Method and device for detecting and identifying cultivated land changes based on time series feature vectors

Abstract

The present application relates to a method and device for detecting and identifying cultivated land changes based on time series feature vectors, which belongs to the field of agricultural cultivated land monitoring technology. It solves the problem that existing remote sensing technology cannot accurately determine the time nodes of cultivated land changes and is difficult to distinguish weak remote sensing signals of some cultivated land changes. The cultivated land change detection and identification method based on time series feature vectors, after acquiring remote sensing image data and multi-source land use data, performs multi-period harmonic fitting of remote sensing image data based on a time domain sliding window, constructs cultivated land change vectors and calculates the relative change vectors of cultivated land change vectors in adjacent time periods to determine potential cultivated land change areas; divides potential cultivated land change areas into cultivated land areas and non-cultivated land areas based on a training sample set, generates corresponding cultivated land classification results and cultivated land probabilities; synthesizes cultivated land classification results to form a continuous time series of cultivated land change results, and corrects the cultivated land change results in combination with cultivated land probabilities. The present application improves the accuracy of detecting cultivated land change dynamics.

Description

Technical Field

[0001] The present application relates to the technical field of agricultural farmland monitoring, and in particular to a method and device for detecting and identifying farmland changes based on time series feature vectors. Background Art

[0002] Arable land is land formed by natural soil development and capable of growing crops. It provides a natural environment for the growth, development, and maturity of crops. Arable land also changes in response to changes in the natural environment. Therefore, accurate and timely information on changes in arable land and the types of changes is of great significance for arable land monitoring, food security, and regional sustainable development.

[0003] Traditional cultivated land monitoring is mostly conducted within administrative regions, combining manual visual interpretation with field verification. While this method can achieve high monitoring accuracy, it is time-consuming and labor-intensive, making it inadequate for large-scale, rapid, and timely agricultural monitoring. Remote sensing technology, on the other hand, can efficiently and rapidly acquire information on cultivated land change over large areas, making it an important tool for identifying cultivated land change and its spatial distribution. Currently, existing methods for identifying cultivated land change mainly include post-classification comparison, image differencing, and time series change detection. The first method generates annual land use classification maps and detects change by comparing the results. The second method, without classification, first detects change by extracting remote sensing information from time series data and then classifies the changed areas. Methods include thresholding, segmentation, and statistical boundary analysis. The third method uses time series data for change monitoring analysis to determine cultivated land change and capture the timing of change. Based on statistical assumptions, these methods construct spatiotemporal change models to achieve change detection, such as the LandTrendr algorithm, the BFAST algorithm, and the CCDC algorithm.

[0004] However, existing methods for identifying cultivated land change focus on confirming change through the deviation of phenological trajectories between predicted and observed values. This inadequately captures weak signals of cultivated land change, such as those with small changes or similar phenology before and after the change. Furthermore, the transition period of land cover change results in a delayed response to spectral change peaks, leading to a lag in change detection. Cultivated land change is influenced by both natural factors and human activities, and cultivated land with different crop types, planting systems, and climatic environments exhibits complex spectral and temporal characteristics, making it difficult to effectively distinguish cultivated land change from environmental noise and spurious changes. Consequently, existing methods for identifying cultivated land change cannot accurately determine the time points of cultivated land change and have difficulty distinguishing some weak remote sensing signals of cultivated land change. Summary of the Invention

[0005] In view of the above analysis, the embodiments of the present invention aim to provide a method and device for detecting and identifying cultivated land changes based on time series feature vectors, so as to solve the technical problems that the existing cultivated land change identification methods cannot accurately determine the time nodes of cultivated land changes and are difficult to distinguish the weak remote sensing signals of some cultivated land changes.

[0006] The present application provides a method for detecting and identifying cultivated land changes based on time series feature vectors, comprising the following steps:

[0007] Acquire remote sensing image data and multi-source land use data; fuse the multi-source land use data to obtain training sample sets of cultivated land and non-cultivated land;

[0008] Multi-period harmonic fitting of remote sensing image data is performed based on a time domain sliding window to construct a cultivated land change vector. The relative change vectors of cultivated land change vectors in adjacent time periods are calculated. The potential cultivated land change areas are determined by comparing the relative change vectors with the preset identification threshold. The year of potential cultivated land change is determined by the peak value of the relative change vector.

[0009] Based on the training sample set, the potential cultivated land change areas are classified into cultivated land areas and non-cultivated land areas, and the corresponding cultivated land classification results and cultivated land probabilities are generated;

[0010] The cultivated land classification results are synthesized to form a continuous time series of cultivated land change results, and the cultivated land change results are corrected in combination with the cultivated land probability.

[0011] The cultivated land change detection and identification method based on time series feature vectors in the embodiment of the present application, after acquiring remote sensing image data and multi-source land use data, performs multi-period harmonic fitting of the remote sensing image data based on a time domain sliding window to construct a cultivated land change vector, and calculates the relative change vector of the cultivated land change vector in adjacent time periods, determines the potential cultivated land change area by comparing the relative change vector with a preset recognition threshold, and discriminates the year of potential cultivated land change by the peak value of the relative change vector. Based on the training sample set, the potential cultivated land change area is classified into cultivated land area and non-cultivated land area, and the corresponding cultivated land classification results and cultivated land probabilities are generated; the cultivated land classification results are synthesized to form a continuous time series of cultivated land change results, and the cultivated land change results are corrected in combination with the cultivated land probability. Based on the time domain sliding window, long time series data fusion is realized to solve problems such as weak signal missed recognition and change detection time deviation, and improve the detection accuracy of cultivated land change dynamics.

[0012] In one embodiment of the present application, multi-period harmonic fitting of remote sensing image data is performed based on a time domain sliding window, using the following formula:

[0013]

[0014] in, Represents any pixel within the maximum cultivated land area; represents any Julian day with remote sensing observations, ranging from all Julian days observed in observation year t to observation year t+w; w is the time window of the function fitting; express Pixel in time The predicted value of NDVI index; For time Remote sensing observation value of NDVI index; is the number of days in a year; is the overall amplitude or intercept, reflecting the average growth status of vegetation in the period from year t to year t + w; is the slope, reflecting the interannual variation of vegetation from year t to year t + w; is the order of the harmonic component, reflecting the annual changes in vegetation at different frequencies, where and 、 and 、 and It reflects the annual changes once a year, twice a year and three times a year respectively.

[0015] In one embodiment of the present application, the following formula is used to calculate the relative change vector of the cultivated land change vector in adjacent time periods:

[0016]

[0017] in, for The normalized relative change vector of pixel t in year, represents the standardized vector difference of pixel i in year t, and They are pixel i from year t to year t+w and The harmonic function fitting vector of the two time periods from year to year t; The calculated parameters include the intercept of the harmonic function from year t to year t+w and seven harmonic fitting parameters of the first, second and third order, namely ; The calculated parameters include The intercept of the harmonic function from year to year t and the first-order, second-order and third-order harmonic fitting parameters are: From t to and The intercept and RMS error of the harmonic function fit for the two time periods from year to year t.

[0018] In one embodiment of the present application, the following two formulas are used to determine the year in which potential cultivated land changes occur:

[0019]

[0020] Among them, CT represents the preset identification threshold, and the area that meets the conditions of the above two formulas is a potential cultivated land change area. are the standardized relative change vectors of years t-2, t-1, t, t+1, and t+2 respectively. When the above two formulas are met, the standardized relative change vector of year t is greater than that of the other years t-2, t-1, t+1, and t+2; then year t is the year when potential cultivated land change occurs.

[0021] In one embodiment of the present application, the multi-source land use data includes CLCD data and GLAD data;

[0022] The process of fusing multi-source land use data to obtain a training sample set of cultivated land and non-cultivated land includes the following steps:

[0023] Reclassify CLCD data into land use types;

[0024] Taking the four years of each GLAD product as the interval time period, the CLCD data of the same time period are superimposed to obtain the area with the same land use type within the period, and then superimposed with the cultivated land and non-cultivated land ranges of the corresponding time period of the GLAD product to generate areas with completely consistent cultivated land and non-cultivated land, and correspondingly construct the training sample set of cultivated land and non-cultivated land.

[0025] In one embodiment of the present application, before constructing the training sample set of cultivated land and non-cultivated land, the following steps are further included:

[0026] Resample the resolution of the areas with completely consistent cultivated land and non-cultivated land to construct a cell network;

[0027] Remote sensing image features are constructed and optimized based on multi-source land use data, and training sample migration is performed.

[0028] In one embodiment of the present application, the process of correcting the cultivated land change result in combination with the cultivated land probability includes the following steps:

[0029] If the cultivated land probability is significantly different from other cultivated land probability segments in the time series, the cultivated land change results are retained; otherwise, they are discarded. The significant difference is determined based on the T-test of the Shapelet algorithm.

[0030] The present application also provides a device for detecting and identifying cultivated land changes based on time series feature vectors, including:

[0031] The data acquisition module is used to acquire remote sensing image data and multi-source land use data; wherein, the multi-source land use data is integrated to obtain a training sample set of cultivated land and non-cultivated land;

[0032] The data fusion module is used to perform multi-period harmonic fitting of remote sensing image data based on a time domain sliding window to construct a cultivated land change vector and calculate the relative change vector of the cultivated land change vector in adjacent time periods. The potential cultivated land change area is determined by comparing the relative change vector with the preset recognition threshold, and the year of potential cultivated land change is determined by the peak value of the relative change vector.

[0033] The data classification module is used to classify the potential cultivated land change areas into cultivated land areas and non-cultivated land areas based on the training sample set, and generate the corresponding cultivated land classification results and cultivated land probabilities;

[0034] The data detection module is used to synthesize the cultivated land classification results, form the cultivated land change results of the continuous time series, and correct the cultivated land change results in combination with the cultivated land probability.

[0035] The cultivated land change detection and identification device based on time series feature vectors of the embodiment of the present application obtains remote sensing image data and multi-source land use data, and then performs multi-period harmonic fitting of the remote sensing image data based on a time domain sliding window to construct a cultivated land change vector, and calculates the relative change vector of the cultivated land change vector in adjacent time periods. The potential cultivated land change area is determined by comparing the relative change vector with a preset recognition threshold, and the year in which the potential cultivated land change occurs is determined by the peak value of the relative change vector. Based on the training sample set, the potential cultivated land change area is classified into cultivated land area and non-cultivated land area, and the corresponding cultivated land classification results and cultivated land probabilities are generated; the cultivated land classification results are synthesized to form a continuous time series of cultivated land change results, and the cultivated land change results are corrected in combination with the cultivated land probability. Based on the time domain sliding window, long time series data fusion is achieved to solve problems such as weak signal missed recognition and change detection time deviation, thereby improving the detection accuracy of cultivated land change dynamics.

[0036] At least one embodiment of the present application further provides a data control device, including:

[0037] one or more memories non-transitorily storing computer-executable instructions;

[0038] One or more processors are configured to run computer-executable instructions, wherein the computer-executable instructions, when executed by the one or more processors, implement the method for detecting and identifying cultivated land changes based on time series feature vectors according to any embodiment of the present application.

[0039] After acquiring remote sensing imagery data and multi-source land use data, the data control device performs multi-period harmonic fitting on the remote sensing imagery data using a time-domain sliding window to construct a cultivated land change vector. The device also calculates the relative change vectors of cultivated land change vectors in adjacent time periods. Potential cultivated land change areas are identified by comparing the relative change vectors with a preset identification threshold, and the year of potential cultivated land change is determined by the peak value of the relative change vector. Based on a training sample set, potential cultivated land change areas are classified into cultivated land and non-cultivated land areas, and corresponding cultivated land classification results and cultivated land probabilities are generated. The cultivated land classification results are synthesized to form a continuous time series of cultivated land change results, which are then corrected based on the cultivated land probabilities. This time-domain sliding window allows for data fusion of long time series, addressing issues such as missed weak signal recognition and time deviation in change detection, thereby improving the accuracy of detecting cultivated land change dynamics.

[0040] At least one embodiment of the present application also provides a non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are executed by a processor, a method for detecting and identifying cultivated land changes based on time series feature vectors according to any embodiment of the present application is implemented.

[0041] After acquiring remote sensing image data and multi-source land use data, the aforementioned non-transient computer-readable storage medium performs multi-period harmonic fitting of the remote sensing image data based on a time-domain sliding window to construct a cultivated land change vector. The relative change vectors of cultivated land change vectors in adjacent time periods are calculated. Potential cultivated land change areas are identified by comparing the relative change vectors with a preset identification threshold, and the year of potential cultivated land change is determined by the peak value of the relative change vector. Based on a training sample set, potential cultivated land change areas are classified into cultivated land and non-cultivated land areas, and corresponding cultivated land classification results and cultivated land probabilities are generated. The cultivated land classification results are synthesized to form a continuous time series of cultivated land change results, which are then corrected based on the cultivated land probabilities. Data fusion over long time series is achieved using a time-domain sliding window to address issues such as missed weak signal recognition and time deviation in change detection, thereby improving the accuracy of detecting cultivated land change dynamics. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 This is a flow chart of a method for detecting and identifying cultivated land changes based on time series feature vectors according to an embodiment of the application;

[0043] Figure 2 This is a module structure diagram of a device for detecting and identifying cultivated land changes based on time series feature vectors according to an embodiment of the application;

[0044] Figure 3 A schematic block diagram of a data control device provided by the present invention;

[0045] Figure 4A schematic diagram of a non-transitory computer-readable storage medium provided by the present invention. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions of the embodiments of the present application will be clearly and completely described below in conjunction with the drawings of the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the described embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0047] Unless otherwise defined, the technical or scientific terms used in this application should have the usual meanings understood by people with ordinary skills in the field to which this application belongs. The "first", "second" and similar words used in this application do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative position relationships. When the absolute position of the object being described changes, the relative position relationship may also change accordingly.

[0048] In order to keep the following description of the embodiments of the present application clear and concise, the present application omits detailed descriptions of some known functions and known components.

[0049] Figure 1 This is a flow chart of a method for detecting and identifying cultivated land changes based on time series feature vectors according to an embodiment of the application. Figure 1 As shown, a method for detecting and identifying cultivated land changes based on time series feature vectors according to an embodiment of the application includes steps S100 to S103:

[0050] S100, acquiring remote sensing image data and multi-source land use data; wherein, fusing the multi-source land use data to obtain a training sample set of cultivated land and non-cultivated land;

[0051] S101, performing multi-period harmonic fitting of remote sensing image data based on a time domain sliding window to construct a cultivated land change vector, and calculating the relative change vector of the cultivated land change vectors in adjacent time periods, determining potential cultivated land change areas by comparing the relative change vectors with a preset identification threshold, and identifying the year in which the potential cultivated land change occurred by using the peak value of the relative change vector;

[0052] S102, classifying potential cultivated land change areas into cultivated land areas and non-cultivated land areas based on the training sample set, and generating corresponding cultivated land classification results and cultivated land probabilities;

[0053] S103, synthesizing the cultivated land classification results to form a continuous time series of cultivated land change results, and correcting the cultivated land change results in combination with the cultivated land probability.

[0054] In the embodiment of the present application, the remote sensing image data is a long-term remote sensing image of the study area. The scope of the cultivated land identification in the present application does not exceed the study area.

[0055] In a preferred embodiment, the remote sensing imagery data includes a long series of orthorectified surface reflectance images from Landsat-5 TM, Landsat-7 ETM+, and Landsat-8 OLI. For example, the selected remote sensing imagery covers the period 1986-2023. To ensure pixel quality, the QA band from the CFMask algorithm is used to remove poor-quality observations caused by clouds, shadows, snow / ice, and other factors. For pixels missing valid observations for four consecutive months, the nearest valid observation is searched for and linear interpolation is used to supplement the missing pixels.

[0056] In a preferred embodiment, multi-source land use data includes GLAD data (a cultivated land product produced every four years using a decision tree algorithm) and CLCD data (an annual cultivated land product generated using a random forest algorithm and the GEE cloud computing platform). The maximum cultivated land area is obtained by extracting the cultivated land distribution range from the CLCD data and overlaying the cultivated land distribution of the CLCD data from 1985 to 2022 with the GLAD data from 2000 to 2019.

[0057] In the embodiment of the present application, in step S101, multi-period harmonic fitting of remote sensing image data is performed based on a time domain sliding window to construct a cultivated land change vector, as shown in the following formula:

[0058]

[0059] in, Represents any pixel within the maximum cultivated land area; represents any Julian day with remote sensing observations, ranging from all Julian days observed in observation year t to observation year t+w; w is the time window of the function fitting; express Pixel in time The predicted value of NDVI index; For time Remote sensing observation value of NDVI index; is the number of days in a year; is the overall amplitude or intercept, reflecting the average growth status of vegetation in the period from year t to year t + w; is the slope, reflecting the interannual variation of vegetation from year t to year t + w; is the order of the harmonic component, reflecting the annual changes in vegetation at different frequencies, where and 、 and 、 and It reflects the annual changes once a year, twice a year and three times a year respectively.

[0060] The method uses a sliding window in the time domain to perform multi-period harmonic fitting of remote sensing image data. Within the defined maximum cultivated land area, the NDVI index is calculated using a long-term Landsat satellite imagery series. A harmonic function, including intercept, slope, and first-, second-, and third-order harmonic parameters, is used to fit the NDVI index. LASSO regression is used to calculate the eight parameters of the harmonic function on a pixel-by-pixel basis.

[0061] Preferably, w is the time window for function fitting. A smaller window width is limited by the number of valid observations and the influence of environmental noise and cannot fit a stable harmonic fitting model. A larger window width can smooth out changes in cultivated land with a shorter duration (such as abandonment). The embodiment of the present application defines a window width of "w" years ("w=5").

[0062] Preferably, is the number of days in a year, or 365.25. Because the harmonic function contains eight parameters, at least 12 valid observations are required before model fitting can begin. Considering that some pixels have fewer than 15 valid observations within a five-year window, the stability of the function fitting decreases significantly. Therefore, the harmonic fitting results from adjacent time periods are used to forward-fill the predicted values.

[0063] In the embodiment of the present application, the process of calculating the relative change vector of the cultivated land change vector in adjacent time periods in step S101 is as follows:

[0064]

[0065] in, for The normalized relative change vector of pixel t in year, represents the standardized vector difference of pixel i in year t, and They are pixel i from year t to year t+w and The harmonic function fitting vector of the two time periods from year to year t; The calculated parameters include the intercept of the harmonic function from year t to year t+w and seven harmonic fitting parameters of the first, second and third order, namely ; The calculated parameters include The intercept of the harmonic function from year to year t and the first-order, second-order and third-order harmonic fitting parameters are: From t to and The intercept and RMS error of the harmonic function fit for the two time periods from year to year t.

[0066] Among them, the cultivated land change vector is constructed based on phenological stability. The standardized relative change vector (SRCV) of adjacent time periods is calculated on an annual scale to detect whether cultivated land has changed.

[0067] In the embodiment of the present application, the process of determining the potential cultivated land change area by comparing the relative change vector with the preset identification threshold in step S101 and determining the year of potential cultivated land change by identifying the peak value of the relative change vector is as follows:

[0068]

[0069] Among them, CT represents the preset identification threshold, and the area that meets the above two conditions is a potential cultivated land change area. are the standardized relative change vectors of years t-2, t-1, t, t+1, and t+2 respectively. When the above formula is met, the standardized relative change vector of year t is greater than that of the other years t-2, t-1, t+1, and t+2. Therefore, year t is the year when potential cultivated land change occurs.

[0070] Preferably, the harmonic vectors of the two five-year windows before and after the cultivated land change should have the largest difference, that is, the SRCV is significantly increased, and the SRCV of that year is greater than the SRCV of the two years before and after. To accurately determine the year when the change occurred, multiple sets of model parameters were tested in different agricultural regions and the results were thresholded, and then the preset identification threshold CT for cultivated land change identification was determined. In theory, different agricultural regions should adopt different preset identification thresholds, but analysis of multiple case regions found that CT = 0.5 is a robust threshold for detecting changes. Therefore, in this embodiment of the application, CT is set to 0.5.

[0071] The potential cultivated land change areas identified based on the above method include pseudo-changes caused by changes in cultivated land planting systems, where phenological trajectories have changed but land cover types have not changed. Therefore, they are called potential cultivated land change areas.

[0072] In the embodiment of the present application, the process of fusing multi-source land use data in step S100 to obtain a training sample set of cultivated land and non-cultivated land includes the following steps:

[0073] Reclassify CLCD data into land use types;

[0074] Taking the four years of each GLAD product as the interval time period, the CLCD data and CLCD data of the same time period are superimposed to obtain the area with the same land use type within the period, and then superimposed on the cultivated land and non-cultivated land ranges of the corresponding time period of the GLAD product to generate areas with completely consistent cultivated land and non-cultivated land, and correspondingly construct the training sample set of cultivated land and non-cultivated land.

[0075] Preferably, multi-source land use data is integrated to obtain a training sample set of cultivated land and non-cultivated land. The CLCD product is reclassified into six land use types: cultivated land, forest land, grassland, water area, construction land, and unused land. Using the four years of each GLAD data period as intervals, the CLCD data from the same time period are superimposed to obtain areas with the same land use type over the four years. The cultivated land and non-cultivated land ranges of the corresponding time period in the GLAD data are then superimposed to generate areas with completely consistent cultivated land and non-cultivated land for each four-year period. Furthermore, for years not covered by the GLAD data, only the four-year, four-period CLCD data are used to obtain areas with completely consistent cultivated land and non-cultivated land. Finally, we obtained completely consistent areas of cultivated land and non-cultivated land for nine time periods: 1986-1991, 1992-1995, 1996-1999, 2000-2003, 2004-2007, 2008-2011, 2012-2015, 2016-2019 and 2020-2023.

[0076] In the embodiment of the present application, before constructing the training sample set of cultivated land and non-cultivated land, the following steps are also included:

[0077] Resample the resolution of the areas with completely consistent cultivated land and non-cultivated land to construct a cell network;

[0078] Remote sensing image features are constructed and optimized based on multi-source land use data, and training sample migration is performed.

[0079] Preferably, to avoid the influence of mixed pixels and edge effects, the present embodiment resamples the resolution of the areas where the cultivated land and non-cultivated land in CLCD and GLAD are completely consistent from 30m to 90m. Pixels with a 3×3 range of 30m pixels and a land cover type of 100% cultivated land or non-cultivated land are extracted, and random sampling is performed on 1°×1° grids, i.e., the initial training samples are collected every four years. Within each 1°×1° grid, 1000 cultivated land samples and 200 samples of each of the other land use types are obtained. If the number of available training samples in the 1°×1° grid is less than the target number, all available training samples are used.

[0080] Preferably, in the process of constructing remote sensing image features, in addition to the blue (Blue), green (Green), red (Red), near infrared (NIR), and shortwave infrared (SWIR-1, SWIR-2) bands of Landsat satellite images, the normalized difference vegetation index (NDVI), enhanced vegetation index (EVI), chlorophyll index (GCVI), normalized burn index (NBR), normalized water body index (NDWI), normalized building index (NDBI), bare soil index (BSI), modified soil adjusted vegetation index (MSAVI2), normalized tillage index (NDTI) and spectral variability vegetation index (SVVI) are also calculated, totaling 16 indicators, as shown in the following formula:

[0081]

[0082] By synthesizing all available Landsat satellite images for the year using multiple temporal phases, we constructed 250 indices, including statistical indicators, time series indicators, harmonic function coefficient indicators, and terrain factor indicators, to fully reflect the differences in remote sensing image characteristics between cultivated and non-cultivated land at different times. The following are the details:

[0083] ① Statistical indicators, namely the minimum value (min), maximum value (max), quantile (1 / 4 quantile (Q1), median (Q2), 3 / 4 quantile (Q3)), range (d max-min ), interquartile range (d Q3-Q1 ), standard deviation (std), a total of 128 indicators; ② time series indicators, that is, sorting the dates from small to large to generate 16 spectral bands and indices of the annual time series, extracting their minimum value (min), maximum value (max), quantile (1 / 4 quantile (Q1), median (Q2), 3 / 4 quantile (Q3)), range (d max-min ), interquartile range (d Q3-Q1 ) and other spectral bands and index data corresponding to 7 dates, totaling 112 indicators; ③ harmonic function coefficient indicators, namely the intercept, slope, first-order, second-order and third-order fitting coefficients of the harmonic function fitting based on the NDVI index time series; ④ terrain factors, namely altitude, slope, aspect and other 3 factors.

[0084] Preferably, during the remote sensing image feature optimization process, a random forest model is used for feature optimization and subsequent model training. First, the training samples obtained every four years are integrated and based on the obtained remote sensing impact feature indicator set. Second, the random forest model is used for training, and the feature importance is evaluated based on the Gini impurity index. Then, the cross-validation recursive feature elimination method in backward feature elimination is used to continuously remove the 10% least important features, and 5-fold cross-validation is used to test the model accuracy until the training model accuracy reaches the highest. The feature set corresponding to the highest model accuracy is used as the final optimized remote sensing image feature set.

[0085] Preferably, during the training sample migration process, based on the initial training samples obtained for each land use type over a four-year period, a principal component analysis is performed on the remote sensing imagery of any target year within these four years. The obtained principal component variables are used to calculate the robust Mahalanobis distance from the initial training sample to the sample center, and a 3σ criterion is established to eliminate outliers in the initial training sample set, thereby effectively identifying outliers and obtaining training samples for the target year. For example, in this embodiment of the present application, the initial training samples from 2000-2003 are migrated to 2000, 2001, 2002, and 2003, respectively, ultimately obtaining training samples for the target years of 2000, 2001, 2002, and 2003.

[0086] Based on this, in step S102, based on the generated training sample set, the extracted potential cultivated land change areas are classified using a local random forest classifier, and the potential cultivated land change areas are classified into cultivated land areas and non-cultivated land areas. The cultivated land classification results and cultivated land probabilities in each 1°×1° cell network are generated year by year.

[0087] Among them, in order to take into account the accuracy and calculation time of the random forest algorithm, the embodiment of the present application adjusted three parameters of the random forest algorithm: 1) the number of decision trees is 500; 2) the sample ratio used in training each tree is set to 2 / 3, that is, each tree contains 2 / 3 of the training samples; 3) the minimum number of samples in the leaf nodes, this parameter is used to control the minimum number of samples in the leaf nodes of each tree to prevent over-segmentation and overfitting, and the minimum number of samples in the leaf nodes is set to 10.

[0088] In the embodiment of the present application, the process of correcting the cultivated land change result in combination with the cultivated land probability includes the following steps:

[0089] If the cultivated land probability is significantly different from other cultivated land probability segments in the time series, the cultivated land change results are retained; otherwise, they are discarded. The significant difference is determined based on the T-test of the Shapelet algorithm.

[0090] Preferably, in order to increase the description of inter-annual dynamic information of cultivated land as much as possible, the embodiment of the present application defines that cultivated land that has not been cultivated for field crops for at least three consecutive years is considered to be abandoned. Cultivated land that has not been cultivated for less than three years is regarded as fallow activity in this study and is not considered in cultivated land changes. The cultivated land classification results of the potential change areas of each year are synthesized to form a continuous time series of potential cultivated land changes. The cultivated land probability in the three-year window and the shapelet method are used to compare whether there is a significant difference (T test) between the cultivated land probability in the window and the other cultivated land probability segments in the time series to determine whether the change is reliable. If there is a significant difference, the change is retained; otherwise, the change is eliminated. As a supplement to the shapelet method, 0.5 is used as the threshold for cultivated land and non-cultivated land. The transition window is set to 2 years, and it is defined that if the average probability of cultivated land in the first 5-year window is greater than 0.5 for 3 consecutive years and the average probability of cultivated land in the last 5-year window is less than 0.5 for 3 consecutive years, it means that cultivated land has been reduced; conversely, if the average probability of cultivated land in the first 5-year window is less than 0.5 for 3 consecutive years and the average probability of cultivated land in the last 5-year window is greater than 0.5 for 3 consecutive years, it means that cultivated land has been added.

[0091] Based on this, in a preferred embodiment, a method for multi-period harmonic fitting of long-time series and identification of cultivated land changes based on a time-domain sliding window can amplify sudden changes, gradual changes, and phenological differences through the stability of segmented fitting trajectories, effectively addressing challenges such as missed weak signal identification and time deviations in change detection that exist in existing cultivated land change identification methods. Furthermore, through key steps such as multi-source land use data fusion, feature optimization, sample migration, machine learning classification, and false change identification and correction, this method not only expands the scope of cultivated land change detection but also provides more accurate detection locations for cultivated land changes, improving the accuracy and robustness of cultivated land change detection compared to traditional methods.

[0092] An embodiment of the present application also provides a device for detecting and identifying cultivated land changes based on time series feature vectors.

[0093] Figure 2 This is a module structure diagram of a device for detecting and identifying cultivated land changes based on time series feature vectors according to an embodiment of the application. Figure 2 As shown, an embodiment of a device for detecting and identifying cultivated land changes based on time series feature vectors includes:

[0094] The data acquisition module 100 is used to acquire remote sensing image data and multi-source land use data; wherein, the multi-source land use data is integrated to obtain a training sample set of cultivated land and non-cultivated land;

[0095] The data fusion module 101 is used to perform multi-period harmonic fitting of remote sensing image data based on a time domain sliding window to construct a cultivated land change vector, and calculate the relative change vector of the cultivated land change vector in adjacent time periods. The potential cultivated land change area is determined by comparing the relative change vector with a preset identification threshold, and the year of potential cultivated land change is determined by the peak value of the relative change vector.

[0096] The data classification module 102 is used to classify the potential cultivated land change area into cultivated land area and non-cultivated land area based on the training sample set, and generate corresponding cultivated land classification results and cultivated land probability;

[0097] The data detection module 103 is used to synthesize the cultivated land classification results to form a continuous time series of cultivated land change results, and to correct the cultivated land change results in combination with the cultivated land probability.

[0098] The cultivated land change detection and identification device based on time series feature vectors of the embodiment of the present application obtains remote sensing image data and multi-source land use data, and then performs multi-period harmonic fitting of the remote sensing image data based on a time domain sliding window to construct a cultivated land change vector, and calculates the relative change vector of the cultivated land change vector in adjacent time periods. The potential cultivated land change area is determined by comparing the relative change vector with a preset recognition threshold, and the year in which the potential cultivated land change occurs is determined by the peak value of the relative change vector. Based on the training sample set, the potential cultivated land change area is classified into cultivated land area and non-cultivated land area, and the corresponding cultivated land classification results and cultivated land probabilities are generated; the cultivated land classification results are synthesized to form a continuous time series of cultivated land change results, and the cultivated land change results are corrected in combination with the cultivated land probability. Based on the time domain sliding window, long time series data fusion is achieved to solve problems such as weak signal missed recognition and change detection time deviation, thereby improving the detection accuracy of cultivated land change dynamics.

[0099] At least one embodiment of the present application further provides a data control device. Figure 3 A schematic block diagram of a data control device provided in at least one embodiment of the present application. Figure 3 As shown, the data control device 20 may include one or more memories 200 and one or more processors 201. The memories 200 are used to non-transiently store computer-executable instructions; the processor 201 is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor 201, the processor 201 may execute one or more steps of the method for detecting and identifying cultivated land changes based on time series feature vectors according to any embodiment of the present application.

[0100] The specific implementation and related explanation of each step of the method for detecting and identifying cultivated land changes based on time series feature vectors can be found in the above-mentioned embodiments of the method for detecting and identifying cultivated land changes based on time series feature vectors, and will not be repeated here. Figure 3The components of the data control device 20 shown are merely exemplary and non-limiting. The data control device 20 may further include other components according to actual application requirements.

[0101] In one embodiment, the processor 201 and the memory 200 can communicate with each other directly or indirectly. For example, the processor 201 and the memory 200 can communicate via a network connection. The network can include a wireless network, a wired network, and / or any combination of wireless and wired networks. This application does not limit the type and function of the network. For another example, the processor 201 and the memory 200 can also communicate via a bus connection. The bus can be a Peripheral Component Interconnect (PCI) bus or an Extended Industrial Standard Architecture (EISA) bus. For example, the processor 201 and the memory 200 can be located on a remote data server (cloud) or a distributed energy system (local), or on a client (e.g., a mobile device such as a mobile phone). For example, the processor 201 can be a device with data processing capabilities and / or instruction execution capabilities, such as a central processing unit (CPU), a tensor processing unit (TPU), or a graphics processing unit (GPU), and can control other components in the data control device 20 to perform the desired functions. The central processing unit (CPU) can be an X86 or ARM architecture, etc.

[0102] In one embodiment, the memory 200 may include any combination of one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and / or cache memory. Non-volatile memory may include, for example, read-only memory (ROM), a hard disk, an erasable programmable read-only memory (EPROM), a portable compact disk read-only memory (CD-ROM), a USB memory, a flash memory, etc. One or more computer-executable instructions may be stored on the computer-readable storage medium, and the processor 201 may execute the computer-executable instructions to implement various functions of the data control device 20. Various applications and various data, as well as various data used and / or generated by the applications, may also be stored in the memory 200.

[0103] It should be noted that the data control device 20 can achieve technical effects similar to those of the aforementioned cultivated land change detection and identification method based on time series feature vectors, and the repeated parts will not be repeated.

[0104] At least one embodiment of the present application also provides a non-transitory computer-readable storage medium. Figure 4 A schematic diagram of a non-transitory computer-readable storage medium provided for at least one embodiment of the present application. For example, Figure 4As shown, one or more computer-executable instructions 301 may be non-transitory stored on a non-transitory computer-readable storage medium 30. For example, when the computer-executable instructions 301 are executed by a computer, the computer may execute one or more steps in the method for detecting and identifying cultivated land changes based on time series feature vectors according to any embodiment of the present application.

[0105] In one embodiment, the non-transitory computer-readable storage medium 30 may be applied to the above-mentioned data control device 20 , for example, it may be the memory 200 in the data control device 20 .

[0106] In one embodiment, the description of the non-transitory computer-readable storage medium 30 may refer to the description of the memory 200 in the embodiment of the data control device 20 , and the repeated parts will be omitted.

[0107] It should be noted that the memory 200 stores different non-transient computer executable instructions, and the data control device 20 corresponds to a firmware upgrade device. When the computer executable instructions are executed by the processor 201, the processor 201 can execute one or more steps in the method for detecting and identifying cultivated land changes based on time series feature vectors according to any embodiment of the present application.

[0108] Regarding this application, the following points need to be explained:

[0109] (1) The drawings of the embodiments of this application only relate to the structures related to the embodiments of this application. Other structures can refer to the general design.

[0110] (2) Unless there is a conflict, the embodiments of this application and the features therein may be combined to form new embodiments. The above are only specific implementation methods of this application, but the scope of protection of this application is not limited thereto. The scope of protection of this application shall be based on the scope of protection of the claims.

[0111] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0112] The above embodiments merely illustrate several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, all of which fall within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be determined by the appended claims.

Claims

1. A method for detecting and identifying cultivated land changes based on time series feature vectors, characterized in that: Including steps: Acquiring remote sensing image data and multi-source land use data; wherein the multi-source land use data are integrated to obtain a training sample set of cultivated land and non-cultivated land; Perform multi-period harmonic fitting of remote sensing image data based on a time domain sliding window to construct a cultivated land change vector, and calculate the relative change vector of the cultivated land change vector in adjacent time periods. Determine the potential cultivated land change area by comparing the relative change vector with a preset recognition threshold, and identify the year when the potential cultivated land change occurs by the peak value of the relative change vector. Classifying the potential cultivated land change area into cultivated land area and non-cultivated land area based on the training sample set, and generating corresponding cultivated land classification results and cultivated land probabilities; synthesizing the cultivated land classification results to form a continuous time series of cultivated land change results, and correcting the cultivated land change results in combination with the cultivated land probability; The relative change vector of the cultivated land change vector in adjacent time periods is calculated using the following formula: ； ； ； ； in, for The normalized relative change vector of pixel t in year, represents the standardized vector difference of pixel i in year t, and They are pixel i from year t to year t+w and The harmonic function fitting vector of the two time periods from year to year t; The calculated parameters include the intercept of the harmonic function from year t to year t+w and seven harmonic fitting parameters of the first, second and third order, namely ; The calculated parameters include The intercept of the harmonic function from year to year t and the first-order, second-order and third-order harmonic fitting parameters are: ; 、 、 、 From t to and The intercept and RMS error of the harmonic function fit for the two time periods from year to year t.

2. The method for detecting and identifying cultivated land changes based on time series feature vectors according to claim 1, characterized in that: The multi-period harmonic fitting of remote sensing image data based on the time domain sliding window is performed using the following formula: ； ； in, Represents any pixel within the maximum cultivated land area; represents any Julian day with remote sensing observations, ranging from all Julian days observed in observation year t to observation year t+w; w is the time window of the function fitting; express Pixel in time The predicted value of NDVI index; For time Remote sensing observation value of NDVI index; is the number of days in a year; is the overall amplitude or intercept, reflecting the average growth status of vegetation in the period from year t to year t + w; is the slope, reflecting the interannual variation of vegetation from year t to year t + w; is the order of the harmonic component, reflecting the annual changes in vegetation at different frequencies, where and 、 and 、 and It reflects the annual changes once a year, twice a year and three times a year respectively.

3. The method for detecting and identifying cultivated land changes based on time series feature vectors according to claim 1, characterized in that: The following two formulas are used to identify the year in which potential cultivated land change occurs: ≥CT&&; max[SRCV(t-2 ),SRCV(t-1 ),SRCV(t ),SRCV(t+1 ),SRCV(t+2 )]=SRCV(t ); Among them, CT represents the preset identification threshold, and the area that meets the above two equations is the potential cultivated land change area; SRCV(t-2 ),SRCV(t-1 ),SRCV(t ),SRCV(t+1 ),SRCV(t+2 ) are the standardized relative change vectors of year t-2, year t-1, year t, year t+1, and year t+2, respectively. When the above two formulas are met, the standardized relative change vector of year t is greater than that of the other years t-2, t-1, t+1, and t+2; Then year t is the year when potential cultivated land change occurs.

4. The method for detecting and identifying cultivated land changes based on time series feature vectors according to claim 1, characterized in that: The multi-source land use data includes CLCD data and GLAD data; The process of fusing the multi-source land use data to obtain a training sample set of cultivated land and non-cultivated land includes the following steps: Reclassify CLCD data into land use types; The CLCD data and CLCD data of the same time period were superimposed to obtain the same land use type area, and then superimposed on the cultivated land and non-cultivated land ranges of the corresponding time period of the GLAD product to generate completely consistent areas of cultivated land and non-cultivated land, and the corresponding training sample sets of cultivated land and non-cultivated land were constructed.

5. The method for detecting and identifying cultivated land changes based on time series feature vectors according to claim 4 is characterized in that: Before constructing the training sample set of cultivated land and non-cultivated land, the following steps are also included: Resampling the areas where the cultivated land and non-cultivated land are completely consistent with each other to construct a cell network; Remote sensing image features are constructed and optimized based on the multi-source land use data, and training sample migration is performed.

6. The method for detecting and identifying cultivated land changes based on time series feature vectors according to claim 1, characterized in that: The process of correcting the cultivated land change result in combination with the cultivated land probability comprises the steps of: If there is a significant difference between the cultivated land probability and other cultivated land probability segments in the time series, the cultivated land change result is retained, otherwise it is eliminated; wherein the significant difference is judged based on the shapelet algorithm.

7. A device for detecting and identifying cultivated land changes based on time series feature vectors, characterized in that: include: A data acquisition module is used to acquire remote sensing image data and multi-source land use data; wherein, the multi-source land use data is integrated to obtain a training sample set of cultivated land and non-cultivated land; A data fusion module is used to perform multi-period harmonic fitting of remote sensing image data based on a time domain sliding window to construct a cultivated land change vector, and calculate the relative change vector of the cultivated land change vector in adjacent time periods, determine the potential cultivated land change area by comparing the relative change vector with a preset recognition threshold, and identify the year when the potential cultivated land change occurs by the peak value of the relative change vector; A data classification module is used to classify the potential cultivated land change area into cultivated land area and non-cultivated land area based on the training sample set, and generate corresponding cultivated land classification results and cultivated land probabilities; a data detection module, configured to synthesize the cultivated land classification results to form a continuous time series of cultivated land change results, and to correct the cultivated land change results in combination with the cultivated land probability; The relative change vector of the cultivated land change vector in adjacent time periods is calculated using the following formula: ； ； ； ； in, for The normalized relative change vector of pixel t in year, represents the standardized vector difference of pixel i in year t, and They are pixel i from year t to year t+w and The harmonic function fitting vector of the two time periods from year to year t; The calculated parameters include the intercept of the harmonic function from year t to year t+w and seven harmonic fitting parameters of the first, second and third order, namely ; The calculated parameters include The intercept of the harmonic function from year to year t and the first-order, second-order and third-order harmonic fitting parameters are: ; 、 、 、 From t to and The intercept and RMS error of the harmonic function fit for the two time periods from year to year t.

8. A non-transitory computer-readable storage medium, characterized in that The non-transitory computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are executed by a processor, the cultivated land change detection and identification method based on time series feature vectors as described in any one of claims 1 to 6 is implemented.

9. A data control device, characterized in that: include: one or more memories non-transitorily storing computer-executable instructions; One or more processors are configured to run computer-executable instructions, wherein the computer-executable instructions, when run by one or more processors, implement the method for detecting and identifying cultivated land changes based on time series feature vectors as described in any one of claims 1 to 6.

Images