] An assessment model module 20 is configured to update the initial assessment model according to the disaster situation prediction result, so as to obtain a disaster situation assessment model.
[0094] The disaster situation prediction result comprises a disaster situation prediction level corresponding to each pixel coordinate of the images, and the initial assessment model is updated by using the disaster situation prediction result. Specifically, cross-entropy is used as a loss function, and the weights of the initial assessment model are updated by minimizing the loss function in the model training process, so as to obtain the trained disaster situation assessment model.
[0095] An assessment result module 30 is configured to input the acquired actual pre-disaster images and actual post-disaster images into the disaster situation assessment model for processing, so as to obtain a disaster situation assessment result.
[0096] When an actual disaster situation occurs, images before and after the disaster situation occurs, i.e., actual pre-disaster images and actual post-disaster images, are acquired by using remote sensing technology. The actual pre-disaster images and the actual post-disaster images are input into the trained disaster situation assessment model, so as to obtain a disaster situation assessment result.
[0097] Further, the disaster situation assessment result comprises a disaster situation prediction level corresponding to each pixel coordinate of the actual disaster situation satellite remote sensing images, so as to improve the accuracy, real-time performance and accuracy of the disaster situation assessment, and provide an effective reference for the staff.
[0098] As an embodiment of the present application, the initial evaluation model comprises an initial feature model and an initial semantic model.
[0099] In the embodiment, as shown in Figure 7 The historical image module 10 comprises:
[0100] The feature map unit 11 is configured to input the historical pre-disaster image and the historical post-disaster image into the preset initial feature model for processing to obtain a first feature map corresponding to the historical pre-disaster image and a second feature map corresponding to the historical post-disaster image.
[0101] The image splicing unit 12 is configured to splice the first feature map and the second feature map by using a channel splicing manner to obtain a feature splicing map.
[0102] The prediction result unit 13 is configured to input the feature splicing map into the preset initial semantic model for processing to obtain a disaster condition prediction result; wherein the disaster condition prediction result comprises a disaster condition prediction level corresponding to each pixel coordinate in the feature splicing map.
[0103] In the embodiment, the evaluation model module is further configured to update the initial feature model and the initial semantic model according to the disaster condition evaluation result and a preset loss function to obtain a feature extraction model and a semantic segmentation model, and use the feature extraction model and the semantic segmentation model as the disaster condition evaluation model.
[0104] In the embodiment, as shown in Figure 8 The evaluation result module 30 comprises:
[0105] The feature extraction unit 31 is configured to input the obtained actual pre-disaster image and actual post-disaster image into the feature extraction model for processing to obtain a pre-disaster feature map and a post-disaster feature map.
[0106] The disaster condition feature map unit 32 is configured to perform feature map splicing processing on the pre-disaster feature map and the post-disaster feature map to obtain a disaster condition feature map.
[0107] The evaluation result unit 33 is configured to input the disaster condition feature map into the semantic segmentation model for processing to obtain a disaster condition evaluation result; wherein the disaster condition evaluation result comprises a disaster condition prediction level corresponding to each pixel coordinate in the disaster condition feature map.
[0108] Based on the same application concept as the above-mentioned disaster condition evaluation method based on multi-temporal images, the present application further provides a disaster condition evaluation device based on multi-temporal images. Since the principle of solving problems of the disaster condition evaluation device based on multi-temporal images is similar to that of the disaster condition evaluation method based on multi-temporal images, the implementation of the disaster condition evaluation device based on multi-temporal images can be referred to the implementation of the disaster condition evaluation method based on multi-temporal images, and the repeated parts will not be described herein.
[0109] The present application trains a model by simultaneously using pre-disaster satellite images and post-disaster satellite images, so that the disaster assessment model can effectively extract the differences before and after the disaster, improve the prediction accuracy of complex disaster assessment, and provide effective reference for staff.
[0110] The present application also provides an electronic device, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the above method when executing the program.
[0111] The present application also provides a computer readable storage medium, which stores a computer program for executing the above method.
[0112] As shown in Figure 9 The electronic device 600 can also include a communication module 110, an input unit 120, an audio processing unit 130, a display 160, and a power supply 170. It is worth noting that the electronic device 600 does not necessarily include all the components shown in Figure 9 In addition, the electronic device 600 can also include components not shown in Figure 9 Reference can be made to the prior art.
[0113] As shown in Figure 9 The central processor 100, also known as a controller or operating control, can include a microprocessor or other processor device and / or logic device, which receives input and controls the operation of various components of the electronic device 600.
[0114] The memory 140, for example, can be one or more of a cache, a flash memory, a hard drive, a removable media, a volatile memory, a non-volatile memory, or other suitable device. Information related to failure can be stored, and programs for executing the information can also be stored. The central processor 100 can execute the programs stored in the memory 140 to achieve information storage or processing, etc.
[0115] The input unit 120 provides input to the central processor 100. The input unit 120 is, for example, a key or touch input device. The power supply 170 is used to provide power to the electronic device 600. The display 160 is used to display display objects such as images and text. The display can be, for example, an LCD display, but is not limited thereto.
[0116] The memory 140 can be a solid state memory such as a read only memory (ROM), a random access memory (RAM), a SIM card, etc. It can also be a memory that retains information even when power is turned off, can be selectively erased, and is provided with more data, an example of which is sometimes referred to as an EPROM, etc. The memory 140 can also be some other type of device. The memory 140 includes a buffer memory 141 (sometimes referred to as a buffer). The memory 140 can include an application / function storage 142 for storing application programs and function programs or a flow for executing the operation of the electronic device 600 by the central processing unit 100.
[0117] The memory 140 can also include a data storage 143 for storing data such as contacts, digital data, pictures, sounds, and / or any other data used by the electronic device. A driver storage 144 of the memory 140 can include various drivers of the electronic device for a communication function and / or for performing other functions of the electronic device such as a messaging application, an address book application, etc.
[0118] The communication module 110 is a transmitter / receiver 110 that transmits and receives signals via an antenna 111. The communication module (transmitter / receiver) 110 is coupled to the central processing unit 100 to provide input signals and receive output signals, which can be the same as in the case of a conventional mobile communication terminal.
[0119] Based on different communication technologies, a plurality of communication modules 110 such as a cellular network module, a Bluetooth module, and / or a wireless local area network module, etc. can be provided in the same electronic device. The communication module (transmitter / receiver) 110 is also coupled to a speaker 131 and a microphone 132 via an audio processor 130 to provide audio output via the speaker 131 and receive audio input from the microphone 132, thereby implementing a general telecommunication function. The audio processor 130 can include any suitable buffer, decoder, amplifier, etc. In addition, the audio processor 130 is also coupled to the central processing unit 100, thereby enabling recording on a local machine through the microphone 132 and enabling playing of a sound stored on the local machine through the speaker 131.
[0120] Those skilled in the art will appreciate that embodiments of the present application can be readily used as a method, a system or a computer program product. Accordingly, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage devices, etc.) embodying computer readable program code thereon.
[0121] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for functionally implementing the steps in one or more flowcharts and / or blocks
[0122] These computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture including instructions which implement the function specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for functionally implementing the steps in one or more flowcharts and / or blocks
[0123] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for functionally implementing the steps in one or more flowcharts and / or blocks
[0124] The principles and implementations of the present application have been described in the specific embodiments. The above description is only used to help understand the method of the present application and its core idea; meanwhile, for those skilled in the art, according to the idea of the present application, the specific implementation and application range can be changed. In summary, the content of the specification should not be understood as a limitation of the present application.
Claims
1. A disaster situation evaluation method based on multi-phase images, characterized by, The method comprises: acquiring historical pre-disaster images and historical post-disaster images, inputting the historical pre-disaster images and the historical post-disaster images into a preset initial feature model for processing to obtain a first feature map corresponding to the historical pre-disaster images and a second feature map corresponding to the historical post-disaster images, splicing the first feature map and the second feature map in a channel splicing manner to obtain a feature splicing map, inputting the feature splicing map into a preset initial semantic model for processing to obtain a disaster situation prediction result; wherein the disaster situation prediction result comprises a disaster situation prediction level corresponding to each pixel coordinate in the feature splicing map; wherein the initial evaluation model comprises an initial feature model and an initial semantic model; the initial feature model is used to extract a feature map of a satellite remote sensing image, the initial semantic model is a semantic segmentation model, and is used to output a label for each pixel point of the satellite remote sensing image to output an accurate disaster situation evaluation result; the initial evaluation model adopts cross entropy as a loss function; updating the initial evaluation model according to the disaster situation prediction result to obtain a disaster situation evaluation model; inputting acquired actual pre-disaster images and actual post-disaster images into the disaster situation evaluation model for processing to obtain a disaster situation evaluation result.
2. The method of claim 1, wherein, The updating the initial evaluation model according to the disaster situation prediction result to obtain a disaster situation evaluation model comprises: updating the initial feature model and the initial semantic model according to the disaster situation evaluation result and a preset loss function to obtain a feature extraction model and a semantic segmentation model, and taking the feature extraction model and the semantic segmentation model as the disaster situation evaluation model.
3. The method of claim 2, wherein, The inputting acquired actual pre-disaster images and actual post-disaster images into the disaster situation evaluation model for processing to obtain a disaster situation evaluation result comprises: inputting the acquired actual pre-disaster images and actual post-disaster images into the feature extraction model for processing to obtain a pre-disaster feature map and a post-disaster feature map; performing feature map splicing processing on the pre-disaster feature map and the post-disaster feature map to obtain a disaster situation feature map; inputting the disaster situation feature map into the semantic segmentation model for processing to obtain the disaster situation evaluation result.
4. The method of claim 3, wherein, The disaster situation evaluation result comprises a disaster situation prediction level corresponding to each pixel coordinate in the disaster situation feature map.
5. A disaster situation evaluation device based on multi-phase images, characterized by, The device comprises: The historical image module is configured to acquire historical pre-disaster images and historical post-disaster images, input the historical pre-disaster images and the historical post-disaster images into a preset initial evaluation model for processing, and obtain a disaster condition prediction result. The initial evaluation model includes an initial feature model and an initial semantic model. The initial feature model is configured to extract a feature map of a satellite remote sensing image. The initial semantic model is a semantic segmentation model configured to output a label for each pixel point of the satellite remote sensing image, so as to output an accurate disaster condition evaluation result. The historical image module includes a feature map unit configured to input the historical pre-disaster images and the historical post-disaster images into the initial feature model of the initial evaluation model for processing, and obtain a first feature map corresponding to the historical pre-disaster images and a second feature map corresponding to the historical post-disaster images. An image splicing unit is configured to splice the first feature map and the second feature map by using a channel splicing method to obtain a feature splicing map. A prediction result unit is configured to input the feature splicing map into the initial semantic model of the initial evaluation model for processing, and obtain the disaster condition prediction result. The disaster condition prediction result includes a disaster condition prediction level corresponding to each pixel coordinate in the feature splicing map. The initial evaluation model uses cross-entropy as a loss function. The evaluation model module is configured to update the initial evaluation model according to the disaster condition prediction result, and obtain a disaster condition evaluation model. The evaluation result module is configured to input acquired actual pre-disaster images and actual post-disaster images into the disaster condition evaluation model for processing, and obtain a disaster condition evaluation result.
6. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the computer program to implement the method of any one of claims 1 to 4.
7. A computer readable storage medium characterized in that, The computer readable storage medium stores a computer program for executing the method of any one of claims 1 to 4.
Citation Information
Patent Citations
Flood disaster assessment system and method
CN105913361A
Disaster evaluation and analysis method based on GIS
CN111144656A