CN115719447A - Building change detection method based on double-time-phase high-resolution remote sensing image - Google Patents
Building change detection method based on double-time-phase high-resolution remote sensing image Download PDFInfo
- Publication number
- CN115719447A CN115719447A CN202211223353.9A CN202211223353A CN115719447A CN 115719447 A CN115719447 A CN 115719447A CN 202211223353 A CN202211223353 A CN 202211223353A CN 115719447 A CN115719447 A CN 115719447A
- Authority
- CN
- China
- Prior art keywords
- remote sensing
- sensing image
- building
- double
- resolution remote
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Abstract
The invention provides a building change detection method based on a double-time-phase high-resolution remote sensing image, which comprises the following steps: acquiring original remote sensing image data; preprocessing original remote sensing image data to obtain a double-temporal high-resolution remote sensing image; obtaining a change area of the building according to the double-time-phase high-resolution remote sensing image; cutting the double-time-phase high-resolution remote sensing image and the change area of the building and dividing a data set according to a preset proportion to obtain a training data set and a test data set; inputting the training data set into the twin network structure for training to obtain a trained twin network structure; inputting the test data set into the trained twin network structure for testing to obtain a tested twin network structure; and inputting the prediction area image into the tested twin network structure for training to obtain a change detection diagram of the building. The disclosure also provides a building change detection device based on the double-time-phase high-resolution remote sensing image, electronic equipment and a storage medium.
Description
Technical Field
The present disclosure relates to the field of building change detection technologies, and in particular, to a method and an apparatus for detecting building change based on a dual-temporal high-resolution remote sensing image, an electronic device, and a storage medium.
Background
The building is an important facility in cities and villages, and the conditions of building new construction, dismantling and reconstruction are closely related to the social and economic development. The change detection of the building has important significance for urban analysis, illegal building identification, disaster assessment and the like. The building change detection based on the high-resolution remote sensing image has the advantages of high efficiency, wide range, low cost and the like compared with the building change detection of a field inspection type, so that the remote sensing technology is more and more applied to the building change detection.
The traditional remote sensing building change detection method is divided into a direct classification method and a classification post-processing method. The main idea of the direct comparison method is to obtain the changed content by direct comparison based on the characteristics of the spectrum, texture, geometric shape and the like of the ground features in the remote sensing image. The method of the comparison after classification requires that the buildings in the same area are extracted from the images in different time phases, and then the change area is detected by comparing the extracted buildings.
In recent years, with the development of big data and the improvement of the performance of related hardware, deep learning becomes an important technical means for detecting the change of a remote sensing building, and related methods can be divided into a single-path network method, a double-path network method and a multi-network integration method through a used network. The single-path network processes the front-rear time phase data through a single data stream, and needs to splice the front-rear time phase data before the data is input into the network. The dual-path network extracts the features of the data of the two time phases respectively, and then processes the extracted features to realize change detection. Network integration refers to the integration of multiple network structures together to achieve change detection with multi-party advantages.
For the traditional machine learning method, the classification precision is generally low, the generalization capability is poor, and the method is not suitable for the detection of the change of the remote sensing building in a large range. The introduction of deep learning solves the problems to a certain extent, and the single-path network method fuses two images in a channel splicing mode, so that once the two images have differences of seasons, illumination, shooting angles and the like, false detection and missing detection can be caused, and the precision is low. The dual path network method needs to extract features from two images respectively and fuse the extracted features, which causes a problem that the extracted features are difficult to align, and does not effectively utilize the scale difference between buildings. Besides the need of training a plurality of models, the multi-network integration method has high calculation cost and labeling cost and low practical application value.
Disclosure of Invention
In view of the above problems, the present invention provides a building change detection method, device, electronic device and storage medium based on a dual-temporal high-resolution remote sensing image, which solves the problem of efficiently and accurately detecting building changes by using remote sensing images.
One aspect of the present disclosure provides a building change detection method based on a dual-temporal high-resolution remote sensing image, including: acquiring original remote sensing image data; preprocessing original remote sensing image data to obtain a double-time-phase high-resolution remote sensing image; the double-time-phase high-resolution remote sensing image is a high-resolution remote sensing image of the same region in different time phases; obtaining a change area of the building according to the double-time-phase high-resolution remote sensing image; cutting the double-time-phase high-resolution remote sensing image and the change area of the building and dividing a data set according to a preset proportion to obtain a training data set and a test data set; inputting the training data set into the twin network structure for training to obtain a trained twin network structure; wherein the twin network structure comprises two branched networks with the same structure; inputting the test data set into the trained twin network structure for testing to obtain a tested twin network structure; and preprocessing the image of the prediction area, inputting the preprocessed image into the tested twin network structure, and training to obtain a change detection diagram of the building.
Further, obtaining a change area of the building according to the double-time-phase high-resolution remote sensing image, comprising: obtaining the geographical ranges of the two corresponding images according to the double time-phase high-resolution remote sensing images, and judging whether the geographical ranges of the two images have an intersection region or not; if yes, calculating the geographic range of the intersection area of the two images, and obtaining front and rear time phase images of the intersection area according to the geographic range of the intersection area; and sequentially carrying out building labeling and XOR operation processing on the front time phase image and the rear time phase image of the intersected area to obtain a change area of the building.
Further, the twin network structure adopts an Encoder-Decoder architecture model; inputting a training data set into the twin network structure for training to obtain a trained twin network structure, wherein the training data set comprises: sequentially carrying out multiple times of downsampling processing on the training data set to extract the features of the double-time-phase high-resolution remote sensing image, and obtaining a plurality of feature maps with different sizes of the double-time-phase high-resolution remote sensing image; fusing a plurality of feature maps with different sizes to obtain a fused feature map with the same size as the double-time-phase high-resolution remote sensing image; respectively inputting the fusion feature map into an attention structure and a semantic segmentation structure in the twin network structure to obtain a large-scale attention mask and a change detection feature; and obtaining a change detection result of the building according to the attention mask with large scale and the change detection characteristics, and finishing the twin network structure training.
Further, the method includes the steps that after preprocessing is conducted on the images of the prediction region, the images of the prediction region are input into the twin network structure after testing to be trained, and a change detection diagram of the building is obtained, and the method includes the following steps: preprocessing the prediction region image to obtain a double-time-phase high-resolution remote sensing image of the prediction region image; forming a plurality of groups of prediction data by using an image tensor obtained by reading a double-temporal high-resolution remote sensing image of the prediction region and adopting different dimensions; sequentially inputting a plurality of groups of prediction data into the tested twin network structure for prediction to obtain a plurality of groups of change detection characteristics and a plurality of groups of attention masks; and according to the multiple groups of change detection characteristics and the multiple groups of attention masks, performing up-sampling to obtain a change detection diagram of the building.
Further, obtaining a change detection map of the building according to the plurality of sets of change detection features and the plurality of sets of attention masks, includes: and combining the multiple groups of change detection characteristics and the multiple groups of attention masks in sequence according to a principle of first-large and second-small to obtain a change detection image of the building.
Further, the method for preprocessing the original remote sensing image data to obtain the double-time-phase high-resolution remote sensing image comprises the following steps: and sequentially performing resampling, projection conversion, radiometric calibration, atmospheric correction and geographic registration processing on the original remote sensing image data to obtain the double-time-phase high-resolution remote sensing image.
Further, the twin network structure is two parallel network structures, and the two parallel network structures have the same structure and share parameters.
Two aspects of the present disclosure provide a building change detection device based on a double-temporal high-resolution remote sensing image, including: the data preprocessing module is used for acquiring original remote sensing image data and preprocessing the original remote sensing image data to obtain a double-time-phase high-resolution remote sensing image; the double-time-phase high-resolution remote sensing image is a high-resolution remote sensing image of the same area in different time phases; the data change area detection module is used for obtaining a change area of the building according to the double-time-phase high-resolution remote sensing image; the data set generating module is used for cutting the double-time-phase high-resolution remote sensing image and the change area of the building and dividing the data set according to a preset proportion to obtain a training data set and a testing data set; the model training module is used for inputting the training data set into the twin network structure for training to obtain a trained twin network structure; wherein the twin network structure comprises two branched networks with the same structure; the model testing module is used for inputting a testing data set into the trained twin network node for testing and training to obtain a tested twin network structure; and the data change detection module is used for preprocessing the image of the prediction region and inputting the preprocessed image into the tested twin network structure for training to obtain a change detection diagram of the building.
A third aspect of the present disclosure provides an electronic device, comprising: the building change detection method based on the double-time-phase high-resolution remote sensing image comprises a memory, a processor and a computer program which is stored on the memory and can run on the processor, wherein when the processor executes the computer program, each step in the building change detection method based on the double-time-phase high-resolution remote sensing image is realized.
A fourth aspect of the present disclosure provides a computer-readable storage medium, on which a computer program is stored, which, when executed by a processor, implements the steps in the method for detecting a change in a building based on a dual-temporal high-resolution remote sensing image.
According to the building change detection method, the building change detection device, the electronic equipment and the storage medium based on the double-time-phase high-resolution remote sensing image, the twin convolution network structure shared by two weights is used, meanwhile, the multi-scale attention mechanism is used, the multi-scale semantic information and the spectrum information of the remote sensing image can be comprehensively utilized, the building change detection extraction precision is improved, and the final building change detection drawing effect is improved.
Drawings
For a more complete understanding of the present disclosure and the advantages thereof, reference is now made to the following descriptions taken in conjunction with the accompanying drawings, in which:
fig. 1 schematically shows a flowchart of a building change detection method based on a dual-temporal high-resolution remote sensing image according to an embodiment of the present disclosure;
FIGS. 2A-2C schematically illustrate a structural diagram, a training process, and a prediction process of a twin convolutional network structure provided by an embodiment of the present disclosure;
fig. 3 is a block diagram schematically illustrating a structure of a building change detection apparatus based on a dual-temporal high-resolution remote sensing image according to an embodiment of the present disclosure;
fig. 4 schematically shows a block diagram of an electronic device provided in an embodiment of the present disclosure.
Detailed Description
Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. It should be understood that these descriptions are illustrative only and are not intended to limit the scope of the present disclosure. In the following detailed description, for purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of the embodiments of the disclosure. It may be evident, however, that one or more embodiments may be practiced without these specific details. Moreover, in the following description, descriptions of well-known structures and techniques are omitted so as to not unnecessarily obscure the concepts of the present disclosure.
The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the disclosure. The terms "comprises," "comprising," and the like, as used herein, specify the presence of stated features, steps, operations, and/or components, but do not preclude the presence or addition of one or more other features, steps, operations, or components.
All terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art unless otherwise defined. It is noted that the terms used herein should be interpreted as having a meaning that is consistent with the context of this specification and should not be interpreted in an idealized or overly formal sense.
Some block diagrams and/or flow diagrams are shown in the figures. It will be understood that some blocks of the block diagrams and/or flowchart illustrations, or combinations thereof, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor, create means for implementing the functions/acts specified in the block diagrams and/or flowchart block or blocks.
Accordingly, the techniques of this disclosure may be implemented in hardware and/or software (including firmware, microcode, etc.). In addition, the techniques of this disclosure may take the form of a computer program product on a computer-readable medium having instructions stored thereon for use by or in connection with an instruction execution system. In the context of this disclosure, a computer-readable medium may be any medium that can contain, store, communicate, propagate, or transport the instructions. For example, the computer readable medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, device, or propagation medium. Specific examples of the computer readable medium include: magnetic storage devices, such as magnetic tape or Hard Disk Drives (HDDs); optical storage devices, such as compact disks (CD-ROMs); a memory, such as a Random Access Memory (RAM) or a flash memory; and/or wired/wireless communication links.
Fig. 1 schematically shows a flowchart of a building change detection method based on a two-time-phase high-resolution remote sensing image according to an embodiment of the present disclosure.
As shown in fig. 1, a method 100 for detecting building changes based on a dual-temporal high-resolution remote sensing image includes: steps S110 to S160.
In operation S110, raw remote sensing image data is acquired. And preprocessing the original remote sensing image data to obtain a double-time-phase high-resolution remote sensing image. The double-time-phase high-resolution remote sensing image is a high-resolution remote sensing image of the same area in different time phases.
In the embodiment of the disclosure, the original remote sensing image data can be obtained by obtaining remote sensing image data corresponding to different resolutions under different wave bands of the high-resolution one-number image. And preprocessing the acquired original remote sensing image data to obtain the double-time-phase high-resolution remote sensing image.
Specifically, the method for preprocessing the original remote sensing image data to obtain the double-time-phase high-resolution remote sensing image comprises the following steps: and sequentially performing resampling, projection conversion, radiometric calibration, atmospheric correction and geographic registration processing on the original remote sensing image data to obtain the double-time-phase high-resolution remote sensing image.
For example, the top-score first image may contain 4 multispectral bands with 8 m resolution and 1 panchromatic band with 2 m resolution, and it is necessary to resample the bands with less than 8 m resolution to 2 m resolution. The projection modes used by the high-resolution images of different scenes are related to the longitude where the high-resolution images are located, which causes different projections among different images, and prevents uniform processing of the images, so that uniform projection conversion needs to be performed on all the images, and in the embodiment of the disclosure, all the images are converted into the WGS-84 coordinate system.
The radiometric calibration processing of the remote sensing image data is a processing process for converting a digital quantization value (DN value) of an image into physical quantities such as a radiance value, a reflectivity and the like, and can automatically perform image radiometric calibration by using a radiometric calibration tool provided in a mature remote sensing data processing platform ENVI.
The atmospheric correction can eliminate radiation errors caused by atmospheric images, better reflect the real situation of the earth surface, and can automatically correct the image atmosphere by utilizing an atmospheric correction tool provided by a mature remote sensing data processing platform ENVI.
The geographic registration is to align the remote sensing images of different time phases in the same region, so that visual interpretation and automatic interpretation are facilitated. Geographic registration work is realized based on ENVI, and registration error is controlled within 1 pixel.
In operation S120, a change area of the building is obtained from the dual-temporal high-resolution remote sensing image.
According to the embodiment of the present disclosure, obtaining the change area of the building according to the two-time-phase high-resolution remote sensing image obtained in step S110 specifically includes: steps S1201 to S1203.
In operation S1201, the geographical ranges of the two corresponding images are obtained according to the two time-phase high-resolution remote sensing images, and it is determined whether there is an intersection region in the geographical ranges of the two images.
In operation S1202, if the image exists, the geographic range of the intersection region of the two images is calculated, and the front-time phase image and the rear-time phase image of the intersection region are obtained according to the geographic range of the intersection region.
In the embodiment of the disclosure, the double-time-phase high-resolution remote sensing image is read first, the geographical range of the two corresponding images is obtained, and whether the geographical range of the two images has an intersection area is judged. If there is no intersection area, it will report error directly. If the intersection area exists, calculating the geographical range of the intersection area, then converting the geographical range of the intersection area into the image range of the front time phase image and the image range of the rear time phase image respectively, and then cutting the data of the front time phase image and the data of the rear time phase image according to the two sets of range information to obtain the front time phase image and the rear time phase image of the intersection area.
In operation S1203, building labeling and xor operation processing are sequentially performed on the time phase images before and after the intersection area, so as to obtain a change area of the building.
In the embodiment of the disclosure, building labeling is sequentially performed on the front time phase image and the rear time phase image of the intersection area, and the labeling can be performed under the interactive interface of image labeling software on the building outline polygon, wherein the labeling software used in the embodiment of the disclosure is ArcMap 10.2, the software supports reading and displaying of the remote sensing image, the drawing of the outline of the ground object by the polygon on the basis of image display is supported, and the drawn polygon is stored as a vector in an shp file format. The building change detection task requires that a marker image of the same size as the original image be provided, with each pixel value on the marker image representing whether the pixel has changed. And (4) realizing polygon-to-image format marking by using a vector-to-grid tool provided by ArcMap to obtain a building outline mark in a grid format.
After the marked images of the double-temporal building are obtained, performing exclusive OR operation on the two groups of images, wherein the calculation mode of the process meets the following relation:
where (x, y) represents the input to the exclusive or function. After the XOR operation, the area value of the changed building is 1, and the area value of the two groups with no building or two stages of buildings with the same appearance is 0, thereby realizing the marking of the changed area of the building.
In operation S130, the double-temporal high-resolution remote sensing image and the change area of the building are cut and the data set is divided according to a preset ratio to obtain a training data set and a test data set.
In the embodiment of the disclosure, for convenience of training and testing, the double-time-phase high-resolution remote sensing image and the change area label of the building are cut into the size of 512 × 512 images, the cut images are divided into proportions according to preset proportions (such as 3:1, 4:1 and the like), and the sample is randomly divided into a training data set and a testing data set. Wherein the training data set is used for network training and the testing data set is used for precision evaluation.
In operation S140, the training data set is input to the twin network structure for training, so as to obtain a trained twin network structure. Wherein the twin network structure comprises two branched networks with the same structure.
In the embodiment of the present disclosure, as shown in fig. 2A, the twin network structure is two parallel branch networks with the same structure, and the two branch networks have the same structure and share parameters. The twin network structure takes a pair of images as input, and the image features are extracted in parallel by two same network paths.
According to the embodiment of the present disclosure, the twin network structure adopts an Encoder-Decoder architecture model, which follows a process of gradually downsampling first and then gradually upsampling, as shown in fig. 2B, step S140 specifically includes: steps 1401 to 1404.
In operation S1401, a training data set is sequentially subjected to multiple downsampling processes to extract features of the dual-temporal high-resolution remote sensing image, and a plurality of feature maps of the dual-temporal high-resolution remote sensing image with different sizes are obtained.
In the embodiment of the disclosure, in the down-sampling stage, the basic structure of EfficientNet-B1 can be adopted. Compared with EfficientNet-B0, efficientNet-B1 expands the depth of the network to 1.1 times of the original depth, expands the image resolution of the network from 224 multiplied by 224 to 240 multiplied by 240, and keeps the network width unchanged. After down-sampling, the size of the output feature map can be 1/32 of the original image. And in the down-sampling stage, the characteristics of the double-time phase images are respectively extracted through a network structure sharing weight, and all the characteristics of down-sampling for multiple times are reserved. For example, the training data set may be sequentially subjected to 5 times of downsampling processing to extract features of the two-time-phase high-resolution remote sensing image, so as to obtain a plurality of feature maps of the two-time-phase high-resolution remote sensing image with different sizes, and then all the features subjected to multiple times of downsampling may be retained.
In operation S1402, a plurality of feature maps with different sizes are fused to obtain a fused feature map with the same size as the dual-temporal high-resolution remote sensing image.
In the embodiment of the disclosure, in the up-sampling stage, the characteristics extracted before and after the fusion of the high-low layer fusion structure of the U-Net can be adopted. And at the same time of upsampling, fusing the feature maps with the same size of the front and rear time phases. The feature fusion adopts a mode of stacking according to channels, then convolution is carried out by a convolution kernel with the size of 3 multiplied by 3, and the upper part, the lower part, the left part and the right part are respectively filled with one line so as to keep the size of the feature graph unchanged.
In operation S1403, the fusion feature map is input into the attention structure and the semantic segmentation structure in the twin network structure, respectively, to obtain a large-scale attention mask and a change detection feature.
In operation S1404, a change detection result of the building is obtained according to the large-scale attention mask and the change detection feature, and the twin network structure training is completed.
In the embodiment of the disclosure, a multi-scale attention mechanism is introduced in the twin network structure training process. For example, first, the two images are respectively reduced by 0.5 times to form a set of data with original size and a set of data with original size 0.5 times. And then, respectively inputting the two groups of data into the twin network, acquiring 5 feature maps with different scales from each group of data, and then fusing the feature maps with different scales of the time-phase images before and after sampling in a channel splicing and convolution mode to obtain a fused feature map with the same size as the original image. And inputting the fusion characteristics of 0.5 times of data into the attention structure to generate an attention mask alpha of 0.5 times, wherein the (1-alpha) is the large-scale attention mask. And respectively inputting the fusion feature maps of the data with the large and small scales into the semantic segmentation structure to generate change detection features. And (3) upsampling the attention mask with the size of 0.5 times and the change detection features to the size of an original image, correspondingly multiplying the attention mask and the fusion features with different scales, and adding the multiplication results to obtain a final building change detection result so as to finish the twin network structure training.
The loss function used by the twin network structure can be a cross entropy loss function and a Dice loss function according to the following ratio of 1: the 1 ratio is weighted. Using Adam optimizer, initial learning rate was set to 0.002, the learning rate strategy selected reduce lronplateau (adaptive learning rate strategy), and Dropout (random drop probability) was 0.1 in training. The self-adaptive adjustment learning rate strategy can set random discarding probability to accelerate training according to the current network training precision and the self-adaptive reduction learning rate of the time staying at the precision, and improves the generalization capability of the network to a certain extent.
In operation S150, the test data set is input to the trained twin network structure for testing, so as to obtain a tested twin network structure.
In the embodiment of the present disclosure, the trained twin network structure obtained in step S140 is tested using the test data set, and the twin network structure is subjected to precision evaluation.
Specifically, four accuracy evaluation parameters of OA (overall classification progress), precision (accuracy), recall (Recall), and F1 score are used as the accuracy evaluation indexes. OA is the ratio of the number of correctly classified category pixels to the total number of category pixels, and is calculated specifically as:
wherein p is i,i Representing pixels classified into and belonging to the ith class; p is a radical of i,j Representing pixels belonging to class i and classified into class j; OA can indicate the overall classification accuracy well. By comparing with the sample label, the true positive TP, false positive FP and false negative FN of the correct extraction total number, false positive FP and the miss extraction total number of the classified pixels can be obtained, so that the calculation accuracy and the recall rate are respectively as follows:
the F1 score is an index for measuring the precision of the classification model in statistics, gives consideration to the precision rate and the recall rate of the classification model at the same time, is a harmonic evaluation of the precision rate and the recall rate, and has the specific calculation formula as follows:
in the embodiment of the disclosure, the twin network structure after training is tested by using the test data set, so that the running accuracy of the twin network structure is detected, and the accuracy of data detection is ensured.
In operation S160, the prediction region image is preprocessed and then input to the tested twin network structure for training, so as to obtain a change detection diagram of the building.
In the embodiment of the disclosure, the change detection diagram of the building can be obtained by predicting the image of the prediction region through the tested twin network structure.
Specifically, step S160 includes: steps S1061 to S1064.
In operation S1601, the prediction region image is preprocessed to obtain a two-time-phase high-resolution remote sensing image of the prediction region image.
In the embodiment of the present disclosure, the prediction region image is subjected to the preprocessing step in step S110, such as resampling, projection conversion, radiometric calibration, atmospheric correction, and geographic registration processing, so as to obtain a two-time-phase high-resolution remote sensing image of the prediction region image.
Because a single high-resolution image is large, direct reading occupies a large amount of memory, and the computing resource is wasted. The embodiment of the disclosure can read the image in a range reading mode, that is, only part of the image is stored in the memory each time, and the image is read in a block circulation mode, so that the memory use in the prediction process can be effectively reduced.
In operation S1602, a plurality of sets of prediction data are formed by using different scale sizes and using an image tensor obtained by reading a two-time-phase high-resolution remote sensing image of a prediction region image.
In operation S1603, multiple sets of prediction data are sequentially input into the tested twin network structure for prediction, so as to obtain multiple sets of change detection features and multiple sets of attention masks.
In operation S1604, a change detection map of the building is up-sampled according to the sets of change detection features and the sets of attention masks.
In the embodiment of the present disclosure, as shown in fig. 2C, an image tensor obtained by reading a two-time-phase high-resolution remote sensing image of a prediction area image is used, and a plurality of sets of prediction data are formed by using different scale sizes. For example, three dimensions of 0.5, 1.0 and 1.5 are adopted to form 3 groups of data, a triple twin network model is called, and three groups of change detection features and three groups of attention masks are output. The combination of the fusion characteristic diagram follows the principle of first big and then small, namely, the change detection characteristics with the size of 1.0 time and 1.5 times are firstly combined and then combined with the change detection characteristics with the size of 0.5 time, and finally the building change detection drawing is generated.
According to the building change detection method based on the double-temporal high-resolution remote sensing image, a twin convolution network structure is used, a multi-scale attention mechanism is used, multi-scale semantic information and spectral information of the remote sensing image can be comprehensively utilized, building change detection and extraction precision is improved, and the final building change detection drawing effect is improved.
Another aspect of the present disclosure provides a building change detection apparatus based on a dual-temporal high-resolution remote sensing image, as shown in fig. 3, the apparatus 300 including: a data preprocessing module 310, a data change region detection module 320, a data set generation module 330, a model training module 340, a model testing module 350, and a data change detection module 360. The apparatus 300 may be used to implement the building change detection method based on the dual-temporal high-resolution remote sensing image described with reference to fig. 1.
The data preprocessing module 310 is used for acquiring original remote sensing image data and preprocessing the original remote sensing image data to obtain a double-time-phase high-resolution remote sensing image; the double-time-phase high-resolution remote sensing image is a high-resolution remote sensing image of the same region in different time phases. According to an embodiment of the present disclosure, the data preprocessing module 310 may be configured to perform the step S110 described above with reference to fig. 1, for example, and is not described in detail herein.
And the data change area detection module 320 is used for obtaining the change area of the building according to the double-time-phase high-resolution remote sensing image. According to an embodiment of the present disclosure, the data change region detection module 320 may be configured to perform the step S120 described above with reference to fig. 1, for example, and is not described in detail herein.
And the data set generating module 330 is configured to cut the double-temporal high-resolution remote sensing image and the change area of the building and divide the data set according to a preset ratio to obtain a training data set and a test data set. According to an embodiment of the present disclosure, the data set generating module 330 may be configured to perform the step S130 described above with reference to fig. 1, for example, and is not described in detail herein.
The model training module 340 is configured to input a training data set to the twin network structure for training, so as to obtain a trained twin network structure; wherein the twin network structure comprises two branched networks with the same structure. According to an embodiment of the present disclosure, the model training module 340 may be, for example, configured to perform the step S140 described above with reference to fig. 1, and details are not repeated here.
And the model testing module 350 is configured to input the test data set to the trained twin network node for test training, so as to obtain a tested twin network structure. According to an embodiment of the present disclosure, the model test module 350 may be used to perform the step S150 described above with reference to fig. 1, for example, and is not described in detail herein.
And the data change detection module 360 is used for preprocessing the image of the prediction region and inputting the preprocessed image into the tested twin network structure for training to obtain a change detection diagram of the building. According to an embodiment of the present disclosure, the data change detection module 360 may be configured to perform the step S160 described above with reference to fig. 1, for example, and is not described in detail herein.
It is understood that the data preprocessing module 310, the data change region detection module 320, the data set generation module 330, the model training module 340, the model testing module 350, and the data change detection module 360 may be combined in one module to be implemented, or any one of them may be split into multiple modules. Alternatively, at least part of the functionality of one or more of these modules may be combined with at least part of the functionality of other modules and implemented in one module. According to an embodiment of the present invention, at least one of the data preprocessing module 310, the data change region detection module 320, the data set generation module 330, the model training module 340, the model testing module 350, and the data change detection module 360 may be implemented at least in part as a hardware circuit, such as a Field Programmable Gate Array (FPGA), a Programmable Logic Array (PLA), a system-on-a-chip, a system-on-a-substrate, a system-on-a-package, an Application Specific Integrated Circuit (ASIC), or may be implemented in hardware or firmware in any other reasonable manner of integrating or packaging a circuit, or in a suitable combination of three implementations of software, hardware, and firmware. Alternatively, at least one of the data preprocessing module 310, the data change region detection module 320, the data set generation module 330, the model training module 340, the model testing module 350, and the data change detection module 360 may be at least partially implemented as a computer program module that, when executed by a computer, performs the functions of the respective module.
Fig. 4 schematically shows a block diagram of an electronic device provided in an embodiment of the present disclosure. The electronic device shown in fig. 4 is only an example, and should not bring any limitation to the functions and the scope of use of the embodiments of the present disclosure.
As shown in fig. 4, the electronic device 400 described in this embodiment includes: a processor 401, which can perform various appropriate actions and processes according to a program stored in a Read Only Memory (ROM) 402 or a program loaded from a storage section 408 into a Random Access Memory (RAM) 403. Processor 401 may include, for example, a general purpose microprocessor (e.g., a CPU), an instruction set processor and/or associated chipset, and/or a special purpose microprocessor (e.g., an Application Specific Integrated Circuit (ASIC)), among others. The processor 401 may also include onboard memory for caching purposes. Processor 401 may include a single processing unit or multiple processing units for performing the different actions of the method flows in accordance with embodiments of the present disclosure.
In the RAM403, various programs and data necessary for the operation of the electronic apparatus 400 are stored. The processor 401, ROM402 and RAM403 are connected to each other by a bus 404. The processor 401 performs various operations of the method flows according to the embodiments of the present disclosure by executing programs in the ROM402 and/or the RAM 403. Note that the programs may also be stored in one or more memories other than the ROM402 and RAM 403. The processor 401 may also perform various operations of the method flows according to embodiments of the present disclosure by executing programs stored in the one or more memories.
According to an embodiment of the present disclosure, electronic device 400 may also include an input/output (I/O) interface 405, input/output (I/O) interface 405 also being connected to bus 404. Electronic device 400 may also include one or more of the following components connected to I/O interface 405: an input section 406 including a keyboard, a mouse, and the like; an output section 407 including a display device such as a Cathode Ray Tube (CRT), a Liquid Crystal Display (LCD), and the like, and a speaker; a storage section 408 including a hard disk and the like; and a communication section 409 including a network interface card such as a LAN card, a modem, or the like. The communication section 409 performs communication processing via a network such as the internet. A driver 410 is also connected to the I/O interface 405 as needed. A removable medium 411 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like is mounted on the drive 410 as necessary, so that a computer program read out therefrom is mounted into the storage section 408 as necessary.
According to an embodiment of the present disclosure, the method flow according to an embodiment of the present disclosure may be implemented as a computer software program. For example, embodiments of the present disclosure include a computer program product comprising a computer program embodied on a computer readable storage medium, the computer program containing program code for performing the method illustrated by the flow chart. In such an embodiment, the computer program may be downloaded and installed from a network through the communication section 409 and/or installed from the removable medium 411. The computer program, when executed by the processor 401, performs the above-described functions defined in the system of the embodiments of the present disclosure. The systems, devices, apparatuses, modules, units, etc. described above may be implemented by computer program modules according to embodiments of the present disclosure.
An embodiment of the present invention further provides a computer-readable storage medium, which may be included in the apparatus/device/system described in the foregoing embodiment; or may exist separately and not be assembled into the device/apparatus/system. The computer-readable storage medium carries one or more programs which, when executed, implement a method for detecting a change in a building based on a two-time-phase high-resolution remote-sensing image according to an embodiment of the present disclosure.
According to embodiments of the present disclosure, the computer-readable storage medium may be a non-volatile computer-readable storage medium, which may include, for example but is not limited to: a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In embodiments of the disclosure, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. For example, according to embodiments of the present disclosure, a computer-readable storage medium may include one or more memories other than the ROM402 and/or RAM403 and/or ROM402 and RAM403 described above.
Embodiments of the present disclosure also include a computer program product comprising a computer program containing program code for performing the method illustrated in the flow chart. When the computer program product runs in a computer system, the program code is used for causing the computer system to realize the building change detection method based on the double time-phase high-resolution remote sensing image provided by the embodiment of the disclosure.
The computer program performs the above-described functions defined in the system/apparatus of the embodiments of the present disclosure when executed by the processor 401. The systems, apparatuses, modules, units, etc. described above may be implemented by computer program modules according to embodiments of the present disclosure.
In one embodiment, the computer program may be hosted on a tangible storage medium such as an optical storage device, a magnetic storage device, or the like. In another embodiment, the computer program may also be transmitted, distributed in the form of a signal on a network medium, downloaded and installed through the communication section 409, and/or installed from the removable medium 411. The computer program containing program code may be transmitted using any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination of the foregoing.
In such an embodiment, the computer program may be downloaded and installed from a network through the communication section 409, and/or installed from the removable medium 411. The computer program, when executed by the processor 401, performs the above-described functions defined in the system of the embodiments of the present disclosure. The systems, devices, apparatuses, modules, units, etc. described above may be implemented by computer program modules according to embodiments of the present disclosure.
In accordance with embodiments of the present disclosure, program code for executing computer programs provided by embodiments of the present disclosure may be written in any combination of one or more programming languages, and in particular, these computer programs may be implemented using high level procedural and/or object oriented programming languages, and/or assembly/machine languages. The programming language includes, but is not limited to, programming languages such as Java, C + +, python, the "C" language, or the like. The program code may execute entirely on the user computing device, partly on the user device, partly on a remote computing device, or entirely on the remote computing device or server. In situations involving remote computing devices, the remote computing devices may be connected to the user computing device through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to external computing devices (e.g., through the internet using an internet service provider).
It should be noted that each functional module in each embodiment of the present disclosure may be integrated into one processing module, or each module may exist alone physically, or two or more modules are integrated into one module. The integrated module can be realized in a hardware mode, and can also be realized in a software functional module mode. The integrated module, if implemented in the form of a software functional module and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be substantially or partially embodied in the form of a software product, or all or part of the technical solution that contributes to the prior art.
The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams or flowchart illustration, and combinations of blocks in the block diagrams or flowchart illustration, can be implemented by special purpose hardware-based systems that perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
It will be appreciated by a person skilled in the art that various combinations or/and combinations of features recited in the various embodiments of the disclosure and/or in the claims may be made, even if such combinations or combinations are not explicitly recited in the disclosure. In particular, various combinations and/or combinations of the features recited in the various embodiments and/or claims of the present disclosure may be made without departing from the spirit or teaching of the present disclosure. All such combinations and/or associations are within the scope of the present disclosure.
While the disclosure has been shown and described with reference to certain exemplary embodiments thereof, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the disclosure as defined by the appended claims and their equivalents. Accordingly, the scope of the present disclosure should not be limited to the above-described embodiments, but should be defined not only by the appended claims, but also by equivalents thereof.
Claims (10)
1. A building change detection method based on double time-phase high-resolution remote sensing images is characterized by comprising the following steps:
acquiring original remote sensing image data;
preprocessing the original remote sensing image data to obtain a double-time-phase high-resolution remote sensing image; the double-time-phase high-resolution remote sensing image is a high-resolution remote sensing image of the same region in different time phases;
obtaining a change area of the building according to the double-time-phase high-resolution remote sensing image;
cutting the double-time-phase high-resolution remote sensing image and the change area of the building and dividing a data set according to a preset proportion to obtain a training data set and a test data set;
inputting the training data set into a twin network structure for training to obtain a trained twin network structure; wherein the twin network structure comprises two branched networks with the same structure;
inputting the test data set into the trained twin network structure for testing to obtain a tested twin network structure;
and preprocessing the image of the prediction area, inputting the preprocessed image into the tested twin network structure, and training to obtain a change detection diagram of the building.
2. The method for detecting the change of the building based on the double-temporal high-resolution remote sensing image according to claim 1, wherein the obtaining of the change area of the building according to the double-temporal high-resolution remote sensing image comprises:
obtaining the geographical ranges of the two corresponding images according to the double time-phase high-resolution remote sensing images, and judging whether the geographical ranges of the two images have an intersection region;
if yes, calculating the geographic range of the intersection area of the two images, and obtaining front and rear time phase images of the intersection area according to the geographic range of the intersection area;
and sequentially carrying out building labeling and XOR operation processing on the front time phase image and the rear time phase image of the intersection area to obtain a change area of the building.
3. The building change detection method based on the double-temporal high-resolution remote sensing image according to claim 1, characterized in that the twin network structure adopts an Encoder-Decoder architecture model; inputting the training data set into a twin network structure for training to obtain a trained twin network structure, wherein the training data set comprises:
sequentially carrying out down-sampling processing on the training data set for multiple times to extract the features of the double-time-phase high-resolution remote sensing image, and obtaining a plurality of feature maps with different sizes of the double-time-phase high-resolution remote sensing image;
fusing the feature maps with different sizes to obtain a fused feature map with the same size as the double-temporal high-resolution remote sensing image;
inputting the fusion feature map into an attention structure and a semantic segmentation structure in the twin network structure respectively to obtain a large-scale attention mask and a change detection feature;
and obtaining a change detection result of the building according to the large-scale attention mask and the change detection characteristics, and finishing the twin network structure training.
4. The building change detection method based on the double-temporal high-resolution remote sensing image according to claim 1, wherein the pre-processing of the prediction area image and the input of the pre-processed prediction area image to the tested twin network structure are trained to obtain a change detection map of the building, and the method comprises the following steps:
preprocessing the image of the prediction area to obtain a double-time-phase high-resolution remote sensing image of the prediction area;
forming a plurality of groups of prediction data by using an image tensor obtained by reading the double-temporal high-resolution remote sensing image of the prediction region image and adopting different dimensions;
sequentially inputting the multiple groups of prediction data into the tested twin network structure for prediction to obtain multiple groups of change detection characteristics and multiple groups of attention masks;
and according to the multiple groups of change detection characteristics and the multiple groups of attention masks, performing up-sampling to obtain a change detection diagram of the building.
5. The method for detecting the change of the building based on the double-temporal high-resolution remote sensing image according to claim 4, wherein the obtaining of the change detection map of the building according to the plurality of sets of change detection features and the plurality of sets of attention masks comprises:
and combining the multiple groups of change detection characteristics and the multiple groups of attention masks in sequence according to a principle of first-large and second-small to obtain a change detection diagram of the building.
6. The method for detecting building change based on double-time-phase high-resolution remote sensing images according to claim 1, wherein the step of preprocessing the original remote sensing image data to obtain double-time-phase high-resolution remote sensing images comprises the following steps:
and sequentially carrying out resampling, projection conversion, radiometric calibration, atmospheric correction and geographic registration processing on the original remote sensing image data to obtain the double-time-phase high-resolution remote sensing image.
7. The building change detection method based on the double-temporal high-resolution remote sensing image according to claim 1, wherein the twin network structure is two parallel network structures, and the two parallel network structures are the same in structure and share parameters.
8. A building change detection device based on double-time-phase high-resolution remote sensing images is characterized by comprising:
the data preprocessing module is used for acquiring original remote sensing image data and preprocessing the original remote sensing image data to obtain a double-time-phase high-resolution remote sensing image; the double-time-phase high-resolution remote sensing image is a high-resolution remote sensing image of the same region in different time phases;
the data change area detection module is used for obtaining a change area of the building according to the double time-phase high-resolution remote sensing image;
the data set generating module is used for cutting the double-time-phase high-resolution remote sensing image and the change area of the building and dividing the data set according to a preset proportion to obtain a training data set and a test data set;
the model training module is used for inputting the training data set into a twin network structure for training to obtain a trained twin network structure; wherein the twin network structure comprises two branched networks with the same structure;
the model testing module is used for inputting the testing data set to the trained twin network node for testing and training to obtain a tested twin network structure;
and the data change detection module is used for preprocessing the image of the prediction region and inputting the preprocessed image into the tested twin network structure for training to obtain a change detection diagram of the building.
9. An electronic device, comprising:
one or more processors;
a storage device to store one or more programs,
wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to perform the method of building change detection based on bi-temporal high resolution remote sensing images according to any one of claims 1 to 7.
10. A computer readable storage medium having stored thereon executable instructions which, when executed by a processor, cause the processor to execute the method for detecting a change in a building based on a two-time-phase high-resolution remote sensing image according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211223353.9A CN115719447A (en) | 2022-10-08 | 2022-10-08 | Building change detection method based on double-time-phase high-resolution remote sensing image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211223353.9A CN115719447A (en) | 2022-10-08 | 2022-10-08 | Building change detection method based on double-time-phase high-resolution remote sensing image |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115719447A true CN115719447A (en) | 2023-02-28 |
Family
ID=85253855
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211223353.9A Pending CN115719447A (en) | 2022-10-08 | 2022-10-08 | Building change detection method based on double-time-phase high-resolution remote sensing image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115719447A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117237823A (en) * | 2023-11-10 | 2023-12-15 | 中国科学院空天信息创新研究院 | Remote sensing basic model migration method and device based on zero sample learning |
-
2022
- 2022-10-08 CN CN202211223353.9A patent/CN115719447A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117237823A (en) * | 2023-11-10 | 2023-12-15 | 中国科学院空天信息创新研究院 | Remote sensing basic model migration method and device based on zero sample learning |
| CN117237823B (en) * | 2023-11-10 | 2024-03-08 | 中国科学院空天信息创新研究院 | Remote sensing basic model migration method and device based on zero sample learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110136170B (en) | Remote sensing image building change detection method based on convolutional neural network | |
| US20200340813A1 (en) | Map feature extraction system for computer map visualizations | |
| CN109447994B (en) | Remote sensing image segmentation method combining complete residual error and feature fusion | |
| Yao et al. | Mapping fine‐scale urban housing prices by fusing remotely sensed imagery and social media data | |
| CN109784283B (en) | Remote sensing image target extraction method based on scene recognition task | |
| CN113780296B (en) | Remote sensing image semantic segmentation method and system based on multi-scale information fusion | |
| CN109508681A (en) | The method and apparatus for generating human body critical point detection model | |
| CN103714339B (en) | SAR image road damaging information extracting method based on vector data | |
| CN108304761A (en) | Method for text detection, device, storage medium and computer equipment | |
| Li et al. | A comprehensive web-based system for flood inundation map generation and comparative analysis based on height above nearest drainage | |
| US11557053B2 (en) | Deep learning to correct map and image features | |
| CN110991430B (en) | Ground feature identification and coverage rate calculation method and system based on remote sensing image | |
| Ye et al. | Urban commerce distribution analysis based on street view and deep learning | |
| CN114758337B (en) | Semantic instance reconstruction method, device, equipment and medium | |
| CN115719447A (en) | Building change detection method based on double-time-phase high-resolution remote sensing image | |
| Xu et al. | Building height calculation for an urban area based on street view images and deep learning | |
| WO2021001631A1 (en) | Method and device for evaluating a geographical zone | |
| CN116611725A (en) | Land type identification method and device based on green ecological index | |
| CN115730718A (en) | Atmospheric NO combining hyper-spectral satellite and artificial intelligence 2 Space-time prediction algorithm | |
| Huang et al. | Simultaneous update of high-resolution land-cover mapping attempt: Wuhan and the surrounding satellite cities cartography using L2HNet | |
| CN114529834A (en) | Small target object detection method and detection device | |
| Krapf et al. | Deep learning for semantic 3D city model extension: Modeling roof superstructures using aerial images for solar potential analysis | |
| Fei et al. | Adapting Public Annotated Data Sets and Low-Quality Dash Cameras for Spatiotemporal Estimation of Traffic-Related Air Pollution: A Transfer-Learning Approach | |
| CN113011294B (en) | Method, computer equipment and medium for identifying circular sprinkling irrigation land based on remote sensing image | |
| CN117128977B (en) | High-quality green road path planning method, device and equipment based on double-image fusion |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |