Building construction year remote sensing inversion method based on medium and low resolution satellite images

Abstract

The invention provides a building construction year remote sensing inversion method based on a medium and low resolution satellite image, which is used for configuring a building construction year remote sensing inversion model by using the medium and low resolution satellite image based on satellite remote sensing, so that a satellite remote sensing technology and a deep learning technology are combined, and the building construction year remote sensing inversion model is constructed. According to the method, efficient, complete and accurate built-up year inversion of the building is realized, the application difficulty of built-up year inversion is reduced, and the adverse effects of limitations such as insufficient historical data, incomplete years and unpublished data on the use requirements of built-up year data can be solved; powerful data support can be provided for urban energy modeling, building safety evaluation, urban sustainable development planning, social ecology and the like.

Description

Technical Field

[0001] The present application relates to the field of geographic information, and in particular to a remote sensing inversion method for the year a building was built based on medium and low resolution satellite images. Background Art

[0002] The extraction of the year of building completion has important application value in the field of geographic information, especially in urban energy management, building safety assessment and urban planning. However, traditional databases of the year of building completion are usually not public, difficult to obtain, and have many limitations. Even if some regions have databases of the year of building completion, these data are usually obtained through manually digitized historical maps, and their time information is often not continuous enough, updated in a timely manner, or the selected year range cannot adapt to specific research or application scenarios. In addition, traditional databases are prone to data omissions and missing problems.

[0003] In contrast, using satellite remote sensing images to invert the year a building was built provides a more flexible and widely applicable solution. Although high-resolution satellite remote sensing images can provide rich spatial detail information, they often have discontinuities in the acquisition of time series data. At the same time, the lack of historical data limits their ability to be applied over long time spans. Medium and low-resolution satellite remote sensing data, due to their characteristics of long-term series acquisition, can be traced back to around the 1980s, and have great potential to cover long-term series information on the year a building was built. This gives medium and low-resolution satellite images a unique advantage in inverting the year a building was built, especially in scenarios where historical changes need to be traced back.

[0004] However, existing remote sensing methods mostly focus on identifying long-term land use changes, such as monitoring changes in natural surface types such as cultivated land and water bodies, or the expansion of artificial surfaces represented by buildings in urban expansion scenarios. Most of these studies focus on urban expansion and land cover changes, and relatively little on identifying the renewal of internal structures. For example, the demolition and renewal of buildings are often overlooked. Existing methods may be biased when estimating the year a building was built, usually making an early estimate of the year a building was built, while ignoring the actual construction or renewal of the building.

[0005] Therefore, there is an urgent need for an efficient and accurate processing method to achieve a complete and precise inversion of the year the building was built. Summary of the invention

[0006] The present application provides a deep learning method for remote sensing inversion of the year of construction of buildings based on medium and low resolution satellite images, which is used to configure a remote sensing inversion model for the year of construction of buildings using medium and low resolution satellite images based on satellite remote sensing. In this way, satellite remote sensing technology and deep learning technology are combined to achieve efficient, complete and accurate inversion of the year of construction of buildings, reducing the application difficulty of the year of construction inversion, and can solve the adverse effects of limitations such as insufficient historical data, incomplete years, and non-disclosure of data on the demand for the use of year of construction data, and can provide strong data support for urban energy modeling, building safety assessment, urban sustainable development planning, and social ecology.

[0007] In the first aspect, the present application provides a remote sensing inversion method for the year of building completion based on medium and low resolution satellite images, the method comprising:

[0008] Obtain medium and low resolution satellite images collected by satellite remote sensing for the target area, wherein the medium and low resolution satellite images are specifically medium and low resolution spectral satellite long time series data;

[0009] Inputting low- and medium-resolution satellite images into a pre-configured remote sensing inversion model of the year a building was built, so that the remote sensing inversion model of the year a building was built carries out corresponding remote sensing inversion processing of the year a building was built, wherein the remote sensing inversion model of the year a building was built is obtained by training an initial deep learning model using sample low- and medium-resolution satellite images configured with corresponding labeling results of the year a building was built;

[0010] Extract the remote sensing inversion results of the building construction year corresponding to the target area output by the remote sensing inversion model of the building construction year.

[0011] In combination with the first aspect of the present application, in a first possible implementation method of the first aspect of the present application, the building construction year marking result is the construction year data at the community granularity captured by the real estate website, integrated with the vector point data formed by the geographical location coordinates of the community.

[0012] In combination with the first aspect of the present application, in a second possible implementation method of the first aspect of the present application, the sample medium and low-resolution satellite images are specifically multi-vector medium and low-resolution satellite images within a preset year range under the band conditions of blue band, green band, red band and near-infrared band.

[0013] In combination with the second possible implementation method of the first aspect of the present application, in the third possible implementation method of the first aspect of the present application, the multi-vector medium and low-resolution satellite images are specifically data obtained by synthesizing the quarterly median, and the multi-vector medium and low-resolution satellite images have also been pre-performed with unified normalization processing.

[0014] In combination with the first aspect of the present application, in a fourth possible implementation of the first aspect of the present application, the remote sensing inversion model for the year of building completion is specifically a deep learning model combining a deep residual network as a front feature extraction unit and a gated recurrent unit as a time series feature learning unit;

[0015] The front feature extraction unit consists of multiple layers of one-dimensional convolution, layer normalization function and ReLU activation function. The image features corresponding to the input image data are extracted through multiple layers of convolution. The output of one-dimensional convolution is 64 dimensions.

[0016] The time series feature learning unit with a dimension of 64 is used to extract the spatiotemporal information of the time series dimension corresponding to the data output by the previous feature extraction unit, and is combined with a fully connected layer to complete the remote sensing inversion processing task of the year the building was built.

[0017] In combination with the first aspect of the present application, in the fifth possible implementation of the first aspect of the present application, the loss function in the model training process specifically adopts the root mean square difference.

[0018] In combination with the first aspect of the present application, in a sixth possible implementation manner of the first aspect of the present application, the method further includes:

[0019] Based on the remote sensing inversion results of the building completion year corresponding to the target area output by the building completion year remote sensing inversion model, mapping processing is performed to obtain mapping results for display;

[0020] According to the preset graphic strategy, the content of the mapping results is displayed through a visual interface.

[0021] In a second aspect, the present application provides a remote sensing inversion device for building completion year based on medium and low resolution satellite images, the device comprising:

[0022] An acquisition unit is used to acquire medium and low resolution satellite images collected by satellite remote sensing for the target area, wherein the medium and low resolution satellite images are specifically medium and low resolution spectral satellite long time series data;

[0023] An inversion unit is used to input medium and low resolution satellite images into a pre-configured remote sensing inversion model of the year when the building was built, so that the remote sensing inversion model of the year when the building was built can carry out corresponding remote sensing inversion processing of the year when the building was built, wherein the remote sensing inversion model of the year when the building was built is obtained by training an initial deep learning model with sample medium and low resolution satellite images configured with corresponding labeling results of the year when the building was built;

[0024] The extraction unit is used to extract the remote sensing inversion result of the building completion year corresponding to the target area output by the remote sensing inversion model of the building completion year.

[0025] In combination with the second aspect of the present application, in a first possible implementation method of the second aspect of the present application, the building construction year marking result is the construction year data at the community granularity captured by the real estate website, integrated with the vector point data formed by the geographical location coordinates of the community.

[0026] In combination with the second aspect of the present application, in a second possible implementation method of the second aspect of the present application, the sample medium and low-resolution satellite images are specifically multi-vector medium and low-resolution satellite images within a preset year range under the band conditions of blue band, green band, red band and near-infrared band.

[0027] In combination with the second possible implementation method of the second aspect of the present application, in the third possible implementation method of the second aspect of the present application, the multi-vector medium and low-resolution satellite images are specifically data obtained by synthesizing the quarterly median, and the multi-vector medium and low-resolution satellite images have also been pre-performed with unified normalization processing.

[0028] In conjunction with the second aspect of the present application, in a fourth possible implementation of the second aspect of the present application, the remote sensing inversion model for the year of building completion is specifically a deep learning model that combines a deep residual network as a front feature extraction unit and a gated recurrent unit as a time series feature learning unit;

[0029] The front feature extraction unit consists of multiple layers of one-dimensional convolution, layer normalization function and ReLU activation function. The image features corresponding to the input image data are extracted through multiple layers of convolution. The output of one-dimensional convolution is 64 dimensions.

[0030] The time series feature learning unit with a dimension of 64 is used to extract the spatiotemporal information of the time series dimension corresponding to the data output by the previous feature extraction unit, and is combined with a fully connected layer to complete the remote sensing inversion processing task of the year the building was built.

[0031] In combination with the second aspect of the present application, in a fifth possible implementation of the second aspect of the present application, the loss function in the model training process specifically adopts the root mean square error.

[0032] In combination with the second aspect of the present application, in a sixth possible implementation manner of the second aspect of the present application, the device further includes:

[0033] A mapping unit is used to perform mapping processing based on the remote sensing inversion result of the building completion year corresponding to the target area output by the building completion year remote sensing inversion model to obtain a mapping result for display;

[0034] The display unit is used to display the content of the mapping result through a visual interface according to a preset diagram strategy.

[0035] In a third aspect, the present application provides a processing device, including a processor and a memory, wherein a computer program is stored in the memory, and when the processor calls the computer program in the memory, the method provided in the first aspect of the present application or any possible implementation method of the first aspect of the present application is executed.

[0036] In a fourth aspect, the present application provides a computer-readable storage medium, which stores multiple instructions, and the instructions are suitable for a processor to load to execute the method provided by the first aspect of the present application or any possible implementation of the first aspect of the present application.

[0037] From the above content, it can be concluded that the present application has the following beneficial effects:

[0038] In order to invert the year of construction of buildings, this application uses low- and medium-resolution satellite images based on satellite remote sensing to configure a remote sensing inversion model for the year of construction of buildings. It can predict the year of construction of corresponding buildings based on low- and medium-resolution satellite images collected by satellite remote sensing in the target area. In this way, it combines satellite remote sensing technology and deep learning technology to achieve efficient, complete and accurate inversion of the year of construction of buildings, reducing the application difficulty of the inversion of the year of construction, and can solve the adverse effects of limitations such as insufficient historical data, incomplete years, and non-disclosure of data on the demand for the use of year of construction data. It can provide strong data support for urban energy modeling, building safety assessment, urban sustainable development planning, and social ecology. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0040] Figure 1 A schematic diagram of a flow chart of a remote sensing inversion method for the year of building completion based on medium and low resolution satellite images in this application;

[0041] Figure 2 A schematic diagram of a scenario of a remote sensing inversion method for the year of building completion based on medium and low resolution satellite images in this application;

[0042] Figure 3 A schematic diagram of a structure of a remote sensing inversion method for building completion year based on medium and low resolution satellite images in this application;

[0043] Figure 4 A schematic diagram of the structure of the processing equipment of this application. DETAILED DESCRIPTION

[0044] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0045] The terms "first", "second", etc. in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described here can be implemented in an order other than that illustrated or described here. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or modules is not necessarily limited to those steps or modules clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products or devices. The naming or numbering of steps in this application does not mean that the steps in the method flow must be executed in the time / logical sequence indicated by the naming or numbering. The process steps that have been named or numbered can change the execution order according to the technical purpose to be achieved, as long as the same or similar technical effects can be achieved.

[0046] The division of modules in this application is a logical division. There may be other division methods when it is implemented in actual applications. For example, multiple modules can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, and the indirect coupling or communication connection between modules can be electrical or other similar forms, which are not limited in this application. In addition, the modules or submodules described as separate components may or may not be physically separated, may or may not be physical modules, or may be distributed in multiple circuit modules, and some or all of the modules may be selected according to actual needs to achieve the purpose of the present application.

[0047] Before introducing the remote sensing inversion method for the year of construction of a building based on medium and low resolution satellite images provided by this application, the background content involved in this application is first introduced.

[0048] The method, device and computer-readable storage medium for remote sensing inversion of the year of construction of buildings based on medium and low resolution satellite images provided in the present application can be applied to processing equipment to configure a remote sensing inversion model for the year of construction of buildings using medium and low resolution satellite images based on satellite remote sensing. In this way, satellite remote sensing technology and deep learning technology are combined to achieve efficient, complete and accurate inversion of the year of construction of buildings, reducing the application difficulty of the inversion of the year of construction, and can solve the adverse effects of limitations such as insufficient historical data, incomplete years, and non-disclosure of data on the demand for the use of year of construction data, and can provide strong data support for urban energy modeling, building safety assessment, urban sustainable development planning, and social ecology.

[0049] The method for remote sensing inversion of the year of construction of a building based on medium and low resolution satellite images mentioned in this application can be performed by a remote sensing inversion device for the year of construction of a building based on medium and low resolution satellite images, or a server, physical host or user equipment (UE) and other different types of processing equipment that integrate the remote sensing inversion device for the year of construction of a building based on medium and low resolution satellite images. Among them, the remote sensing inversion device for the year of construction of a building based on medium and low resolution satellite images can be implemented in hardware or software, and the UE can be a terminal device such as a smart phone, tablet computer, laptop computer, desktop computer or personal digital assistant (PDA), and the processing equipment can be set in the form of a device cluster.

[0050] It is easy to understand that, considering that the focus of the present application is on data processing based on the acquired data, the processing equipment that executes the remote sensing inversion method for the year of construction of a building based on low and medium resolution satellite images provided in the present application, or the processing equipment equipped with the corresponding application service of the remote sensing inversion method for the year of construction of a building based on low and medium resolution satellite images provided in the present application, only needs to meet the data processing capability requirements, and therefore, it can be configured as different types of hardware devices such as servers, physical hosts, or even UEs according to the specific application requirements in actual situations.

[0051] If the details involve the real-time collection of data such as medium and low resolution satellite images, then data sources such as satellite remote sensing systems can also be included in the category of processing equipment. The processing equipment here can also be called a processing system, which may also happen in some cases.

[0052] Of course, considering that this application does not involve corresponding optimization and improvement on the data source side, in practical applications, obtaining ready-made data from data sources such as satellite remote sensing systems can meet the data usage requirements of this application.

[0053] Next, we will introduce the remote sensing inversion method for the year of building completion based on medium and low resolution satellite images provided by this application.

[0054] First, see Figure 1 , Figure 1 A schematic flow chart of a method for remote sensing inversion of the year of building completion based on medium and low resolution satellite images provided in the present application is shown. The method for remote sensing inversion of the year of building completion based on medium and low resolution satellite images provided in the present application may specifically include the following steps S101 to S103:

[0055] Step S101, obtaining medium and low resolution satellite images collected by satellite remote sensing for the target area, wherein the medium and low resolution satellite images are specifically medium and low resolution spectral satellite long time series data;

[0056] It can be understood that in order to achieve the goal of inverting the year a building was built, this application considers using satellite remote sensing technology to carry out remote sensing inversion processing of the year a building was built through low and medium resolution satellite images collected by satellite remote sensing (technology / system / processing) to invert the year a building was built.

[0057] Among them, for medium and low resolution satellite images, they are themselves a kind of time series data with time series characteristics, which are collected by one or more spectral sensors carried on the satellite. In addition, this application specifically limits them to medium and low resolution spectral satellite long time series data. The medium and low resolution and long time series involved are common concepts in this field, so they are not specifically explained here.

[0058] It is easy to understand that for medium and low resolution satellite images, they are usually ready-made data, which can be directly obtained from the corresponding satellite remote sensing system through open source, or can be obtained from the corresponding satellite remote sensing system in a subscription / business manner when the system actively pushes it or has data pulling permissions, or can be obtained from other sources (which are actually obtained from the satellite remote sensing system in the beginning). Considering how to collect medium and low resolution satellite images and how to obtain medium and low resolution satellite images are not the focus of this application, it will not be elaborated here.

[0059] Among them, in specific applications, the remote sensing inversion processing of the year a building is built is usually initiated in the form of a task. The remote sensing inversion task of the year a building is built can be initiated manually, autonomously according to the corresponding initiation strategy, or it can be a task forwarded by other devices. This can be flexibly configured.

[0060] In the task information of the remote sensing inversion task of the year of building completion, the target area for remote sensing inversion of the year of building completion will be specified. The target area is usually described in terms of geographical location. Of course, the method of using information other than geographical location to indirectly indicate the geographical location is not excluded, which is also configurable in some cases.

[0061] In addition, as an exemplary embodiment, for the currently acquired medium and low resolution satellite images, data preprocessing can be performed on them before the corresponding building construction year remote sensing inversion model is developed to enhance their data quality, thereby helping to promote more efficient and high-precision building construction year remote sensing inversion effects.

[0062] Specifically, the data preprocessing involved here may include specific operations such as image correction, declouding and geometric correction.

[0063] It can be understood that in actual situations, the specific preprocessing operations adopted can usually meet the needs by using existing operating methods. The embodiment here focuses on pointing out that preprocessing can be applied to improve data quality. Of course, it is not ruled out that in specific operations, optimized or even novel preprocessing methods can be used to meet higher quality data preprocessing needs.

[0064] Step S102, inputting medium and low resolution satellite images into a pre-configured remote sensing inversion model of the year when the building was built, so that the remote sensing inversion model of the year when the building was built carries out corresponding remote sensing inversion processing of the year when the building was built, wherein the remote sensing inversion model of the year when the building was built is obtained by training an initial deep learning model with sample medium and low resolution satellite images configured with corresponding labeling results of the year when the building was built;

[0065] It is understandable that the present application also considers introducing deep learning technology to perform remote sensing inversion processing of the year a building is built. In this regard, the present application can pre-configure low-resolution satellite images in the sample, and corresponding to the output of the model as the remote sensing inversion result of the year a building is built, configure the corresponding building construction year annotation result (i.e., the true value, used to assist / guide model training). In this way, model training can be carried out on the initial deep learning model that has completed the initialization configuration work, and a remote sensing inversion model of the year a building is built that can be put into practical use.

[0066] Among them, for the remote sensing inversion model / initial deep learning model of the year of construction of the building, it can adopt the existing deep learning model architecture, or it can be a deep learning model architecture obtained by improving the prediction logic / model architecture on the basis of the existing deep learning model architecture, or it can be a self-developed, novel deep learning model architecture, which can be adaptively adjusted according to actual needs.

[0067] Model training can include the following general training principles:

[0068] After configuring the sample low-resolution satellite images with the corresponding building completion year annotation results, in each round of model training, the sample low-resolution satellite images are input into the model for training, so that the model can carry out remote sensing inversion processing of the building completion year to realize forward propagation, and then the loss function is calculated according to the remote sensing inversion results of the building completion year output by the model combined with the corresponding building completion year annotation results, and the model parameters are optimized according to the loss function calculation results to realize reverse propagation. After a large number of rounds of model training, if the preset model training requirements such as prediction accuracy, number of trainings or training time are met, the model training can be completed, and the model at this time is a remote sensing inversion model of the building completion year that can be put into practical use.

[0069] In this way, after obtaining the medium and low resolution satellite images of the target area collected by satellite remote sensing that currently require remote sensing inversion processing of the building's completion year, they can be input into the remote sensing inversion model of the building's completion year according to the preset model input method, so that the remote sensing inversion model of the building's completion year triggers and promotes the corresponding remote sensing inversion processing of the building's completion year.

[0070] Step S103, extracting the remote sensing inversion result of the building construction year corresponding to the target area output by the building construction year remote sensing inversion model.

[0071] After the remote sensing inversion model of the building construction year completes the remote sensing inversion processing of the building construction year, it can be understood that the remote sensing inversion model of the building construction year will output the corresponding remote sensing inversion result of the building construction year, and at this time the remote sensing inversion result of the building construction year can be extracted.

[0072] When the remote sensing inversion results of the year a building was built are extracted, they can be stored locally, stored in the cloud, forwarded, their contents displayed, corresponding prompt information indicating the completion of the inversion can be output, and the next step of data analysis and processing can be carried out (for example, the next step of data analysis and processing in terms of urban energy modeling, building safety assessment, urban sustainable development planning, and social ecology).

[0073] As an example, reservoir management work may involve the need to check the buildings around the reservoir, which may involve the need to check the year of construction. It is understandable that older buildings around the reservoir are often abandoned and can be considered as a kind of construction waste. The existence of the buildings will affect the environment around the reservoir. At the same time, buildings of different construction years may also bring differences in different aspects such as the flow of people. Therefore, in the reservoir management work involving the environment around the reservoir, if the remote sensing inversion results of the construction year of the buildings around the reservoir can be obtained, the work can be carried out better.

[0074] Of course, it can be understood that the above examples are only examples. The specific data application processing carried out for the remote sensing inversion results of the year of construction of the buildings in the target area can be carried out according to a pre-configured data application strategy or a real-time configured data application strategy. This application does not make specific limitations on this link.

[0075] from Figure 1 It can be seen from the illustrated embodiments that, in order to invert the year of construction of buildings, the present application utilizes medium and low-resolution satellite images based on satellite remote sensing to configure a remote sensing inversion model for the year of construction of buildings, which can predict the year of construction of corresponding buildings based on medium and low-resolution satellite images collected by satellite remote sensing in the target area. In this way, satellite remote sensing technology and deep learning technology are combined to achieve efficient, complete and accurate inversion of the year of construction of buildings, reducing the application difficulty of the inversion of the year of construction, and can solve the adverse effects of limitations such as insufficient historical data, incomplete years, and non-disclosure of data on the demand for the use of year of construction data, and can provide strong data support for urban energy modeling, building safety assessment, urban sustainable development planning, and social ecology.

[0076] Continue to the above Figure 1 Each step of the illustrated embodiment and possible implementation methods thereof in practical applications are described in detail.

[0077] Specifically, the labeling results of the year of construction of buildings configured in the low-resolution satellite images in the samples used for model training can be obtained by manual labeling or using corresponding automatic labeling tools, or they can be obtained through secondary processing based on other data.

[0078] As an exemplary embodiment, the building construction year labeling result may specifically be the construction year data at the cell granularity captured by the real estate website, integrated with the vector point data formed by the geographical location coordinates of the cell.

[0079] Among them, real estate websites can be understood as websites that publicly display information about different properties or communities on their web pages. Their form is relatively flexible, and they are mainly real estate agency websites / housing agency websites. The year of construction data can be extracted from these websites through tools such as crawlers, or through business cooperation. The specific acquisition methods involved are relatively flexible.

[0080] The data on the year of construction are recorded in units of small communities to form a small community granularity. Then, in order to correspond to the low-resolution satellite images of the samples with geographical location characteristics, the geographical location coordinates of the small communities involved in the data on the year of construction also need to be configured.

[0081] As for the geographic coordinate system involved, it can be understood that a coordinate system such as longitude and latitude can be used, which can be adjusted according to actual conditions. Moreover, when the geographic coordinate system involved in the construction year data is different from the geographic coordinate system involved in the low-resolution satellite images in the sample, conversion between geographic coordinate systems may also be involved to align the geographic coordinate systems.

[0082] As an example, the constructed vector point data may be configured with a label of the year in which the corresponding building was built based on the geographic location coordinates.

[0083] In addition, the constructed vector point data can also be marked with the positional relationship from the province to the community, for example, arranged according to the name of province-city-district-community. This is also helpful to locate the area based on this positional relationship. At the same time, when performing remote sensing inversion of the year of construction of a building based on the geographical location, the name of the corresponding location can also be output, which has a more intuitive and vivid expression effect.

[0084] Of course, this location relationship from province to community can itself be treated as a geographical location coordinate, which is also possible in actual situations.

[0085] At the same time, for the sample medium and low-resolution satellite images, as another exemplary embodiment, it can specifically be a multi-vector medium and low-resolution satellite image within a preset year range under the band conditions of four bands, namely, blue band, green band, red band and near-infrared band.

[0086] It can be understood that the embodiment here specifically limits the sample low-resolution satellite images to four bands. The multi-vector low-resolution satellite images formed by the low-resolution satellite images in these four bands can bring better processing accuracy and processing efficiency, and can meet the needs of higher-quality remote sensing inversion of the year of construction of buildings.

[0087] In addition, the low-resolution satellite images in the sample are historical data within a preset year range, and the preset year range can be adjusted.

[0088] Taking Landsat satellite remote sensing products as an example, Landsat data from Landsat 5 (1990 to 2011), Landsat 7 (2012 to 2013) and Landsat 8 (2014 to 2023) can be obtained in the blue band, green band, red band and near-infrared band.

[0089] At the same time, for multi-vector medium and low-resolution satellite images, based on the need for data simplification, as another exemplary embodiment, it can be specifically data synthesized according to the quarterly median. That is to say, in the time dimension, each quarter is based on the median to form a group of multi-vector medium and low-resolution satellite images, and then combined with the multi-vector medium and low-resolution satellite images of multiple quarters to form an overall multi-vector medium and low-resolution satellite image.

[0090] In addition, also based on the need for data simplification and to achieve data standardization, multi-vector medium and low-resolution satellite images can also be pre-performed with unified normalization processing, or in other words, the present application can also perform unified normalization processing on multi-vector medium and low-resolution satellite images to further improve data quality and facilitate use in subsequent model training.

[0091] It should be understood that the normalization processing involved here can be performed in any link of processing the low-resolution satellite images in the sample, rather than after obtaining the multi-vector low-resolution satellite images. For example, the normalization processing can be performed before the data is obtained by synthesizing according to the quarterly median, or after the data is obtained by synthesizing according to the quarterly median.

[0092] In addition, in addition to normalization processing, it is easy to understand that the present application may also involve other aspects of data preprocessing in any link of processing low-resolution satellite images in the sample, such as image correction, cloud removal, geometric correction and other specific operations.

[0093] As mentioned earlier, the present application can make corresponding optimization improvements based on the existing network architecture. The following will continue to introduce the optimization improvements made by the present application in the model architecture of the remote sensing inversion model / initial deep learning model of the year of construction of the building.

[0094] As another exemplary embodiment, the remote sensing inversion model / initial deep learning model of the building completion year of the present application is specifically a deep learning model combining a deep residual network (ResNet) as a front feature extraction unit and a gated recurrent unit (GRU) as a time series feature learning unit;

[0095] The front feature extraction unit (i.e., ResNet) consists of multiple layers of one-dimensional convolution, layer normalization function, and ReLU activation function. The image features corresponding to the input image data are extracted through multiple layers of convolution. The output of the one-dimensional convolution is 64 dimensions.

[0096] The time series feature learning unit (i.e., GRU) with a dimension of 64 is used to extract the spatiotemporal information of the time series dimension corresponding to the data output by the previous feature extraction unit (i.e., ResNet), and is combined with a fully connected layer (for output) to complete the remote sensing inversion processing task of the year the building was built.

[0097] Among them, the gated recurrent unit can specifically adopt a bidirectional gated recurrent unit (Bidirectional Gated Recurrent Unit, Bi-GRU), and the fully connected layer can specifically adopt a linear layer (Linear).

[0098] It can be understood that in the embodiment here, a specific implementation scheme of CNN+RNN is provided for the model architecture of the remote sensing inversion model / initial deep learning model of the building construction year of this application.

[0099] At the same time, for the loss function involved in the model training process, as another exemplary embodiment, the loss function in the model training process can specifically adopt the root mean square error (RMSE).

[0100] The RMSE indicator is the square root of the ratio of the square of the deviation between the observed value and the true value to the number of observations n. It reflects the degree to which the measured data deviates from the true value and can well evaluate the predictive ability of the model.

[0101] It can be understood that RMSE itself is an existing evaluation indicator, and the embodiment here only points out that this indicator can be used as a specific loss function in the model training process.

[0102] Of course, in specific operations, it is not ruled out that other types of evaluation indicators are used as the loss function specifically used in the model training process, such as Mean Squared Error (MSE), Cross-Entropy Loss, Exponential Loss, Log Loss and other types of evaluation indicators.

[0103] Even in some cases, other optimized or self-developed evaluation indicators can be used, which can be flexibly configured according to actual needs.

[0104] In addition, in the application link of the remote sensing inversion model of the year a building was built, in addition to the basic application, this application also provides a practical application solution.

[0105] As another exemplary embodiment, in order to enhance the data presentation effect through visualization, the present application may also introduce the mapping process of the year of building completion. Specifically, the method of the present application may also include:

[0106] Based on the remote sensing inversion results of the building completion year corresponding to the target area output by the building completion year remote sensing inversion model, mapping processing is performed to obtain mapping results for display;

[0107] According to the preset graphic strategy, the content of the mapping results is displayed through a visual interface.

[0108] It can be understood that the target area contains different buildings (usually large buildings with an area greater than a threshold. In this case, the present application scheme can specifically only target large buildings), and the remote sensing inversion results of the building construction years corresponding to the target area will respectively mark the corresponding building construction years of these buildings. In this way, mapping can be carried out based on the target area according to the preset graphical strategy to achieve large-area mapping, which can be displayed in the corresponding configured visualization interface for users to view intuitively and vividly.

[0109] Among them, the preset graphic strategy can correspond to both a static graphic mechanism and a dynamic graphic mechanism. The static graphic mechanism is easy to understand, which is to use a static image to display the construction year of the buildings in the entire area, or to use a static image to display the construction year of the buildings in a local area. The specific local area needs to be manually selected / adjusted, and the dynamic image mechanism can automatically display the construction year of the buildings in the entire area or the local area according to a more flexible dynamic display effect. It can be understood that the two have different focuses and correspond to different application needs in actual situations.

[0110] In addition, it can be understood that the mapping processing involved in the embodiments here can be targeted at not only the target area, but also multiple areas including the target area, which requires the application of remote sensing inversion results of the year of construction of buildings corresponding to areas other than the target area, or the application of mapping results corresponding to areas other than the target area.

[0111] Among them, it is easy to understand that the visual interface for displaying the mapping results can be realized by the processing device through its own display screen, or by a display screen device external to the processing device, or by other on-site devices with display screens other than the processing device, or by remote / cloud-based devices with display screens. These can be flexibly configured according to application requirements.

[0112] For the above exemplary embodiments, it is also possible to combine Figure 2 A schematic diagram of a scenario of the remote sensing inversion method of the building construction year based on medium and low resolution satellite images of the present application is shown to help a more vivid understanding.

[0113] In order to fully verify the application value of this application scheme, the following set of examples can also be used to fully demonstrate it:

[0114] In a specific experiment, the present application scheme was deployed on a computer equipped with an Nvidia RTX 4080Ti GPU and 16GB of memory to perform model training and building construction year mapping. A building construction year inversion experiment was performed using a medium- and low-resolution satellite image from 1990 to 2020. The building construction year data of 11 megacities crawled from the XXX real estate website was used as a reference for model training. The model achieved model data fitting after about 25 rounds of training, and the model parameters were saved when the model data was in the best fitting state. At this time, the MAE was 2.07 and the RMSE was 2.67, which fully proved that the present application scheme is very effective for the inversion of the construction year.

[0115] The above is an introduction to the remote sensing inversion method for the year of construction of buildings based on medium and low resolution satellite images provided in this application. In order to facilitate better implementation of the remote sensing inversion method for the year of construction of buildings based on medium and low resolution satellite images provided in this application, this application also provides a remote sensing inversion device for the year of construction of buildings based on medium and low resolution satellite images from the perspective of functional modules.

[0116] See also Figure 3 , Figure 3This is a schematic diagram of a structure of a remote sensing inversion device for the year of building construction based on medium and low resolution satellite images in this application. In this application, the remote sensing inversion device 300 for the year of building construction based on medium and low resolution satellite images may specifically include the following structure:

[0117] The acquisition unit 301 is used to acquire medium- and low-resolution satellite images collected by satellite remote sensing for the target area, wherein the medium- and low-resolution satellite images are specifically spectral satellite long time series data;

[0118] The inversion unit 302 is used to input the medium and low resolution satellite images into a pre-configured remote sensing inversion model of the year when the building was built, so that the remote sensing inversion model of the year when the building was built can carry out the corresponding remote sensing inversion processing of the year when the building was built, wherein the remote sensing inversion model of the year when the building was built is obtained by training an initial deep learning model with sample medium and low resolution satellite images configured with the corresponding building construction year labeling results;

[0119] The extraction unit 303 is used to extract the remote sensing inversion result of the building construction year corresponding to the target area output by the remote sensing inversion model of the building construction year.

[0120] In an exemplary embodiment, the building construction year labeling result is the construction year data at the cell granularity captured from the real estate website, integrated with the vector point data formed by the geographical location coordinates of the cell.

[0121] In another exemplary embodiment, the sample low-resolution satellite images are specifically multi-vector low-resolution satellite images within a preset year range under the band conditions of a blue band, a green band, a red band and a near-infrared band.

[0122] In another exemplary embodiment, the multi-vector low-resolution satellite images are data synthesized according to the quarterly median, and the multi-vector low-resolution satellite images have been pre-processed with unified normalization.

[0123] In another exemplary embodiment, the remote sensing inversion model of the year of building completion is specifically a deep learning model combining a deep residual network as a front feature extraction unit and a gated recurrent unit as a time series feature learning unit;

[0124] The front feature extraction unit consists of multiple layers of one-dimensional convolution, layer normalization function and ReLU activation function. The image features corresponding to the input image data are extracted through multiple layers of convolution. The output of one-dimensional convolution is 64 dimensions.

[0125] The time series feature learning unit with a dimension of 64 is used to extract the spatiotemporal information of the time series dimension corresponding to the data output by the previous feature extraction unit, and is combined with a fully connected layer to complete the remote sensing inversion processing task of the year the building was built.

[0126] In yet another exemplary embodiment, the loss function in the model training process specifically adopts the root mean square error.

[0127] In yet another exemplary embodiment, the apparatus further comprises:

[0128] A mapping unit 304 is used to perform mapping processing based on the remote sensing inversion result of the building completion year corresponding to the target area output by the building completion year remote sensing inversion model to obtain a mapping result for display;

[0129] The display unit 305 is used to display the content of the mapping result through a visual interface according to a preset diagram strategy.

[0130] This application also provides a processing device from the perspective of hardware structure, see Figure 4 , Figure 4 401, a computer program stored in the memory 402, and an input / output device 403. The processor 401 is used to execute the computer program stored in the memory 402. Figure 1 The steps of the remote sensing inversion method for the year of building construction based on medium and low resolution satellite images in the corresponding embodiment; or, the processor 401 is used to execute the computer program stored in the memory 402 to implement the following Figure 3 Corresponding to the functions of each unit in the embodiment, the memory 402 is used to store the processor 401 executing the above Figure 1 The computer program required for the remote sensing inversion method of the year of building completion based on medium and low resolution satellite images in the corresponding embodiment.

[0131] Exemplarily, the computer program may be divided into one or more modules / units, one or more modules / units are stored in the memory 402, and executed by the processor 401 to complete the present application. One or more modules / units may be a series of computer program instruction segments capable of completing specific functions, and the instruction segments are used to describe the execution process of the computer program in the computer device.

[0132] The processing device may include, but is not limited to, a processor 401, a memory 402, and an input / output device 403. Those skilled in the art will appreciate that the illustration is merely an example of a processing device and does not constitute a limitation on the processing device, and may include more or fewer components than shown in the illustration, or a combination of certain components, or different components. For example, the processing device may also include a network access device, a bus, etc., and the processor 401, the memory 402, the input / output device 403, etc. are connected via a bus.

[0133] The processor 401 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the processing device, and uses various interfaces and lines to connect various parts of the entire device.

[0134] The memory 402 can be used to store computer programs and / or modules. The processor 401 implements various functions of the computer device by running or executing the computer programs and / or modules stored in the memory 402 and calling the data stored in the memory 402. The memory 402 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required for a function, etc.; the data storage area can store data created according to the use of the processing device, etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0135] When the processor 401 is used to execute the computer program stored in the memory 402, the following functions can be implemented:

[0136] Obtain medium and low resolution satellite images collected by satellite remote sensing for the target area, wherein the medium and low resolution satellite images are specifically medium and low resolution spectral satellite long time series data;

[0137] Inputting low- and medium-resolution satellite images into a pre-configured remote sensing inversion model of the year a building was built, so that the remote sensing inversion model of the year a building was built carries out corresponding remote sensing inversion processing of the year a building was built, wherein the remote sensing inversion model of the year a building was built is obtained by training an initial deep learning model using sample low- and medium-resolution satellite images configured with corresponding labeling results of the year a building was built;

[0138] Extract the remote sensing inversion results of the building construction year corresponding to the target area output by the remote sensing inversion model of the building construction year.

[0139] Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working process of the remote sensing inversion device, processing equipment and corresponding units of the building completion year based on medium and low resolution satellite images described above can refer to the following Figure 1 The description of the remote sensing inversion method for the year of building completion based on medium and low resolution satellite images in the corresponding embodiment will not be repeated here.

[0140] A person of ordinary skill in the art will appreciate that all or part of the steps in the various methods of the above embodiments may be completed by instructions, or by controlling related hardware through instructions. The instructions may be stored in a computer-readable storage medium and loaded and executed by a processor.

[0141] To this end, the present application provides a computer-readable storage medium, in which a plurality of instructions are stored, and the instructions can be loaded by a processor to execute the present application as follows: Figure 1 The steps of the remote sensing inversion method for the year of building completion based on medium and low resolution satellite images in the corresponding embodiment, the specific operation can be referred to as follows Figure 1 The description of the remote sensing inversion method for the year of building completion based on medium and low resolution satellite images in the corresponding embodiment will not be repeated here.

[0142] The computer-readable storage medium may include: a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0143] Due to the instructions stored in the computer-readable storage medium, the present application can be executed. Figure 1 The steps of the remote sensing inversion method for the year of building completion based on medium and low resolution satellite images in the corresponding embodiment, therefore, the present application can be realized as follows Figure 1 The beneficial effects that can be achieved by the remote sensing inversion method for the year of building completion based on medium and low resolution satellite images in the corresponding embodiment are detailed in the previous description and will not be repeated here.

[0144] The above is a detailed introduction to the remote sensing inversion method, device, processing equipment and computer-readable storage medium for the year of building construction based on medium and low resolution satellite images provided by the present application. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea; at the same time, for technical personnel in this field, according to the idea of ​​the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A remote sensing inversion method for the year of building completion based on medium and low resolution satellite images, characterized in that: The method comprises: Acquire medium- and low-resolution satellite images collected by satellite remote sensing for the target area, wherein the medium- and low-resolution satellite images are specifically medium- and low-resolution spectral satellite long-time series data; Inputting the low- and medium-resolution satellite images into a pre-configured remote sensing inversion model of the year a building was built, so that the remote sensing inversion model of the year a building was built carries out corresponding remote sensing inversion processing of the year a building was built, wherein the remote sensing inversion model of the year a building was built is obtained by training an initial deep learning model with sample low- and medium-resolution satellite images that are configured with corresponding building construction year labeling results; The remote sensing inversion result of the building construction year corresponding to the target area output by the remote sensing inversion model of the building construction year is extracted.

2. The method according to claim 1, characterized in that The building construction year labeling result is the construction year data of the community granularity captured by the real estate website, integrated with the vector point data formed by the geographical location coordinates of the community.

3. The method according to claim 1, characterized in that The sample medium and low-resolution satellite images are specifically multi-vector medium and low-resolution satellite images within a preset year range under the band conditions of a blue band, a green band, a red band and a near-infrared band.

4. The method according to claim 3, characterized in that The multi-vector low-resolution satellite images are specifically data synthesized according to the quarterly median, and the multi-vector low-resolution satellite images have also been pre-processed with unified normalization.

5. The method according to claim 1, characterized in that The remote sensing inversion model for the year of building completion is specifically a deep learning model that combines a deep residual network as a front feature extraction unit and a gated recurrent unit as a time series feature learning unit; The front feature extraction unit is composed of multiple layers of one-dimensional convolution, layer normalization function and ReLU activation function, and extracts image features corresponding to the input image data through multiple layers of convolution, wherein the one-dimensional convolution output is 64 dimensions; The time series feature learning unit with a dimension of 64 is used to extract the spatiotemporal information of the time series dimension corresponding to the data output by the previous feature extraction unit, and is combined with a fully connected layer to complete the remote sensing inversion processing task of the year the building was built.

6. The method according to claim 1, characterized in that The loss function in the model training process specifically adopts the root mean square error.

7. The method according to claim 1, characterized in that The method further comprises: Based on the remote sensing inversion result of the building completion year corresponding to the target area output by the remote sensing inversion model of the building completion year, a mapping process is performed to obtain a mapping result for display; According to the preset graphic strategy, the content of the mapping result is displayed through a visual interface.

8. A remote sensing inversion device for the year of building completion based on medium and low resolution satellite images, characterized in that: The device comprises: An acquisition unit is used to acquire medium and low resolution satellite images collected by satellite remote sensing for the target area, wherein the medium and low resolution satellite images are specifically medium and low resolution spectral satellite long time series data; An inversion unit is used to input the medium and low resolution satellite images into a pre-configured remote sensing inversion model of the year when the building was built, so that the remote sensing inversion model of the year when the building was built carries out corresponding remote sensing inversion processing of the year when the building was built, wherein the remote sensing inversion model of the year when the building was built is obtained by training an initial deep learning model with sample medium and low resolution satellite images configured with corresponding labeling results of the year when the building was built; The extraction unit is used to extract the remote sensing inversion result of the building construction year corresponding to the target area output by the remote sensing inversion model of the building construction year.

9. A processing device, characterized in that: The method comprises a processor and a memory, wherein the memory stores a computer program, and the processor executes the method according to any one of claims 1 to 7 when calling the computer program in the memory.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor to execute the method according to any one of claims 1 to 7.

Images