Crop classification method and system based on adaptive time window time sequence characteristics

Abstract

The invention provides a crop classification method and system based on self-adaptive time window time sequence characteristics, and the method comprises the steps: carrying out the preprocessing of long-time sequence data, constructing the optimal time sequence characteristics based on a self-adaptive strategy, and generating a crop distribution diagram through the characteristics. The method focuses on the winter crops and the summer crops in different growth seasons, the time windows of the vigorous growth periods of the crops are automatically positioned, so that the time sequence characteristics capable of representing the difference between the crops to the maximum degree are respectively customized and constructed, and the accuracy and the effectiveness of classification drawing are improved.

Description

Technical Field

[0001] The present invention relates to the field of agricultural remote sensing technology, and in particular to a crop classification method and system based on adaptive time window temporal characteristics. Background Art

[0002] Different types of crops show unique temporal variation patterns according to their inherent biological characteristics and growth cycles. Time series remote sensing images can capture and effectively depict the characteristics of different types of land objects that change with seasons and years, greatly improving the information dimension. This enhanced information dimension can not only make up for the defects of image intermittence and spectral signal saturation caused by cloudy and rainy weather, but also effectively alleviate the classic "same object, different spectrum" and "different objects, same spectrum" problems in remote sensing classification. Therefore, how to determine the optimal time window to accurately capture the key growth stages of various crops has become the key to improving the accuracy and practicality of crop classification mapping.

[0003] Since crops in different geographical locations are affected by factors such as climate conditions and planting habits, their growth cycles and growth rates vary. Existing studies that use a fixed and unified time window are often insufficient to reflect the dynamic characteristics of crop growth, and it is challenging to find a universal and efficient time window for various crops. Summary of the invention

[0004] The present invention provides a crop classification method and system based on adaptive time window temporal characteristics, which are used to solve the defects in the prior art.

[0005] In a first aspect, the present invention provides a crop classification method based on adaptive time window temporal features, comprising: Acquire a long-time sequence optical image within the temporal and spatial range of the target area, and preprocess the long-time sequence optical image to obtain preprocessed long-time sequence data; Performing window initialization and time series feature extraction on the preprocessed long time series data to obtain preliminary time series extraction features; Using the preliminary time series extracted features, the separability of different crops is measured and evaluated to obtain an initial separability evaluation result; Iteratively optimizing and updating the time series features that do not reach the maximum separability between crops in the initial separability evaluation results to obtain final time series features; The random forest model is trained based on the final time series features to obtain a trained crop classification model, and the crop classification result is obtained using the crop classification model.

[0006] According to a crop classification method based on adaptive time window temporal characteristics provided by the present invention, the long time series optical image is preprocessed to obtain preprocessed long time series data, including: The long time series optical images are subjected to cropping, mean synthesis, linear interpolation and SG filtering processing in sequence to obtain the preprocessed long time series data.

[0007] According to a crop classification method based on adaptive time window time series features provided by the present invention, window initialization and time series feature extraction are performed on the preprocessed long time series data to obtain preliminary time series extraction features, including: Sorting the preprocessed long time series data, determining a growth peak definition index within a specified percentile range, and using the long time series data corresponding to the growth peak definition index as an initial growth peak data threshold; The first time point in the preprocessed long time series data that exceeds the initial growth peak data threshold is used as the time window starting point, and the last time point that exceeds the initial growth peak data threshold is used as the time window ending point to obtain the initial time window; In the initial time window, the time series extreme value difference and the data point standard deviation of the preprocessed long time series data are calculated, the time series dispersion is obtained from the time series extreme value difference and the data point standard deviation, and the preliminary time series extraction features are obtained based on the time series dispersion and the duration of the time series.

[0008] According to a crop classification method based on adaptive time window time series features provided by the present invention, the separability of different crops is measured and evaluated by using the preliminary time series extraction features to obtain an initial separability evaluation result, including: The preliminary time series extraction features are calculated using Mahalanobis distance and Manhattan distance respectively to obtain a first distance measurement value and a second distance measurement value; An average of the first distance metric value and the second distance metric value is calculated as the initial separability evaluation result.

[0009] According to a crop classification method based on adaptive time window time series features provided by the present invention, the time series features that do not reach the maximum separability between crops in the initial separability evaluation results are iteratively optimized and updated to obtain the final time series features, including: According to the initial separability evaluation result, adjusting the growth peak definition index to obtain a new optimal time window, and re-extracting the time series extraction features within the optimal time window; Recalculating the Mahalanobis distance and the Manhattan distance of the time series extraction features within the optimal time window to obtain a new separability evaluation result; Using a genetic algorithm to adjust the growth peak definition index and the optimal time window until the separability evaluation result reaches or approaches a maximum value; When the iterative optimization no longer significantly improves the crop separability evaluation index, the iteration is stopped to determine the final time series feature.

[0010] According to a crop classification method based on adaptive time window time series features provided by the present invention, a random forest model is trained based on the final time series features to obtain a trained crop classification model, including: The final time series features are used as input data of the random forest model, and the category label data are used as output data, and are randomly divided into a training set and a test set according to a preset division ratio; The random forest model is trained, parameter optimized and verified using the training set and the test set, and the trained crop classification model is output.

[0011] In a second aspect, the present invention further provides a crop classification system based on adaptive time window temporal features, comprising: A preprocessing module, used for acquiring a long-time sequence optical image within the temporal and spatial range of a target area, and preprocessing the long-time sequence optical image to obtain preprocessed long-time sequence data; An extraction module, used to perform window initialization and time series feature extraction on the preprocessed long time series data to obtain preliminary time series extraction features; An evaluation module, used to measure and evaluate the separability of different crops by using the preliminary time series extraction features to obtain an initial separability evaluation result; An updating module, used for iteratively optimizing and updating the time series features that do not reach the maximum separability between crops in the initial separability evaluation results, to obtain final time series features; A training module is used to train the random forest model based on the final time series features to obtain a trained crop classification model, and obtain a crop classification result using the crop classification model.

[0012] In a third aspect, the present invention further provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for classifying crops based on adaptive time window timing features as described above is implemented.

[0013] In a fourth aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-described crop classification methods based on adaptive time window temporal features.

[0014] The crop classification method and system based on adaptive time window temporal features provided by the present invention automatically locate the time windows of the crop growth peak periods for winter crops and summer crops in different growing seasons, thereby customizing and constructing temporal features that can maximize the characterization of differences between crops, thereby improving the accuracy and effectiveness of classification mapping. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0016] Figure 1 This is one of the flow charts of the crop classification method based on the adaptive time window temporal characteristics provided by the present invention; Figure 2 This is the second flow chart of the crop classification method based on the adaptive time window temporal characteristics provided by the present invention; Figure 3 It is a structural schematic diagram of a crop classification system based on adaptive time window temporal characteristics provided by the present invention; Figure 4 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] Figure 1 is one of the flow charts of the crop classification method based on adaptive time window temporal features provided by an embodiment of the present invention, such as Figure 1 As shown, including: Step 100: acquiring a long time series optical image within the time and space range of the target area, and preprocessing the long time series optical image to obtain preprocessed long time series data; Step 200: performing window initialization and time series feature extraction on the preprocessed long time series data to obtain preliminary time series extraction features; Step 300: using the preliminary time series extraction features, measuring and evaluating the separability of different crops to obtain an initial separability evaluation result; Step 400: iteratively optimizing and updating the time series features that do not reach the maximum separability between crops in the initial separability evaluation results to obtain final time series features; Step 500: training a random forest model based on the final time series features to obtain a trained crop classification model, and obtaining a crop classification result using the crop classification model.

[0019] The embodiment of the present invention selects a province in China as the research area, and selects five field crops, namely wheat, corn, peanuts, soybeans, and rice in 2019 as research objects.

[0020] like Figure 2 As shown, step 100 includes: The long time series optical images are subjected to cropping, mean synthesis, linear interpolation and SG filtering processing in sequence to obtain the preprocessed long time series data.

[0021] Specifically, the long-time optical sequence images in the embodiment of the present invention are derived from Sentinel-2 and downloaded through the GEE (Google Earth Engine) cloud platform.

[0022] The long time series Sentinel-2 optical images are cropped, mean synthesized, linearly interpolated, and SG filtered for smoothing. Specifically, the Sentinel-2 images are cropped to extract the required provincial area. The 2019 full year images are synthesized for 10 days. Invalid pixels are filled using linear interpolation with a window size of 9, specifically, the invalid pixels are interpolated using valid pixels of the nearest date before and after the invalid pixels; the time series images are smoothed using SG filtering with a window size of 2 and a polynomial degree of 3. The processed image data is made more consistent and smooth, thereby reducing the impact of noise and data loss, and improving the quality and continuity of time series data.

[0023] Step 200 includes: Sorting the preprocessed long time series data, determining a growth peak definition index within a specified percentile range, and using the long time series data corresponding to the growth peak definition index as an initial growth peak data threshold; The first time point in the preprocessed long time series data that exceeds the initial growth peak data threshold is used as the time window starting point, and the last time point that exceeds the initial growth peak data threshold is used as the time window ending point to obtain the initial time window; In the initial time window, the time series extreme value difference and the data point standard deviation of the preprocessed long time series data are calculated, the time series dispersion is obtained from the time series extreme value difference and the data point standard deviation, and the preliminary time series extraction features are obtained based on the time series dispersion and the duration of the time series.

[0024] Specifically, the initial growth peak index is set first: the time series data NDVI is calculated from the preprocessed Sentinel-2 optical image, and the NDVI is sorted to set the growth peak index (Growth Peak Index, GPI), and the time series value corresponding to the GPI value is recorded as The GPI is a metric defined based on a specific percentile in time series data, ranging from 50 to 100.

[0025] Then the initial time window is determined based on the initial GPI: in the original time series data, the first The time point is taken as the start of the time window, and the last one exceeds The time point of is taken as the end of the window, and this time period is defined as the time window. The time period where the initial GPI value is positioned is called the initial time window. Specifically, in this embodiment, the initial GPI value is set to 50, and the initial time window and the initial timing characteristics are determined based on the GPI value.

[0026] Finally, the initial time series characteristics are calculated: Timing Dispersion (TDI) is a statistic that measures the magnitude and rate of change in the NDVI time series. It calculates the time series extreme value difference ( ) and the standard deviation of these data points ( ), to characterize the volatility of the data. Timing duration (TDU) is used to describe the duration of the NDVI time series. The specific calculation is as follows:

[0027]

[0028] Step 300 includes: The preliminary time series extraction features are calculated using Mahalanobis distance and Manhattan distance respectively to obtain a first distance measurement value and a second distance measurement value; An average of the first distance metric value and the second distance metric value is calculated as the initial separability evaluation result.

[0029] Specifically, by adopting the separability index calculation, two measurement methods, Mahalanobis distance and Manhattan distance, are introduced, and the average of these two distance metrics is used as the quantitative indicator of crop separability.

[0030] Step 400 includes: According to the initial separability evaluation result, adjusting the growth peak definition index to obtain a new optimal time window, and re-extracting the time series extraction features within the optimal time window; Recalculating the Mahalanobis distance and the Manhattan distance of the time series extraction features within the optimal time window to obtain a new separability evaluation result; Using a genetic algorithm to adjust the growth peak definition index and the optimal time window until the separability evaluation result reaches or approaches a maximum value; When the iterative optimization no longer significantly improves the crop separability evaluation index, the iteration is stopped to determine the final time series feature.

[0031] Specifically, the first step is to adjust the GPI value and redefine the time window. According to the current separability evaluation result, the GPI value is adjusted to redefine the optimal time window and re-extract the timing features within the adaptive time window, including timing discreteness and timing continuity.

[0032] Secondly, the separability evaluation is repeatedly performed. The Mahalanobis distance and Manhattan distance are recalculated for the new time series characteristics, and their resolution for different crops is evaluated. The GPI value and time window are adjusted using a genetic algorithm until the separability index reaches or approaches the maximum value.

[0033] Finally, the iteration process is terminated. When further iterative optimization no longer significantly improves the separability of crops, the iteration is stopped. Specifically, for corn, peanuts, soybeans, rice, and wheat, when the GPI values ​​are set to 63, 62, 69, 53, and 87, respectively, the corresponding time windows can maximize the spectral differences between crops, thereby constructing effective time series discreteness and time series continuity features. Confirm the final time series features to provide optimized input data for subsequent model training and distribution map production.

[0034] Step 500 includes: The final time series features are used as input data of the random forest model, and the category label data are used as output data, and are randomly divided into a training set and a test set according to a preset division ratio; The random forest model is trained, parameter optimized and verified using the training set and the test set, and the trained crop classification model is output.

[0035] Specifically, the obtained time series features are used as input data to generate the random forest model, and the category label data is used as output data, and the training set and test set are randomly divided into two sets in a ratio of 8:2.

[0036] Construct a random forest model, train, optimize parameters and verify the random forest model to obtain a trained random forest model.

[0037] Use the trained model to create a crop distribution map of a certain province in 2019.

[0038] The following is a description of the crop classification system based on adaptive time window timing features provided by the present invention. The crop classification system based on adaptive time window timing features described below and the crop classification method based on adaptive time window timing features described above can be referenced to each other.

[0039] Figure 3 is a schematic diagram of the structure of a crop classification system based on adaptive time window temporal features provided by an embodiment of the present invention. Figure 3 As shown, it includes: a preprocessing module 31, an extraction module 32, an evaluation module 33, an update module 34 and a training module 35, wherein: The preprocessing module 31 is used to obtain long-time series optical images within the time and space range of the target area, and preprocess the long-time series optical images to obtain preprocessed long-time series data; the extraction module 32 is used to perform window initialization and time series feature extraction on the preprocessed long-time series data to obtain preliminary time series extraction features; the evaluation module 33 is used to measure and evaluate the separability of different crops using the preliminary time series extraction features to obtain initial separability evaluation results; the updating module 34 is used to iteratively optimize and update the time series features that have not reached the maximum separability between crops in the initial separability evaluation results to obtain final time series features; the training module 35 is used to train the random forest model based on the final time series features to obtain a trained crop classification model, and obtain crop classification results using the crop classification model.

[0040] Figure 4 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 4As shown, the electronic device may include: a processor (processor) 410, a communication interface (Communications Interface) 420, a memory (memory) 430 and a communication bus 440, wherein the processor 410, the communication interface 420, and the memory 430 communicate with each other through the communication bus 440. The processor 410 can call the logic instructions in the memory 430 to execute a crop classification method based on adaptive time window timing features, the method comprising: acquiring a long time series optical image within the time and space range of the target area, preprocessing the long time series optical image to obtain preprocessed long time series data; performing window initialization and timing feature extraction on the preprocessed long time series data to obtain preliminary timing extraction features; using the preliminary timing extraction features, measuring and evaluating the separability of different crops to obtain initial separability evaluation results; iteratively optimizing and updating the timing features that do not reach the maximum separability between crops in the initial separability evaluation results to obtain final timing features; training a random forest model based on the final timing features to obtain a trained crop classification model, and using the crop classification model to obtain a crop classification result.

[0041] In addition, the logic instructions in the above-mentioned memory 430 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0042] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the crop classification method based on adaptive time window time series features provided by the above methods, the method including: acquiring a long time series optical image within the time and space range of the target area, preprocessing the long time series optical image to obtain preprocessed long time series data; performing window initialization and time series feature extraction on the preprocessed long time series data to obtain preliminary time series extraction features; using the preliminary time series extraction features, measuring and evaluating the separability of different crops to obtain initial separability evaluation results; iteratively optimizing and updating the time series features that do not reach the maximum separability between crops in the initial separability evaluation results to obtain final time series features; training a random forest model based on the final time series features to obtain a trained crop classification model, and using the crop classification model to obtain a crop classification result.

[0043] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which is implemented when the processor executes the crop classification method based on adaptive time window time series features provided by the above methods, the method comprising: acquiring a long time series optical image within the time and space range of the target area, preprocessing the long time series optical image to obtain preprocessed long time series data; performing window initialization and time series feature extraction on the preprocessed long time series data to obtain preliminary time series extraction features; using the preliminary time series extraction features, measuring and evaluating the separability of different crops to obtain initial separability evaluation results; iteratively optimizing and updating the time series features that do not reach the maximum separability between crops in the initial separability evaluation results to obtain final time series features; training a random forest model based on the final time series features to obtain a trained crop classification model, and using the crop classification model to obtain a crop classification result.

[0044] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0045] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0046] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A crop classification method based on adaptive time window temporal features, characterized in that: include: Acquire a long-time sequence optical image within the temporal and spatial range of the target area, and preprocess the long-time sequence optical image to obtain preprocessed long-time sequence data; Performing window initialization and time series feature extraction on the preprocessed long time series data to obtain preliminary time series extraction features; Using the preliminary time series extracted features, the separability of different crops is measured and evaluated to obtain an initial separability evaluation result; Iteratively optimizing and updating the time series features that do not reach the maximum separability between crops in the initial separability evaluation results to obtain final time series features; The random forest model is trained based on the final time series features to obtain a trained crop classification model, and the crop classification result is obtained using the crop classification model.

2. The crop classification method based on adaptive time window temporal features according to claim 1 is characterized in that: Preprocessing the long time series optical images to obtain preprocessed long time series data includes: The long time series optical images are subjected to cropping, mean synthesis, linear interpolation and SG filtering processing in sequence to obtain the preprocessed long time series data.

3. The crop classification method based on adaptive time window temporal features according to claim 1 is characterized in that: Perform window initialization and time series feature extraction on the preprocessed long time series data to obtain preliminary time series extraction features, including: Sorting the preprocessed long time series data, determining a growth peak definition index within a specified percentile range, and using the long time series data corresponding to the growth peak definition index as an initial growth peak data threshold; The first time point in the preprocessed long time series data that exceeds the initial growth peak data threshold is used as the time window starting point, and the last time point that exceeds the initial growth peak data threshold is used as the time window ending point to obtain the initial time window; In the initial time window, the time series extreme value difference and the data point standard deviation of the preprocessed long time series data are calculated, the time series dispersion is obtained from the time series extreme value difference and the data point standard deviation, and the preliminary time series extraction features are obtained based on the time series dispersion and the duration of the time series.

4. The crop classification method based on adaptive time window temporal features according to claim 3 is characterized in that: The preliminary time series extraction features are used to measure and evaluate the separability of different crops, and the initial separability evaluation results are obtained, including: The preliminary time series extraction features are calculated using Mahalanobis distance and Manhattan distance respectively to obtain a first distance measurement value and a second distance measurement value; An average of the first distance metric value and the second distance metric value is calculated as the initial separability evaluation result.

5. The crop classification method based on adaptive time window temporal features according to claim 4 is characterized in that: Iteratively optimize and update the time series features that do not reach the maximum separability between crops in the initial separability evaluation results to obtain the final time series features, including: According to the initial separability evaluation result, adjusting the growth peak definition index to obtain a new optimal time window, and re-extracting the time series extraction features within the optimal time window; Recalculating the Mahalanobis distance and the Manhattan distance of the time series extraction features within the optimal time window to obtain a new separability evaluation result; Using a genetic algorithm to adjust the growth peak definition index and the optimal time window until the separability evaluation result reaches or approaches a maximum value; When the iterative optimization no longer significantly improves the crop separability evaluation index, the iteration is stopped to determine the final time series feature.

6. The crop classification method based on adaptive time window temporal features according to claim 1, characterized in that: The random forest model is trained based on the final time series features to obtain a trained crop classification model, including: The final time series features are used as input data of the random forest model, and the category label data are used as output data, and are randomly divided into a training set and a test set according to a preset division ratio; The random forest model is trained, parameter optimized and verified using the training set and the test set, and the trained crop classification model is output.

7. A crop classification system based on adaptive time window temporal features, characterized in that: include: A preprocessing module, used for acquiring a long-time sequence optical image within the temporal and spatial range of a target area, and preprocessing the long-time sequence optical image to obtain preprocessed long-time sequence data; An extraction module, used to perform window initialization and time series feature extraction on the preprocessed long time series data to obtain preliminary time series extraction features; An evaluation module, used to measure and evaluate the separability of different crops by using the preliminary time series extraction features to obtain an initial separability evaluation result; An updating module, used for iteratively optimizing and updating the time series features that do not reach the maximum separability between crops in the initial separability evaluation results, to obtain final time series features; A training module is used to train the random forest model based on the final time series features to obtain a trained crop classification model, and obtain a crop classification result using the crop classification model.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the crop classification method based on adaptive time window temporal features as described in any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the crop classification method based on adaptive time window temporal features as claimed in any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the crop classification method based on adaptive time window temporal features as claimed in any one of claims 1 to 6 is implemented.

Images