Change Detection Device

Abstract

To provide a difference detection device and method capable of improving robustness in detection of a change in a building, by matching attribute information (distributions) between multi-time satellite images.SOLUTION: A difference detection device comprises: a data collection module configured to collect a satellite image set comprising a plurality of satellite images; a pre-processing module configured to generate a pre-processing image set by matching attribute information of the satellite images included in the satellite image set; and a difference determination module configured to generate difference information between the plurality of satellite images by determining a difference between pre-processing images included in the pre-processing image set, the attribute information comprising at least one of illumination information and color information of the satellite image.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a device and method for change detection.

[0002] Specifically, the present invention relates to a change detection device and method that can improve robustness in detecting changes in buildings, etc., by combining attribute information (distribution) between multi-time satellite images.

[0003] The present invention also relates to a change detection device and method that can improve the robustness and performance of change detection by utilizing an adaptive change detection algorithm that adds an attention module to the change detection algorithm. [Background technology]

[0004] The content of this section merely provides background information regarding the present invention and does not constitute prior art.

[0005] This invention is being carried out with the support of the Ministry of Trade, Industry and Energy and the Defense Acquisition Program Administration (Civil-Military Cooperation Agency), and the project announcement number is 2022-DC-2, the research project name is National Defense Technology Commercialization Support Project, the research and development project number is 22-DC-IN-17, the research project name is Feasibility study of 'Aerial imagery-based AI smart construction service' commercialization, and the research period is from 2022.12.01 to 2024.11.30.

[0006] Currently, hundreds of satellites around the world are performing their specific tasks in the Earth's orbit for purposes such as meteorology, communications, broadcasting, agriculture, space development, and military affairs. Satellite images taken by these satellites are used in a variety of ways as public data in a variety of fields.

[0007] Because such satellite images are taken at very high altitudes, technologies for processing, analyzing, and / or storing these satellite images using deep learning techniques have recently been attracting attention.

[0008] One of the applications of satellite image analysis is change detection technology, which compares and analyzes satellite images from multiple periods to confirm that buildings, etc. have been constructed or destroyed within a given time period.

[0009] However, existing change detection technologies have only attempted to improve the robustness of change detection models through methods such as data augmentation, and there have been no attempts to perform accurate change detection by matching attribute information (distribution) between multi-temporal satellite images.

[0010] Therefore, there has been a significant need for a technique that can improve the robustness of change detection through preprocessing such as matching distributions between multi-temporal satellite images. Summary of the Invention [Problem to be solved by the invention]

[0011] An object of the present invention is to provide a change detection device and method that can improve robustness in detecting changes in buildings, etc. by matching attribute information between multi-temporal satellite images. Specifically, an object of the present invention is to provide a change detection device and method that can more accurately detect changes in buildings, etc. by matching illumination information and color information between multi-temporal satellite images using a reference image.

[0012] Another object of the present invention is to provide a change detection apparatus and method that can improve the robustness and performance of change detection by using an adaptive change detection algorithm that adds an attention module to the change detection algorithm. Specifically, an object of the present invention is to provide a change detection apparatus and method that can perform an exchange process using the importance of each channel and spatial by additionally using attention modules related to a channel exchange module that exchanges any of RGB information between images and a spatial exchange module that exchanges pixels at specific coordinates, thereby improving the robustness and performance of change detection.

[0013] The objects of the present invention are not limited to those mentioned above, and other objects and advantages of the present invention not mentioned above can be understood from the following description and will become more clearly understood by the examples of the present invention. Furthermore, it will be easily understood that the objects and advantages of the present invention can be realized by the means and combinations thereof as claimed. [Means for solving the problem]

[0014] A change detection device according to some embodiments of the present invention includes a data collection module that collects a satellite image set including a plurality of satellite images, a pre-processing module that matches attribute information of each of the satellite images included in the satellite image set to generate a pre-processed image set, and a difference determination module that determines differences between the pre-processed images included in the pre-processed image set to generate change information between the plurality of satellite images, where the attribute information may include at least one of illumination information and color information of the satellite images.

[0015] Furthermore, the satellite image set includes a first satellite image taken from a first viewpoint and a second satellite image taken from a second viewpoint different from the first viewpoint, and the first satellite image and the second satellite image may be satellite images taken of the same coordinate area.

[0016] The pre-processing module may also generate the set of pre-processed images based on pre-defined reference images.

[0017] The preprocessing module can also extract the attribute information of the reference image and determine it as reference attribute information, and change the attribute information of each satellite image included in the satellite image set according to the determined reference attribute information.

[0018] The pre-processing module may also use a predefined image processing algorithm to modify attribute information of each satellite image included in the set of satellite images.

[0019] Additionally, the image processing algorithm may include at least one of image histogram matching and feature distribution matching.

[0020] In addition, the difference determination module can generate the change information using deep learning technology.

[0021] The difference determination module may also generate the change information based on a predefined change detection algorithm related to change detection between a plurality of images.

[0022] The image processing device may further include a learning module that learns an adaptive change detection algorithm in which an attention module is added to the change detection algorithm, and the difference determination module may generate the change information using the adaptive change detection algorithm.

[0023] The adaptive change detection algorithm may also include a first attention module which is an attention module related to a channel exchange module that exchanges at least one of RGB information between the preprocessed images of the preprocessed image set, and a second attention module which is an attention module related to a spatial exchange module that exchanges at least one pixel at a specific coordinate between the preprocessed images of the preprocessed image set. [Effects of the Invention]

[0024] The change detection apparatus and method according to some embodiments of the present invention can improve the robustness of change detection of buildings, etc. by matching attribute information between multi-temporal satellite images. Specifically, the change detection apparatus and method according to some embodiments of the present invention can enable more accurate change detection of buildings, etc. by matching illumination information and color information between multi-temporal satellite images using a reference image.

[0025] In addition, some embodiments of the change detection apparatus and method may further improve the robustness and performance of change detection by using an adaptive change detection algorithm that adds an attention module to the change detection algorithm. That is, some embodiments of the change detection apparatus and method may further use an attention module related to a channel exchange module that exchanges any of RGB information between images, or an attention module related to a spatial exchange module that exchanges pixels at specific coordinates, thereby performing an exchange process using the importance of each channel and spatial, thereby further improving the robustness and performance of change detection.

[0026] In addition to the above, the specific effects of the present invention will be explained together with the following description of specific matters for carrying out the invention. [Brief explanation of the drawings]

[0027] [Figure 1] FIG. 1 illustrates a change detection system according to some embodiments of the present invention. [Figure 2] FIG. 2 is a block diagram of a change detection device according to some embodiments of the present invention. [Figure 3] FIG. 3 is a diagram illustrating a set of satellite images according to some embodiments of the present invention. [Figure 4a] FIG. 4a is a diagram illustrating the operation of a pre-processing module according to some embodiments of the present invention. [Figure 4b] FIG. 4b is a diagram illustrating the operation of a pre-processing module according to some embodiments of the present invention. [Figure 5a] FIG. 5a is a diagram illustrating the neural network structure of a deep learning module according to some embodiments of the present invention. [Figure 5b] FIG. 5b is a diagram illustrating an autoencoder according to some embodiments of the present invention. [Figure 6] FIG. 6 is a diagram illustrating a predefined change detection algorithm. [Figure 7] FIG. 7 is a diagram illustrating an adaptive change detection algorithm according to some embodiments of the present invention. [Figure 8] FIG. 8 is a flowchart of a change detection method according to some embodiments of the present invention. [Figure 9] FIG. 9 is a diagram illustrating a hardware implementation of a change detection device that executes a change detection method according to some embodiments of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0028] The terms and words used in this specification and claims should not be interpreted as being limited to their general or dictionary meanings. They should be interpreted as meanings and concepts consistent with the technical idea of ​​the present invention, based on the principle that an inventor can define the concept of a term or word to best describe his or her invention. In addition, the embodiments described in this specification and the configurations shown in the drawings are merely examples in which the present invention is realized, and do not represent the entire technical idea of ​​the present invention. It should be understood that there may be various equivalents, modifications, and applicable examples that can replace them at the time of filing this application.

[0029] Terms such as "first," "second," "A," and "B" used in this specification and claims may be used to describe various elements, but the elements should not be limited by these terms. These terms are used solely to distinguish one element from another. For example, a first element can be designated as a second element, and similarly, a second element can be designated as a first element, without departing from the scope of the present invention. The term "and / or" includes a combination of multiple related listed items or any of multiple related listed items.

[0030] The terms used in the present specification and claims are merely used to describe specific embodiments and are not intended to limit the present invention. The singular expressions include the plural expressions unless the context clearly indicates otherwise. In this application, the terms "comprise" or "have" and the like should be understood not to preclude the presence or possibility of additional features, numbers, steps, operations, components, parts, or combinations thereof described in the specification.

[0031] Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0032] Terms as defined in commonly used dictionaries should be construed to have a meaning consistent with the contextual meaning of the relevant art, and not to be construed as idealized or overly formal unless expressly defined in this application.

[0033] Furthermore, the configurations, processes, steps, or methods included in the embodiments of the present invention may be shared to the extent that they are not technically inconsistent with each other.

[0034] In the following, change detection apparatus and methods according to some embodiments of the present invention will be described with reference to FIGS.

[0035] FIG. 1 illustrates a change detection system according to some embodiments of the present invention.

[0036] Referring to FIG. 1, a change detection system (1) according to some embodiments of the present invention may include an external database (100), a change detection device (200), and a communication network (300).

[0037] The external database (100) may be a database that stores, manages, and / or transmits raw data that is the subject of change detection.

[0038] In some examples, the external database 100 may be in the form of various types of electronic devices, such as a computer, a laptop, a mobile device, a wearable device, a workstation, a data center, an internet data center (IDC), a direct attached storage (DAS) system, a storage area network (SAN) system, a network attached storage (NAS) system, and a redundant array of independent disks (RAID) system, although embodiments of the present invention are not limited thereto.

[0039] The external database 100 can transmit raw data for the change detection device 200 to perform change detection to the change detection device 200. In other words, the change detection device 200 can receive raw data related to change detection from the external database 100.

[0040] The original data may refer to a data set that is the target of change detection. For example, the original data may include an image, an image frame, an image, a panorama, etc., for a specific coordinate region. In other words, the original data may include image data, image frame data, image data, panorama data, etc., for a specific coordinate region. However, the embodiment of the present invention is not limited thereto.

[0041] In some examples, the raw data may include a set of satellite images for a particular coordinate region.

[0042] The satellite image set may refer to a set of images taken from a satellite located at a predetermined height above a specific coordinate region, or in other words, may include multiple satellite images taken from a satellite above a specific coordinate region.

[0043] For example, the satellite image set may include a first satellite image taken from a first viewpoint for a specific coordinate region and a second satellite image taken from a second viewpoint for the coordinate region. In this case, the first viewpoint and the second viewpoint may be different viewpoints from each other, for example, the second viewpoint may be a viewpoint located later in time than the first viewpoint. Furthermore, the first satellite image and the second satellite image may be satellite images for the same coordinate region.

[0044] Taken together, a satellite image set may include multiple satellite images taken at different times for the same coordinate region, and such a satellite image set may be referred to as a "multi-time satellite image."

[0045] For the sake of convenience, the following description will be given assuming that the original data stored and managed by the external database (100) is a set of satellite images.

[0046] The change detection device (200) can generate change information between each satellite image included in the set of satellite images, i.e., the change detection device (200) can perform change detection from the set of satellite images received from the external database (100).

[0047] In this case, the change detection device (200) may be in the form of a workstation, a data center, an internet data center (IDC), a DAS (direct attached storage) system, a SAN (storage area network) system, a NAS (network attached storage) system, and a RAID (redundant array of inexpensive disks, or redundant array of independent disks) system, but the embodiment of the present invention is not limited thereto.

[0048] In some examples, the change detection device (200) can pre-process a set of satellite images and then generate change information between each of the pre-processed satellite images.

[0049] Specifically, the change detection device 200 can first generate a set of preprocessed images by matching attribute information (distribution) between satellite images included in the set of satellite images.

[0050] For example, the change detection device (200) can match attribute information of a first satellite image and a second satellite image included in a satellite image set to generate a first preprocessed image and a second preprocessed image, and then determine the generated first preprocessed image and second preprocessed image as a preprocessed image set.

[0051] The attribute information may include illumination information, color information, etc. In other words, the change detection device 200 may perform preprocessing to match illumination information, color information, etc. between the first satellite image and the second satellite image.

[0052] The illumination information may include information regarding the amount of light and / or intensity of light reaching a particular coordinate region determined within the satellite image. In other words, the illumination information may include information regarding the amount and intensity of light reaching a particular coordinate region of the satellite image.

[0053] The color information may include information about the color characteristics of each pixel of the satellite image, in other words, the color information may include information about the channels (Red, Green, Blue, RGB) that each pixel of the satellite image contains.

[0054] Through this pre-processing process, the change detection device 200 of the present invention can perform more accurate change detection and improve the robustness of change detection.

[0055] In other words, if the illuminance and color information of each satellite image included in a satellite image set differ, change detection, which aims to determine the construction or disappearance of structures such as buildings, may not be performed properly. Specifically, from the perspective of "illuminance information," in multiple satellite images of the same coordinate area, the first satellite image may have been taken at a specific time with high illuminance, while the second satellite image may have been taken at a different time with low illuminance. These differences in illuminance can change image characteristics such as the length of a building's shadow or the reflective properties of an object, and attempts to detect changes based on these characteristics may result in false recognition. Furthermore, from the perspective of "color information," if the color information of the first and second satellite images is inconsistent, for example, a building may appear bright in the first satellite image but dark in the second satellite image. This can lead to confusion when determining the presence or absence of a building in each satellite image.

[0056] Thus, unless the consistency of the illumination information and / or color information of each satellite image included in the satellite image set is guaranteed, there is a high possibility that an actual change (construction or demolition of a building) will be confused with a change due to such external factors (illumination and / or color). Therefore, the change detection device (200) of the present invention can increase the robustness of change detection by performing preprocessing aimed at reducing the influence of such external factors.

[0057] At this time, the change detection device (200) can use the reference image to match the attribute information of each satellite image.

[0058] For example, the change detection device 200 can determine reference attribute information from the reference image and match the attribute information of each satellite image according to the determined reference attribute information. In other words, the change detection device 200 can process the attribute information of each satellite image so that it matches the reference attribute information.

[0059] In this case, the change detection device 200 can change the attribute information of each satellite image using a predefined image processing algorithm. In other words, the change detection device 200 can use a predefined image processing algorithm when processing the attribute information of each satellite image so that it matches the reference attribute information.

[0060] The image processing algorithms used by the change detection device 200 may include image histogram matching, which matches the brightness of each satellite image, and / or feature distribution matching, which compares and matches the distribution of feature points in the images. However, the present invention is not limited to this, and the change detection device 200 may use other algorithms such as color transfer, gamma correction, and white balance adjustment.

[0061] After completing the preprocessing, the change detection device 200 can generate change information between each satellite image in the set of satellite images by comparing the preprocessed images. In other words, the change detection device 200 can determine change information between a first satellite image taken at a first time point and a second satellite image taken at a second time point through the comparison between the preprocessed images.

[0062] In this case, the change information may include information about structures (for example, buildings) that have been newly constructed or destroyed at the second time point compared to the first time point.

[0063] In some examples, the change detection device (200) may use deep learning techniques to generate the change information. In other words, the change detection device (200) may use deep learning techniques to compare a first preprocessed image and a second preprocessed image included in the set of preprocessed images and generate the change information based on the comparison results.

[0064] For example, the change detection device 200 may generate change information based on a predefined change detection algorithm. For example, the change detection device 200 may train an adaptive change detection algorithm in which an attention module is added to the change detection algorithm, and generate change information using the trained adaptive change detection algorithm.

[0065] In this case, the adaptive change detection algorithm may include a first attention module which is an attention module related to a Channel Exchange Module that exchanges any of the RGB information between preprocessed images included in the preprocessed image set, and / or a second attention module which is an attention module related to a Spatial Exchange Module that exchanges pixels at specific coordinates between preprocessed images included in the preprocessed image set.

[0066] The detailed operation process of the change detection device 200 will be described later.

[0067] Meanwhile, the communication network 300 serves to connect the external database 100 and the change detection device 200. That is, the communication network 300 refers to a communication network that provides a connection path so that the change detection device 200 can send and receive data from the external database 100. The communication network 300 may include, for example, wired networks such as LANs (Local Area Networks), WANs (Wide Area Networks), MANs (Metropolitan Area Networks), and ISDNs (Integrated Service Digital Networks), as well as wireless networks such as wireless LANs, CDMA, Bluetooth, and satellite communications, but the scope of the present invention is not limited thereto.

[0068] The change detection device (200) will now be described in more detail with reference to FIG.

[0069] FIG. 2 is a block diagram of a change detection device according to some embodiments of the present invention.

[0070] Referring to Figures 1 and 2, the change detection device (200) receives a satellite image set (hereinafter referred to as "SI_set"), and then performs change detection based on the received satellite image set (SI_set) to generate difference information (hereinafter referred to as "DI").

[0071] Specifically, the change detection device (200) may include a data collection module (210), a preprocessing module (220), a difference determination module (230), and a learning module (240).

[0072] The data collection module (210) can receive the satellite image set (SI_set), in other words, the data collection module (210) can receive the satellite image set (SI_set) from the external database (100).

[0073] The satellite image set (SI_set) may refer to a set of images taken from a satellite located at a predetermined height above a specific coordinate region. In other words, the satellite image set (SI_set) may include multiple satellite images (SI_1, SI_2) taken from a satellite above a specific coordinate region.

[0074] The satellite image set (SI_set) will now be described in more detail with reference to FIG.

[0075] FIG. 3 is a diagram illustrating a set of satellite images according to some embodiments of the present invention.

[0076] 3, the satellite image set (SI_set) may refer to a set of images taken from a satellite located at a predetermined height above a specific coordinate region. In other words, the satellite image set (SI_set) may include multiple satellite images (SI_1, SI_2) taken from a satellite above a specific coordinate region.

[0077] For example, a satellite image set (SI_set) may include a first satellite image (SI_1) taken from a first viewpoint for a specific coordinate region and a second satellite image (SI_2) taken from a second viewpoint for the coordinate region. In this case, the first viewpoint and the second viewpoint may be different from each other, and for example, the second viewpoint may be a viewpoint located later in time than the first viewpoint.

[0078] In this case, the first satellite image (SI_1) and the second satellite image (SI_2) may be satellite images of the same coordinate region. In other words, the first satellite image (SI_1) and the second satellite image (SI_2) may be satellite images taken of the same coordinate region.

[0079] Taken together, a satellite image set (SI_set) can include multiple satellite images (SI_1, SI_2) taken at different times for the same coordinate region, and such a satellite image set (SI_set) can be called a "multi-time satellite image."

[0080] 1 and 2, the data collection module 210 can transmit the received satellite image set (SI_set) to other components within the change detection device 200. For example, the data collection module 210 can transmit the satellite image set (SI_set) to the pre-processing module 220, but the present invention is not limited thereto.

[0081] The data collection module (210) can be implemented using various communication modules, allowing data exchange between the external database (100) and the change detection device (200) via a communication network (300 in FIG. 1).

[0082] The pre-processing module 220 can pre-process the satellite image set (SI_set). In other words, the pre-processing module 220 can pre-process the satellite image set (SI_set) to generate a pre-processed image set (hereinafter referred to as "PI_set"). That is, the pre-processing module 220 can pre-process a first satellite image (SI_1) and a second satellite image (SI_2) to generate a first pre-processed image (PI_1) and a second pre-processed image (PI_2), respectively.

[0083] In some examples, the preprocessing module 220 may generate a preprocessed image set (PI_set) by matching attribute information (distribution) between each satellite image included in the satellite image set (SI_set). In other words, the preprocessing module 220 may generate a first preprocessed image (PI_1) and a second preprocessed image (PI_2) by matching attribute information between a first satellite image (SI_1) and a second satellite image (SI_2). That is, the attribute information between the first preprocessed image (PI_1) and the second preprocessed image (PI_2) generated by the preprocessing module 220 may be identical or similar within a predetermined critical range.

[0084] The attribute information may include illumination information, color information, etc. In other words, the pre-processing module 220 may perform pre-processing operations to match illumination information, color information, etc. between the first satellite image SI_1 and the second satellite image SI_2.

[0085] The illumination information may include information regarding the amount of light and / or intensity of light reaching a particular coordinate region determined within the satellite image. In other words, the illumination information may include information regarding the amount and intensity of light reaching a particular coordinate region of the satellite image.

[0086] The color information may include information about the color characteristics of each pixel of the satellite image, in other words, the color information may include information about the channels (Red, Green, Blue, RGB) that each pixel of the satellite image contains.

[0087] Through this pre-processing process, the change detection device 200 of the present invention can perform more accurate change detection and improve the robustness of change detection.

[0088] In other words, if the illuminance and color information of each satellite image (SI_1, SI_2) included in a satellite image set (SI_set) differ, change detection, which aims to determine the construction or disappearance of structures such as buildings, may not be performed properly. Specifically, from the perspective of "illuminance information," among multiple satellite images (SI_1, SI_2) of the same coordinate area, the first satellite image (SI_1) may have been captured at a specific time with high illuminance, while the second satellite image (SI_2) may have been captured at a different time with low illuminance. These differences in illuminance can change image characteristics such as the length of a building's shadow or the reflective properties of an object, and attempts to detect changes based on these characteristics may result in false recognition. Furthermore, from the perspective of "color information," if the color information of the first satellite image (SI_1) and the second satellite image (SI_2) is inconsistent, for example, a building may appear bright in the first satellite image (SI_1) but dark in the second satellite image (SI_2). In this case, it may be confusing to determine whether or not a building exists in each satellite image (SI_1, SI_2).

[0089] Thus, unless the consistency of the illumination information and / or color information of each satellite image (SI_1, SI_2) included in the satellite image set (SI_set) is guaranteed, there is a high possibility that an actual change (construction or demolition of a building) will be confused with a change due to such external factors (illumination and / or color). Therefore, the preprocessing module (220) of the present invention can increase the robustness of change detection by performing preprocessing aimed at reducing the influence of such external factors.

[0090] At this time, the pre-processing module (220) can use the reference image to match the attribute information of each satellite image (SI_1, SI_2).

[0091] Hereinafter, with reference to FIGS. 4a and 4b, the process by which the pre-processing module (220) of the present invention generates a pre-processed image set (PI_set) using reference images will be described in more detail.

[0092] 4a and 4b are diagrams illustrating the operation of a pre-processing module according to some embodiments of the present invention. Specifically, FIG. 4a is a conceptual diagram illustrating the process in which the pre-processing module (220) performs pre-processing using a reference image (RI), and FIG. 4b is a block diagram illustrating the process in which the pre-processing module (220) performs pre-processing using a reference image (RI).

[0093] Referring to Figures 4a and 4b, the pre-processing module (220) can pre-process each satellite image (SI_1, SI_2) included in the satellite image set (SI_set) using a reference image (RI).

[0094] In some examples, the preprocessing module (220) can generate a preprocessed image set (PI_set) by matching attribute information of each satellite image (SI_1, SI_2) included in the satellite image set (SI_set) using a reference image (RI).

[0095] In this case, as described above, the attribute information may include illuminance information, color information, etc. The illuminance information may include information regarding the amount of light and / or intensity of light reaching a specific coordinate area determined within the satellite image. In other words, the illuminance information may include information regarding the amount and intensity of light reaching a specific coordinate area within the satellite image. The color information may include information regarding the color characteristics of each pixel of the satellite image. In other words, the color information may include information regarding the channels (red, green, blue, RGB) included in each pixel of the satellite image.

[0096] Specifically, the pre-processing module 220 may first extract reference characteristic information (hereinafter referred to as "RCI") from the reference image (RI). For example, the pre-processing module 220 may extract the reference characteristic information (RCI) using a predetermined attribute information determination algorithm. In this case, the attribute information determination algorithm may include known algorithms based on, for example, histogram analysis, feature analysis, etc., but the present invention is not limited thereto.

[0097] Then, the pre-processing module (220) can match the attribute information of each satellite image (SI_1, SI_2) according to the reference attribute information (RCI). In other words, the change detection device (200) can process the attribute information of each satellite image (SI_1, SI_2) so that it has the same value as the reference attribute information (RCI).

[0098] In this case, the change detection device 200 can change the attribute information of each satellite image (SI_1, SI_2) using a predefined image processing algorithm. In other words, the change detection device 200 can use a predefined image processing algorithm when processing the attribute information of each satellite image (SI_1, SI_2) so that it matches the reference attribute information (RCI).

[0099] The image processing algorithms used by the change detection device 200 may include image histogram matching, which matches the brightness of each satellite image, and / or feature distribution matching, which compares and matches the distribution of feature points in the images. However, the present invention is not limited to this, and the change detection device 200 may use other algorithms such as color transfer, gamma correction, and white balance adjustment.

[0100] The attribute information between the first preprocessed image (PI_1) and the second preprocessed image (PI_2) of the preprocessed image set (PI_set) generated by the preprocessing module (220) may be identical or similar within a predetermined critical range.

[0101] Referring again to FIGS. 1 and 2, the preprocessing module (220) can transmit the generated set of preprocessed images to the difference determination module (230).

[0102] The difference determination module 230 can generate change information (DI) between each satellite image SI_1, SI_2 in the satellite image set SI_set by comparing each preprocessed image PI_1, PI_2 in the preprocessed image set PI_set. In other words, the change detection device 200 can determine change information (DI) between a first satellite image SI_1 captured at a first time point and a second satellite image SI_2 captured at a second time point through comparison between the preprocessed images PI_1, PI_2.

[0103] In this case, the change information (DI) may include information about structures (for example, buildings) that have been newly constructed or disappeared at the second time point compared to the first time point.

[0104] In some examples, the difference determination module (230) may utilize deep learning techniques to generate the change information (DI).

[0105] At this time, the difference determination module (230) can learn based on machine learning.

[0106] More specifically, deep learning technology, a type of machine learning, is a method of learning at multiple levels based on data. In other words, deep learning refers to a set of machine learning algorithms that extract key data from multiple data by increasing the level.

[0107] In some examples, the difference determination module 230 may utilize various well-known deep learning architectures, such as a convolutional neural network (CNN), a recurrent neural network (RNN), a deep belief network (DBN), a graph neural network (GNN), a generative adversarial network (GAN), a transformer, an autoencoder, or the like.

[0108] Specifically, a CNN (Convolutional Neural Network) is a model that mimics the function of the human brain, based on the assumption that when humans recognize an object, they extract basic features of the object, then perform complex calculations in the brain, and recognize the object based on the results. CNNs can include, but are not limited to, well-known structures such as LeNet, AlexNet, VGGNet, GoogleNet, and ResNet.

[0109] RNNs (Recurrent Neural Networks) are often used in natural language processing and are an effective structure for processing time-series data that changes over time, and can be configured as an artificial neural network structure by stacking layers at each moment.

[0110] A DBN (Deep Belief Network) is a deep learning structure that consists of multiple layers of RBM (Restricted Boltzmann Machine), a deep learning technology. By repeatedly learning RBM (Restricted Boltzmann Machine) to a certain number of layers, a DBN (Deep Belief Network) with the corresponding number of layers can be constructed.

[0111] A GNN (Graphic Neural Network, hereinafter referred to as "GNN") is an artificial neural network structure implemented in a manner that derives similarities and feature points between modeling data using modeling data that is modeled based on data mapped between specific parameters.

[0112] A GAN (Generative Adversarial Network, hereinafter referred to as "GAN") is an artificial neural network structure that generates new data with a similar form to input data using a generative neural network and a discriminative neural network. GANs may include well-known deep convolutional GANs (DCGANs), conditional GANs (CGANs), Wasserstein GANs (WGANs), style-based GANs (StyleGANs), and cycle GANs, but the embodiments of the present invention are not limited thereto.

[0113] The Transformer is an attention-driven, encoder-decoder artificial neural network that can grasp the overall meaning between input and output sequences. The Transformer uses an attention mechanism to ensure that all elements of the input sequence affect the output sequence, allowing both the encoder and decoder to consider the entire sequence. The Transformer can use natural language, time series data, and even image patches as input.

[0114] An autoencoder is a deep learning architecture that extracts and reconstructs data features. Typically, an autoencoder includes an encoder that compresses input values ​​and a decoder that restores the compressed data. The encoder converts the input values ​​into a low-dimensional latent representation, and the decoder restores the latent representation to the same dimensions as the input values. The encoder and decoder can each be configured as a multilayer perceptron (MLP). To train an autoencoder, input data is input and weights and biases are trained to minimize the difference between the output value and the input value. Trained autoencoders can effectively extract features from the input data and restore noisy input data. Autoencoders are primarily used in data compression, dimensionality reduction, noise removal, and data generation, as well as in image recognition, natural language processing, and speech recognition.

[0115] Meanwhile, the artificial neural network training of the difference determination module (230) can be performed by adjusting the weights of the connecting lines between nodes (and adjusting the bias values, if necessary) so that a desired output is obtained for a given input. Furthermore, the artificial neural network can continuously update the weight values ​​through training. Also, methods such as back propagation can be used to train the artificial neural network.

[0116] In this case, the machine learning method of the artificial neural network can be unsupervised learning, semi-supervised learning, supervised learning, etc. Furthermore, the difference determination module (230) can be controlled to automatically update the artificial neural network structure for outputting analysis data after learning according to settings.

[0117] Hereinafter, with reference to FIG. 5a, the neural network structure of the difference determination module (230) implemented by deep learning technology according to some embodiments of the present invention will be described in more detail.

[0118] FIG. 5a is a diagram illustrating the neural network structure of a deep learning module according to some embodiments of the present invention.

[0119] Referring to Figures 2 and 5a, the difference determination module (230) according to some embodiments of the present invention may include an input layer (input) with a preprocessed image set (PI_set) as an input node, an output layer (Output) with change information (DI) as an output node, and M hidden layers arranged between the input layer and the output layer.

[0120] Here, weights can be set for the edges connecting the nodes in each layer. These weights or the presence or absence of edges can be added, deleted, or updated during the learning process. Therefore, the weights of the nodes and edges located between the k input nodes and the i output nodes can be updated during the learning process.

[0121] Before the difference determination module 230 performs learning, all nodes and edges can be set to initial values. However, when cumulative information is input, the weights of the nodes and edges are changed, and during this process, a match may be made between the parameters input to the learning factor (preprocessed image set (PI_set)) and the values ​​assigned to the output node (change information (DI)).

[0122] Additionally, when using a cloud server, the difference determination module 230 can receive and process a large number of parameters, thereby enabling the difference determination module 230 to perform learning based on a large amount of data.

[0123] The weights of the nodes and edges between the input nodes and output nodes constituting the difference determination module 230 can be updated through the learning process of the difference determination module 230. It goes without saying that the parameters input or output by the difference determination module 230 can be expanded to various data other than the preprocessed image set (PI_set) and change information (DI).

[0124] The autoencoder used in the present invention will now be described in more detail with reference to FIG. 5b.

[0125] FIG. 5b is a diagram illustrating an autoencoder according to some embodiments of the present invention.

[0126] Referring to Figures 2 and 5b, an autoencoder (hereinafter referred to as "AE") includes an encoder neural network (hereinafter referred to as encoder unit (EN)) and a decoder neural network (hereinafter referred to as decoder unit (DN)), and may include a middle layer (ML) disposed between the encoder unit (EN) and the decoder unit (DN).

[0127] An autoencoder (AE) is a type of deep neural network model that compresses data input through an encoder (EN) to reduce the size of the data, and then converts the reduced data using a decoder (DN) to the same size as the input data in the encoder (EN) and outputs it, thereby making the output data of the autoencoder (AE) the same as the input data.

[0128] An autoencoder (AE) learns the features of input data in an unsupervised manner. To do this, it converts the input data through the encoder (EN) into low-dimensional data (latent representation) that effectively represents the features, and the converted data can then be restored to the original data through the decoder (DN).

[0129] An autoencoder (AE) can learn patterns inherent in the original data with the goal of minimizing the reconstruction error, which corresponds to the difference between the original data (X1, X2, X3, X4; i.e., input data (preprocessed image set (PI_set))) and the restored data (X1', X2', X3', X4'; i.e., output data (transformation information (DI))).

[0130] Referring again to Figures 1 and 2, the difference determination module (230) using such deep learning technology can generate change information (DI) based on a predefined change detection algorithm (hereinafter referred to as "CDA").

[0131] In some examples, the learning module (240) can train an adaptive change detection algorithm (CDA_adj) in which an attention module (Attention Module) is added to the change detection algorithm (CDA), and the difference determination module (230) can generate change information (DI) using the adaptive change detection algorithm (CDA_adj) learned and received from the learning module (240).

[0132] In this case, the adaptive change detection algorithm may include a first attention module which is an attention module related to a channel exchange module that exchanges any of RGB (Red, Green, Blue) information between preprocessed images (PI_1, PI_2) included in the preprocessed image set (PI_set), and / or a second attention module which is an attention module related to a spatial exchange module that exchanges pixels at specific coordinates between preprocessed images (PI_1, PI_2) included in the preprocessed image set (PI_set).

[0133] A predefined change detection algorithm will now be described with reference to FIG. 6, and an adaptive change detection algorithm utilized by the difference determination module (230) according to some embodiments of the present invention will now be described with reference to FIG.

[0134] FIG. 6 is a diagram illustrating a predefined change detection algorithm.

[0135] 6, the change detection algorithm (CDA) may be an algorithm that searches for changed portions between multiple preprocessed images (PI_1, PI_2) with different viewpoints. In this case, the change detection algorithm (CDA) may be an algorithm based on binary semantic segmentation.

[0136] To explain in detail how the change detection algorithm (CDA) works, first, input images (PI_1, PI_2) can be input to the initial layer (stem).

[0137] Next, features can be extracted through the initial layer (stem), the first stage (stage1), and the second stage (stage2).

[0138] Next, a spatial exchange module (hereinafter referred to as "SEM") can execute a process of exchanging pixels at specific coordinates between the preprocessed images (PI_1, PI_2). For example, the spatial exchange module (SEM) can change and / or replace the values ​​of pixels at any positions (e.g., 100x100, 100x102, 100x104, etc.) between the preprocessed images (PI_1, PI_2) consisting of 256x256 pixels.

[0139] In this case, the number of pixels changed between the pre-processed images (PI_1, PI_2) by the spatial exchange module (SEM) may vary. In other words, the spatial exchange module (SEM) may change and / or replace the values ​​of one or more pixels between the pre-processed images (PI_1, PI_2).

[0140] Next, features can be extracted through the third stage (stage 3).

[0141] Next, a Channel Exchange Module (hereinafter referred to as "CEM") can execute a process of exchanging one or more channel information such as RGB (Red, Green, Blue) between the preprocessed images (PI_1, PI_2). In other words, the Channel Exchange Module (CEM) can change and / or replace the values ​​of one channel (e.g., any one of RGB) or multiple channels (e.g., any two of RGB) between the preprocessed images (PI_1, PI_2).

[0142] In this case, the number of channels exchanged between the pre-processed images (PI_1, PI_2) by the channel exchange module (CEM) may vary. In other words, the channel exchange module (CEM) may exchange and / or replace one or more channels between the pre-processed images (PI_1, PI_2).

[0143] Next, features can be extracted through the fourth stage (stage 4).

[0144] The channel exchange module (CEM) can then re-execute the process of exchanging any of the RGB information in each preprocessed image (PI_1, PI_2).

[0145] Then, the MLP decoder can restore the results of the above process into the form of input data (each pre-processed image (PI_1, PI_2)).

[0146] The change detection module (Flow Dual Alignment Fusion, FDAF module) then merges the features recovered by the decoder (MLP decoder) into one and can detect the changed parts between the two preprocessed images (PI_1, PI_2).

[0147] In some examples, the change detection module (FDAF module) can perform pre-processing to detect changes between the pre-processed images (PI_1, PI_2), including correcting registration errors between features extracted from each of the pre-processed images (PI_1, PI_2).

[0148] For example, the change detection module (FDAF module) can input the feature maps extracted from each of the preprocessed images (PI_1, PI_2) into a small neural network consisting of two convolutional layers to obtain corrected features, and then calculate the distances by comparing the corrected features with the original feature maps.The change detection module (FDAF module) can then correct the distortion error between the features extracted from each of the preprocessed images (PI_1, PI_2) using the obtained distances, and then detect changed areas.

[0149] The Projection Head can then display the altered portions detected by the Alteration Detection Module (FDAF) in one image.

[0150] FIG. 7 is a diagram illustrating an adaptive change detection algorithm according to some embodiments of the present invention.

[0151] 7, the adaptive change detection algorithm (CDA_adj) may be an algorithm that searches for changed portions between multiple preprocessed images (PI_1, PI_2) with different viewpoints. In this case, the adaptive change detection algorithm (CDA_adj) may be an algorithm based on binary semantic segmentation.

[0152] Compared to the change detection algorithm (CDA) described above in FIG. 6, the adaptive change detection algorithm (CDA_adj) may further include a channel attention module (hereinafter referred to as "CAM"), which is an attention module for the channel exchange module (CEM), and a spatial attention module (hereinafter referred to as "SAM"), which is an attention module for the spatial exchange module (SEM).

[0153] The attention mechanism implemented by the attention module assigns weights to various parts of the data, allowing the algorithm to focus more on important parts. Therefore, the attention module can calculate weights that indicate the importance or relevance of specific parts or values ​​within the preprocessed images (PI_1, PI_2). For example, the channel attention module (CAM) can determine which channel values ​​of RGB information are important to exchange between the preprocessed images (PI_1, PI_2) for generating difference information (DI), and the spatial attention module (SAM) can determine which pixel values ​​are important to exchange between the preprocessed images (PI_1, PI_2) for generating difference information (DI).

[0154] To specifically explain the operation method of the adaptive change detection algorithm (CDA_adj), first, the input images (PI_1, PI_2) can be input to the channel attention module (CAM).

[0155] The channel attention module (CAM) may determine which channel values ​​of RGB (Red, Green, Blue) information are important for generating difference information (DI) to be exchanged between the preprocessed images (PI_1, PI_2). In other words, the channel attention module (CAM) may transmit exchange information regarding which channel values ​​of RGB information are to be exchanged between the preprocessed images (PI_1, PI_2) to the channel exchange module (CEM) based on the importance determination result. In this case, the number of channels to be exchanged determined by the channel attention module (CAM) may be one or two or more.

[0156] Then, the channel exchange module (CEM) can execute a process of exchanging any one or more channel information such as RGB information between the preprocessed images (PI_1, PI_2). For example, the channel exchange module (CEM) can change or replace any one or more channels (one or more of RGB) between the preprocessed images (PI_1, PI_2) according to the exchange information sent from the channel attention module (CAM).

[0157] Then, encoding can be performed through a first encoder (Encoder1).

[0158] The encoding results can then be input into a spatial attention module (SAM).

[0159] The spatial attention module (SAM) can determine which pixel values ​​between the preprocessed images (PI_1, PI_2) are important for generating difference information (DI). In other words, the spatial attention module (SAM) can transmit exchange information regarding which coordinates of pixels to exchange between the preprocessed images (PI_1, PI_2) to the spatial exchange module (SEM) based on the importance determination result. In this case, the number of pixels to be exchanged determined by the spatial attention module (SAM) can be one or two or more.

[0160] Then, the spatial exchange module (SEM) can execute a process of exchanging pixels at specific coordinates between the included preprocessed images (PI_1, PI_2). For example, the spatial exchange module (SEM) can change or replace the values ​​of pixels (e.g., 150×150, 100×100, 100×102, 100×104, etc.) in the preprocessed images (PI_1, PI_2) consisting of 256×256 pixels according to the exchange information sent from the spatial attention module (SAM).

[0161] Then, encoding can be performed through a second encoder (Encoder2).

[0162] The encoding result can then be input to the channel attention module (CAM), and the channel exchange module (CEM) can perform an exchange process according to the exchange information output by the channel attention module (CAM). A detailed description of this has been given above, so it will be omitted here.

[0163] Then, decoding can be performed through a decoder. For example, the decoder can perform a process of restoring the results of the above-described process into the form of input data (each preprocessed image (PI_1, PI_2)). In this case, the decoder can play the same role as the decoder (MLP decoder) of FIG. 6.

[0164] The difference module then merges the values ​​restored by the decoder into one and detects the changed parts between the two preprocessed images (PI_1, PI_2). In this case, the difference module can play the same role as the change detection module (Flow Dual Alignment Fusion, FDAF module) in Figure 6.

[0165] Then, the classifier can generate change information (DI) by displaying the changed parts detected by the difference module in one image. In this case, the classifier can play the same role as the projection head in Figure 6.

[0166] Of course, the embodiment of the generated change information (DI) is not limited to that shown in FIG.

[0167] Figure 8 is a flowchart of a change detection method according to some embodiments of the present invention. Each step (S100 to S300) in Figure 8 can be performed by the change detection device (200 in Figure 1) in Figures 1 and 2. Below, we will briefly explain the details, omitting overlapping content.

[0168] Referring to Figures 1, 2 and 8, first, a satellite image set (SI_set) can be collected (S100).

[0169] In some examples, the data collection module (210) can receive a satellite image set (SI_set) from an external database (100).

[0170] The satellite image set (SI_set) may refer to a set of images taken from a satellite located at a predetermined height above a specific coordinate region. In other words, the satellite image set (SI_set) may include multiple satellite images (SI_1, SI_2) taken from a satellite above a specific coordinate region.

[0171] For example, a satellite image set (SI_set) may include a first satellite image (SI_1) taken from a first viewpoint for a specific coordinate region and a second satellite image (SI_2) taken from a second viewpoint for the coordinate region. In this case, the first viewpoint and the second viewpoint may be different from each other, and for example, the second viewpoint may be a viewpoint located later in time than the first viewpoint.

[0172] In this case, the first satellite image (SI_1) and the second satellite image (SI_2) may be satellite images of the same coordinate region. In other words, the first satellite image (SI_1) and the second satellite image (SI_2) may be satellite images taken of the same coordinate region.

[0173] A pre-processed image set (PI_set) can then be generated (S200).

[0174] In some examples, the preprocessing module (220) can preprocess the set of satellite images (SI_set) to generate a set of preprocessed images (PI_set), i.e., the preprocessing module (220) can preprocess a first satellite image (SI_1) and a second satellite image (SI_2) to generate a first preprocessed image (PI_1) and a second preprocessed image (PI_2), respectively.

[0175] In some examples, the preprocessing module (220) may generate the preprocessed image set (PI_set) by matching attribute information (distribution) between each satellite image included in the satellite image set (SI_set). In other words, the preprocessing module (220) may generate the first preprocessed image (PI_1) and the second preprocessed image (PI_2) by matching attribute information between the first satellite image (SI_1) and the second satellite image (SI_2). That is, the attribute information between the first preprocessed image (PI_1) and the second preprocessed image (PI_2) generated by the preprocessing module (220) may be identical or similar within a predetermined critical range.

[0176] The attribute information may include illumination information, color information, etc. In other words, the pre-processing module 220 may perform pre-processing operations to match illumination information, color information, etc. between the first satellite image SI_1 and the second satellite image SI_2.

[0177] The illumination information may include information regarding the amount of light and / or intensity of light reaching a particular coordinate region determined within the satellite image. In other words, the illumination information may include information regarding the amount and intensity of light reaching a particular coordinate region of the satellite image.

[0178] The color information may include information about the color characteristics of each pixel of the satellite image, in other words, the color information may include information about the channels (Red, Green, Blue, RGB) that each pixel of the satellite image contains.

[0179] At this time, the pre-processing module (220) can use the reference image to match the attribute information of each satellite image.

[0180] For example, the pre-processing module 220 can determine reference attribute information from the reference image and match the attribute information of each satellite image according to the determined reference attribute information. In other words, the pre-processing module 220 can process the attribute information of each satellite image so that it matches the reference attribute information.

[0181] In this case, the pre-processing module 220 can change the attribute information of each satellite image using a pre-defined image processing algorithm. In other words, the pre-processing module 220 can use a pre-defined image processing algorithm when processing the attribute information of each satellite image so that it matches the reference attribute information.

[0182] The image processing algorithms used by the pre-processing module 220 may include image histogram matching, which matches the brightness of each satellite image, and / or feature distribution matching, which compares and matches the distribution of feature points in the images. However, the present invention is not limited to this, and the pre-processing module 220 may use other algorithms such as color transfer, gamma correction, and white balance adjustment.

[0183] Change information (DI) between the multiple satellite images can then be generated (S300).

[0184] In some examples, the difference determination module 230 may generate change information (DI) between each satellite image (SI_1, SI_2) in the set of satellite images (SI_set) by comparing each preprocessed image (PI_1, PI_2) included in the set of preprocessed images (PI_set). In other words, the change detection device 200 may determine change information (DI) between a first satellite image (SI_1) taken at a first time point and a second satellite image (SI_2) taken at a second time point by comparing the preprocessed images (PI_1, PI_2).

[0185] In this case, the change information (DI) may include information about structures (for example, buildings) that have been newly constructed or disappeared at the second time point compared to the first time point.

[0186] In some examples, the difference determination module (230) may use deep learning techniques to generate the change information (DI), i.e., the difference determination module (230) may use deep learning techniques to compare the first preprocessed image (PI_1) and the second preprocessed image (PI_2) included in the preprocessed image set (PI_set), and generate the change information (DI) based on the comparison result.

[0187] For example, the difference determination module 230 may generate change information (DI) based on a predefined change detection algorithm. For example, the difference determination module 230 may train an adaptive change detection algorithm (CDA_adj) in which an attention module is added to the change detection algorithm, and generate change information using the trained adaptive change detection algorithm (CDA_adj).

[0188] In this case, the adaptive change detection algorithm (CDA_adj) may include a first attention module which is an attention module related to a channel exchange module (Channel Exchange Module) that exchanges any of the RGB information between the preprocessed images (PI_1, PI_2) included in the preprocessed image set (PI_set), and / or a second attention module which is an attention module related to a spatial exchange module (Spatial Exchange Module) that exchanges pixels at specific coordinates between the preprocessed images (PI_1, PI_2) included in the preprocessed image set (PI_set).

[0189] FIG. 9 is a diagram illustrating a hardware implementation of a change detection device that executes a change detection method according to some embodiments of the present invention.

[0190] 9, a change detection device 200 for performing a change detection method according to some embodiments of the present invention may be implemented as an electronic device 1000. The electronic device 1000 may include a controller 1010, an input / output device 1020, a memory device 1030, an interface 1040, and a bus 1050. The controller 1010, the input / output device 1020, the memory device 1030, and / or the interface 1040 may be coupled to each other via a bus 1050. In this case, the bus 1050 corresponds to a path through which data travels.

[0191] Specifically, the controller (1010) may include at least one of a CPU (Central Processing Unit), an MPU (Micro Processor Unit), an MCU (Micro Controller Unit), a GPU (Graphic Processing Unit), a microprocessor, digital signal processing, a microcontroller, an application processor (AP), and logic elements capable of performing similar functions.

[0192] The input / output device (1020) may include at least one of a keypad, a keyboard, a touch screen, and a display device.

[0193] The memory device (1030) can store data and / or programs.

[0194] The interface 1040 may transmit data to or receive data from a communication network. The interface 1040 may be wired or wireless. For example, the interface 1040 may include an antenna or a wired / wireless transceiver. Although not shown, the memory device 1030 is an operating memory for improving the operation of the controller 1010 and may further include high-speed DRAM and / or SRAM. The memory device 1030 may store programs or applications therein.

[0195] The change detection device 200 and the external database 100 according to an embodiment of the present invention may each be a system formed by connecting a plurality of electronic devices 1000 to each other via a network. In this case, each module or a combination of modules may be embodied as the electronic device 1000. However, the present embodiment is not limited thereto.

[0196] Furthermore, the change detection device 200 may be embodied in at least one of a workstation, a data center, an internet data center (IDC), a DAS (direct attached storage) system, a SAN (storage area network) system, a NAS (network attached storage) system, a RAID (redundant array of inexpensive disks, or redundant array of independent disks) system, and an EDMS (Electronic Document Management) system, but this embodiment is not limited thereto.

[0197] The change detection device 200 can also transmit data to the external database 100 via a network. The network can include a network using wired internet technology, wireless internet technology, and short-range communication technology. The wired internet technology can include, for example, at least one of a local area network (LAN) and a wide area network (WAN).

[0198] The wireless internet technology may include at least one of, for example, Wireless LAN (WLAN), Digital Living Network Alliance (DMNA), Wireless Broadband (Wibro), World Interoperability for Microwave Access (Wimax), High Speed ​​Downlink Packet Access (HSDPA), High Speed ​​Uplink Packet Access (HSUPA), IEEE 802.16, Long Term Evolution (LTE), Long Term Evolution-Advanced (LTE-A), Wireless Mobile Broadband Service (WMBS), and 5G New Radio (NR) technologies, but is not limited thereto.

[0199] The short-range communication technology may include at least one of, for example, Bluetooth (registered trademark), RFID (Radio Frequency Identification), Infrared Data Association (IrDA), UWB (Ultra-Wideband), ZigBee (registered trademark), Near Field Communication (NFC), Ultra Sound Communication (USC), Visible Light Communication (VLC), Wi-Fi, Wi-Fi Direct, and 5G New Radio (NR). However, the present embodiment is not limited thereto.

[0200] The change detection device 200 communicating via a network may comply with technical standards and standard communication methods for mobile communication. For example, the standard communication methods may include at least one of Global System for Mobile communication (GSM), Code Division Multi Access (CDMA), Code Division Multi Access 2000 (CDMA2000), Enhanced Voice-Data Optimized or Enhanced Voice-Data Only (EV-DO), Wideband CDMA (WCDMA), High Speed ​​Downlink Packet Access (HSDPA), High Speed ​​Uplink Packet Access (HSUPA), Long Term Evolution (LTE), Long Term Evolution-Advanced (LTEA), and 5G New Radio (NR). However, this embodiment is not limited thereto.

[0201] The above description merely exemplifies the technical concept of the present embodiment, and various modifications and variations may be made by a person skilled in the art without departing from the essential characteristics of the present embodiment. Therefore, the present embodiment is intended to illustrate, not limit, the technical concept of the present embodiment, and the scope of the technical concept of the present embodiment is not limited by such an embodiment. The scope of protection of the present embodiment should be interpreted by the following claims, and all technical concepts within the scope equivalent thereto should be interpreted as being included in the scope of the present embodiment.

Claims

1. A data collection module for collecting a satellite image set including multiple satellite images taken at different times for the same coordinate region; a preprocessing module that matches attribute information of each satellite image included in the set of satellite images to generate a set of preprocessed images; and a difference determination module that determines differences between preprocessed images included in the set of preprocessed images to generate change information between the plurality of satellite images; The attribute information includes at least one of illumination information and color information of the satellite image. Change detection device.

2. The satellite image set includes a first satellite image taken at a first time point and a second satellite image taken at a second time point different from the first time point, The first satellite image and the second satellite image are satellite images captured in the same coordinate area. The change detection device according to claim 1 .

3. The pre-processing module includes: generating the pre-processed image set based on a pre-defined reference image; The change detection device according to claim 1 .

4. The pre-processing module includes: extracting the attribute information of the reference image and determining it as reference attribute information; changing attribute information of each satellite image included in the set of satellite images in accordance with the determined reference attribute information; The change detection device according to claim 3 .

5. The pre-processing module includes: modifying attribute information of each satellite image included in the set of satellite images using a predefined image processing algorithm; The change detection device according to claim 4 .

6. The image processing algorithm at least one of Image Histogram Matching and Feature Distribution Matching; The change detection device according to claim 5 .

7. The difference determination module: generating the change information using deep learning technology; The change detection device according to claim 1 .

8. The difference determination module: generating the change information based on a predefined change detection algorithm relating to change detection between a plurality of images; The change detection device according to claim 7 .

9. Further comprising a learning module that learns an adaptive change detection algorithm in which an attention module is added to the change detection algorithm, The difference determination module: generating the change information utilizing the adaptive change detection algorithm; The change detection device according to claim 8 .

10. The adaptive change detection algorithm comprises: a first attention module, which is an attention module related to a channel exchange module that exchanges at least one of RGB information between the preprocessed images of the preprocessed image set; a second attention module which is an attention module related to a spatial exchange module that exchanges pixels at at least one specific coordinate between the preprocessed images of the preprocessed image set; The change detection device according to claim 9 .

Images