Building construction year identification method based on remote sensing waveband similarity matrix clustering

Abstract

The invention provides a building construction year identification method based on remote sensing waveband similarity matrix clustering, and the method is characterized in that the method comprises the steps: obtaining a multi-stage multispectral remote sensing image covering a target area, carrying out the preprocessing and annual synthesis, and constructing a long-time-sequence multiband time sequence; dividing the target area into analysis units; for each analysis unit, calculating the wave band similarity measurement between the annual synthetic images of any two different years in the time dimension to form a similarity matrix; regarding the similarity matrix as a weighted graph, and segmenting the weighted graph by adopting an unsupervised graph clustering algorithm to obtain a time continuous cluster distributed along a main diagonal line; and determining the construction year of the building corresponding to the analysis unit based on the starting year of the end time cluster including the latest year.

Description

Technical Field

[0001] This invention belongs to the field of intelligent interpretation of remote sensing images and temporal analysis of the land surface, specifically involving a building construction year identification technology based on long-time remote sensing band similarity matrix clustering. Background Technology

[0002] The year of a building's construction is one of the important attribute parameters of the urban built environment, directly reflecting the building's service life, material composition, energy-saving characteristics, and seismic performance. This information is not only a basic element in building databases, but also an important supporting data for carrying out building material stock estimation, carbon emission accounting, old community renovation, and urban renewal monitoring. In the fields of urban planning, ecological assessment, and building life cycle management, the accurate identification of the year of a building's construction is of key significance for improving the scientific nature of urban spatial analysis and policy formulation. However, the currently available data on the year of a building's construction is still very limited, with problems such as narrow coverage, long update cycles, and poor spatiotemporal consistency. Traditional methods for obtaining the year of a building's construction mainly include the following three categories: (1) crowdsourced geographic information data method, (2) single / multi-temporal image interpretation method, and (3) time series remote sensing inversion method.

[0003] First, methods based on crowdsourced geographic information data can provide relatively accurate building year information; for example, cadastral databases or real estate registration systems can contain building completion year attributes. However, the openness and accessibility of such data vary significantly globally, with only a few countries or regions achieving partial openness. In recent years, some studies have attempted to extract building year information from real estate websites by combining web crawling and geographic matching. However, crowdsourced data generally suffers from uneven spatiotemporal distribution, unstable metadata quality, and privacy and copyright restrictions, making it difficult to support wide-area and long-term mapping of building completion years.

[0004] Secondly, manual or semi-automatic interpretation methods based on single / multi-temporal high-resolution imagery can identify building changes by comparing historical aerial imagery, commercial satellite imagery, or street view data. These methods can achieve high accuracy in local areas, but they generally rely on manual intervention and discrete time-point imagery, resulting in low automation, long data update cycles, and difficulty in capturing the continuous evolution of buildings over time. Existing change detection algorithms mostly focus on spatial texture or morphological differences, failing to fully utilize the dynamic spectral information of remote sensing time-series data.

[0005] Furthermore, long-term satellite remote sensing data (such as Landsat, Sentinel, and MODIS) offer new technological pathways for dynamic monitoring of the built environment due to their advantages of long time spans, wide coverage, and rich spectral information. In recent years, some studies have used long-term land use / impervious surface products to indirectly infer the year of urban construction. However, these products are mostly based on the principle of "consistency correction" (i.e., assuming that changes in impervious surfaces are irreversible). When complex urban processes such as demolition, reconstruction, and functional transformation occur, this can lead to errors in year identification.

[0006] To improve the temporal identification capabilities of the built environment, various continuous change detection algorithms have been proposed, such as VCT (Vegetation Change Tracker), BFAST (Breaks For Additive Seasonal Trend), and CCDC (Continuous Change Detection and Classification). These methods have achieved good results in forest ecology and land cover monitoring and are gradually being introduced into urban change analysis. However, factors such as the nonlinear characteristics of building spectral changes, ground cover mixing effects, shadow interference, and high reflectivity backgrounds limit the applicability of these algorithms in building temporal identification. In addition, traditional change detection algorithms usually rely on specific band thresholds or classification models, lacking a unified description of spectral similarity and temporal continuity, resulting in identification results being significantly affected by temporal resolution, image noise, and threshold settings. Summary of the Invention

[0007] The purpose of this invention is to provide a method and system for identifying building construction years based on long-time remote sensing band similarity matrix clustering. By constructing a unified long-time multi-band sequence across sensors, the method automatically identifies continuous temporal structures using inter-band similarity matrices and unsupervised graph clustering algorithms. It outputs only the starting year of the last time cluster as the building's construction year, achieving construction year identification and update without the need for preset thresholds or determining the number of clusters. This method possesses high robustness, interpretability, and wide applicability.

[0008] The technical solution of this invention is a method for identifying the construction year of buildings based on remote sensing band similarity matrix clustering, comprising: Acquire multi-period multispectral remote sensing images covering the target area, perform preprocessing and annual synthesis, and construct a long-term multi-band time series. The target region is divided into analysis units; For each of the aforementioned analysis units, the band similarity measure between annual composite images of any two different years is calculated in the time dimension to form a similarity matrix; The similarity matrix is ​​treated as a weighted graph, and an unsupervised graph clustering algorithm is used to segment the weighted graph to obtain time-continuous clusters distributed along the main diagonal. The construction year of the building corresponding to the analysis unit is determined based on the starting year of the end time cluster containing the latest year.

[0009] Furthermore, the construction of the long-term multi-band time series includes: performing cross-sensor spectral response difference correction and radiometric normalization processing on multiple periods of remote sensing images; aggregating multiple images of the same year on a calendar year basis to generate an annual composite image; and stacking the composite images of each year to form the long-term multi-band time series.

[0010] Furthermore, the analysis unit is the result of a sliding window division or a land parcel object; the side length of the sliding window is adaptively adjusted according to the local spectral variance.

[0011] Furthermore, the calculation of band similarity measurement includes: for any two years, calculating the weighted correlation coefficient of the feature vectors corresponding to the analysis unit; introducing a coefficient that decays with increasing time distance, converting the correlation coefficient into a distance value, and thus forming the similarity matrix.

[0012] Moreover, the unsupervised graph clustering algorithm is a graph segmentation algorithm based on minimizing the average cut cost, which can automatically determine the time segmentation point without pre-setting the number of clusters.

[0013] Furthermore, determining the construction year of a building also includes: classifying the building status corresponding to the analysis unit as unchanged, newly built, or updated based on the shape and distribution of the time continuous cluster; for the updated status, outputting both the initial construction year and the update year.

[0014] On the other hand, the present invention also provides a building construction year identification system based on remote sensing band similarity matrix clustering, comprising the following modules: The preprocessing module is used to acquire multi-period multispectral remote sensing images covering the target area, perform preprocessing and annual synthesis, and construct long-term multi-band time series. An analysis unit generation module is used to divide the target region into analysis units; The similarity calculation module is used to calculate the band similarity measure between annual composite images of any two different years in the time dimension for each of the analysis units, forming a similarity matrix; The graph clustering module is used to treat the similarity matrix as a weighted graph, and to segment the weighted graph using an unsupervised graph clustering algorithm to obtain time-continuous clusters distributed along the main diagonal. The year determination module is used to determine the construction year of the building corresponding to the analysis unit based on the starting year of the end time cluster containing the latest year.

[0015] On the other hand, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, it implements a method for identifying the construction year of a building based on clustering of remote sensing band similarity matrices as described above.

[0016] On the other hand, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements a method for identifying the construction year of a building based on clustering of remote sensing band similarity matrices as described above.

[0017] On the other hand, the present invention also provides a computer program product, including a computer program that, when executed by a processor, implements a method for identifying the construction year of a building based on clustering of remote sensing band similarity matrices as described above.

[0018] Compared with the prior art, the beneficial effects achieved by the present invention are as follows: This invention offers significant innovation and technological advantages over existing methods for identifying building construction years. Traditional methods often rely on crowdsourced geographic information data or high-resolution image interpretation from a single time phase, which suffers from challenges such as difficulty in data acquisition, uneven spatial coverage, and significant human intervention, making it difficult to achieve large-scale, continuous, and automated identification. In contrast, this invention constructs a complete annual image sequence based on long-term multi-source remote sensing data and utilizes band similarity matrices and parametric graph clustering algorithms to achieve automated identification of building construction years. By focusing on long-term observations of multi-source, multispectral remote sensing data, this invention significantly improves the completeness of the time series and the continuity of spatial coverage, providing a solid data and algorithmic foundation for long-term, wide-area identification of building construction years.

[0019] Secondly, existing change detection techniques based on multi-temporal imagery mostly focus on comparing differences between limited time points, neglecting the continuity of temporal information and the internal structural features of the change process. Furthermore, they are limited by the spectral mixing effects of low-to-medium resolution imagery and the complexity of ground features, making it difficult to accurately characterize the gradual and abrupt changes in buildings. This invention constructs a time-series band similarity matrix to systematically quantify the similarity of spectral features across years, thereby revealing the spectral response patterns of buildings during stable, abrupt, and reconstructive periods over time. This method effectively improves the sensitivity of identifying "newly built," "updated," and "unchanged" building scenarios. While ensuring long-term temporal continuity, it significantly improves the accuracy and robustness of change breakpoint detection, achieving refined identification of the dynamic evolution process of buildings.

[0020] Furthermore, traditional clustering and change detection algorithms often rely on manually set parameters or empirical thresholds, resulting in limited stability and versatility, making them unsuitable for diverse regions and complex terrain environments. This invention innovatively applies the MeanCut parametric graph clustering algorithm to building construction year identification, achieving adaptive segmentation of time series by minimizing the average cut cost function, eliminating the need for pre-setting the number of clusters or manual parameter tuning. This algorithm automatically extracts continuous time clusters distributed along the main diagonal, thus distinguishing between stable and changing periods in the temporal feature space. Compared to traditional threshold-based or fitting-based change detection methods, the clustering method of this invention is more adaptive and generalizable, exhibiting excellent stability and consistency across different remote sensing platforms and diverse landform types, making it suitable for wide-area applications in complex terrain environments.

[0021] Furthermore, traditional methods often struggle to accurately identify the "new construction" and "renovation" processes of buildings in Chinese cities due to the complex changes such as demolition and reconstruction, expansion, renovation, and functional transformation, leading to significant errors in construction year estimation. This invention, through continuous monitoring of long-term time-series imagery and similarity breakpoint identification, can simultaneously identify both the initial construction year and the renovation year of a building, distinguishing between "first construction" and "reconstruction" changes over time. This method effectively solves the problem of difficulty in identifying complex phenomena such as "repeated construction" and "demolition and reconstruction" in traditional change detection, significantly improving the timeliness and reliability of building construction year data.

[0022] In summary, this invention not only achieves fully automated and high-precision identification of building construction years, but also distinguishes between the construction and renovation stages of buildings, providing crucial time-dimensional data support for urban material stock assessment, building lifecycle management, and the construction of digital twin cities. The mapping results of building construction and renovation years generated by this method can directly serve urban planning, spatial optimization, carbon emission accounting, and building energy efficiency analysis, providing important technical support for promoting the scientific, dynamic, and refined management of urban construction. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating a method for identifying the construction year of a building based on long-time remote sensing band similarity matrix clustering, according to an exemplary embodiment. Figure 2 A schematic diagram of the building construction year identification system based on long-time remote sensing band similarity matrix clustering in this embodiment of the invention; Figure 3 A schematic diagram of Landsat band clustering in an embodiment of the present invention; Figure 4 This is a schematic diagram illustrating the method of the present invention for identifying the construction year of buildings in a certain urban cluster. Detailed Implementation

[0024] To better illustrate the technical content and practical value of this invention, the following describes in detail the building construction year identification method and system based on long-term remote sensing band similarity matrix clustering proposed in this invention, in conjunction with the accompanying drawings and embodiments. The method and system of this invention are based on multi-source long-term remote sensing data, deeply mine the spectral temporal similarity structure of building plots, and automatically identify continuous time clusters through unsupervised graph clustering. Only the starting year of the last time cluster containing the most recent year is used as the building construction year, achieving large-scale, long-term automatic identification and mapping.

[0025] It should be noted that the present invention follows applicable laws, regulations and industry standards in the process of acquiring, storing, processing and using remote sensing images and geographic information data.

[0026] It should be further noted that general-purpose or open-source tools (such as GDAL / Rasterio for data processing, and Python / NumPy / SciPy / scikit-learn for matrix calculations and clustering) can be used during implementation. The tools mentioned in the specification are only for feasibility and do not constitute a limitation on the internal implementation; in the absence of explicit limitations, they should be understood as functional purposes rather than structural limitations. The scope of protection of this invention is determined by the claims.

[0027] Example 1 like Figure 1 As shown, this embodiment of the invention provides a method for identifying the construction year of buildings based on remote sensing band similarity matrix clustering, including: Acquire multi-period multispectral remote sensing images covering the target area, perform preprocessing and annual synthesis, and construct a long-term multi-band time series. The target region is divided into analysis units; For each of the aforementioned analysis units, the band similarity measure between annual composite images of any two different years is calculated in the time dimension to form a similarity matrix; The similarity matrix is ​​treated as a weighted graph, and an unsupervised graph clustering algorithm is used to segment the weighted graph to obtain time-continuous clusters distributed along the main diagonal. The construction year of the building corresponding to the analysis unit is determined based on the starting year of the end time cluster containing the latest year.

[0028] In some embodiments, the preprocessing includes acquiring multi-period homogeneous or cross-sensor multispectral remote sensing images covering the target area, performing radiometric normalization and spectral response matching, and then performing annual synthesis to form a multi-band time series covering T years, where T is a preset number of years.

[0029] In some embodiments, the target area is divided into sliding windows or plot objects as analysis units, each containing a time series of several pixels.

[0030] In some embodiments, for each analysis unit, the band similarity measure between any two years is calculated in the time dimension, preferably using the Pearson correlation coefficient and converting it into a distance measure to form a T×T similarity matrix.

[0031] In some embodiments, the construction year of a building is determined based on the starting year of the end time cluster, and the analysis unit is classified as "unchanged", "newly built" or "updated" in combination with the time cluster pattern.

[0032] Example 2 like Figure 1 As shown, this embodiment of the invention provides a method for identifying the construction year of buildings based on remote sensing band similarity matrix clustering, including: like Figure 1 As shown, this invention provides a method and system for identifying the construction year of buildings based on long-term remote sensing band similarity matrix clustering, including the following steps: As an optional embodiment, in the first implementation stage of the present invention, multi-period homogeneous or cross-sensor multispectral remote sensing images covering the target area are acquired, and radiometric normalization, spectral response matching, and annual composite processing are performed to construct a long-term multi-band time series covering year T. The specific implementation process of step 1 includes the following sub-steps: Step 1.1: Acquire multiple clear-sky condition observation images covering the target area, preferably using multispectral images such as Landsat 5, Landsat 7, Landsat 8, Landsat 9, or Sentinel-2. For years with scarce data, cross-platform data can be introduced to achieve continuous temporal supplementation.

[0033] Step 1.2: Reproject all images to a unified coordinate system (such as WGS84 / UTM) and align them to a unified raster grid and spatial resolution. Perform spectral response difference correction and radiometric normalization on cross-platform images, preferably using methods such as histogram matching, quantile matching, or robust regression, and combine spectral response function mapping to achieve equivalence matching between channels.

[0034] Step 1.3: Using the Gregorian calendar year as the unit, multiple images from the same year are composited using a composite strategy (such as median, quantile, or quality-weighted average) to generate an annual composite image representing the state of that year. Typically, 20-30 cloud-free images (i.e., clear-sky observation images) can be obtained in the same year, so composite work is necessary.

[0035] Step 1.4: For strip, track, or block stitching areas, perform overlapping area weight fusion or seamless stitching processing to eliminate track boundary and seam effects and form a spatially continuous annual composite image.

[0036] Step 1.5: By Year Stacking images from different years to form a coverage Long-term, multi-band time series data spanning years. When time ranges differ across sensors, a timescale normalization mechanism ensures seamless integration of dissimilar time periods. Simultaneously, an annual index table and data integrity identifiers are established to guarantee the temporal continuity, spatial consistency, and traceability of the time series. For example, Landsat series data covers the period from 1986 to the present; Sentinel-2 data covers the period from 2016 to the present, and cross-platform fusion can be used to create a comprehensive time series from 1986 to the present.

[0037] As an optional embodiment, in the second stage, the target area is divided into analysis units. These analysis units can be sliding windows or land parcel objects, used to describe the time-series characteristics of local areas. The specific implementation process of step 2 includes the following sub-steps: Step 2.1: Based on the characteristics of the target area, divide the image of the target area into analysis units using an adaptive sliding window, or divide it into plots of land as analysis units using an object-oriented approach. The side length of the sliding window can be adjusted. , , The system adaptively adjusts across multiple scales, and the optimal scale is determined based on minimizing the local spectral variance, thereby ensuring statistical stability.

[0038] Step 2.2: Expand the pixels within each analysis unit in row-major order. The column vectors are then normalized by performing standardization or principal component transformation on the band features.

[0039] in, Indicates the first Year, This indicates a standardization or principal component transformation operation.

[0040] As an optional embodiment, the specific implementation process of step 3 includes the following sub-steps: For each analysis unit, the band similarity measure between any two years is calculated in the time dimension. The preferred method is to use a weighted Pearson correlation coefficient, which is then converted into a distance metric to form... The similarity matrix, This represents the observed value of each pixel in the spatial analysis unit at year t. This represents the number of pixels within the analysis unit, i.e., the dimension of the column vector expanded in row-major order.

[0041] Each window corresponds to a time series of length T. , can be represented as: (1) in, This represents each year in a long-term, multi-band time series of T years, such as 2016, 2017, etc. This spatial analysis unit is located in the year. The corresponding normalized feature vector.

[0042] As an optional embodiment, in the third stage, for each analysis unit, the band similarity measure between any two annual composite images is calculated in the time dimension to obtain a T×T similarity matrix. The specific implementation process of step 3 includes the following sub-steps: Step 3.1: Calculate the weighted Pearson correlation coefficient. For any two years... Calculate the weighted Pearson correlation coefficient from the feature vectors:

[0043]

[0044] in, and These represent the spatial analysis unit in the year. and The normalized value of the k-th pixel. =1,2, … ; It is an eigenvector The weighted mean, It is an eigenvector The weighted mean, The pixel weights are determined based on time series stability, cloud mask ratio, or brightness anomaly factor.

[0045] Step 3.2: Time Attenuation and Distance Measurement. To suppress noise caused by differences in distance, a time attenuation coefficient is introduced: in, It is an exponential function. It is a time-scale parameter used to control the decay rate.

[0046] Based on this definition, the weighted Pearson distance reflects the degree of difference in the band characteristics of the analysis unit over two years in the time dimension:

[0047] Step 3.3: Similarity Matrix Generation and Filtering. A time-series similarity matrix is ​​generated by weighted distances between all years. : (2) in, This indicates that the analysis unit is in the first... Year and the The weighted Pearson distance between years.

[0048] To assess the smoothness and temporal continuity of similar structures. If the proportion of distance values ​​greater than 1 in any row or column exceeds a threshold. If the analysis unit is an anomaly, it will be marked as an anomaly and removed to improve the robustness and reliability of subsequent clustering. This is an empirical threshold, typically ranging from 0.1 to 0.5, and needs to be flexibly adjusted based on factors such as data noise, image quality, and heterogeneity of the study area. For areas with high quality and relatively uniform land cover types, A value of 0.3–0.5 is acceptable; for areas with high noise levels or strong spatial variability of ground features, it is recommended to use 0.1–0.3 to increase the stringency of anomaly detection.

[0049] As an optional embodiment, in the fourth stage, based on the aforementioned similarity matrix, a weighted undirected graph structure is generated and unsupervised graph clustering is performed to obtain time-continuous clusters distributed along the main diagonal. The specific implementation process of step 4 includes the following sub-steps: Step 4.1: Suppose a weighted undirected graph is formed based on the similarity matrix. Node set Representing each year's time point, edge set This represents the similarity connections between time points. (Edge weight matrix) The edge weight represents the similarity strength between time nodes, and is defined as follows:

[0050] The first term is a similarity term based on spectral differences, and the second term is a smoothing attenuation term based on temporal proximity, to ensure the continuity of the temporal structure and local constraints. When At that time, the edge weight is set to zero to suppress pseudo-correlation in distant years; It is a spectral distance scale parameter. It is the maximum allowed time span threshold.

[0051] Step 4.2: Perform unsupervised clustering on the weighted graph, preferably using the MeanCut clustering algorithm. The embodiment performs unsupervised MeanCut clustering on the weighted graph. MeanCut automatically determines the time split points by minimizing the average cut cost function. Defined as:

[0052] Among them, the cut weight function yes:

[0053] The MeanCut algorithm does not require manual pre-setting of the number of clusters. The algorithm automatically determines the time segmentation point by minimizing the average cut cost. That is, the algorithm can directly identify continuous time clusters distributed along the main diagonal.

[0054] Step 4.3: Clustering Stability Constraints. The clustering results are typically represented as a series of time-continuous clusters distributed along the main diagonal of the similarity matrix, where the later clusters (i.e., the terminal clusters containing the most recent year) are... This corresponds to a period of time after a building is completed that remains stable. This invention focuses only on the starting year of the final time cluster as the building's completion year, and does not rely on the total number of clusters. The accuracy of clustering is high; scattered or early clustering has little impact on the identification results and requires no human intervention.

[0055] To ensure temporal continuity, the time cluster corresponding to each cluster is limited to a continuous range of years. And calculate the mean intra-cluster similarity. Mean inter-cluster similarity Define the cluster stability index:

[0056] when Or the cluster width is less than the minimum threshold When needed, adjacent clusters are automatically merged to enhance stability.

[0057] As an optional embodiment, in the fifth stage, based on the continuous cluster structure along the time axis in the clustering results, the starting year of the final time cluster is determined and the building status is classified. The specific implementation process of step 5 includes the following sub-steps: Step 5.1: First, analyze the cluster structure that is continuous along the time axis in the clustering results to identify the last time clusters containing the most recent year. Its starting year This refers to the year the building was constructed. If the terminal cluster... There was another stable time cluster previously. Then its starting year Recorded as the year of initial construction.

[0058] Step 5.2: State classification rules.

[0059] The building status is determined based on the morphology and distribution characteristics of time clusters: When the number of clusters And the unique cluster covers no less than Year, or the starting year of the terminal cluster. With the start year of the time series The difference The analysis unit is determined to be in an "unchanged" state. When end clusters exist And its starting year And in that starting year Previously, there was no minimum cluster width that met the requirements. When a cluster is found to be valid, it is classified as "new". When another valid cluster exists before the terminal cluster. When this occurs, it is determined to be an "update". Among them, The annual tolerance threshold is set (preferably 1–2 years). The minimum cluster width threshold (preferably set to 2 years or more).

[0060] Step 5.3: After the determination is completed, the system outputs the corresponding building construction year and tag information according to different states. For the analysis unit in the "unchanged" state, the output construction year is the starting year of the sequence. The year a building was completed can be set as the starting year of a time series. Or the earliest available year; for analysis units in the "New" state, output The year the building was completed is set as the starting year of the end time cluster; for the analysis unit in the "updated" state, double labels are output. ,in The year of initial construction. For updating the year; For units that cannot be clearly classified, output "Undetermined" along with confidence level and quality indicator. This means that the time cluster features of the analyzed unit do not meet the above conditions and cannot be clearly classified. It can be determined and used later.

[0061] See Figure 3 Taking Landsat data from 1990 to 2024 as an example, the band clustering similarity matrix of three building update scenarios was obtained using the implementation method.

[0062] See Figure 4 Based on the method proposed in this invention, using time-series Landsat data from 1990 to 2024, and taking a certain urban cluster in China as the study area, we can realize the mapping of the construction year of buildings.

[0063] Example 3 like Figure 2 As shown, this embodiment of the invention provides a building construction year identification system based on remote sensing band similarity matrix clustering, including the following modules: The preprocessing module is used to acquire multi-period multispectral remote sensing images covering the target area, perform preprocessing and annual synthesis, and construct long-term multi-band time series. An analysis unit generation module is used to divide the target region into analysis units; The similarity calculation module is used to calculate the band similarity measure between annual composite images of any two different years in the time dimension for each of the analysis units, forming a similarity matrix; The graph clustering module is used to treat the similarity matrix as a weighted graph, and to segment the weighted graph using an unsupervised graph clustering algorithm to obtain time-continuous clusters distributed along the main diagonal. The year determination module is used to determine the construction year of the building corresponding to the analysis unit based on the starting year of the end time cluster containing the latest year.

[0064] In specific implementation, the method proposed in the technical solution of this invention can be automatically executed by those skilled in the art using computer software technology. System devices for implementing the method, such as computer-readable storage media storing the corresponding computer program of the technical solution of this invention and computer equipment including the computer program running the corresponding computer program, should also be within the protection scope of this invention.

[0065] The following embodiments describe the electronic device provided by the present invention. The electronic device described below can be referred to in correspondence with the building construction year identification method based on remote sensing band similarity matrix clustering described above.

[0066] The electronic device may include a processor, a communications interface, a memory, and a communication bus. The processor, communications interface, and memory communicate with each other via the communication bus. The processor can call logical instructions from the memory to execute a building construction year identification method based on remote sensing band similarity matrix clustering, mainly including the software processing part described above.

[0067] Furthermore, the logical instructions in the aforementioned memory can be implemented as software functional units and sold or used as independent products, and can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0068] On the other hand, embodiments of the present invention also provide a computer program product, the computer program product including a computer program, the computer program being stored on a non-transitory computer-readable storage medium, and when the computer program is executed by a processor, the computer is able to execute the software processing part of the building construction year identification method based on remote sensing band similarity matrix clustering provided by the above methods.

[0069] In another aspect, embodiments of the present invention also provide a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the software processing portion of the building construction year identification method based on remote sensing band similarity matrix clustering provided by the above methods.

[0070] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0071] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0072] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for identifying the construction year of buildings based on remote sensing band similarity matrix clustering, characterized in that, include: Acquire multi-period multispectral remote sensing images covering the target area, perform preprocessing and annual synthesis, and construct a long-term multi-band time series. The target region is divided into analysis units; For each of the aforementioned analysis units, the band similarity measure between annual composite images of any two different years is calculated in the time dimension to form a similarity matrix; The similarity matrix is ​​treated as a weighted graph, and an unsupervised graph clustering algorithm is used to segment the weighted graph to obtain time-continuous clusters distributed along the main diagonal. The construction year of the building corresponding to the analysis unit is determined based on the starting year of the end time cluster containing the latest year.

2. The method for identifying the construction year of a building based on remote sensing band similarity matrix clustering according to claim 1, characterized in that, The construction of the long-term multi-band time series includes: performing cross-sensor spectral response difference correction and radiometric normalization processing on multiple periods of remote sensing images; aggregating multiple images of the same year on a calendar year basis to generate an annual composite image; and stacking the composite images of each year to form the long-term multi-band time series.

3. The method for identifying the construction year of buildings based on remote sensing band similarity matrix clustering according to claim 1, characterized in that, The analysis unit is the result of the sliding window division or the land parcel object; the side length of the sliding window is adaptively adjusted according to the local spectral variance.

4. The method for identifying the construction year of a building based on remote sensing band similarity matrix clustering according to claim 1, characterized in that, The calculation of band similarity measurement includes: for any two years, calculating the weighted correlation coefficient of the feature vectors corresponding to the analysis unit; introducing a coefficient that decays with increasing time distance, converting the correlation coefficient into a distance value, and thus forming the similarity matrix.

5. The method for identifying the construction year of a building based on remote sensing band similarity matrix clustering according to claim 1, characterized in that, The unsupervised graph clustering algorithm is a graph segmentation algorithm based on minimizing the average cut cost, which can automatically determine the time segmentation point without pre-setting the number of clusters.

6. The method for identifying the construction year of a building based on remote sensing band similarity matrix clustering according to claim 1, characterized in that, Determining the construction year of a building also includes: classifying the building status corresponding to the analysis unit as unchanged, newly built, or updated based on the shape and distribution of the time continuous cluster; for the updated status, outputting both the initial construction year and the update year.

7. A building construction year identification system based on remote sensing band similarity matrix clustering, characterized in that, Includes the following modules: The preprocessing module is used to acquire multi-period multispectral remote sensing images covering the target area, perform preprocessing and annual synthesis, and construct long-term multi-band time series. An analysis unit generation module is used to divide the target region into analysis units; The similarity calculation module is used to calculate the band similarity measure between annual composite images of any two different years in the time dimension for each of the analysis units, forming a similarity matrix; The graph clustering module is used to treat the similarity matrix as a weighted graph, and to segment the weighted graph using an unsupervised graph clustering algorithm to obtain time-continuous clusters distributed along the main diagonal. The year determination module is used to determine the construction year of the building corresponding to the analysis unit based on the starting year of the end time cluster containing the latest year.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, it implements a method for identifying the building construction year based on remote sensing band similarity matrix clustering as described in any one of claims 1 to 7.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements a method for identifying the construction year of a building based on clustering of remote sensing band similarity matrices as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that: When the computer program is executed by the processor, it implements a method for identifying the construction year of a building based on clustering of remote sensing band similarity matrices as described in any one of claims 1 to 6.

Images