A Remote Sensing Extraction Method for Abandoned Cultivated Lands Based on the Change Characteristics of Time Series

Abstract

The present invention discloses a method for remotely sensing and extracting abandoned farmland based on the change characteristics of time series, including: performing data preprocessing on long-time series remote sensing image data; based on the preprocessed long-time series remote sensing image data, statistically analyzing the threshold intervals of various vegetation indices and calculating the annual summary indicators of various vegetation indices; based on the annual summary indicators of various vegetation indices, establishing extraction rules for the conversion of cultivated land into abandoned farmland; based on historical and current high-resolution images, establishing a sample library of cultivated land and abandoned farmland, inputting the index values and operation rules extracted from the long-time series remote sensing data of the sample library into a random forest classifier for training, and performing remote sensing extraction of abandoned farmland in the study area based on the training results. The present invention aims at the problems of remote sensing monitoring errors caused by the confusion of the spectral characteristics of abandoned farmland with other ground objects and the spectral differences at different stages of cultivated land abandonment, studies a remote sensing extraction strategy for abandoned farmland based on the time series difference characteristics of vegetation indices, and realizes precise monitoring of large-area abandoned farmland.

Description

Technical Field

[0001] The present invention relates to the field of geographic information and remote sensing, and particularly to a method for remotely sensing and extracting abandoned farmland based on time-series change characteristics. Background Art

[0002] For efficient, accurate, and large-scale monitoring of abandoned farmland, it is urgent to combine spatial information with social information and establish a hierarchical scientific planning scheme according to local conditions for the overall utilization of abandoned farmland. The ground survey method consumes a large amount of manpower and material resources and is not suitable for large-scale surveys. Remote sensing means are mostly used for large-area monitoring of abandoned farmland, mainly using the vegetation spectral differences between abandoned farmland and other ground objects for pixel-based or object-oriented classification or change detection. Due to the confusion of abandoned farmland, traditional remote sensing monitoring methods have large misjudgments. With the continuous maturity of machine learning algorithms, classification algorithms such as random forest and support vector machine have gradually been applied to the monitoring of abandoned farmland. However, on the one hand, there are confusion problems between abandoned farmland and land use categories such as grassland, bare land, shrubs, and low-intensity cultivated land; on the other hand, different spectral characteristics are presented at each stage from semi-abandoned to completely abandoned, resulting in inconsistent remote sensing classification standards. All of the above cause errors in remote sensing monitoring of abandoned farmland. Summary of the Invention

[0003] (I) Technical Problems to be Solved

[0004] In view of this, the main object of the present invention is to provide a method for remotely sensing and extracting abandoned farmland based on time-series change characteristics to achieve accurate monitoring of large-area abandoned farmland.

[0005] (II) Technical Solutions

[0006] To achieve the above object, the present invention provides a method for remotely sensing and extracting abandoned farmland based on time-series change characteristics, and the method includes:

[0007] Step 1: Perform data preprocessing on long-time-series remote sensing image data;

[0008] Step 2: Based on the preprocessed long-time-series remote sensing image data, statistically analyze the threshold intervals of each vegetation index and calculate the annual summary indicators of each vegetation index;

[0009] Step 3: Based on the annual summary indicators of each vegetation index, establish extraction rules for the conversion of cultivated land into abandoned farmland;

[0010] Step 4: Based on historical and current high-resolution images, establish a sample library of cultivated land and abandoned farmland, input the index values and operation rules extracted from the long-time-series remote sensing data of the sample library into a random forest classifier for training, and perform remote sensing extraction of abandoned farmland in the study area based on the training results to achieve remote sensing extraction of abandoned farmland based on time-series change characteristics.

[0011] In the above solution, step 1 includes: using the Google Earth Engine cloud platform, and screening the long-term time series remote sensing image data by using the image quality parameter Cloud attached to the satellite to exclude cloud, cloud shadow and snow / ice images.

[0012] In the above solution, when performing image screening, the screening index is to retain the images with the cloud amount area less than 30% in the current period. A total of 39 regions are selected, with 3 regions selected for each of the 13 major grain-producing areas in the country, and processed to obtain at least 10 images per year for all research regions from 1985 to 2021.

[0013] In the above solution, step 2 includes: based on the preprocessed long-term time series remote sensing image data, statistically analyzing the threshold intervals of the vegetation indices of the four types of land cover, namely abandoned cultivated land, cultivated land, natural forest land, and bare land, and calculating the annual summary indicators of the vegetation indices.

[0014] In the above solution, the vegetation indices of the four types of land cover, namely abandoned cultivated land, cultivated land, natural forest land, and bare land, include: Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), Bare Land Index (SI), as well as Built-up Index (IBI), Dryness Index (NDBSI) and Wetness Index (WET); based on the preprocessed long-term time series remote sensing image data, statistically analyzing the threshold intervals of the vegetation indices of the four types of land cover, namely abandoned cultivated land, cultivated land, natural forest land, and bare land, is to use the band operation tool provided by Google Earth Engine to perform the following calculations:

[0015] (1)

[0016] EVI = 2.5 * ((NIR - RED) / (NIR + 6 * RED - 7.5 * BLUE + 1)) (2)

[0017] SI = (RED+MIR)-(RED+MIR) / (RED+MIR)+(RED+MIR) (3)

[0018] IBI = (((2*MIR) / (NIR+MIR))-((NIR / (RED+NIR))+(GREEN / (GREEN+MIR)))) /

[0019] ((((2*MIR) / (NIR+MIR))+((NIR / (RED+NIR))+(GREEN / (GREEN+MIR)))) (4)

[0020] NDBSI=(IBI+SI) / 2 (5)

[0021] WET = C1*BLUE + C2*GREEN + C3*RED + C4*NIR + C5*MIR1 + C6*MIR2 (6)

[0022] Among them, for the TM sensor, C1 - C6 are 0.0315, 0.2021, 0.3012, 0.1594, -0.6806, -0.6109 respectively; for the OLI sensor, C1 - C6 are 0.1511, 0.1973, 0.3283, 0.3407, -0.7117, -0.4559 respectively.

[0023] In the above solution, the inter - annual summary index for calculating each vegetation index is to calculate the inter - annual values and time nodes of the maximum value, minimum value, average value, median value, and 20% and 80% quantiles of each vegetation index, and generate layers respectively.

[0024] In the above solution, the inter - annual summary index of each vegetation index is used as an input variable for classification.

[0025] In the above solution, the extraction rule for the conversion of cultivated land to abandoned land in step 3 is as follows:

[0026] When the numerical difference of the annual summary index exceeds 30% of the previous year, and the time - node difference of the numerical value exceeds 1 month, it is considered that the index has changed;

[0027] Among the six - year summary indexes of the same vegetation index, if more than three of the indexes have changed, it is considered that the annual periodicity of the vegetation index has changed;

[0028] If more than three vegetation indexes have changed, it is considered that the vegetation growth periodicity of the pixel has changed, that is, it is considered to have changed from cultivated land to abandoned land.

[0029] In the above solution, in step 4, the remote - sensing extraction of abandoned land in the study area based on the training result, the extraction result includes the spatial distribution of abandoned - land plots and the conversion time, and the rotation - fallow cultivated land and the already - reused abandoned land are excluded based on the conversion time.

[0030] In the above solution, after the method realizes the remote - sensing extraction of abandoned land based on the time - series change characteristics, it further includes: collecting the ground - survey data of the main study area, and using the confusion - matrix method to give the classification accuracy of abandoned land.

[0031] (III) Beneficial effects

[0032] As can be seen from the above technical solutions, the remote - sensing extraction method of abandoned land based on time - series change characteristics provided by the present invention has the following beneficial effects:

[0033] 1. Aiming at the problem of remote sensing monitoring errors caused by the confusion of the spectral characteristics of abandoned farmland with other ground objects and the spectral differences in different stages of cultivated land abandonment, the method for remotely extracting abandoned farmland based on the time series change characteristics provided by the present invention studies the strategy for remotely extracting abandoned farmland based on the time series difference characteristics of vegetation indices. By making full use of the characteristics that the vegetation spectrum changes from periodic to aperiodic, it extracts abandoned farmland, realizes the high-precision extraction of abandoned farmland, and further realizes the precise monitoring of large-area abandoned farmland, providing data support for the management and scientific prevention of cultivated land abandonment.

[0034] 2. At present, there is no nationwide spatio-temporal monitoring result of abandoned farmland in China. The ground survey method consumes a large amount of manpower and material resources and is not suitable for large-scale surveys. Due to the confusion of abandoned farmland, traditional remote sensing monitoring methods have large misjudgments. By using the time series characteristics that the vegetation growth of abandoned farmland changes from periodic to aperiodic, the present invention has invented a method for remotely monitoring large-area abandoned farmland, monitored the abandoned farmland in 13 major grain-producing areas in China, obtained the spatio-temporal pattern of abandoned farmland since 2000 in 13 major grain-producing areas in the country, realized the precise monitoring of large-area abandoned farmland, and provided data support and decision-making support for finding out the stock of abandoned farmland in each region and scientifically planning and utilizing abandoned farmland. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a flowchart of the method for remotely extracting abandoned farmland based on the time series change characteristics according to an embodiment of the present invention;

[0036] Figure 2 It is a comparison diagram of the surface forms of abandoned farmland and cultivated land according to an embodiment of the present invention;

[0037] Figure 3 It is a vegetation index curve of abandoned farmland, cultivated land, and natural forest land in a long time series according to an embodiment of the present invention;

[0038] Figure 4 It is a roadmap of the extraction rule for cultivated land turning into abandoned farmland according to an embodiment of the present invention;

[0039] Figure 5 It is an interface diagram of the raster calculator tool in ArcGIS software according to an embodiment of the present invention;

[0040] Figure 6 It is a spatial distribution map of abandoned farmland determined by the method for remotely extracting abandoned farmland based on the time series change characteristics according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0041] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present disclosure. In the following detailed description, for the purpose of explanation, numerous specific details are set forth in order to provide a comprehensive understanding of the embodiments of the present disclosure. However, it is obvious that one or more embodiments can be implemented without these specific details. In addition, in the following description, descriptions of well-known structures and technologies are omitted to avoid unnecessarily confusing the concepts of the present disclosure.

[0042] When cultivated land is converted into abandoned land, the most prominent feature reflected in remote sensing information is the transition of the vegetation spectrum from the periodic characteristics of cultivated land to the aperiodic characteristics of bare land or natural vegetation. Based on the 35-year long-term medium-resolution image data of the remote sensing cloud platform and the strategy for remotely sensing the extraction of abandoned land based on the time series difference characteristics of vegetation indices, the present invention quantifies the characteristics of the change of vegetation indices from periodic characteristics to aperiodic characteristics, and uses the "big data + cloud platform" model to accurately monitor abandoned land, providing a method for remotely sensing the extraction of abandoned land based on the time series change characteristics.

[0043] As Figure 1 shown, Figure 1 FIG. is a flowchart of a method for remotely sensing the extraction of abandoned land based on the time series change characteristics according to an embodiment of the present invention. The method includes the following steps:

[0044] Step 1: Perform data preprocessing on the long-term time series remote sensing image data;

[0045] The characteristic that the vegetation index of cultivated land changes from periodic to aperiodic of abandoned land, Figure 2 FIG. is a comparison chart of the surface morphologies of abandoned land and cultivated land according to an embodiment of the present invention, Figure 3 FIG. is a vegetation index curve of abandoned land, cultivated land, and natural forest land in the long time series according to an embodiment of the present invention.

[0046] In the embodiment of the present invention, the reflectance data of the medium-resolution remote sensing satellite Landsat series satellites (TM and OLI) from 1985 to 2021 are used as the long-term time series remote sensing image data.

[0047] When performing data preprocessing on the long-term time series remote sensing image data, using the Google Earth Engine cloud platform, the long-term time series remote sensing image data is screened by using the image quality parameter Cloud attached to the satellite to exclude cloud, cloud shadow, and snow / ice images. The screening index is to retain the images with the cloud area less than 30% in the current period. A total of 39 regions are selected from 3 regions in each of the 13 major grain-producing areas in the country for processing, and at least 10 images are obtained for each research region from 1985 to 2021 every year.

[0048] Step 2: Based on the preprocessed long - time - series remote - sensing image data, statistically analyze the threshold intervals of various vegetation indices for four types of land cover, namely abandoned cultivated land, cultivated land, natural forest land, and bare land, and calculate the annual summary indicators of various vegetation indices.

[0049] Among them, the various vegetation indices for the four types of land cover, namely abandoned cultivated land, cultivated land, natural forest land, and bare land, include: Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), Bare - land Index (SI), as well as Built - up Index (IBI), Dryness Index (NDBSI), and Wetness Index (WET);

[0050] Based on the preprocessed long - time - series remote - sensing image data, statistically analyze the threshold intervals of various vegetation indices for the four types of land cover, namely abandoned cultivated land, cultivated land, natural forest land, and bare land, by using the band operation tool provided by Google Earth Engine for the following calculations:

[0051] (1)

[0052] EVI = 2.5 * ((NIR - RED) / (NIR + 6 * RED - 7.5 * BLUE + 1)) (2)

[0053] SI = (RED+MIR)-(RED+MIR) / (RED+MIR)+(RED+MIR) (3)

[0054] IBI = (((2*MIR) / (NIR+MIR))-((NIR / (RED+NIR))+(GREEN / (GREEN+MIR)))) /

[0055] (((2*MIR) / (NIR+MIR))+((NIR / (RED+NIR))+(GREEN / (GREEN+MIR)))) (4)

[0056] NDBSI=(IBI+SI) / 2 (5)

[0057] WET = C1*BLUE+C2*GREEN+C3*RED+C4*NIR+C5*MIR1+C6*MIR2 (6)

[0058] Among them, for the TM sensor, C1 - C6 are 0.0315, 0.2021, 0.3012, 0.1594, - 0.6806, - 0.6109 respectively; for the OLI sensor, C1 - C6 are 0.1511, 0.1973, 0.3283, 0.3407, - 0.7117, - 0.4559 respectively.

[0059] Calculate the annual summary indicators of each vegetation index, which are the annual values and time nodes of the maximum, minimum, average, median, and 20% and 80% quantiles of each vegetation index among the vegetation indices, and generate layers respectively.

[0060] In this step, the annual summary indicators of each vegetation index are used as input variables for classification.

[0061] Step 3: Based on the annual summary indicators of each vegetation index, establish the extraction rules for the conversion of cultivated land into abandoned land.

[0062] The extraction rule for the conversion of cultivated land into abandoned land is that when the numerical difference of the annual summary indicators exceeds 30% of the previous year, and the time node difference of the values exceeds more than 1 month, it is considered that the indicator has changed. Among the six annual summary indicators of the same vegetation index, if more than three of the indicators have changed, it is considered that the annual periodicity of the vegetation index has changed; if more than three vegetation indices have changed, it is considered that the vegetation growth periodicity of the pixel has changed, that is, it is considered that the cultivated land has been converted into abandoned land.

[0063] Figure 4 It is the roadmap for the extraction rules of the conversion of cultivated land into abandoned land according to the embodiments of the present invention. The tool used is the raster calculator tool of ArcGIS software, as Figure 5 shown, Figure 5 It is the interface diagram of the raster calculator tool of ArcGIS software according to the embodiments of the present invention.

[0064] Step 4: Based on historical and current high-resolution images, establish a sample library of cultivated land and abandoned land. Input the index values and operation rules extracted from the long-time series remote sensing data of the sample library into a random forest classifier for training, and conduct remote sensing extraction of abandoned land in the study area based on the training results to achieve remote sensing extraction of abandoned land based on the time series change characteristics.

[0065] In this step, based on historical and current high-resolution images, establish a sample library of cultivated land and abandoned land. Input the index values and operation rules extracted from the long-time series remote sensing data of the sample library into a random forest classifier for training, and conduct remote sensing extraction of abandoned land in the study area based on the training results; the extraction results include the spatial distribution and conversion time (unit: year) of the abandoned land plots, as Figure 6 shown, which is the extraction result map of abandoned land, that is, the spatial distribution map of abandoned land determined by the remote sensing extraction method of abandoned land based on the time series change characteristics according to the embodiments of the present invention. Exclude fallow cultivated land and reclaimed abandoned land based on the conversion time.

[0066] Step 5: Ground verification.

[0067] After implementing the remote sensing extraction of fallow land based on the time series change characteristics, the method for remote sensing extraction of fallow land based on the time series change characteristics provided by the present invention further includes the following steps: collecting the ground survey data of the main research area, and giving the classification accuracy of fallow land by using the confusion matrix method.

[0068] As can be seen from the above embodiments, compared with the existing automatic extraction scheme for fallow land, the method for remote sensing extraction of fallow land based on the time series change characteristics provided by the present invention uses detailed remote sensing big data, adopts the Landsat data with the longest time series as the basic data for fallow land identification, and realizes the precise monitoring of fallow land in large areas, which is an example of the application of remote sensing big data in production practice.

[0069] In addition, the method for remote sensing extraction of fallow land based on the time series change characteristics provided by the present invention is applicable to the selection of planting areas for different crops. It only needs to change the candidate area screening conditions according to the growth conditions of crops. For example, the irrigation convenience degree and terrain conditions can be comprehensively considered to obtain a more scientific and accurate candidate area for rice.

[0070] The above specific embodiments have further detailed the purpose, technical solution and beneficial effects of the present invention. It should be understood that the above are only specific embodiments of the present invention and are not used to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for remotely sensing and extracting abandoned farmland based on the change characteristics of time series, characterized in that, The method includes: Step 1: Perform data preprocessing on long - time - series remote sensing image data; Step 2: Based on the preprocessed long - time - series remote sensing image data, statistically analyze the threshold intervals of each vegetation index and calculate the annual summary indicators of each vegetation index; Step 3: Based on the annual summary indicators of each vegetation index, establish extraction rules for the conversion of cultivated land into abandoned land; Step 4: Based on historical and current high - resolution images, establish a sample library of cultivated land and abandoned land. Input the index values and operation rules extracted from the long - time - series remote sensing data of the sample library into a random forest classifier for training, and perform remote sensing extraction of abandoned land in the study area based on the training results to achieve remote sensing extraction of abandoned land based on time - series change characteristics; Among them, Step 2 includes: Based on the preprocessed long - time - series remote sensing image data, statistically analyze the threshold intervals of each vegetation index of four land covers, namely abandoned cultivated land, cultivated land, natural forest land, and bare land, and calculate the annual summary indicators of each vegetation index; The vegetation indices of the four land covers of abandoned cultivated land, cultivated land, natural forest land, and bare land include: Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), Bare Land Index (SI), Built - up Index (IBI), Dryness Index (NDBSI), and Wetness Index (WET); The statistical analysis of the threshold intervals of each vegetation index of the four land covers of abandoned cultivated land, cultivated land, natural forest land, and bare land based on the preprocessed long - time - series remote sensing image data is calculated as follows using the band operation tool provided by Google Earth Engine: （1） EVI = 2.5 * ((NIR - RED) / (NIR + 6 * RED - 7.5 * BLUE + 1)) (2) SI = (RED+MIR)-(RED+MIR) / (RED+MIR)+(RED+MIR) (3) IBI = (((2*MIR) / (NIR+MIR))-((NIR / (RED+NIR))+(GREEN / (GREEN+MIR)))) / (((2*MIR) / (NIR+MIR))+((NIR / (RED+NIR))+(GREEN / (GREEN+MIR))))(4) NDBSI=(IBI+SI) / 2 (5) WET = C1*BLUE+C2*GREEN+C3*RED+C4*NIR+C5*MIR1+C6*MIR2(6) Among them, for the TM sensor, C1 - C6 are 0.0315, 0.2021, 0.3012, 0.1594, - 0.6806, - 0.6109 respectively; for the OLI sensor, C1 - C6 are 0.1511, 0.1973, 0.3283, 0.3407, - 0.7117, - 0.4559 respectively; The extraction rules for the conversion of cultivated land into abandoned land described in Step 3 are: If the numerical difference of the annual summary indicators exceeds 30% of the previous year and the time node difference of the numerical values exceeds 1 month, it is considered that the annual summary indicators have changed; Among the six annual summary indicators of the same vegetation index, if more than three indicators have changed, it is considered that the annual periodicity of the vegetation index has changed; If more than three vegetation indices have occurred, it is considered that the vegetation growth periodicity of the remote sensing data pixel has changed, that is, it is considered that the cultivated land has been converted into abandoned land.

2. The method for remotely sensing and extracting abandoned farmland based on time series change features according to claim 1, wherein Step 1 includes: Using the Google Earth Engine cloud platform, the Cloud of the image quality parameters attached to the satellite is used to screen the long-term time series remote sensing image data to exclude cloud, cloud shadow and snow / ice images.

3. The method for remotely sensing and extracting abandoned farmland based on the time series change characteristics according to claim 2, wherein When performing image screening, the screening index is to retain the images with the cloud amount area less than 30% in the current period. A total of 39 regions are selected from 3 regions in each of the 13 major grain-producing areas in the country for processing, and at least 10 images per year in all research regions from 1985 to 2021 are obtained.

4. The method for remotely sensing and extracting abandoned farmland based on time series change characteristics according to claim 1, characterized in that, Calculating the inter-annual summary indicators of each vegetation index is to calculate the inter-annual numerical values and time nodes of the maximum value, minimum value, average value, median value, and 20% and 80% quantiles of each vegetation index, and generate layers respectively.

5. The method for remotely sensing and extracting abandoned farmland based on the time series change characteristics according to claim 4, characterized in that, The inter-annual summary indicators of each vegetation index are used as the input variables for classification.

6. The method for remotely sensing and extracting abandoned farmland based on the time series change characteristics according to claim 1, wherein In step 4, the remote sensing extraction of abandoned land in the study area based on the training results, and the extraction results include the spatial distribution of abandoned land plots and the conversion time. The rotation cultivated land and the reused abandoned land are excluded based on the conversion time.

7. The method for remotely sensing and extracting abandoned farmland based on time series change features according to claim 1, wherein After the method realizes the remote sensing extraction of abandoned land based on the time series change characteristics, it further includes: Collecting the ground survey data of the main study area and using the confusion matrix method to give the classification accuracy of abandoned land.

Images