A remote sensing method for extracting information on abandoned farmland in mountainous and hilly areas

Abstract

This invention relates to the field of remote sensing technology, specifically to a remote sensing method for extracting information on abandoned arable land in mountainous and hilly areas. The method includes: collecting data and performing union and masking processing to form potential arable land extent data; detecting and processing the potential arable land extent data based on a band-time series change detection algorithm; classifying arable land use data into different change stages; calculating the normalized vegetation index (NVI) to classify arable land use change type data; extracting information on abandoned arable land in mountainous and hilly areas based on the arable land use change type data; and verifying accuracy and effectiveness. This invention, through union and masking processing of arable land data, extracts arable land use data into different change stages based on a band-time series change detection algorithm, and uses the NVI to classify arable land use change type data, extracting information on abandoned arable land and verifying effectiveness. The method is simple to operate, highly efficient, and accurate, which is beneficial for the detection and control of abandoned arable land.

Description

Technical Field

[0001] This invention relates to the field of remote sensing technology, and in particular to a remote sensing method for extracting information on abandoned farmland in mountainous and hilly areas. Background Technology

[0002] Remote sensing technology has significant advantages in acquiring large-scale information on abandoned farmland, better reflecting its spatiotemporal evolution and becoming the mainstream method for such information acquisition. However, this method also has certain drawbacks: 1. The temporal consistency of land cover classification results is difficult to guarantee. Image noise and classification algorithms lead to uncertainty in the classification results for each period. This uncertainty is amplified in the land cover change trajectory, making it difficult to determine whether farmland abandonment is real or due to misclassification in a particular image. 2. The remote sensing image classification process involves considerable human intervention, and the classification results are largely affected by human interference. However, commonly used supervised classification algorithms rely on high-quality labeled samples to train the classification model to obtain reliable results. Furthermore, intra-year image synthesis strategies blur the time of land cover type transitions, losing valuable phenological information for distinguishing vegetation types, resulting in classification results that do not conform to the evolutionary patterns of abandoned farmland. Therefore, extracting information on abandoned farmland in mountainous and hilly areas presents significant challenges. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of existing technologies. This invention addresses the problems of inconsistent temporal consistency in land cover classification results, human interference with remote sensing image classification results, and loss of phenological information due to intra-year image synthesis strategies, leading to classification results that do not conform to the evolutionary patterns of abandoned farmland, in the current application of remote sensing technology for acquiring information on abandoned farmland. It provides a remote sensing extraction method for abandoned farmland information in mountainous and hilly areas. This method uses union and masking processing of farmland data, extracts farmland use data at different stages of change based on a band-time series change detection algorithm, and uses the normalized vegetation index to classify farmland use change types. Abandoned farmland information is extracted and its effectiveness is verified. The method is simple to operate, highly efficient, and accurate, which is beneficial for the detection and control of abandoned farmland.

[0004] This invention proposes a remote sensing method for extracting information on abandoned farmland in mountainous and hilly areas, comprising:

[0005] Collect multi-source land use data and cultivated land data within the study area, and perform union and masking processing on the multi-source land use data and cultivated land data to form potential cultivated land range data;

[0006] The potential arable land area data is detected and processed based on the band time series change detection algorithm to obtain the band fitting harmonic function fitting result of the potential arable land area data;

[0007] Based on the band fitting harmonic function fitting results of the potential arable land range data, calculate the arable land use change breakpoints of the potential arable land range data, and divide the arable land use data of different change stages based on the arable land use change breakpoints.

[0008] The normalized vegetation index (NDI) of the remote sensing images corresponding to the farmland use data at different stages of change is calculated. The NDI is then processed based on the band time series change detection algorithm to obtain the band fitting harmonic function fitting result of the NDI. Based on the band fitting harmonic function fitting result of the NDI, the farmland use change type data is classified.

[0009] Based on the data on changes in cultivated land use, information on abandoned cultivated land in mountainous and hilly areas is extracted by combining data on the evolution of abandoned cultivated land with data on vegetation phenological characteristics.

[0010] The multi-source land use data and cultivated land data collected within the study area include:

[0011] Based on the China Life Cycle Basic Database, the Global 30-meter Land Cover Dataset, the Global High-Resolution Land Cover Dataset, the Global 30-meter Land Cover Fine Classification Dataset, the China Terraced Fields Map, and global farmland data, this study collected multi-source land use data and arable land data for a specified time period within the research area.

[0012] The process of combining and masking the multi-source land use data and cultivated land data includes:

[0013] The multi-source land use data and cultivated land data are combined to obtain the maximum potential cultivated land area data;

[0014] Masking is applied to the built-up area data and water body data in the maximum potential arable land area data to obtain the potential arable land area data.

[0015] The band-based temporal change detection algorithm is used to detect and process the potential arable land area data, and the band-fitted harmonic function fitting results are as follows:

[0016] Based on the band time-series change detection algorithm, cloud and cloud shadow masking is performed on the remote sensing images of Landsat 4, 5, 7 and 8 in the potential cultivated land range data for a specified time period. The remote sensing images after cloud and cloud shadow masking are divided into seven intervals based on six spectral bands and one temperature band.

[0017] Based on continuous change detection and classification algorithms, the band fitting harmonic function is processed after cloud and cloud shadow masking to obtain the band fitting harmonic function fitting result. The calculation formula includes:

[0018]

[0019] In the formula, x represents the date sequence within the year, i represents the Landsat band, k represents the number of bands, and a o,i Let a be the fitting efficiency parameter for band i. 1,i b 1,i c 1,i Let i be the interannual variation efficiency parameter for band i. For the k-th mutation point, The value for the i-th band is the model-based prediction on day x.

[0020] The process of calculating the breakpoints of farmland use change in the potential farmland area data based on the fitting results, and dividing farmland use data into different stages of change based on the breakpoints, includes:

[0021] Based on the fitting results, the pixel time-series value deviation is calculated, the breakpoints of farmland use change in the potential farmland area data are identified, and farmland use data for different change stages are divided based on the breakpoints. The calculation formula includes:

[0022]

[0023] In the formula, ρ(i,x) is the observed value of band i on day x of the year. The predicted value for band i on day x of the year, RMSE i This represents the root mean square error of the i-band.

[0024] The predicted value of band i on day x of the year is calculated based on the ordinary least squares model.

[0025] The calculation of the normalized vegetation index (NDI) of the remote sensing images corresponding to the farmland use data at different stages of change includes:

[0026]

[0027] In the formula, the Normalized Difference Vegetation Index (NDVI) is between 0 and 1, NIR is the reflectance of the near-infrared band of the remote sensing image, and Red is the reflectance of the red band of the remote sensing image.

[0028] The normalized vegetation index (NVI) is detected and processed using a band-based temporal change detection algorithm to obtain the band-fitted harmonic function fitting result of the NVI. Based on the fitting result, the cultivated land use change type data is classified as follows:

[0029] Based on continuous change detection and classification algorithms, the normalized vegetation index (NDI) of remote sensing images corresponding to farmland use data at different change stages is fitted to classify farmland use types. The calculation formula includes:

[0030]

[0031] In the formula, This represents the predicted value at time t, where t is the image capture time, slope is the vegetation index trend, intercept is a constant term, N is the phenological frequency, and a n b n Representing n respectively th Fitting efficiency of seasonal harmonic functions.

[0032] The extraction of information on abandoned farmland in mountainous and hilly areas based on the farmland use change type data, combined with farmland abandonment evolution data and vegetation phenological characteristic data, includes:

[0033] Classification based on the decision tree method, the calculation formula includes:

[0034] (time end -time start (*slope>threshold_1)

[0035] Overall s2 -Overall S1 >threshold_2,

[0036]

[0037] In the formula, time start The starting time for farmland use data at different stages of change is denoted as time. end The endpoint for farmland use data at different stages of change is represented by "slope" (the trend of vegetation index change) and "Overall". s1 With Overall s2 The base values ​​of vegetation index are before and after the breakpoint of farmland use change, threshold_1 and threshold_2 are the judgment thresholds, and intercept is a constant term.

[0038] The method further includes:

[0039] Based on data collected from Earth, Google Earth, and time series observers, stratified sampling was performed from the farmland use change type data. Combined with vegetation index time series, the accuracy and effectiveness of the information on abandoned farmland in the mountainous and hilly areas were verified from three aspects: producer accuracy, user accuracy, and overall accuracy.

[0040] This invention collects multi-source land use and arable land data through various channels, enabling cross-verification of the collected data and improving accuracy. It performs union and masking processing on the multi-source land use and arable land data to obtain potential arable land extent data, eliminating interference from other land activities and improving accuracy. Based on a band-time series change detection algorithm, it detects and processes the potential arable land extent data, accurately identifying breakpoints in arable land use changes and rationally dividing arable land use data into different change stages. It classifies arable land use change types based on the normalized vegetation index and extracts abandoned arable land information by combining arable land abandonment evolution data with prepared phenological characteristic data, ensuring it conforms to the arable land abandonment evolution pattern and is more reasonable and accurate. The accuracy and effectiveness of the abandoned arable land information are verified to guarantee the effectiveness of the method. This invention is simple to operate and has high extraction efficiency, realizing remote sensing extraction of abandoned arable land information in mountainous and hilly areas where high spatiotemporal resolution imagery is lacking and under limited human subjective conditions. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 This is a flowchart of a remote sensing method for extracting information on abandoned farmland in mountainous and hilly areas according to an embodiment of the present invention.

[0043] Figure 2 This is a flowchart of the process of combining and masking multi-source land use data and cultivated land data in an embodiment of the present invention.

[0044] Figure 3 This is a flowchart illustrating the detection and processing of potential arable land area data using a band-time series change detection algorithm in an embodiment of the present invention.

[0045] Figure 4 This is a schematic diagram of farmland use data divided into different stages of change based on the farmland use change breakpoints in an embodiment of the present invention.

[0046] Figure 5 This is a schematic diagram illustrating the classification results of cultivated land use change type data in an embodiment of the present invention. Detailed Implementation

[0047] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0048] In this invention, it should be understood that terms such as “comprising” or “having” are intended to indicate the presence of features, figures, steps, behaviors, components, portions or combinations thereof disclosed in this specification, and are not intended to exclude the possibility that one or more other features, figures, steps, behaviors, components, portions or combinations thereof are present or added.

[0049] It should also be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0050] This invention relates to a remote sensing method for extracting information on abandoned arable land in mountainous and hilly areas, comprising: collecting multi-source land use data and arable land data within a study area, and performing union and masking processing on the multi-source land use data and arable land data to form potential arable land extent data; detecting and processing the potential arable land extent data based on a band-time series change detection algorithm to obtain the band fitting harmonic function fitting result of the potential arable land extent data; calculating the arable land use change breakpoints of the potential arable land extent data based on the band fitting harmonic function fitting result of the potential arable land extent data, and dividing the arable land use data into different change stages based on the arable land use change breakpoints; and calculating the remote sensing shadows corresponding to the arable land use data in different change stages. The Normalized Difference Vegetation Index (NDV) is used to detect and process the NDC based on a band-time series change detection algorithm, resulting in a band-fitted harmonic function of the NDC. Based on this result, farmland use change type data is categorized. Based on this farmland use change type data, combined with farmland abandonment evolution data and vegetation phenological characteristic data, information on abandoned farmland in mountainous and hilly areas is extracted. Using data from Earth, Google Earth, and a time series observer, stratified sampling is performed from the farmland use change type data. Combined with the vegetation index time series, the accuracy and effectiveness of the information on abandoned farmland in mountainous and hilly areas are verified from three aspects: producer accuracy, user accuracy, and overall accuracy.

[0051] In one optional implementation of this embodiment, such as Figure 1 As shown, Figure 1 The flowchart of the remote sensing extraction method for abandoned farmland information in mountainous and hilly areas according to an embodiment of the present invention is shown, including the following steps:

[0052] S101. Collect multi-source land use data and cultivated land data within the study area;

[0053] In one optional implementation of this embodiment, multi-source land use data and cultivated land data for a specified time period within the study area are collected based on the China Life Cycle Basic Database (CLCD), the Global 30-meter Land Cover Dataset (GlobeLand30), the Global High-Resolution Land Cover Dataset (FROM_GLC), the Global 30-meter Land Cover Fine Classification Dataset (GLC_FCS30), the China Terrace Map, and the Global Cropland.

[0054] Specifically, the Chinese Life Cycle Database (CLCD) is a localized life cycle database for China, jointly developed by the School of Architecture and Environment of Sichuan University and Yike Environment. The data comes from industry statistics and literature, represents the average of the Chinese market, and includes multiple indicators related to resource consumption and energy conservation and emission reduction.

[0055] Furthermore, the Global 30-meter Land Cover Dataset (GlobeLand30) is a significant achievement of the National High Technology Research and Development Program of China (863 Program) project on Global Land Cover Remote Sensing Mapping and Key Technologies. This dataset includes ten major land cover types: cultivated land, forest, grassland, shrubland, wetland, water bodies, tundra, artificial land, bare land, glaciers, and permanent snow.

[0056] Furthermore, the Global High-Resolution Land Cover Dataset (FROM_GLC, Finer Resolution Observation and Monitoring Global Land Cover) was created by Professor Peng Gong's team at Tsinghua University using the Google Earth Engine platform. It utilized 130,000 sample points globally and employed a random forest classification method to map the 2017 Sentinel-2 imagery at a 10-meter resolution. This dataset is divided into 10 classes: cultivated land, forest, grassland, shrubland, wetland, water bodies, tundra, artificial surfaces, bare land, glaciers, and permanent snow.

[0057] Furthermore, the Global Land Cover with Fine Classification System in 30m (GLC_FCS30) dataset was developed by Liu Liangyun's team at the Aerospace Information Research Institute of the Chinese Academy of Sciences. It utilizes all Landsat data (Landsat TM (Thematic mapper), ETM+ (Enhanced Thematic Mapper), and OLI (Operational Land Imager)) produced since 1984 to provide a dynamic monitoring product of global 30-meter fine land cover from 1985 to the present. This product includes 29 land cover types and is updated every 5 years.

[0058] Furthermore, the China Terrace Map, a 30-meter resolution map of China's terraced fields released in 2018, was developed using pixel-based supervised classification of multi-source and multi-temporal data based on the Google Earth engine platform. Its overall accuracy and Kappa coefficient reached 94% and 0.72, respectively. This first 30-meter map of China's terraced fields can be used for research on soil erosion, food security, biogeochemical cycles, biodiversity, and ecosystem service assessments.

[0059] More importantly, Global Cropland is a 30-meter high-resolution interactive map of farmland around the world, published by the U.S. Geological Survey. The map data comes from Landsat's global food security support analysis data.

[0060] This method obtains land use and arable land data from six different sources, allowing the data to be cross-referenced and improving accuracy.

[0061] S102. The multi-source land use data and cultivated land data are combined and masked to form potential cultivated land range data.

[0062] In one optional implementation of this embodiment, such as Figure 2 As shown, Figure 2 This diagram illustrates a flowchart of the process for unifying and masking multi-source land use data and cultivated land data according to an embodiment of the present invention, including the following steps:

[0063] S201. Perform union processing on the multi-source land use data and cultivated land data to obtain the maximum potential cultivated land area data;

[0064] In an optional implementation of this embodiment, the multi-source land use data and cultivated land data obtained in step S101 are processed by union according to time and space correspondence.

[0065] S202. Mask the built-up area data and water body data in the maximum potential arable land area data to obtain the potential arable land area data.

[0066] In one optional implementation of this embodiment, the built-up area data and water body data in the maximum potential arable land range data are masked to remove changes in arable land use caused by non-abandonment, thereby obtaining the potential arable land range data.

[0067] This method involves performing union and masking processing on the multi-source land use data and cultivated land data, which can accurately locate the abandoned cultivated land objects to be extracted, exclude cultivated land use changes caused by non-abandonment, and improve the accuracy of the data.

[0068] S103. The potential arable land range data is detected and processed based on the band time series change detection algorithm to obtain the band fitting harmonic function fitting result of the potential arable land range data.

[0069] In one optional implementation of this embodiment, such as Figure 3 As shown, Figure 3 The flowchart illustrates the detection and processing of potential arable land area data using a band-time series change detection algorithm according to an embodiment of the present invention, including the following steps:

[0070] S301. Based on the band time series change detection algorithm, perform cloud and cloud shadow masking on the remote sensing images of Landsat 4, 5, 7 and 8 in the potential cultivated land range data for a specified time period.

[0071] In an optional implementation of this embodiment, the band temporal change detection algorithm is a pixel-based temporal change detection method. In this embodiment, the spectral-time trajectory method is used. The spectral-time trajectory method determines the change type of each pixel by setting a threshold or parameter based on the spectral-time trajectory characteristics of the change detection index.

[0072] In one optional implementation of this embodiment, the cloud and cloud shadow masking is an image processing method that can effectively detect the coverage of clouds and cloud shadows in remote sensing images, while reducing the influence of built-up areas and water bodies, and improving the detection accuracy of remote sensing images.

[0073] S302. Divide the remote sensing image after cloud and cloud shadow masking into seven intervals based on six spectral bands and one temperature band.

[0074] In one optional implementation of this embodiment, the six spectral bands are the blue light band of the remote sensing image, the green light band of the remote sensing image, the red light band of the remote sensing image, the near-infrared band of the remote sensing image, the first short-wave infrared band of the remote sensing image, and the second short-wave infrared band of the remote sensing image.

[0075] In one optional implementation of this embodiment, the shortest time interval between the seven intervals is 1.33 years.

[0076] S303. Based on the continuous change detection and classification algorithm, the band fitting harmonic function is processed by cloud and cloud shadow masking to obtain the band fitting harmonic function fitting result.

[0077] In one optional implementation of this embodiment, the calculation formula includes:

[0078]

[0079] In the formula, x represents the date sequence within the year, i represents the Landsat band, k represents the number of bands, and a o,i Let a be the fitting efficiency parameter for band i. 1,i b 1,i c 1,i Let i be the interannual variation efficiency parameter for band i. For the k-th mutation point, The value for the i-th band is the model-based prediction on day x.

[0080] Specifically, the Continuous Change Detection and Classification (CCDC) algorithm is a time-series-based change detection algorithm specifically designed for the field of remote sensing. It uses all available Landsat data to model time-spectral characteristics, simulates trends, seasonal changes, abrupt changes, etc., can detect various types of land cover change (phenological changes, slow interannual changes, abrupt changes), and can draw land use maps at any given time.

[0081] S104. Based on the band fitting harmonic function fitting results of the potential arable land range data, calculate the arable land use change breakpoints of the potential arable land range data, and divide the arable land use data of different change stages based on the arable land use change breakpoints.

[0082] In an optional implementation of this embodiment, the pixel time series value deviation is calculated based on the band fitting harmonic function fitting result of the potential arable land range data, the arable land use change discontinuities of the potential arable land range data are identified, and arable land use data for different change stages are divided based on the arable land use change discontinuities. The calculation formula includes:

[0083]

[0084] In the formula, ρ(i,x) is the observed value of band i on day x of the year. The predicted value for band i on day x of the year, RMSE i This represents the root mean square error of the i-band.

[0085] In an optional implementation of this embodiment, the predicted value of band i on day x of the year is calculated based on the ordinary least squares (OLS) model.

[0086] In an optional implementation of this embodiment, the identification basis for the breakpoint of farmland use change is the predicted value. The error between the observed value ρ(i,x) and the actual observed value exceeds the root mean square error (RMSE). i If the above behavior occurs four times consecutively, it is considered that the use of cultivated land has changed, and the corresponding time point can be regarded as the breakpoint of the change in cultivated land use.

[0087] In one optional implementation of this embodiment, such as Figure 4 As shown, BLUE represents the blue light band reflectance of the remote sensing image, GREEN represents the green light band reflectance, RED represents the red light band reflectance, NIR represents the near-infrared band reflectance, SWIR1 represents the first short-wave infrared band of the remote sensing image, SWIR2 represents the second short-wave infrared band of the remote sensing image, and TEMP represents the temperature band. Figure 4 As shown, any band identifies the same time point as a breakpoint in farmland use change, and the base values ​​and fluctuations of the bands before and after this time point show significant differences, which proves the rationality of setting breakpoints for farmland use change.

[0088] S105. Calculate the normalized vegetation index of the remote sensing images corresponding to the cultivated land use data at different stages of change.

[0089] In an optional implementation of this embodiment, the formula for calculating the normalized vegetation index is:

[0090]

[0091] In the formula, the Normalized Difference Vegetation Index (NDVI) is between 0 and 1, NIR is the reflectance of the near-infrared band of the remote sensing image, and Red is the reflectance of the red band of the remote sensing image.

[0092] S106. The normalized vegetation index is detected and processed based on the band time series change detection algorithm to obtain the band fitting harmonic function fitting result of the normalized vegetation index, and the cultivated land use change type data is divided based on the band fitting harmonic function fitting result of the normalized vegetation index.

[0093] In one optional implementation of this embodiment, based on a continuous change detection and classification algorithm, the cultivated land use type data is classified by fitting the normalized vegetation index of the remote sensing images corresponding to cultivated land use data at different change stages. The calculation formula includes:

[0094]

[0095] In the formula, This represents the predicted value at time t, where t is the image capture time, slope is the vegetation index trend, intercept is a constant term, N is the phenological frequency, and a n b n Representing n respectively th Fitting efficiency of seasonal harmonic functions.

[0096] S107. Based on the data on changes in cultivated land use, combined with data on the evolution of cultivated land abandonment and data on vegetation phenological characteristics, information on abandoned cultivated land in mountainous and hilly areas is extracted.

[0097] In one optional implementation of this embodiment, classification is based on a decision tree method, and the calculation formula includes:

[0098] (time end -time start (*slope>threshold_1)

[0099] Overall s2 -Overall S1 >threshold_2,

[0100]

[0101] In the formula, time start The starting time for farmland use data at different stages of change is denoted as time. end The endpoint for farmland use data at different stages of change is represented by "slope" (the trend of vegetation index change) and "Overall". s1 With Overall s2 The base values ​​of vegetation index are before and after the breakpoint of farmland use change, threshold_1 and threshold_2 are the judgment thresholds, and intercept is a constant term.

[0102] In an optional implementation of this embodiment, based on randomly selected land category conversion sample points, after threshold setting sensitivity analysis, the value of threshold_1 is set to any value between 0.05 and 0.2, and the value of threshold_2 is set to any value between 0.05 and 0.2.

[0103] Preferably, the value of threshold_1 is set to 0.1, and the value of threshold_2 is set to 0.1.

[0104] In an optional implementation of this embodiment, if the farmland use change type data satisfies (time end -time start )*slope>threshold_1 and Overall s2 -Overall S1 If the threshold value is greater than 2, meaning the change in the base value between two stable use stages in the cultivated land use change data is greater than the threshold, and the difference in the base values ​​of the stable land use types before and after the change is greater than the threshold, then the cultivated land is considered to have evolved from bare land without vegetation to perennial weeds, shrubs, and woodland, representing a typical form of abandonment. Figure 5 As shown in Scenario II.

[0105] Furthermore, if the data on farmland use change types only satisfy Overall s2 -Overall S1 >threshold_2, meaning the difference between the baseline values ​​of stable land use types before and after fallow in the cultivated land use change data is greater than the judgment threshold, is considered that the cultivated land has evolved from stable use before fallow to stable shrub vegetation, which is a form of fallow lacking stable and continuous observation. Figure 5 As shown in scenario III.

[0106] Furthermore, if the data on the type of arable land use change does not meet the (time) requirement... end -time start *slope>threshold_1 or Overall s2 -Overall S1 >threshold_2, meaning that the change in the base value between two stable use stages in the cultivated land use change type data is not greater than the judgment threshold, or the difference in the base values ​​of the stable land use types before and after is not greater than the judgment threshold, then the cultivated land is considered to be stably used cultivated land. Figure 5 As shown in Scenario I, the farmland before and after the breakpoint in farmland use change is either in stable use or, for example... Figure 5 As shown in scenario IV.

[0107] S108. Based on the data collected from Earth, Google Earth, and time series observers, stratified sampling is performed from the farmland use change type data. Combined with the vegetation index time series, the accuracy and effectiveness of the information on abandoned farmland in the mountainous and hilly areas are verified from three aspects: producer accuracy, user accuracy, and overall accuracy.

[0108] In one optional implementation of this embodiment, based on Collect Earth, Google Earth, and Time Series Viewer software, stratified sampling is performed from the farmland use change type data, and combined with vegetation index time series, the accuracy and effectiveness of the abandoned farmland information in the mountainous and hilly areas are verified from three aspects: producer accuracy, user accuracy, and overall accuracy.

[0109] In summary, this invention collects multi-source land use data and cultivated land data through multiple channels, enabling cross-verification of the collected data and improving accuracy. It performs union and masking processing on the multi-source land use data and cultivated land data to obtain potential cultivated land area data, eliminating interference from other land activities and improving accuracy. Based on a band-time series change detection algorithm, it detects and processes the potential cultivated land area data, accurately identifying cultivated land use change breakpoints and rationally dividing cultivated land use data into different change stages. It classifies cultivated land use change types based on the normalized vegetation index and extracts abandoned cultivated land information by combining cultivated land abandonment evolution data with prepared phenological characteristic data, ensuring it conforms to the cultivated land abandonment evolution law and is more reasonable and accurate. Verifying the accuracy and effectiveness of abandoned cultivated land information guarantees the effectiveness of the method. This invention is simple to operate, highly efficient in extraction, and achieves remote sensing extraction of abandoned cultivated land information in mountainous and hilly areas lacking high spatiotemporal resolution imagery and under limited human subjective conditions.

[0110] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, which may include: read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0111] Furthermore, the embodiments of the present invention have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

[0112] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, which may include: read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0113] Furthermore, the embodiments of the present invention have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for extracting information of abandoned cultivated land in mountain and hilly areas by remote sensing, characterized in that, The method comprises the following steps: Collecting multi-source land use data and cultivated land data in a research area, and performing set union and mask processing on the multi-source land use data and cultivated land data to form potential cultivated land range data; Performing detection processing on the potential cultivated land range data based on a band time series change detection algorithm to obtain a band fitting harmonic function fitting result of the potential cultivated land range data; Calculating cultivated land use change breakpoints of the potential cultivated land range data based on the band fitting harmonic function fitting result of the potential cultivated land range data, and dividing cultivated land use data in different change stages based on the cultivated land use change breakpoints; Calculating normalized vegetation indices of remote sensing images corresponding to the cultivated land use data in different change stages, performing detection processing on the normalized vegetation indices based on a band time series change detection algorithm to obtain a band fitting harmonic function fitting result of the normalized vegetation indices, and dividing cultivated land use change type data based on the band fitting harmonic function fitting result of the normalized vegetation indices; Based on the cultivated land use change type data, combined with cultivated land abandonment evolution rule data and vegetation phenology feature data, extracting mountain and hilly region abandoned cultivated land information; Based on the collection of Earth, Google Earth and time series observer, layer sampling is performed from the cultivated land use change type data, combined with vegetation index time series, and the mountain and hilly region abandoned cultivated land information is verified in terms of producer accuracy, user accuracy and overall accuracy.

2. The method for extracting information of abandoned cultivated land in mountain and hilly areas from remote sensing images according to claim 1, wherein, The method comprises the following steps: Based on the Chinese life cycle database, the global 30-meter land cover dataset, the global high-resolution land cover dataset, the global 30-meter land cover fine classification dataset, the Chinese terrace map and the global farmland, the multi-source land use data and cultivated land data in the research area within a specified time period are collected.

3. The method according to claim 2, wherein the mountain and hilly region abandoned farmland information is extracted by using the remote sensing image of the mountain and hilly region. The method comprises the following steps: Performing set union processing on the multi-source land use data and cultivated land data to obtain maximum potential cultivated land range data; Performing mask processing on built-up area data and water body data in the maximum potential cultivated land range data to obtain potential cultivated land range data.

4. The method of claim 1, wherein the mountainous hilly region abandoned farmland information is extracted from the remote sensing image. The method comprises the following steps: Based on the band time series change detection algorithm, the remote sensing images of Landsat 4, 5, 7 and 8 in the specified time period in the potential cultivated land range data are subjected to cloud and cloud shadow mask processing, and the remote sensing images subjected to cloud and cloud shadow mask processing are divided into seven intervals based on six spectral bands and one temperature band; Based on the continuous change detection and classification algorithm, the band fitting harmonic function subjected to cloud and cloud shadow mask processing is processed to obtain the band fitting harmonic function fitting result, and the calculation formula comprises: where x is the sequence of dates within a year, i is the Landsat band, k is the number of bands, a o,i is the fitting efficiency parameter for band i, a 1,i , b 1,i , c 1,i is the interannual variability efficiency parameter for band i, is the kth break point, is the predicted value for band i at day x based on the model.

5. The method of claim 4, wherein the mountainous hilly region abandoned farmland information is extracted from the remote sensing image. The method comprises the following steps: Calculate pixel timing value deviation based on the fitting result, identify the cultivated land use change breakpoint of the potential cultivated land range data, and divide the cultivated land use data of different change stages based on the cultivated land use change breakpoint, and the calculation formula comprises: where p(i, x) is the observed value of the i-th band on the x-th day of the year, is the predicted value of the i-th band on the x-th day of the year, and RMSE i is the root mean square error of the i-th band.

6. The method of claim 5, wherein the mountainous hilly region abandoned farmland information is extracted from the remote sensing image. calculating a predicted value for the ith waveband on the xth day of the year based on an ordinary least squares model 7. The method of claim 1, wherein the mountainous hilly area abandoned farmland information is extracted from the remote sensing image. The calculation of the normalized vegetation index of the remote sensing image corresponding to the cultivated land use data of different change stages comprises: In the formula, the normalized vegetation index NDVI is between 0 and 1, NIR is the near-infrared band reflectivity of the remote sensing image, and Red is the red light band reflectivity of the remote sensing image.

8. The method of claim 1, wherein the mountainous hilly region abandoned farmland information is extracted from the remote sensing image. The band fitting harmonic function fitting result of the normalized vegetation index is obtained by detecting and processing the normalized vegetation index based on the band timing change detection algorithm, and the cultivated land use change type data is divided based on the fitting result, and the calculation formula comprises: Based on the continuous change detection and classification algorithm, the normalized vegetation index of the remote sensing image corresponding to the cultivated land use data of different change stages is fitted, the cultivated land use type data is divided, and the calculation formula comprises: In the formula, represents the prediction value at time t, t is the image shooting time, slope is the vegetation index change trend, intercept is the constant term, N is the phenology frequency, a n , b n respectively represent the n th order seasonal harmonic function fitting efficiency.

9. The method of claim 1, wherein the method is characterized by: Based on the cultivated land use change type data, the abandoned cultivated land information in the mountain hilly area is extracted in combination with the cultivated land abandonment evolution rule data and the vegetation phenology feature data, and the calculation formula comprises: Classification is performed based on the decision tree method, and the calculation formula comprises: (time end -time start )*slope>threshold_1, Overall s2 -Overall S1 >threshold_2, In the formula, time start is the starting time of the cultivated land use data of different change stages, time end is the ending time of the cultivated land use data of different change stages, slope is the vegetation index change trend, Overall s1 is the vegetation index of the Overall s2 is the vegetation index base value before and after the cultivated land use change breakpoint, threshold_1 and threshold_2 are the judgment thresholds, and intercept is the constant term.

Images