Urban renewal area identification method based on the fusion of time series remote sensing data and geographic big data

Abstract

The present invention discloses an urban renewal area identification method by integrating time series remote sensing data with geographic big data. The method comprises the following steps: S1, studying a quantile-based method, integrating POI and NPPNL data, and extracting the scope of urban built-up areas in the study area in year X; S2, by processing multi-source remote sensing data from year X-8 to year X, identifying urban vacant land, urban green space and urban construction land, and then obtaining an annual land cover map within the scope of urban built-up areas. Based on POI and night light data, the present invention uses the quantile method to extract the scope of urban built-up areas in the end year of the study area, processes Google Earth HRRS time series data and Landsat time series data based on semantic segmentation algorithms, remote sensing index calculation methods, etc., can accurately classify urban land cover types, thereby further carrying out urban renewal monitoring, formulating land cover type conversion rules, monitoring urban renewal dynamics, and having important significance for the academic development of urban renewal and future urban development.

Description

Technical Field

[0001] The present invention relates to the technical field of urban renewal area renewal, and specifically to an urban renewal area identification method that integrates time-series remote sensing data with geographic big data. Background Art

[0002] The spatiotemporal distribution information of urban renewal is the premise for conducting research on driving factors of urban renewal, urban renewal paths and strategies, soil heavy metal risk control, urban thermal environment analysis, etc. Therefore, timely and accurate spatiotemporal distribution information of urban renewal is of great significance for urban planning and management, smart city construction and urban sustainable development.

[0003] Deficiencies of existing technology:

[0004] Although the spatiotemporal distribution information of urban renewal is very important, the dynamics of urban renewal are currently difficult to monitor. There are usually two methods to obtain urban renewal information. The first is the official statistical data released by the relevant urban planning departments, or the information of urban renewal project units in a small area. However, this type of method cannot capture the detailed changes in urban renewal demolition and reconstruction. In order to study the spatiotemporal characteristics of urban renewal, many studies have adopted the second method, that is, researchers monitor through field surveys or based on visual interpretation of high-resolution remote sensing images. However, this type of method has high labor or time costs and is more suitable for small-scale research.

[0005] Remote sensing data with high spatial resolution and high temporal resolution can reveal the details of urban changes. Using satellite remote sensing images for dynamic monitoring of urban renewal is more suitable for capturing the spatiotemporal pattern of urban renewal. The commonly used method is to extract and compare building information (such as building spectrum, building height, etc.) in two periods through satellite remote sensing technology, and then draw the spatial layout of urban renewal. However, the building spectra before and after urban renewal are very similar, and information such as building height in medium-scale or large-scale urban dense areas has the disadvantages of high acquisition cost and difficult to ensure accuracy. Therefore, it is difficult to accurately draw the spatial layout of urban renewal by comparing the changes in building information alone. In addition, the time span of urban renewal is usually long, and the dynamic information of urban renewal in a long time series cannot usually be accurately identified by comparing remote sensing images of two periods. Due to these difficulties, the current time series dynamic monitoring of urban renewal is greatly limited in both research scale and temporal resolution.

[0006] In recent years, time series algorithms have shown great potential in understanding the dynamic changes of land cover with high temporal resolution at different research scales. Currently, the commonly used time series algorithms include LandTrendr, VCT, BFAST, and CCDC. These time series methods monitor the changes of a certain type of land cover category by depicting the temporal change trajectory of a certain spectral index. Commonly used spectral indices are usually used to explore the dynamic changes of natural ecosystems, such as NDVI index, NBR index, and TCW index. However, the above time series algorithms are rarely used in urban monitoring research. This is mainly because there are complex and high-frequency spatiotemporal changes between urban land cover types, and a single indicator or rule cannot fully monitor the dynamic changes of the city. For example: urban expansion includes land cover type conversion in various situations, such as the conversion of vegetation, bare land, and water to construction land. In addition, the surface reflectance of artificial facilities built with different materials is very different, so the spectral reflectance of the land before and after the expansion will be very different. Therefore, a single building indicator, vegetation indicator, or water body indicator cannot fully monitor the dynamics of urban expansion.

[0007] In order to solve this problem, many studies have obtained multi-temporal land cover types through multi-temporal classification methods, and monitored urban expansion, green space loss, and abandoned farmland by analyzing the land cover type trajectory of each pixel. However, it is currently difficult to apply multi-temporal classification methods to dynamic monitoring of urban renewal, mainly because the land cover type conversion rules for urban renewal are unclear. Urban renewal usually includes two stages: demolition and reconstruction. Before demolition and after reconstruction, the land cover type is construction land. However, the land cover types after demolition are diverse, which may be bare land, hardened ground after demolition, and land that becomes overgrown with weeds due to construction delays after demolition. Therefore, the current research simply defines the land cover type conversion rules for urban renewal as "conversion from construction land to bare land", which is obviously incomplete. Urban vacant land can well describe the land cover types after demolition, mainly including bare land, demolished land, land in the early stage of construction, overgrown land, and bare land covered with green film. At present, the semantic segmentation method based on high-resolution remote sensing images can achieve high-precision urban vacant land identification, which provides a research basis for formulating complete land cover type conversion rules for urban renewal. Summary of the invention

[0008] The purpose of the present invention is to provide an urban renewal area identification method by fusing time series remote sensing data with geographic big data, so as to solve the problems raised in the above-mentioned background technology.

[0009] The urban renewal area identification method based on the fusion of time series remote sensing data and geographic big data includes the following steps:

[0010] S1. Study the quantile-based method to integrate POI and NPPNL data to extract the scope of urban built-up areas in the study area in year X;

[0011] S2. By processing multi-source remote sensing data from X-8 to X, urban vacant land, urban green land and urban construction land are identified, and then the annual land cover map within the urban built-up area is obtained;

[0012] S3. Formulate rules for land cover type conversion, monitor urban renewal dynamics, draw urban renewal scope maps, urban renewal demolition time maps, and urban renewal reconstruction time maps.

[0013] As a further improvement of the present invention, the step S1 extracts the urban built-up area scope in the study area in year X to limit the monitoring research scope, and the extraction of the urban built-up area scope in the study area in year X constructs a new built-up area scope identification index DNPOI.

[0014] As a further improvement of the present invention, the acquisition of the DNPOI quantile curve comprises the following steps:

[0015] a1. Arrange the DNPOI values ​​from small to large, obtain the corresponding DNPOI values ​​at different quantile levels, and draw the DNPOI quantile curve with the quantile level as the horizontal axis and the DNPOI value corresponding to the quantile level as the vertical axis. The calculation formula of the DNPOI value is:

[0016] DNPOI=DN×PD

[0017] Among them, DN is the brightness value of NPP-VIIRS, and PD is the normalized value of POI kernel density;

[0018] a2. The urban built-up areas extracted by the quantile method are trimmed by combining visual identification with manual modification to ensure that complete and continuous urban built-up areas can be obtained. The urban built-up areas need to deduct the areas within the basic ecological control line.

[0019] As a further improvement of the present invention, step S2 includes the following steps:

[0020] b1. Automatic identification method of urban vacant land based on semantic segmentation. Process Google Earth HRRS from X-8 to X to obtain urban vacant land in the study area.

[0021] b2. Based on the AWEI water index and NDVI normalized vegetation index, the study extracted the water bodies and vegetation in the study area based on Landsat data, and used ArcGIS to erase the urban vacant land and urban green space within the urban built-up area to obtain urban construction land;

[0022] b3. Integrate urban vacant land, urban green land and urban construction land to obtain urban land cover type data within the urban built-up area from X-8 to X.

[0023] As a further improvement of the present invention, the urban open space comprises:

[0024] Bare land, completely exposed land;

[0025] Weedy land, land overgrown with weeds and bare soil;

[0026] Green film, land covered with green film in the early stage of construction;

[0027] Demolition land, hardened ground after demolition of buildings;

[0028] In the early stage of construction, there is a lot of bare soil in the early stage of building construction.

[0029] As a further improvement of the present invention, the acquisition method of the city is to study 14 training modules selected from different years, which can widely cover urban vacant land and non-vacant land, and maintain a high vacancy rate, thereby reducing model bias and enhancing model robustness. A training module includes training samples and corresponding training labels. The training module is converted to 1.6m spatial resolution and divided into 1694 image blocks (training patches) (256px×256px). Secondly, the currently commonly used image segmentation method DeepLabv3 model is selected for model training, and the image blocks are divided into training samples (75%, 1271) and verification samples (25%, 423). The training samples are used to train the model, and the F2 score is used to judge the accuracy of the training model to obtain the optimal training model. The results show that the accuracy of the verification sample of the optimal training model is F2=0.864. Finally, the Google Earth image of the study area from X-8 to X years is used. HRRS was used as the input data of the optimal training model to identify urban vacant land in the study area. After ArcGIS binarization and raster-to-surface conversion, the raster urban vacant land range (30 m × 30 m) was obtained. If the image quality of a smaller range in a certain year was poor, the recognition result was manually modified based on the images of adjacent time periods.

[0030] As a further improvement of the present invention, the urban green space comprises:

[0031] Water bodies, both inland and artificial;

[0032] vegetation, woodland and cultivated land.

[0033] As a further improvement of the present invention, the urban green space uses AWEIsh and NDVI to extract water bodies and vegetation from year X-8 to year X within the urban built-up area. The thresholds of AWEIsh and NDVI are determined by selecting fixed sample points. The fixed sample points are stable vegetation or water bodies. The extraction results are raster data (30m×30m). If the urban vacant land overlaps with the urban green land, the overlapping area is identified as the urban vacant land, wherein:

[0034] AWEIsh=BLUE+2.5GREEN-1.5(NIR+SWR1)-0.25SWIR2

[0035] NDVI=(NIR-RED) / (NIR+RED)

[0036] Among them, BLUE represents the blue light band, GREEN represents the green light band, NIR represents the near infrared band, SWIR1 and SWIR2 represent the shortwave infrared bands, and RED represents the red light band.

[0037] As a further improvement of the present invention, the urban construction land includes:

[0038] Buildings, including residential, commercial, factory and other buildings;

[0039] Late-stage construction land, where there is no bare soil in the late stage of building construction;

[0040] Hardened surfaces, roads, parking lots, and hardened surfaces with coverings.

[0041] As a further improvement of the present invention, the land cover type conversion rule is the process of converting construction land into urban vacant land, and the reconstruction time is recorded as T1. Since some construction is stagnant for several years, during this period, the urban vacant land may be continuously converted between urban green land and urban vacant land, and the reconstruction time is recorded as T2.

[0042] Before the monitoring termination year, the monitored area is converted from urban vacant land / urban green land to construction land, and the urban renewal completes the complete process of demolition and reconstruction. The reconstruction time is recorded as T3;

[0043] Before the monitoring termination year, the monitoring area was not converted into construction land, and the monitoring area was demolished but not rebuilt, and the reconstruction time was recorded as 9999.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] Based on POI and night light data, the present invention uses the quantile method to extract the scope of the urban built-up area in the end year of the study area, and processes the Google Earth high-resolution remote sensing (HRRS) time series data and the Landsat time series data based on semantic segmentation algorithm, remote sensing index calculation and other methods to generate an annual land cover map within the scope of the urban built-up area. The research results show that the method proposed in this study can accurately classify urban land cover types, further carry out urban renewal monitoring, formulate land cover type conversion rules, and monitor urban renewal dynamics. The research results show that the method of the present invention can more accurately draw urban renewal scope maps (average f2=81.94%), urban renewal demolition time maps (average f2=73.39%), and urban renewal reconstruction time maps (average f2=66.37%), which is of great significance to the academic development of urban renewal and future urban development. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 is a flow chart of the method of the present invention;

[0047] Figure 2 To study the overall roadmap for this invention;

[0048] Figure 3 This is a schematic diagram of the land cover type conversion rules for urban renewal in the present invention;

[0049] Figure 4 A schematic diagram of the inspection accuracy of demolition / reconstruction time for urban renewal;

[0050] Figure 5 This is a statistical diagram of urban renewal monitoring results. DETAILED DESCRIPTION

[0051] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0052] See also Figure 1-5 The present invention provides the following technical solution: a method for identifying urban renewal areas by integrating time series remote sensing data with geographic big data, the method comprising the following steps:

[0053] S1. Study the quantile-based method to integrate POI and NPPNL data to extract the scope of urban built-up areas in the study area in year X;

[0054] S2. By processing multi-source remote sensing data from X-8 to X, urban vacant land, urban green land and urban construction land are identified, and then the annual land cover map within the urban built-up area is obtained;

[0055] S3. Formulate rules for land cover type conversion, monitor urban renewal dynamics, draw urban renewal scope maps, urban renewal demolition time maps, and urban renewal reconstruction time maps.

[0056] Shenzhen is one of China's economic centers, and its international influence is also expanding. It has a total of 10 district-level administrative units. In 1979, Shenzhen's population was only 310,000. In 2019, Shenzhen's permanent population has grown to 13.44 million. At present, the total area of ​​Shenzhen is only 1,997.47 square kilometers, of which 974.5 square kilometers (48.76%) of the land is within the basic ecological control line. In recent years, Shenzhen has faced the contradiction between land shortage and urban development. From 2012 to 2020, urban renewal projects have gradually been carried out in Shenzhen;

[0057] The method of the present invention is used to study Shenzhen. Based on the map software API, 18 types of Point of Interest (POI) data (71,409) in Shenzhen in 2020 are obtained, and the NPP-VIIRSnighttime light (NPPNL) data with a spatial resolution of 500m in 2020 is obtained and cropped to the Shenzhen range. From 2012 to 2020, Google Earth high-resolution remote sensing (HRRS) has three bands: red, green, and blue bands, with a spatial resolution of 1.1m. The acquisition time is December of each year. If the images of certain areas are covered by a large amount of clouds, it is necessary to replace the images of these areas with images that can clearly show objects in adjacent times. The specific method is as follows:

[0058] The urban built-up area (UBA) of the end year (2020) was selected as the research scope. The fusion of night light data and POI data significantly improved the accuracy of urban built-up area extraction. The new built-up area identification index DNPOI was constructed to extract the urban built-up area. The idea of ​​obtaining the DNPOI quantile curve mainly includes two points: first, arrange the DNPOI values ​​from small to large, and obtain the corresponding DNPOI values ​​at different quantile levels; secondly, use the quantile level as the horizontal axis and the DNPOI value corresponding to the quantile level as the vertical axis to draw the DNPOI quantile curve. This study extracts UBA based on the 2020 NPPNL data and POI data;

[0059] DNPOI=DN×PD

[0060] Among them, DN is the brightness value of NPP-VIIRS, and PD is the normalized value of POI kernel density;

[0061] Secondly, the UBA extracted by the quantile method is trimmed by combining visual judgment with manual modification to ensure that a complete and continuous UBA can be obtained. The UBA needs to deduct the area within the basic ecological control line. The basic ecological control line is the area designated by the planning department to prevent the disorderly spread of urban construction. Construction activities are prohibited within this area.

[0062] The urban land cover types are divided into three primary classes and 10 secondary classes. Firstly, the automatic identification method of urban vacant land based on semantic segmentation is studied to process the Google Earth HRRS from 2012 to 2020 to obtain the urban vacant land in the study area. Secondly, the water body and vegetation in the study area are extracted based on Landsat data according to the AWEI water body index and NDVI normalized vegetation index. ArcGIS is used to erase the urban vacant land and urban green land within the UBA to obtain the urban construction land. Finally, the urban vacant land, urban green land and urban construction land are integrated to obtain the urban land cover type data (30m×30m) within the UBA from 2012 to 2020.

[0063] Urban vacant land is divided into five types: bare land, weeds, green film, demolished land, and pre-construction land. The automatic identification method of urban vacant land based on semantic segmentation is mainly divided into three parts: data labeling, model training, and result prediction.

[0064] First, the study selected 14 training tiles (4km×4km) from different years, which can widely cover urban vacant and non-vacant land and maintain a high vacancy rate, thereby reducing model bias and enhancing model robustness. A training tile includes training samples and corresponding training labels. The training tile is converted to 1.6m spatial resolution and segmented into 1694 image patches (256px×256px).

[0065] Secondly, the DeepLabv3 model, a commonly used image segmentation method, was selected for model training. The image blocks were divided into training samples (75%, 1271) and validation samples (25%, 423). The training samples were used to train the model, and the F2 score was used to judge the accuracy of the training model to obtain the optimal training model. The results showed that the accuracy of the validation sample of the optimal training model was F2=0.864.

[0066] Finally, the Google Earth HRRS of the study area from 2012 to 2020 was used as the input data of the optimal training model to identify the urban vacant lots in the study area. After ArcGIS binarization and raster-to-surface conversion, the raster urban vacant lot range (30 m × 30 m) was obtained. If the image quality of a smaller range in a certain year was poor, the recognition results were manually modified according to the images of adjacent times.

[0067] Urban green space includes two types: water bodies and vegetation. The AWEI index can improve the accuracy of water extraction in the presence of various environmental noises, while providing a stable threshold. Among them, the improved AWEIsh index improves the separability of water pixels and non-water pixels (more importantly, shadow pixels), is more suitable for surface water extraction, and has been widely used in water body extraction research. NDVI is widely used in surface cover monitoring, agricultural yield estimation and other research, and is an effective index for extracting urban vegetation. This paper uses AWEIsh and NDVI to extract water bodies and vegetation within the UBA range from 2012 to 2020. The thresholds of AWEIsh and NDVI are determined by selecting fixed sample points. Fixed sample points are stable vegetation or water bodies. The final extraction result is raster data (30m×30m). If urban vacant land overlaps with urban green land, the overlapping area is identified as urban vacant land;

[0068] AWEIsh=BLUE+2.5GREEN-1.5(NIR+SWR1)-0.25SWIR2

[0069] NDVI=(NIR-RED) / (NIR+RED)

[0070] Among them, BLUE represents the blue light band; GREEN represents the green light band; NIR represents the near infrared band; SWIR1 and SWIR2 represent the shortwave infrared band; RED represents the red light band.

[0071] Formulate appropriate conversion rules to convert construction land into urban vacant land, and the reconstruction time is recorded as T1 (e.g. Figure 3 Secondly, due to the stagnation of some construction for several years, during which time, urban vacant land may be constantly converted between urban green space and urban vacant land (such as Figure 3 Finally, if the monitoring area is converted from urban vacant land / urban green land to construction land before the monitoring termination year, then the urban renewal completes the complete process of demolition and reconstruction, and the reconstruction time is recorded as T3. If the monitoring area is not converted into construction land before the monitoring termination year, then the monitoring area completes the demolition but not the reconstruction, and the reconstruction time is recorded as 9999. If an area completes the demolition, regardless of whether it is rebuilt, it is considered to have carried out urban renewal activities. By combining the annual demolition range, the urban renewal range of the study area can be obtained;

[0072] Formulate area constraints, set area constraints for urban renewal in accordance with the provisions of the "Implementation Rules of Shenzhen Urban Renewal Measures", and set the minimum annual demolition area to 5,000m 2 The minimum area of ​​each urban renewal area is 10,000m 2 ,According to the above land cover type conversion rules and area constraints, ,the study monitors the urban renewal situation within the UBA from 2012 to 2020, ,and draws the urban renewal scope map, the urban renewal demolition ,time map, and the urban renewal reconstruction time map.

[0073] The study evaluated the accuracy of the annual urban land cover type map and urban renewal map, and selected the validation samples based on the following principles:

[0074] Evenly distributed within the UBA range;

[0075] Wide coverage of various land cover types and multiple urban renewal areas;

[0076] (3) Ensure sufficient number of validations. A total of 11 3 km × 3 km validation areas were selected for this study, containing a total of 99,276 grids (30 m × 30 m);

[0077] Ten experts with experience in remote sensing data analysis were invited to visually discriminate the annual Google Earth HRRS in the validation area, marking the real urban land cover type, urban renewal extent, demolition time, and reconstruction time, following the same sample labeling criteria. When a decision was difficult to make, a second opinion was obtained and the samples were co-labeled;

[0078] Using the confusion matrix, four parameters including producer accuracy (PA), user accuracy (UA), overall accuracy (OA) and Kappa coefficient (Kappa) are constructed to evaluate the recognition accuracy of urban land cover types. The F2 score (F2=5UA×PA / (4UA+PA)) can comprehensively reflect PA and UA, and better reflect the reliability of the urban renewal monitoring method. Therefore, this study uses F2 as an indicator to evaluate the results of urban renewal mapping.

[0079] According to the method of the present invention, the spatial distribution of DNPOI has a spatial resolution of 500m×500m. According to the DNPOI quantile curve, two threshold screenings were performed, and the results are as follows:

[0080] In the first screening, the quantile inflection point appeared at DNPOI=8.23. The areas with DNPOI>8.23 were mainly concentrated in Futian District and Luohu District, which are the economically developed areas in Shenzhen.

[0081] After deducting the area with DNPOI>8.23 for the second screening, the quantile inflection point appears at DNPOI=1.50. The area with DNPOI<1.50 is mostly covered by green vegetation. Based on the range of DNPOI>1.50, the UBA is visually interpreted and manually corrected according to Google images to make it continuous. Finally, after deducting the ecological control area, the UBA is 81680.33hm 2 Among them, Longgang District has the largest UBA area (20268.42hm 2 ), Baoan District is second (17606.38hm 2 ), Dapeng New District has the smallest UBA area (1547.574hm 2 );

[0082] The urban land cover type map within the UBA from 2012 to 2020 was drawn and the accuracy of the result was verified. The verification results show that:

[0083] During the period from 2012 to 2020, the average PA and UA of urban vacant land were 77.21% and 80.29%, the average PA and UA of urban green land were 83.69% and 84.55%, and the average PA and UA of urban construction land were 96.41% and 96.14%;

[0084] During the period from 2012 to 2020, the average OA of urban land cover types was 93.68%, and the average kappa coefficient was 80.69%. The results show that the proposed multi-temporal land cover type mapping method can ensure the accuracy and reliability of subsequent change analysis;

[0085] The urban renewal scope from 2012 to 2020 was extracted, and the demolition time and reconstruction time of urban renewal were obtained;

[0086] First, the accuracy of the urban renewal range was tested. By visually distinguishing the urban renewal range, it was found that the UA of the entire area was 0.83 (True Positive (TP) = 2537.42hm 2 , false positive (FP) = 506hm 2 ), since the real urban renewal data of the entire area is difficult to obtain, it is impossible to obtain the PA and F2 scores of the entire area. In addition, the real urban renewal scope of the 11 verification areas is obtained. The research results show that the accuracy of urban renewal scope mapping in the 11 verification areas is F2=81.94%;

[0087] Secondly, the results of urban renewal demolition time and urban renewal reconstruction time were tested for accuracy, and the actual urban renewal demolition / reconstruction time of the 11 verification areas was obtained. In the 11 verification areas, the average F2 of demolition time from 2013 to 2020 was 73.39%; the average F2 of reconstruction time from 2013 to 2020 was 66.37% (e.g. Figure 4 The results show that the proposed method can better monitor the dynamics of urban renewal.

[0088] According to the results of urban renewal demolition mapping, the total urban renewal area within the entire UBA is 3466.08hm 2 Longgang District has the largest urban renewal area (792.77hm 2 ), followed by Baoan District (773.22 hm 2 ), Dapeng New District has the smallest urban renewal area (17.73hm 2 )(like Figure 5 As shown in the table above, from the results of the urban renewal demolition / reconstruction time mapping, the demolition area and reconstruction area in 2019 were the largest, at 571.77hm 2 , 513.63hm 2 (like Figure 5 As shown in Figure 2, the area that has not been rebuilt by 2020 (9999) is 1216.44hm 2 Among them, Dapeng New District, Nanshan District, Guangming District and Longgang District mainly focus on industrial transformation, and their industrial transformation areas account for 47.74%, 35.73%, 41.91% and 23.13% of the urban renewal scope of each district respectively. Industrial transformation is mainly because some high-pollution, high-energy-consuming and low-efficiency industrial land is difficult to form a benign interactive and compatible relationship with residential public facilities and service land, which in turn leads to various environmental problems. Yantian District (32.26%), Luohu District (29.59%), Futian District (29.51%), Pingshan District (39.53%), Longhua District (39.04%) and Bao'an District (27.98%) are mainly based on residential transformation. The residential transformation areas are characterized by high population density, low facility abundance, and a large number of urban villages, which leads to a decrease in the quality of production and life of residents. Commercial transformation is mainly concentrated in the economically highly developed areas of Futian District (18.73%) and Luohu District (13.29%).

[0089] In the first step of the present invention, a quantile-based method is studied to integrate POI and NPPNL data to extract the scope of the urban built-up area (UBA) in the study area in 2020. In the second step, the multi-source remote sensing data from 2012 to 2020 are processed to identify urban vacant land, urban green space and urban construction land, and then the annual land cover map within the UBA is obtained. In the third step, land cover type conversion rules are formulated, and the dynamics of urban renewal are monitored to draw urban renewal scope maps, urban renewal demolition time maps, urban renewal reconstruction time maps, and carry out accuracy verification and analysis (such as Figure 2 As shown in the figure), based on POI and night light data, the quantile method was used to extract the urban built-up area scope of the study area in the termination year. Based on semantic segmentation algorithm, remote sensing index calculation and other methods, Google Earth high-resolution remote sensing (HRRS) time series data and Landsat time series data were processed to generate an annual land cover map within the urban built-up area. The research results show that the method proposed in this study can accurately classify urban land cover types, further carry out urban renewal monitoring, formulate land cover type conversion rules, and monitor urban renewal dynamics. The research results show that the method of the present invention can more accurately draw urban renewal scope maps (average f2=81.94%), urban renewal demolition time maps (average f2=73.39%), and urban renewal reconstruction time maps (average f2=66.37%), which is of great significance to the academic development of urban renewal and future urban development.

[0090] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0091] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for identifying urban renewal areas by integrating time series remote sensing data with geographic big data, characterized by: The method comprises the following steps: S1. Study the quantile-based method to integrate POI and NPPNL data to extract the scope of urban built-up areas in the study area in year X; S2. By processing multi-source remote sensing data from X-8 to X, urban vacant land, urban green land and urban construction land are identified, thereby obtaining an annual land cover map within the scope of the urban built-up area; the urban vacant land includes bare land, weeds, green film, demolished land and pre-construction land, and the urban green land includes water bodies and vegetation; S3. Formulate land cover type conversion rules, monitor urban renewal dynamics, draw urban renewal scope maps, urban renewal demolition time maps, and urban renewal reconstruction time maps; The step S2 comprises the following steps: b1. The automatic identification method of urban vacant land based on semantic segmentation processes the HRRS data from X-8 to X to obtain the urban vacant land in the study area; b2. Based on the AWEI water index and NDVI normalized vegetation index, the study extracted the water bodies and vegetation in the study area based on Landsat data, and used ArcGIS to erase the urban vacant land and urban green space within the urban built-up area to obtain urban construction land; b3. Integrate urban vacant land, urban green land and urban construction land to obtain urban land cover type data within the urban built-up area from X-8 to X; The acquisition method of the city is to select 14 training modules from different years for research. A training module includes training samples and corresponding training labels. The image segmentation method DeepLabv3 model is selected for model training. The HRRS of the study area from year X-8 to year X is used as the input data of the optimal training model to identify the urban vacant land in the study area. After ArcGIS processing, the range of urban vacant land in 30m×30m grid is obtained; The urban green space uses AWEIsh and NDVI to extract water bodies and vegetation from year X-8 to year X within the urban built-up area. The thresholds of AWEIsh and NDVI are determined by selecting fixed sample points. The fixed sample points are stable vegetation or water bodies. The extraction results are 30m×30m grid data. If the urban vacant land overlaps with the urban green land, the overlapping area is identified as the urban vacant land, where: ; ; Among them, BLUE represents the blue light band, GREEN represents the green light band, NIR represents the near infrared band, SWIR1 and SWIR2 represent the short-wave infrared band, and RED represents the red light band; In step S3, the land cover type conversion rule is the process of converting construction land into urban vacant land, and the reconstruction time is recorded as T1. Since some construction has been stagnant for several years, during this period, the urban vacant land may be continuously converted between urban green land and urban vacant land, and the reconstruction time is recorded as T2; Before the monitoring termination year, the monitored area is converted from urban vacant land / urban green land to construction land, and the urban renewal completes the entire process of demolition and reconstruction, and the reconstruction time is recorded as T3.

2. The urban renewal area identification method based on the fusion of time series remote sensing data and geographic big data according to claim 1 is characterized by: In step S1, the scope of the urban built-up area in the study area in year X is extracted to limit the research scope of monitoring. The scope of the urban built-up area in the study area in year X is extracted to construct a new built-up area scope identification index DNPOI. The calculation formula of the DNPOI value is: ; Among them, DN is the brightness value of NPP-VIIRS, and PD is the normalized value of POI kernel density.

3. The urban renewal area identification method based on the fusion of time series remote sensing data and geographic big data according to claim 2 is characterized by: The acquisition of the DNPOI quantile curve comprises the following steps: a1. Arrange the DNPOI values ​​from small to large, obtain the corresponding DNPOI values ​​at different quantile levels, and draw the DNPOI quantile curve with the quantile level as the horizontal axis and the DNPOI value corresponding to the quantile level as the vertical axis; a2. The urban built-up areas extracted by the quantile method are trimmed by combining visual identification with manual modification to ensure that complete and continuous urban built-up areas can be obtained. The urban built-up areas need to deduct the areas within the basic ecological control line.

4. The urban renewal area identification method based on the fusion of time series remote sensing data and geographic big data according to claim 1 is characterized by: The urban construction land includes: Buildings, including residential, commercial and factory buildings; Late-stage construction land, where there is no bare soil in the late stage of building construction; Hardened surfaces, roads, parking lots, and hardened surfaces with coverings.

5. The urban renewal area identification method based on the fusion of time series remote sensing data and geographic big data according to claim 1 is characterized by: Before the monitoring termination year, the monitoring area was not converted into construction land, and the monitoring area was demolished but not rebuilt, and the reconstruction time was recorded as 9999.

Images