Building change online detection method based on deep learning

Abstract

The invention discloses a building change online detection method based on deep learning, and relates to the technical field of data processing, target building information is determined, dynamic data are collected corresponding to a target building, and the dynamic data comprise building dynamic images and corresponding collection time; and setting an analysis platform, setting a mirror image layer corresponding to the target building in the analysis platform, and transmitting the dynamic data to the corresponding mirror image layer in the analysis platform for storage. And analyzing the dynamic data based on deep learning in the mirror image layer to obtain feature information, and comparing the feature information of the target building according to the acquisition time to obtain change information. According to the method, the dynamic data can be adaptively hidden and combined according to the type of the access network, the data storage security is ensured while normal extraction and use of the data are ensured, and a good data protection effect is achieved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and more specifically to an online method for detecting building changes based on deep learning. Background Technology

[0002] In recent years, artificial intelligence technology has developed rapidly, and deep learning-based building change detection methods have been widely applied in many fields due to their efficiency and accuracy. By rapidly analyzing changes in building structure and location, this technology provides crucial information for decision-making in industries such as urban planning and disaster monitoring. Regular building inspections ensure the normal use of buildings, and the security of building data storage directly affects subsequent maintenance and analysis decisions.

[0003] Currently, common data protection methods have many limitations in practical applications. While data encryption technology can effectively ensure data security during transmission, in scenarios such as building change detection where data requires frequent access and real-time analysis, repeated encryption and decryption operations severely reduce system efficiency and fail to meet the real-time requirements of online detection. Furthermore, traditional data storage methods often preserve data in its entirety; if accessed by an unauthorized network, all building information can be easily obtained, leading to a persistently high risk of data leakage. Summary of the Invention

[0004] The purpose of this invention is to provide an online building change detection method based on deep learning to address the shortcomings of the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an online building change detection method based on deep learning, comprising the following steps: Determine the target building information and collect dynamic data for the corresponding target building. The dynamic data includes dynamic images of the building and the corresponding collection time. Set up an analysis platform, set up a mirror layer for the target building in the analysis platform, and transfer dynamic data to the corresponding mirror layer in the analysis platform for storage. In the mirror layer, deep learning is used to analyze dynamic data to obtain feature information, and the feature information of the target building is compared with the collection time to obtain change information.

[0006] In a preferred embodiment, the step of determining the target building and collecting dynamic data corresponding to the target building includes: The buildings that need to be monitored for changes are identified as target buildings. The geographic information of the target buildings is data-bound to the buildings, which includes the geographic coordinates of the buildings and the ground area they occupy. Set a collection period, and collect images of the target building from multiple angles according to the collection period. The dynamic images of the building and the corresponding collection time are used as dynamic data.

[0007] In a preferred embodiment, the step of setting up an analysis platform, establishing a mirror layer corresponding to the target building within the analysis platform, and transmitting dynamic data to the corresponding mirror layer in the analysis platform for storage includes: Set up an analysis platform, and set up a first mirror plane and a second mirror plane in the analysis platform. Set up a transport plane between the first mirror plane and the second mirror plane. The same number of mirror blocks are set for the corresponding transport surface, the first mirror surface and the second mirror surface. The mirror blocks in the first mirror surface are matched one-to-one with the mirror blocks in the second mirror surface and connected. The mirror blocks in the transport surface are connected one-to-one with the corresponding mirror blocks in the first mirror surface and the second mirror surface. The first mirror surface, the second mirror surface, and the transport surface are used as mirror layers; The corresponding mirror layer is activated for the target building, and the dynamic data is transmitted to the corresponding mirror layer in the analysis platform for storage.

[0008] In a preferred embodiment, the step of connecting the mirror blocks in the transport surface to the corresponding mirror blocks in the first and second mirror surfaces includes: The same number of ports are set for both the first and second mirror surfaces. The same number of mirror blocks are set for the transport surface, the first mirror surface, and the second mirror surface. The number of mirror blocks in the first and second mirror surfaces is the same as the number of corresponding ports. In the first mirror image, the mirror blocks are mapped one-to-one with the ports and connected. In the second mirror plane, the mirror blocks are mapped one-to-one with the ports and connected. Connect the ports in the first and second mirror planes one by one to obtain the port connection relationship. Then connect the corresponding mirror blocks in the first and second mirror planes according to the port connection relationship. Based on the port connection relationship, the mirror blocks in the transport surface are connected to the corresponding mirror blocks in the first mirror surface and the second mirror surface respectively; Configure corresponding replacement blocks for each of the multiple mirror blocks in the first and second mirror planes.

[0009] In a preferred embodiment, the step of activating the corresponding mirror layer for the target building and transmitting dynamic data to the corresponding mirror layer in the analysis platform for storage includes: Multiple mirror layers are set up in the analysis platform, and mirror layers are assigned to the target building; The dynamic data is transmitted to the analysis platform, and the dynamic data is divided into two data clusters. The two data clusters are divided into multiple data segments, wherein the number of data segments in a single data cluster is the same as the number of mirror blocks in the first or second mirror plane; Multiple data segments from the two data clusters are stored in mirror blocks of the first and second mirror planes, respectively. When unauthorized network access to the mirror layer occurs, the mirror block is disconnected from the replacement block, and the replacement block takes over the connection to the corresponding port. When accessing the mirror layer through an authorized network, dynamic data is obtained based on the mirror block.

[0010] In a preferred embodiment, the step of disconnecting the mirror block from the replacement block when unauthorized network access to the mirror layer exists, connecting the replacement block to the corresponding port in place of the mirror block, and obtaining dynamic data based on the mirror block when accessing the mirror layer through an authorized network includes: An access zone is set up between the authorized network and the mirror layer. The access zone includes multiple pairs of associated terminals. The number of associated terminals is the same as the number of mirror blocks in the first mirror or the second mirror. Each pair of associated terminals is connected to a mirror block in the first mirror or the second mirror that has a connection relationship. A first verification code is set for the corresponding access zone. The corresponding image layer is set with an authorized network, and the corresponding authorized network is set with a second verification code. The first verification code and the second verification code are matched with each other. After matching, they are used to enable multiple pairs of associated terminals in the access area. When an authorized network accesses the mirror layer, the authorized network and the access area are verified by the second verification code. Paired associated terminals are opened and connected to the mirror blocks in the first and second mirror surfaces that have a connection relationship. The mirror blocks in the first and second mirror surfaces are transferred to the mirror blocks corresponding to the transport surface for combination to obtain dynamic data. The authorized network continues to connect with the transport surface to acquire dynamic data; When there is unauthorized network access to the mirror layer, the associated end in the access area cannot be enabled. When directly accessing the port in the first mirror or / and second mirror, the mirror block corresponding to the accessed port is disconnected from the port. The mirror block is replaced by connecting the replacement block to the port and the connection between the replacement block and the corresponding mirror block is disconnected.

[0011] In a preferred embodiment, the step of analyzing dynamic data based on deep learning in the mirror layer to obtain feature information, and comparing the feature information of the target building with the data acquisition time to obtain change information, includes: The convolutional neural network is trained in the mirror layer based on historical data to obtain a trained deep learning model; Dynamic data from the mirror layer is input into a deep learning model to obtain change information.

[0012] In a preferred embodiment, the step of training the convolutional neural network in the mirror layer based on historical data to obtain a trained deep learning model includes: Historical data is set into paired images according to the continuity of the collection time; The paired images and their corresponding change information are used as the training set and input into the convolutional neural network for training to obtain a deep learning model.

[0013] The technical effects and advantages provided by the present invention in the above technical solution are as follows: This invention enables the storage of dynamic data in blocks through a mirror layer. Each block of data is incomplete, making it difficult to understand the complete data even if it is obtained, resulting in low reference value. Block storage and protection of data can ensure the security of data storage. In addition, dynamic data can be adaptively hidden and combined according to the type of network access, ensuring both normal data retrieval and use while also ensuring the security of data storage, thus providing good data protection. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0015] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Example 1, please refer to Figure 1 As shown in this embodiment, a deep learning-based online building change detection method includes the following steps: S1. Determine the target building information and collect dynamic data for the target building. The dynamic data includes dynamic images of the building and the corresponding collection time. S2. Set up the analysis platform, set up the mirror layer of the corresponding target building in the analysis platform, and transfer the dynamic data to the corresponding mirror layer in the analysis platform for storage. S3. In the mirror layer, dynamic data is analyzed based on deep learning to obtain feature information, and the feature information of the target building is compared with the collection time to obtain change information.

[0018] In one embodiment, step S1, which involves determining the target building and collecting dynamic data corresponding to the target building, includes: S11. Identify the buildings that need to be monitored for changes as target buildings, and bind the geographic information of the target buildings to the building data as target building information. The geographic information includes the geographic coordinates of the building and the ground area it occupies. S12. Set the acquisition period, and acquire images of the target building from multiple angles according to the acquisition period. Use the dynamic images of the building and the corresponding acquisition time as dynamic data.

[0019] As described in steps S11 and S12 above, during the detection of abnormal changes in buildings (target buildings) in sensitive areas such as radar stations, images or videos of the target buildings can be collected using satellite or UAV remote sensing imagery. It is necessary to determine the geographical coordinates of the target building, which can be the geographical coordinates of the corner of the target building's landing point. At the same time, it is also necessary to determine the ground area occupied by the target building. This can serve as a reference for subsequent image or video collection of the target building. Based on the target building information, the corresponding building can be located for image or video collection. A collection period is set, and images of the target building are collected from multiple angles according to the collection period to form dynamic images of the building (which can be images or videos). The dynamic images of the building and the corresponding collection time are used as dynamic data, which can facilitate subsequent monitoring of changes in the building.

[0020] In one embodiment, step S2, which involves setting up an analysis platform, establishing a mirror layer corresponding to the target building within the analysis platform, and transmitting dynamic data to the corresponding mirror layer in the analysis platform for storage, includes: S21. Set up an analysis platform, set up a first mirror plane and a second mirror plane in the analysis platform, and set up a transport plane between the first mirror plane and the second mirror plane; S22. Set the same number of mirror blocks (virtual machines) for the corresponding transport surface, the first mirror surface and the second mirror surface (data space), and connect the mirror blocks in the first mirror surface to the mirror blocks in the second mirror surface one by one. Connect the mirror blocks in the transport surface one by one to the corresponding connected mirror blocks in the first mirror surface and the second mirror surface. S23. The first mirror surface, the second mirror surface, and the transport surface are used as mirror layers; S24. Activate the corresponding mirror layer for the target building and transmit the dynamic data to the corresponding mirror layer in the analysis platform for storage.

[0021] In one embodiment, step S22, which connects the mirror blocks in the transport surface one by one to the corresponding mirror blocks in the first mirror surface and the second mirror surface, includes: S221. The same number of ports are set for both the first mirror surface and the second mirror surface. The same number of mirror blocks are set for both the transport surface, the first mirror surface and the second mirror surface. The number of mirror blocks in the first mirror surface and the second mirror surface is the same as the number of corresponding ports. S222. In the first mirror plane, the mirror blocks and ports are matched one by one and connected. S223. In the second mirror plane, the mirror blocks and ports are matched one by one and connected. S224. Connect the ports in the first mirror and the second mirror one by one to obtain the port connection relationship (the port connection relationship here is used to verify whether the accessed port is the corresponding port when the external network accesses the first mirror and the second mirror through the associated terminal of the access area and the mutual interaction between the ports in the first mirror and the second mirror). Connect the corresponding mirror blocks in the first mirror and the second mirror according to the port connection relationship. S225. Based on the port connection relationship, connect the mirror blocks in the transport surface to the corresponding mirror blocks in the first mirror surface and the second mirror surface respectively; S226. Configure corresponding replacement blocks for each of the multiple mirror blocks in the first and second mirror planes.

[0022] In one embodiment, step S24, where the corresponding target building activates a corresponding mirror layer and transmits dynamic data to the corresponding mirror layer in the analysis platform for storage, includes: S241. Set up multiple mirror layers in the analysis platform and assign mirror layers to the target building; S242. Transmit the dynamic data to the analysis platform, divide the dynamic data, and obtain two data clusters (equivalent to dividing the building dynamic image into the upper and lower or left and right parts of the image to obtain two data information, and combining the corresponding acquisition time with the two data information to obtain two data clusters). S243. Divide the two data clusters into multiple data segments (divide them according to the acquisition time, that is, the duration of a single acquisition, to obtain multiple data segments. For example, the duration of a single acquisition is from the start of acquiring the building dynamic image to the end of the acquisition. Divide the data cluster within this duration into a preset time period to obtain multiple data segments. For example, if the duration is one hour and the preset time period is 15 minutes, then the data cluster is divided into four parts, that is, four data segments are obtained). The number of data segments in a single data cluster is the same as the number of mirror blocks in the first mirror surface or the second mirror surface. S244. Store multiple data segments of the two data clusters in the mirror blocks of the first and second mirror planes, respectively. S245. When there is unauthorized network access to the mirror layer, disconnect the mirror block from the replacement block, and connect the mirror block to the corresponding port through the replacement block. When accessing the mirror layer through an authorized network, obtain dynamic data based on the mirror block.

[0023] In one embodiment, step S245, which involves disconnecting the mirror block from the replacement block when unauthorized network access to the mirror layer occurs, connecting the replacement block to the corresponding port in place of the mirror block, and obtaining dynamic data based on the mirror block when accessing the mirror layer via an authorized network, includes: S2451. An access area is set up between the authorized network and the mirror layer. The access area includes multiple pairs of associated terminals (each pair of associated terminals is connected to each other). The number of associated terminals is the same as the number of mirror blocks in the first mirror surface or the second mirror surface. Each pair of associated terminals is connected to mirror blocks in the first mirror surface and the second mirror surface that have a connection relationship. A first verification code is set for the corresponding access area. S2452. Set up an authorized network for the corresponding mirror layer, set up a second verification code for the corresponding authorized network, and match the first verification code and the second verification code with each other. After matching, the matching is used to enable multiple pairs of associated terminals in the access area. S2453. When the authorized network accesses the mirror layer, the authorized network and the access area are verified by the second verification code. The paired associated terminals are connected to the mirror blocks in the first and second mirror surfaces that have a connection relationship (in this way, the corresponding mirror blocks in the first and second mirror surfaces are connected to the paired associated terminals, and the ports corresponding to the mirror blocks in the first and second mirror surfaces that have a connection relationship are also interconnected, thus forming a closed loop channel. This means that the mirror blocks are accessed simultaneously and correspondingly, which means that the access is performed by the authorized network). The mirror blocks in the first and second mirror surfaces are transferred (transmitted) to the mirror blocks corresponding to the transport surface for combination (the combination here is the reverse operation of dividing dynamic data into data clusters and then dividing them into multiple data segments respectively) to obtain dynamic data. S2454. The authorized network continues to connect with the transport surface to obtain dynamic data; S2455. When there is unauthorized network access to the mirror layer, the associated end in the access area cannot be enabled. When directly accessing the port in the first mirror surface or / and the second mirror surface, the mirror block corresponding to the accessed port is disconnected from the port. The mirror block is replaced by connecting the replacement block to the port and the connection between the replacement block and the corresponding mirror block is disconnected. (Dynamic data can only be obtained through the transport surface. The ports on the first and second mirror surfaces are used for triggering and cannot be used for data transmission. The mirror block can learn about the corresponding situation of external network access through the port. The corresponding situation includes not triggering the mirror block simultaneously and not through a pair of associated ends.)

[0024] As described in steps S21-S24 above, an analysis platform is set up. This analysis platform is also an IoT platform, capable of connecting with drones or other devices corresponding to the collected dynamic data of the building, and receiving data from drones and other devices. Since the dynamic data of some sensitive areas, such as radar stations, needs to be kept confidential, it is necessary to ensure the security of the dynamic data storage on the analysis platform. First, multiple mirror layers are set up in the analysis platform. Each mirror layer is used to store the dynamic data of a target building. Taking the technical construction of a single mirror layer as an example, the transport surface, the first mirror surface, and the second mirror surface are first determined in the storage space of the analysis platform. These three are all data spaces with a certain amount of data storage capacity (sufficient to support the dynamic data collected from the target building in all cycles). Mirror blocks with the same data are set up in the transport surface, the first mirror surface, and the second mirror surface. This image block is a virtual machine, and the image block is stored after the three mirror blocks to store subsequent dynamic data. Each image block in the transport plane, the first image plane, and the second image plane is assigned a corresponding image block. For example, there are three image blocks in the transport plane: b1, b2, and b3; the first image plane has image blocks: y1, y2, and y3; and the second image plane has image blocks: e1, e2, and e3. Then, b1 and y1 are matched with e1; b2 and y2 are matched with e2; and b3 and y3 are matched with e3. Communication connections are established between the corresponding ports of the image blocks in the first image plane and the image blocks in the second image plane, and communication connections are established between the image blocks in the first image plane and the image blocks in the second image plane and their corresponding ports. (Dynamic data cannot be transmitted; this is only used for communication triggering to facilitate subsequent prompts for the image blocks, which are then transferred to the transport plane for combination to obtain dynamic data.) Connect b1 to y1 and e1 respectively; connect b2 to y2 and e2 respectively; connect b3 to y3 and e3 respectively. These communication connections are used for the subsequent transfer and assembly of mirror blocks from the first and second mirror surfaces to the transport surface. Multiple ports in the first and second mirror surfaces are connected one-to-one with their corresponding mirror blocks. An access zone is established between the mirror layer and the external network. This access zone can be bypassed, but the authorized network cannot. This access zone is a data space within the mirror layer, where multiple pairs of associated terminals are set up. These associated terminals are communication connection ports, and each pair is interconnected. For example, if the number of associated terminals is g1 and G1, g1 and G1, and g1 and G1, then the port corresponding to g1 and y1 is connected, the port corresponding to G1 and e1 is connected, and so on. A first verification code is set for the corresponding access zone, and a second verification code is set for the corresponding authorized network. The first and second verification codes are matched for verification. After establishing the above technical architecture, the specific operations are as follows: When an authorized network accesses the mirror layer, it first verifies and matches the second verification code with the first verification code in the access area. After verification and matching, multiple pairs of associated terminals in the access area are activated. These terminals then trigger ports in the corresponding first and second mirror layers according to pre-set connection relationships, forming a closed loop. This indicates that the mirror blocks are accessed simultaneously and correspondingly, signifying access by the authorized network. The mirror blocks are then transferred from the corresponding first and second mirror layers to the corresponding mirror blocks in the transport layer. The data is then restored to dynamic data and stored in the mirror blocks in the transport layer through a reverse operation of dividing the dynamic data into data clusters and then further dividing them into multiple data segments. The authorized network then leaves the access area to access the transport layer and obtains the dynamic data from the transport layer for the administrator (authorized network) to understand.

[0025] When an unauthorized network accesses a mirror layer, it may or may not cross the access zone. If it crosses the access zone, it will acquire a single port from either the first or second mirror layer. This is a standalone access port, not a paired port corresponding to the first and second mirror layers, thus indicating an unauthorized network. If it doesn't cross the access zone, it cannot establish an authentication relationship with the access zone. Therefore, it will still acquire a single port from either the first or second mirror layer, again a standalone access port, and again indicating an unauthorized network. When an unauthorized network is identified, the mirror block corresponding to the accessed port is disconnected from the port. A replacement block is then connected to the port to replace the mirror block, and the connection between the replacement block and the corresponding mirror block is broken. This effectively hides the mirror block. The system is configured so that the connection between the mirror block, replacement block, and corresponding port is not used to transmit dynamic data, but rather to indicate and trigger the status of the external network (whether it is an authorized or unauthorized network). However, given the existence of external network access, there is still a risk of data breach and interception. Therefore, the mirror block is hidden here. The mirror layer allows for the storage of dynamic data in blocks, and each block is incomplete, making it difficult to obtain complete data even if it is acquired, resulting in low reliability. Block storage and protection ensure data storage security. Furthermore, the system can adaptively hide and combine dynamic data based on the type of network access, ensuring both normal data retrieval and storage security, thus providing good data protection. The security status during the data protection process is represented by a security index, specifically: ,in, For safety index, The total number of replacement blocks, To replace the frequency, This represents the cumulative number of times the mirror block has been replaced. and It is a constant greater than zero. It should be noted that the smaller the security index, the lower the security of the mirror layer.

[0026] In one embodiment, step S3, which involves analyzing dynamic data based on deep learning in the mirror layer to obtain feature information and comparing the feature information of the target building with the feature information based on the acquisition time to obtain change information, includes: S31. Train the convolutional neural network in the mirror layer based on historical data to obtain a trained deep learning model. S32. Use the dynamic data in the mirror layer to input into the deep learning model to obtain change information; In one embodiment, step S32, which trains the convolutional neural network in the mirror layer based on historical data to obtain a trained deep learning model, includes: S321. Set historical data into paired images according to the continuity of the acquisition time; S322. Input the paired images and their corresponding change information as the training set into the convolutional neural network for training to obtain a deep learning model.

[0027] As described in steps S31 and S32 above, historical baseline images (such as dynamic data from the past quarter as historical data) are acquired and input into a convolutional neural network in pairs. Here, the pairs of images represent dynamic data collected in the previous time period and dynamic data collected in the next consecutive time period. Features are extracted from each pair. A feature difference map is calculated, and a change probability map is output after the dimensions are compressed by the convolutional layer. The change probability map is compared, and the network parameters are updated by backpropagation. Iterative training is performed to obtain a deep learning model. Then, the dynamic data in the mirror layer is input into the deep learning model to obtain change information.

[0028] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A deep learning-based online building change detection method, characterized in that, Includes the following steps: Determine the target building information and collect dynamic data for the corresponding target building. The dynamic data includes dynamic images of the building and the corresponding collection time. Set up an analysis platform, set up a mirror layer for the target building in the analysis platform, and transfer dynamic data to the corresponding mirror layer in the analysis platform for storage. In the mirror layer, deep learning is used to analyze dynamic data to obtain feature information, and the feature information of the target building is compared with the collection time to obtain change information.

2. The online building change detection method based on deep learning according to claim 1, characterized in that: The step of determining the target building and collecting dynamic data corresponding to the target building includes: The buildings that need to be monitored for changes are identified as target buildings. The geographic information of the target buildings is data-bound to the buildings, which includes the geographic coordinates of the buildings and the ground area they occupy. Set a collection period, and collect images of the target building from multiple angles according to the collection period. The dynamic images of the building and the corresponding collection time are used as dynamic data.

3. The online building change detection method based on deep learning according to claim 1, characterized in that: The step of setting up an analysis platform, configuring a mirror layer for the target building within the analysis platform, and transmitting dynamic data to the corresponding mirror layer for storage includes: Set up an analysis platform, and set up a first mirror plane and a second mirror plane in the analysis platform. Set up a transport plane between the first mirror plane and the second mirror plane. The same number of mirror blocks are set for the corresponding transport surface, the first mirror surface and the second mirror surface. The mirror blocks in the first mirror surface are matched one-to-one with the mirror blocks in the second mirror surface and connected. The mirror blocks in the transport surface are connected one-to-one with the corresponding mirror blocks in the first mirror surface and the second mirror surface. The first mirror surface, the second mirror surface, and the transport surface are used as mirror layers; The corresponding mirror layer is activated for the target building, and the dynamic data is transmitted to the corresponding mirror layer in the analysis platform for storage.

4. The online building change detection method based on deep learning according to claim 3, characterized in that: The step of connecting the mirror blocks in the transport surface to the corresponding mirror blocks in the first and second mirror surfaces one by one includes: The same number of ports are set for both the first and second mirror surfaces. The same number of mirror blocks are set for the transport surface, the first mirror surface, and the second mirror surface. The number of mirror blocks in the first and second mirror surfaces is the same as the number of corresponding ports. In the first mirror image, the mirror blocks are mapped one-to-one with the ports and connected. In the second mirror plane, the mirror blocks are mapped one-to-one with the ports and connected. Connect the ports in the first and second mirror planes one by one to obtain the port connection relationship. Then connect the corresponding mirror blocks in the first and second mirror planes according to the port connection relationship. Based on the port connection relationship, the mirror blocks in the transport surface are connected to the corresponding mirror blocks in the first mirror surface and the second mirror surface respectively; Configure corresponding replacement blocks for each of the multiple mirror blocks in the first and second mirror planes.

5. The online building change detection method based on deep learning according to claim 4, characterized in that: The step of activating the corresponding mirror layer for the target building and transmitting dynamic data to the corresponding mirror layer in the analysis platform for storage includes: Multiple mirror layers are set up in the analysis platform, and mirror layers are assigned to the target building; The dynamic data is transmitted to the analysis platform, and the dynamic data is divided into two data clusters. The two data clusters are divided into multiple data segments, wherein the number of data segments in a single data cluster is the same as the number of mirror blocks in the first or second mirror plane; Multiple data segments from the two data clusters are stored in mirror blocks of the first and second mirror planes, respectively. When unauthorized network access to the mirror layer occurs, the mirror block is disconnected from the replacement block, and the replacement block takes over the connection to the corresponding port. When accessing the mirror layer through an authorized network, dynamic data is obtained based on the mirror block.

6. The online building change detection method based on deep learning according to claim 5, characterized in that: The steps of disconnecting the mirror block from the replacement block when unauthorized network access to the mirror layer occurs, connecting the replacement block to the corresponding port in place of the mirror block, and obtaining dynamic data based on the mirror block when accessing the mirror layer through an authorized network include: An access zone is set up between the authorized network and the mirror layer. The access zone includes multiple pairs of associated terminals. The number of associated terminals is the same as the number of mirror blocks in the first mirror or the second mirror. Each pair of associated terminals is connected to a mirror block in the first mirror or the second mirror that has a connection relationship. A first verification code is set for the corresponding access zone. The corresponding image layer is set with an authorized network, and the corresponding authorized network is set with a second verification code. The first verification code and the second verification code are matched with each other. After matching, they are used to enable multiple pairs of associated terminals in the access area. When an authorized network accesses the mirror layer, the authorized network and the access area are verified by the second verification code. Paired associated terminals are opened and connected to the mirror blocks in the first and second mirror surfaces that have a connection relationship. The mirror blocks in the first and second mirror surfaces are transferred to the mirror blocks corresponding to the transport surface for combination to obtain dynamic data. The authorized network continues to connect with the transport surface to acquire dynamic data; When there is unauthorized network access to the mirror layer, the associated end in the access area cannot be enabled. When directly accessing the port in the first mirror or / and second mirror, the mirror block corresponding to the accessed port is disconnected from the port. The mirror block is replaced by connecting the replacement block to the port and the connection between the replacement block and the corresponding mirror block is disconnected.

7. The online building change detection method based on deep learning according to claim 1, characterized in that: The steps of analyzing dynamic data based on deep learning in the mirror layer to obtain feature information, and comparing the feature information of the target building with the data acquisition time to obtain change information, include: The convolutional neural network is trained in the mirror layer based on historical data to obtain a trained deep learning model; Dynamic data from the mirror layer is input into a deep learning model to obtain change information.

8. The online building change detection method based on deep learning according to claim 1, characterized in that: The step of training the convolutional neural network in the mirror layer based on historical data to obtain a trained deep learning model includes: Historical data is set into paired images according to the continuity of the collection time; The paired images and their corresponding change information are used as the training set and input into the convolutional neural network for training to obtain a deep learning model.

Images