A cross-regional crop harvest period monitoring method based on deep learning

Abstract

The present invention provides a cross-regional crop harvest period monitoring method based on deep learning, including: obtaining high-temporal and spatial resolution remote sensing data within the corresponding crop growing season in the target area, and generating multiple groups of vegetation index time series; extracting pixels from each group of vegetation index time series based on a classification model to obtain the time series characteristics of the pixels, and obtaining classification results based on the time series characteristics; based on the deep learning algorithm and the classification results, performing deep feature extraction on the space and spectral space of the high-temporal and spatial resolution remote sensing data to obtain feature-level data; constructing an initial neural network model, and training the initial neural network model based on the feature-level data and the classification results; inputting the data to be tested into the cross-regional crop harvest period monitoring model to obtain monitoring results. The present invention can use large-scale remote sensing images to perform high-precision monitoring of crop harvest periods, thereby improving the automation and intelligence of cross-regional crop harvest period monitoring.

Description

Technical Field

[0001] The present invention relates to the field of remote sensing monitoring technology, and in particular to a cross-regional crop harvest period monitoring method based on deep learning. Background Art

[0002] Agricultural production is highly seasonal, and timely operations have a great impact on yields, and the timing of harvesting operations has a particularly significant impact on yields. For example, if rice is harvested too early, the crop will not fully develop, the grain moisture content is high but the dry matter content such as starch is low, the thousand-grain weight is low, and the rate of broken rice during processing is high; if the grain is harvested too late, the thousand-grain weight of the grain will decrease, and the probability of reduced or even total yields due to bad weather will increase. It will be more vulnerable to attacks from insects, birds, beasts and microorganisms, resulting in reduced yields. Timely harvesting is crucial to reducing losses in the harvesting process of crops.

[0003] (1) Current status of research on timeliness loss theory

[0004] Since ancient times, my country's agricultural production has always been about keeping the farming season. "Lüshi Chunqiu" mentioned that "the way of farming is to be careful about the season", and timely operation is the prerequisite for stable production. The loss caused by delayed operation is called timely loss, which consists of three parts: agronomy, nature and operation process. Among them, the timely loss of planting and harvesting is the focus.

[0005] In terms of planting, Wang Jinwu et al. (2004) set up 22 plots in the same paddy field, planted rice in one plot every day for 22 consecutive days, measured the yield of each plot after the rice matured, and calculated the quantitative relationship between rice yield and transplanting period, but he did not reveal the mechanism of yield change. The rice transplanting period has a significant effect on yield, mainly because there are significant differences in the number of effective tillers, panicles, population growth rate, temperature and light utilization efficiency, and photosynthetic material production characteristics of rice at different transplanting periods. In order to explore the suitable sowing period for regional rice, Huang Zheng et al. (2021) used five hybrid rice varieties such as Chuanyou 6203 as experimental materials, and adopted a two-factor split-plot design to study the effects of staged direct seeding on the growth process and yield of hybrid rice. It was found that the rice yield showed a trend of increasing first and then decreasing with the delay of transplanting period.

[0006] In terms of harvesting, Wang Jinwu (2004) used a randomized block experiment to measure the yield of rice planted on the same day in the same field, and measured it every other day, and obtained a quantitative relationship between rice yield loss and harvesting period. Huyen et al. (2010) measured the chemical components of various organs of Miscanthus that can be converted into fermentable sugars at different harvesting periods, and found that early harvesting can obtain greater benefits than late harvesting. Godin et al. (2013) evaluated the impact of harvesting period on the yield of various energy crops converted into biofuels, and found that if harvested too early or too late, biofuel yield would decrease. Heidari et al. (2012) determined the impact of harvesting period on fennel yield and other factors through experiments, and the results showed that fennel yield was the highest during maturity. Qiao Jinyou et al. (2017) studied the relationship between soybean actual yield and harvesting date, and obtained the law of loss of soybean mechanized harvesting timeliness. Du Zhimin et al. (2018) found through experiments that with the delay of the harvest period, the brown rice rate and polished rice rate of japonica rice first gradually increased, reached the maximum value 65 days after full heading, and then decreased significantly. The polished rice rate fluctuated greatly before 65 days after full heading, with no obvious regularity, and decreased significantly after 65 days after full heading.

[0007] (2) Research progress on methods for determining the appropriate harvest period

[0008] Li Zhenqing (1986) proposed that the harvesting day with the lowest loss of timely harvest of crops is the best harvesting day, and the harvesting period with the smallest total loss of timely harvest of rice is the suitable harvesting period. Since the loss of rice harvest changes with the harvesting period in a quadratic equation, the best harvesting day should be in the middle of the suitable harvesting period. At present, the main methods for determining the suitable harvesting period of crops are: determination based on crop canopy characterization and determination based on the content of specific substances in agricultural products.

[0009] Based on the judgment of crop canopy surface features. Hyperspectral remote sensing is widely used in agricultural crops, and has been maturely applied in crop biological and biochemical parameter inversion, agricultural production monitoring, and agricultural information monitoring. Wan et al. (2018) designed a maturity detection device based on computer vision technology, which determined the maturity by extracting the color feature values ​​of the tomato surface, with an average accuracy of 99.31%. Pereira et al. (2018) used digital imaging technology and random forest algorithm to predict the maturity of papaya fruit, and verified the prediction results by measuring the hardness of papaya pulp. This type of judgment method is fast and convenient, non-destructive to crops, and the equipment cost is relatively high. Shen Yu et al. (2020) used hyperspectral imaging technology to identify the effective band of the automatic grading instrument for slight damage to apples. The support vector machine (SVM) damage recognition model achieved an identification rate of 90.63% for slight damage to apples in the band of 811nm. Ge Yonghui et al. (2022) used hyperspectral imaging technology (400-1000nm) to study the non-destructive identification of kiwifruit chilling damage. The continuous projection algorithm (SPA) selected a better model for characteristic wavelengths than the full wavelength model, with the accuracy of the modeling set and prediction set being 100% and 94.2%, respectively. Guo Jingjing et al. (2022) proposed that the model based on the SG+FD+CARS+LSSVM combination method is the optimal lettuce greenness determination model, which can achieve 64.59% of the extracted sensitive wavelengths of the full wavelength. Compared with the original hyperspectral (1.25%), the number of extracted sensitive wavelengths increased by 63.34%.

[0010] Determination based on the content of specific substances in agricultural products. Wei Changling et al. (2017) extracted volatile oil from perilla leaves by steam distillation and determined the relative content of its characteristic components, perilla aldehyde, perilla ketone or perilla ene, to determine the maturity of perilla. This type of discrimination method has low cost, low efficiency and poor timeliness

[53] . Pan Yihong et al. (2017) took different varieties of mid-term tobacco leaves as the research object and determined the chlorophyll content and SPAD value of tobacco leaves. They found that from moderate maturity, maturity to over-maturity, the chlorophyll a, chlorophyll b, total chlorophyll content and SPAD value of tobacco leaves decreased significantly. Zhang Yuqin et al. (2018) used a fully automatic colorimeter to measure the change characteristics of color parameters (L*, a*, C*) and color difference parameters (△L*, △a*, △C*) of tobacco leaves of different maturity. The Fisher discriminant function established with color parameters as indicators can distinguish tobacco leaves of different maturity. Ren Erfang et al. (2019) compared and analyzed the results of using electronic nose and electronic tongue to measure the soluble solids content, Vc content and pH value of freshly squeezed strawberry juice, and determined the maturity of the strawberries based on this. Feng Jianying et al. (2020) conducted a follow-up study on the working principles and results of three typical intelligent sensory technologies: electronic nose technology, electronic tongue technology and texture analysis technology, and analyzed the accuracy of determining the maturity of strawberries based on soluble solids content, Vc content and pH value.

[0011] At present, my country's crop harvest monitoring methods are inefficient and costly, and the data are scattered and not in a uniform format, so it is impossible to achieve large-scale centralized management and updating and subsequent decision-making. Therefore, how to use large-scale remote sensing images to monitor the crop harvest period and obtain a series of corresponding effective crop growth data sets has become a key research goal and direction, and it is also a key technology for popularizing precision agriculture. Therefore, in response to the above problems and the subsequent impacts such as the informatization and intelligence of farmland agricultural machinery in my country and the low decision-making dependence caused by them, how to realize the automation and intelligence of cross-regional crop harvest monitoring in my country is the primary technical problem to be solved. Summary of the invention

[0012] In order to overcome the shortcomings of the prior art, the object of the present invention is to provide a cross-regional crop harvest period monitoring method based on deep learning.

[0013] To achieve the above object, the present invention provides the following solutions:

[0014] A cross-regional crop harvest period monitoring method based on deep learning, comprising:

[0015] Obtain high temporal and spatial resolution remote sensing data in the target area during the corresponding crop growing season, and generate multiple sets of vegetation index time series;

[0016] Extracting pixels from each group of vegetation index time series based on a pixel-level crop biological growth classification model to obtain time series features of the pixels, and obtaining classification results based on the time series features;

[0017] Based on the deep learning algorithm and the classification result, deep feature extraction is performed on the spatial and spectral space of the high temporal and spatial resolution remote sensing data to obtain the feature-level data;

[0018] Constructing an initial neural network model, and training the initial neural network model based on the feature-level data and the classification results to obtain a trained cross-regional crop harvest period monitoring model;

[0019] The data to be tested is input into the cross-regional crop harvest period monitoring model to obtain monitoring results.

[0020] Preferably, the high temporal and spatial resolution remote sensing data includes microwave data Sentinel-1 data set and multispectral Sentinel-2 data set.

[0021] Preferably, pixels of each group of vegetation index time series are extracted based on a pixel-level crop biological growth classification model to obtain time series features of pixels, and the time series features are classified to obtain classification results, including:

[0022] Construct pixel-level crop biological growth classification model;

[0023] Extracting the time series features of each pixel from the vegetation index time series using the pixel-level crop biological growth classification model;

[0024] According to the time series features, the corresponding crops in the target area are classified at the pixel level, and fusion extraction is performed according to a single-color channel, a two-color channel or a multi-color channel fusion method to extract crop pixels and form the corresponding classification result.

[0025] Preferably, after extracting pixels from each group of vegetation index time series based on the pixel-level crop biological growth classification model to obtain the time series features of the pixels, and classifying the time series features to obtain the classification results, the method further includes:

[0026] Image denoising, image enhancement and image registration are performed on the high temporal and spatial resolution remote sensing data to obtain preprocessed high temporal and spatial resolution remote sensing data.

[0027] Preferably, based on the deep learning algorithm and the classification result, deep feature extraction is performed on the space and spectral space of the high temporal and spatial resolution remote sensing data to obtain the feature-level data, including:

[0028] Obtain training remote sensing dataset;

[0029] quantifying the feature information of the high spatiotemporal resolution remote sensing data based on the vegetation information, spatial information and spectral spatial information in the training remote sensing data set to determine the key factors of the interaction between the high spatiotemporal resolution remote sensing data and each feature;

[0030] Constructing a feature extraction model for the high spatiotemporal resolution remote sensing data based on the key factors and the training remote sensing data set;

[0031] Optimizing the feature extraction model based on the temporal evolution characteristics of the classification results to obtain an optimized model;

[0032] The high spatiotemporal resolution remote sensing data is subjected to feature extraction according to the optimization model to obtain the feature-level data.

[0033] Preferably, the initial neural network model is a UNet convolutional neural network structure.

[0034] Preferably, the pixel-level crop biological growth condition classification model is any one of a long short-term memory model or a Transformer model.

[0035] Preferably, constructing a feature extraction model for the high spatiotemporal resolution remote sensing data based on the key factors and the training remote sensing data set includes:

[0036] constructing a feature estimation multidimensional data set based on the key factors;

[0037] The feature extraction model is established based on the comprehensive impact of each element in the feature estimation multidimensional data set on the high temporal and spatial resolution remote sensing data; the feature extraction model includes the numerical value, time series information and spatial information of each feature element.

[0038] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0039] The present invention provides a cross-regional crop harvest period monitoring method based on deep learning, including: obtaining high-temporal and spatial resolution remote sensing data within the corresponding crop growing season in the target area, and generating multiple groups of vegetation index time series; extracting pixels of each group of vegetation index time series based on a pixel-level crop biological growth classification model to obtain the time series characteristics of pixels, and obtaining classification results according to the time series characteristics; based on a deep learning algorithm and the classification results, performing deep feature extraction on the space and spectral space of the high-temporal and spatial resolution remote sensing data to obtain the feature-level data; constructing an initial neural network model, and training the initial neural network model based on the feature-level data and the classification results to obtain a trained cross-regional crop harvest period monitoring model; inputting the data to be tested into the cross-regional crop harvest period monitoring model to obtain monitoring results. The present invention can use large-scale remote sensing images to perform high-precision monitoring of crop harvest periods, thereby improving the automation and intelligence of cross-regional crop harvest period monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0041] Figure 1 A flow chart of a method provided by an embodiment of the present invention;

[0042] Figure 2 A schematic diagram of a technical route provided for an embodiment of the present invention;

[0043] Figure 3 A schematic diagram of an HSI model provided in an embodiment of the present invention;

[0044] Figure 4 A UNet network structure diagram provided for an embodiment of the present invention. DETAILED DESCRIPTION

[0045] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0046] The purpose of the present invention is to provide a cross-regional crop harvest period monitoring method based on deep learning, which can use large-scale remote sensing images to perform high-precision monitoring of the crop harvest period, thereby improving the automation and intelligence level of cross-regional crop harvest period monitoring.

[0047] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0048] Figure 1 A flow chart of a method provided by an embodiment of the present invention, such as Figure 1 As shown, the present invention provides a cross-regional crop harvest period monitoring method based on deep learning, comprising:

[0049] Step 100: obtaining high temporal and spatial resolution remote sensing data corresponding to the crop growing season in the target area, and generating multiple sets of vegetation index time series;

[0050] Step 200: extracting pixels from each group of vegetation index time series based on a pixel-level crop biological growth classification model to obtain time series features of the pixels, and obtaining classification results based on the time series features;

[0051] Step 300: Based on the deep learning algorithm and the classification result, deep feature extraction is performed on the space and spectral space of the high temporal and spatial resolution remote sensing data to obtain the feature-level data;

[0052] Step 400: constructing an initial neural network model, and training the initial neural network model based on the feature-level data and the classification results to obtain a trained cross-regional crop harvest period monitoring model;

[0053] Step 500: input the data to be tested into the cross-regional crop harvest period monitoring model to obtain monitoring results.

[0054] Preferably, the high temporal and spatial resolution remote sensing data includes microwave data Sentinel-1 data set and multispectral Sentinel-2 data set.

[0055] Preferably, pixels of each group of vegetation index time series are extracted based on a pixel-level crop biological growth classification model to obtain time series features of pixels, and the time series features are classified to obtain classification results, including:

[0056] Construct pixel-level crop biological growth classification model;

[0057] Extracting the time series features of each pixel from the vegetation index time series using the pixel-level crop biological growth classification model;

[0058] According to the time series features, the corresponding crops in the target area are classified at the pixel level, and fusion extraction is performed according to a single-color channel, a two-color channel or a multi-color channel fusion method to extract crop pixels and form the corresponding classification result.

[0059] like Figure 2 As shown, the present invention is mainly based on the monitoring research of remote sensing time series. Taking full account of the temporal accumulation process and dynamic change characteristics of crop growth, it is planned to build a cross-regional crop monitoring model based on remote sensing time series based on deep learning time series models such as Long Short-Term Memory (LSTM) and Transformer model, combined with remote sensing time series data of crop growing season. It mainly includes the following contents:

[0060] ① Acquisition of high-temporal and spatial resolution remote sensing time series data: Based on the field geographic information and main crop distribution in the study area, high-temporal and spatial resolution remote sensing data of the corresponding crop growing season are obtained, and multiple sets of vegetation index time series are generated to analyze the growth characteristics of each crop from sowing to maturity, preliminarily evaluate the growth status and development speed of crops, and form a complete set of crop growth data, which provides the corresponding data basis for the construction of the crop biological growth classification model mentioned below, and also provides data support for model training.

[0061] ② Pixel-level classification of crop biological growth conditions: Based on deep learning methods such as the Long Short-Term Memory model (LSTM) and the Transformer model, a pixel-level crop biological growth classification model is constructed to extract the temporal features of each pixel from the vegetation index time series, and to classify the crops in the study area at the pixel level. Fusion extraction experiments are performed based on R, G, B single-color channels, nR-G, nR-B, nG-B and other dual-color channel or multi-color channel fusion formulas to extract crop pixels and form corresponding pixel data. At the same time, based on the obtained RGB color channel pixels, the HSI pixel model data is obtained, the two color channels are analyzed and compared, and the appropriate color channel pixels are selected. The data of the above two color channel pixels provide data support for subsequent model training. The specific description of the HSI color channel is as follows: Figure 3 As shown:

[0062] Figure 3 The top and bottom ends of the cone in the middle model represent white and black respectively, the vertical axis represents the intensity of light, and the circumference represents the hue.

[0063] In color images, there is a strict mathematical relationship between the HSI model and the RGB model. In this embodiment, RGB can be converted to HSI. The following is the conversion method:

[0064] (A) First, normalize the values ​​of R, G, and B to the range of [0,255], and then obtain three r, g, and b values ​​in the range of [0,1]:

[0065]

[0066] (B) The calculation formulas for the corresponding three components H, S, and I obtained in this embodiment are as follows:

[0067]

[0068] s=1-3·min(r,g,b),S∈[0,1];

[0069]

[0070]

[0071] (C) The above formula calculates that the value range of h is [0, 2π], the value range of s is [0, 1], and the value range of i is [0, 1]. However, for ease of understanding and calculation, this is converted to [0°, 360°], [0, 100], [0, 255]:

[0072]

[0073] By transforming the above calculation formula, this embodiment can obtain a calculation method for the HSI color model and apply it to actual image operations.

[0074] Preferably, after extracting pixels from each group of vegetation index time series based on the pixel-level crop biological growth classification model to obtain the time series features of the pixels, and classifying the time series features to obtain the classification results, the method further includes:

[0075] Image denoising, image enhancement and image registration are performed on the high temporal and spatial resolution remote sensing data to obtain preprocessed high temporal and spatial resolution remote sensing data.

[0076] Preferably, based on the deep learning algorithm and the classification result, deep feature extraction is performed on the space and spectral space of the high temporal and spatial resolution remote sensing data to obtain the feature-level data, including:

[0077] Obtain training remote sensing dataset;

[0078] quantifying the feature information of the high spatiotemporal resolution remote sensing data based on the vegetation information, spatial information and spectral spatial information in the training remote sensing data set to determine the key factors of the interaction between the high spatiotemporal resolution remote sensing data and each feature;

[0079] Constructing a feature extraction model for the high spatiotemporal resolution remote sensing data based on the key factors and the training remote sensing data set;

[0080] Optimizing the feature extraction model based on the temporal evolution characteristics of the classification results to obtain an optimized model;

[0081] The high spatiotemporal resolution remote sensing data is subjected to feature extraction according to the optimization model to obtain the feature-level data.

[0082] Furthermore, there are relatively mature technical solutions in the prior art for the feature quantification of spatial information and spectral information, which will not be elaborated in this embodiment. Through the interaction between various quantitative information and corresponding features, this embodiment can extract key factors and use the key factors as labeled samples, thereby constructing a feature extraction model together with the training remote sensing data set.

[0083] Preferably, the initial neural network model is a UNet convolutional neural network structure.

[0084] Preferably, the pixel-level crop biological growth condition classification model is any one of a long short-term memory model or a Transformer model.

[0085] Preferably, constructing a feature extraction model for the high spatiotemporal resolution remote sensing data based on the key factors and the training remote sensing data set includes:

[0086] constructing a feature estimation multidimensional data set based on the key factors;

[0087] The feature extraction model is established based on the comprehensive impact of each element in the feature estimation multidimensional data set on the high temporal and spatial resolution remote sensing data; the feature extraction model includes the numerical value, time series information and spatial information of each feature element.

[0088] The last step in the technical route of this embodiment is specifically:

[0089] ③ Cross-regional crop harvest period monitoring: Based on the pixel-level crop biological growth classification results, combined with remote sensing image data, based on UNet and other convolutional neural network structures, through the training of crop growth data and pixel-level crop biological growth classification models, a cross-regional crop harvest period monitoring model is constructed to form a cross-regional crop harvest period monitoring form. The specific UNet convolutional neural network structure is as follows: Figure 4As shown, UNet uses a fully convolutional neural network. The network on the left is a feature extraction network: using conv and pooling; the network on the right is a feature fusion network: using the feature map generated by upsampling to concatenate with the feature map on the left. (The pooling layer will lose image information and reduce image resolution and is permanent. It has some impact on image segmentation tasks, but not much impact on image classification tasks. Upsampling can make low-resolution images containing high-level abstract features become high-resolution while retaining high-level abstract features, and then concatenate with the high-resolution image of low-level surface features on the left); finally, after two more convolution operations, a feature map is generated, and then two convolution kernels with a size of 1*1 are used for classification to obtain the final two heatmaps. For example, the first one represents the score of the first category, and the second one represents the score heatmap of the second category. Then, it is used as the input of the softmax function to calculate the softmax with a relatively large probability, and then the loss and back propagation calculations are performed.

[0090] The beneficial effects of the present invention are as follows:

[0091] The present invention can utilize large-scale remote sensing images to carry out high-precision monitoring of the suitable harvest period of crops, thereby improving the automation and intelligence level of cross-regional monitoring of the suitable harvest period of crops.

[0092] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0093] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A cross-regional crop harvest period monitoring method based on deep learning, characterized in that: include: Obtain high temporal and spatial resolution remote sensing data in the target area during the corresponding crop growing season, and generate multiple sets of vegetation index time series; Extracting pixels from each group of vegetation index time series based on a pixel-level crop biological growth classification model to obtain time series features of the pixels, and obtaining classification results based on the time series features; Based on the deep learning algorithm and the classification result, deep feature extraction is performed on the spatial and spectral space of the high temporal and spatial resolution remote sensing data to obtain the feature-level data; Constructing an initial neural network model, and training the initial neural network model based on the feature-level data and the classification results to obtain a trained cross-regional crop harvest period monitoring model; Inputting the data to be tested into the cross-regional crop harvest period monitoring model to obtain monitoring results; Based on the deep learning algorithm and the classification result, deep feature extraction is performed on the space and spectral space of the high temporal and spatial resolution remote sensing data to obtain the feature-level data, including: Obtain training remote sensing dataset; quantifying the feature information of the high spatiotemporal resolution remote sensing data based on the vegetation information, spatial information and spectral spatial information in the training remote sensing data set to determine the key factors of the interaction between the high spatiotemporal resolution remote sensing data and each feature; Based on the key factors and the training remote sensing data set, a feature extraction model for the high spatiotemporal resolution remote sensing data is constructed; through the interaction between each quantitative information and the corresponding features, the key factors are used as labeled samples to jointly construct the feature extraction model with the training remote sensing data set; Optimizing the feature extraction model based on the temporal evolution characteristics of the classification results to obtain an optimized model; The high spatiotemporal resolution remote sensing data is subjected to feature extraction according to the optimization model to obtain the feature-level data.

2. The cross-regional crop harvest period monitoring method based on deep learning according to claim 1 is characterized in that: The high temporal and spatial resolution remote sensing data include microwave data Sentinel-1 data set and multispectral Sentinel-2 data set.

3. The cross-regional crop harvest period monitoring method based on deep learning according to claim 1 is characterized in that: Based on the pixel-level crop biological growth classification model, pixels are extracted from each group of vegetation index time series to obtain the time series characteristics of the pixels, and the time series characteristics are classified to obtain classification results, including: Construct pixel-level crop biological growth classification model; Extracting the time series features of each pixel from the vegetation index time series using the pixel-level crop biological growth classification model; According to the time series features, the corresponding crops in the target area are classified at the pixel level, and fusion extraction is performed according to a single-color channel, a two-color channel or a multi-color channel fusion method to extract crop pixels and form the corresponding classification result.

4. The cross-regional crop harvest period monitoring method based on deep learning according to claim 1 is characterized in that: After extracting pixels from each group of vegetation index time series based on the pixel-level crop biological growth classification model to obtain the time series features of the pixels, and classifying the time series features to obtain the classification results, the method further includes: Image denoising, image enhancement and image registration are performed on the high temporal and spatial resolution remote sensing data to obtain preprocessed high temporal and spatial resolution remote sensing data.

5. The cross-regional crop harvest period monitoring method based on deep learning according to claim 1 is characterized in that: The initial neural network model is a UNet convolutional neural network structure.

6. The cross-regional crop harvest period monitoring method based on deep learning according to claim 1 is characterized in that: The pixel-level crop biological growth classification model is any one of a long short-term memory model or a Transformer model.

7. The cross-regional crop harvest period monitoring method based on deep learning according to claim 1 is characterized in that: Building a feature extraction model for the high spatiotemporal resolution remote sensing data based on the key factors and the training remote sensing data set includes: constructing a feature estimation multidimensional data set based on the key factors; The feature extraction model is established based on the comprehensive impact of each element in the multidimensional data set on the high temporal and spatial resolution remote sensing data according to the feature estimation; the feature extraction model is used to integrate the numerical value, time series information and spatial information of each feature element.

Images