Remote sensing image-based crop planting plot monitoring method and system

Abstract

The invention relates to the technical field of image recognition, in particular to a crop planting plot monitoring method and system based on a remote sensing image so as to improve the accuracy of crop planting plot recognition. Acquiring a remote sensing image, DEM data and LiDAR point cloud data of a crop planting plot, and preprocessing the remote sensing image, the DEM data and the LiDAR point cloud data to obtain a preprocessed remote sensing image, preprocessed DEM data and an elevation difference edge; inputting the preprocessed data into a multi-task crop planting land parcel identification model for identification, and obtaining crop features, non-crop features and land parcel features in combination with enhanced land parcel boundary features and non-crop land feature mask features; obtaining the health degree of the crop planting plot according to the crop features, the non-crop features and the plot features; comparing the health degree of the crop planting plot with a preset threshold value, and if the health degree exceeds the preset threshold value, outputting early warning information; otherwise, inputting the remote sensing image and the health degree of the crop planting plot into a crop planting plot yield prediction model for processing to obtain plot yield prediction information.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition, and particularly to a method and system for monitoring crop planting plots based on remote sensing images. Background Art

[0002] The method for monitoring crop planting plots based on remote sensing images usually preprocesses the remote sensing image data of the monitoring area to obtain high-quality image data; then, a neural network is used to identify the features extracted from the high-quality image data and output the recognition results.

[0003] However, the existing technology still has deficiencies in the monitoring of crop planting plots based on remote sensing images; on the one hand, the existing technology does not consider that non-crop features on the crop planting plots will interfere with crop recognition; on the other hand, the existing technology is affected by the irregularity of the planting plots, resulting in relatively blurred extraction of plot boundaries, which directly affects the accurate recognition of crop planting areas, thus reducing the accuracy and reliability of crop planting plot monitoring.

[0004] Therefore, a method and system for monitoring crop planting plots based on remote sensing images are proposed. Summary of the Invention

[0005] The purpose of the present invention is to provide a method and system for monitoring crop planting plots based on remote sensing images. First, obtain the remote sensing image, DEM data, and LiDAR point cloud data of the crop planting plot and preprocess them to obtain the preprocessed remote sensing image, preprocessed DEM data, and elevation difference edge; input the preprocessed data into a multi-task crop planting plot recognition model for processing, and combine the enhanced plot boundary features and non-crop feature masks to obtain crop features, non-crop features, and plot features; obtain the health degree of the crop planting plot according to the crop features, non-crop features, and plot features; compare the health degree of the crop planting plot with a preset threshold, and if it exceeds the preset threshold, output a warning message; otherwise, input the remote sensing image and the health degree of the crop planting plot into a crop planting plot yield prediction model for processing to obtain plot yield prediction information. The present invention can effectively improve the accuracy and reliability of crop planting plot monitoring.

[0006] To achieve the above object, the present invention provides the following technical solutions: A method for monitoring crop planting plots based on remote sensing images, comprising: Obtain the remote sensing image, DEM data, and LiDAR point cloud data of the crop planting plot; Preprocess the remote sensing image, the DEM data, and the LiDAR point cloud data to obtain a preprocessed remote sensing image, preprocessed DEM data, and an elevation difference edge; Build a multi-task crop planting plot recognition model, input the preprocessed remote sensing image, the preprocessed DEM data, and the elevation difference edge into the multi-task crop planting plot recognition model to obtain enhanced plot boundary features and non-crop feature masks; combine the enhanced plot boundary information, the non-crop feature mask information, and the multi-task crop planting plot recognition model to obtain crop features, non-crop features, and plot features; Obtain the health degree of the crop planting plot according to the crop features, the non-crop features, and the plot features; Compare the health degree of the crop planting plot with a preset threshold. If it exceeds the preset threshold, output a warning message; otherwise, input the remote sensing image and the health degree of the crop planting plot into a crop planting plot yield prediction model for processing to obtain plot yield prediction information.

[0007] Further, the process of preprocessing the remote sensing image, the DEM data, and the LiDAR point cloud data includes: performing correction, enhancement, and normalization processing on the remote sensing image to obtain the preprocessed remote sensing image; filtering the DEM data to obtain the preprocessed DEM data; filtering and normalizing the LiDAR point cloud data to obtain preprocessed LiDAR point cloud data; performing edge extraction on the elevation difference map obtained by processing the preprocessed LiDAR point cloud data to obtain the elevation difference edge.

[0008] Further, the multi-task crop planting plot recognition model includes: an input layer, a feature extraction layer, a plot boundary edge enhancement layer, a non-crop feature mask extraction layer, a multi-task plot recognition layer, and an output layer.

[0009] Further, the process of obtaining the enhanced plot boundary features, the non-crop feature masks, the crop features, the non-crop features, and the plot features includes: Input the preprocessed remote sensing image, the preprocessed DEM data, and the elevation difference edge into the input layer and the feature extraction layer of the multi-task crop planting plot recognition model to obtain remote sensing features, DEM features, and elevation difference edge features; Input the remote sensing features, the DEM features, and the elevation difference edge features into the plot boundary edge enhancement layer of the multi-task crop planting plot recognition model to obtain the enhanced plot boundary features; input the remote sensing features into the non-crop feature mask extraction layer to obtain the non-crop feature masks; Input the remote sensing features, the enhanced plot boundary features, and the non-crop feature masks into the multi-task plot recognition layer and the output layer of the multi-task crop planting plot recognition model to obtain the crop features, the non-crop features, and the plot features.

[0010] Further, the calculation process of the health degree of the crop planting plot includes: Obtaining a crop index, a non-crop index, and a plot index according to the crop characteristics, the non-crop characteristics, and the plot characteristics respectively; Performing weighted summation on the crop index, the non-crop index, and the plot index to obtain the health degree of the crop planting plot; wherein, the calculation formula of the health degree of the crop planting plot is: ; Wherein, is the health degree of the crop planting plot; is the crop weight factor; is the crop index; is the non-crop weight factor; is the non-crop index; is the plot weight factor; is the plot index.

[0011] Further, the process of obtaining the plot yield prediction information includes: Training the crop planting plot yield prediction model by using historical remote sensing images, historical health degrees, and historical yields to obtain historical plot yield influence coefficients; Inputting the remote sensing image, the health degree of the crop planting plot, and the historical plot yield influence coefficients into the trained crop planting plot yield prediction model for parameter update, and outputting the plot yield prediction information.

[0012] A crop planting plot monitoring system based on remote sensing images includes: a data acquisition module, a data preprocessing module, a plot recognition module, a plot health degree module, a plot yield prediction module, and an output module; The data acquisition module is used to obtain remote sensing images, DEM data, and LiDAR point cloud data of the crop planting plot; The data preprocessing module is used to preprocess the remote sensing images, the DEM data, and the LiDAR point cloud data to obtain preprocessed remote sensing images, preprocessed DEM data, and preprocessed LiDAR point cloud data; The plot recognition module is used to input the preprocessed remote sensing images, the preprocessed DEM data, and the preprocessed LiDAR point cloud data into a multi-task crop planting plot recognition model to obtain enhanced plot boundary features and non-crop ground object mask features; combining the enhanced plot boundary features, the non-crop ground object mask features, and the multi-task crop planting plot recognition model to obtain crop characteristics, non-crop characteristics, and plot characteristics; The plot health module is used to obtain the health of the crop planting plot according to the crop characteristics, the non-crop characteristics, and the plot characteristics; The plot yield prediction module is used to input the remote sensing image and the health of the crop planting plot into a crop planting plot yield prediction model for processing to obtain plot yield prediction information; The output module is used to compare the health of the crop planting plot with a preset threshold. If it exceeds the preset threshold, warning information is output; otherwise, the plot yield prediction information is output.

[0013] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. The present invention proposes a multi-task crop planting plot recognition method for accurately recognizing different types of characteristics of plots; this method combines multi-source data combination and a multi-task crop planting plot recognition model, which can utilize the differences between remote sensing data and elevation data to enhance the edges of plot boundaries. At the same time, by using the output features of the plot boundary edge enhancement layer and the non-crop feature mask extraction layer to assist the multi-object recognition of the multi-task plot recognition layer, the accuracy and reliability of crop planting plot monitoring are effectively improved.

[0014] 2. The present invention proposes a crop planting plot health for monitoring the actual health of plots; this health is obtained by weighting different types of characteristics output by the multi-task crop planting plot recognition model, comprehensively measuring the health of the planting plot from crops, non-crops, and the plot itself, providing a basis for subsequent crop planting plot yield prediction, and thus effectively improving the accuracy and reliability of crop planting plot monitoring.

[0015] 3. The present invention proposes a crop planting plot yield prediction method for predicting the yield trend of planting plots; this plot yield prediction method uses the historical plot yield influence coefficient obtained by training the crop planting plot yield prediction model as prior knowledge, and combines real-time remote sensing images and the health of the crop planting plot to obtain plot yield prediction information; using the plot yield prediction information can further adjust crop planting plot monitoring, thereby improving the accuracy and reliability of crop planting plot monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is a schematic flowchart of a method for monitoring crop planting plots based on remote sensing images according to the present invention; Figure 2 is a schematic structural diagram of a multi-task crop planting plot recognition model according to the present invention; Figure 3 is a schematic structural diagram of a system for monitoring crop planting plots based on remote sensing images according to the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0017] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0018] Please refer to Figures 1 to 3 , the present invention provides a method and system for monitoring crop planting plots based on remote sensing images, and the technical solutions are as follows: Embodiment 1 A certain company, in order to improve the accuracy and reliability of monitoring crop planting plots based on remote sensing images, uses a method for monitoring crop planting plots based on remote sensing images proposed by the present invention. The process schematic of this method is as Figure 1 shown, and specifically includes: Obtain remote sensing images, DEM data, and LiDAR point cloud data of crop planting plots; Furthermore, the remote sensing images and LiDAR point cloud data are collected by remote sensing sensors and LiDAR devices carried by unmanned aerial vehicles; the DEM data, that is, digital elevation model data, can be directly called through a dedicated data platform.

[0019] Preprocess the remote sensing images, DEM data, and LiDAR point cloud data to obtain preprocessed remote sensing images, preprocessed DEM data, and elevation difference edges; Furthermore, the process of preprocessing the remote sensing images, DEM data, and LiDAR point cloud data includes: correcting, enhancing, and normalizing the remote sensing images to obtain preprocessed remote sensing images; filtering the DEM data to obtain preprocessed DEM data; filtering and normalizing the LiDAR point cloud data to obtain preprocessed LiDAR point cloud data; extracting the edges of the elevation difference map obtained by processing the preprocessed LiDAR point cloud data to obtain elevation difference edges; Furthermore, the process of extracting the edges of the elevation difference map obtained by processing the preprocessed LiDAR point cloud data to obtain elevation difference edges includes: Use progressive triangular mesh filtering to extract ground points and generate a digital terrain model (DTM); Divide the monitoring area into multiple sub-plots and calculate the average elevation of the DTM inside each sub-plot; Visualize the calculated plot elevation difference values between sub-plots to generate an elevation difference map; Use an edge algorithm to process the elevation difference map to obtain elevation difference edges.

[0020] By using pre - processed remote sensing images, DEM data, and LiDAR point cloud data, it is possible to provide rich spectral information, spatial and topographic information for the subsequent multi - task crop planting plot recognition model. It can improve the recognition accuracy of crop planting plots while solving the problem of blurred boundary recognition of complex and irregular plots, thereby enhancing the accuracy and reliability of crop planting plot monitoring.

[0021] Construct a multi - task crop planting plot recognition model; Furthermore, the structure of the multi - task crop planting plot recognition model is as Figure 2 shown, including: an input layer, a feature extraction layer, a plot boundary edge enhancement layer, a non - crop feature mask extraction layer, a multi - task plot recognition layer, and an output layer.

[0022] In this embodiment, the multi - task crop planting plot recognition model integrates the functions of edge enhancement, mask extraction, and plot recognition. At the same time, using the boundary edge enhancement information and non - crop mask information can enable the multi - task plot recognition layer to accurately identify the plot boundary and reduce the interference of non - crop features on crop recognition, thus improving the accuracy of plot recognition and further enhancing the accuracy and reliability of crop planting plot monitoring.

[0023] Input the pre - processed remote sensing image, pre - processed DEM data, and elevation difference edge into the multi - task crop planting plot recognition model to obtain enhanced plot boundary features and non - crop feature mask features; combine the enhanced plot boundary information, non - crop feature mask information, and the multi - task crop planting plot recognition model to obtain crop features, non - crop features, and plot features; Furthermore, the acquisition process of the enhanced plot boundary features, non - crop feature mask features, crop features, non - crop features, and plot features includes: Input the pre - processed remote sensing image, pre - processed DEM data, and elevation difference edge into the input layer and feature extraction layer of the multi - task crop planting plot recognition model to obtain remote sensing features, DEM features, and elevation difference edge features; Input the remote sensing features, DEM features, and elevation difference edge features into the plot boundary edge enhancement layer of the multi - task crop planting plot recognition model to obtain enhanced plot boundary features; input the remote sensing features into the non - crop feature mask extraction layer to obtain non - crop feature mask features; Furthermore, the process of inputting the remote sensing features, DEM features, and elevation difference edge features into the plot boundary edge enhancement layer of the multi - task crop planting plot recognition model to obtain enhanced plot boundary features includes: Merge the remote sensing features and DEM features in channels to obtain merged features; Use the improved UNet network to process the merged features to obtain plot boundary features; Taking the elevation difference edge feature as the guiding feature, combining the max-pooling and average-pooling operations to generate the spatial weight matrix; Using the spatial weight matrix to weight the plot boundary feature, and obtaining the enhanced plot boundary feature through the residual connection; Furthermore, the non-crop object mask extraction layer uses the GBFNet network to output the binary mask feature of the non-crop object.

[0024] Inputting the remote sensing feature, the enhanced plot boundary feature and the non-crop object mask feature into the multi-task plot recognition layer and the output layer of the multi-task crop planting plot recognition model to obtain the crop feature, the non-crop feature and the plot feature.

[0025] Furthermore, the process of inputting the remote sensing feature, the enhanced plot boundary feature and the non-crop object mask feature into the multi-task plot recognition layer and the output layer of the multi-task crop planting plot recognition model to obtain the crop feature, the non-crop feature and the plot feature includes: Using the non-crop object mask feature to divide the remote sensing feature into the crop remote sensing feature and the non-crop remote sensing feature; Combining the crop remote sensing feature, the non-crop remote sensing feature and the remote sensing feature with the enhanced plot boundary feature respectively, and then inputting the merged features into each independent branch of the multi-task plot recognition layer for recognition and detection to obtain the recognition results of each branch; Using the output layer to process the recognition results of each branch to obtain the crop feature, the non-crop feature and the plot feature; Furthermore, the multi-task plot recognition layer has three independent branches, and each branch uses a convolutional neural network for feature recognition.

[0026] To illustrate the effectiveness of the plot boundary edge enhancement method, 3 groups of historical datasets from different plots were selected for the plot boundary edge enhancement effectiveness test, which were respectively recorded as the first dataset, the second dataset and the third dataset; each group of historical datasets included 400 pieces of acquisition data from different times; two different schemes were used to process the 3 groups of data. Among them, Scheme 1 was to use remote sensing images, DEM data and LiDAR point cloud data and process them, combine the multi-source data features with the plot boundary edge enhancement layer to obtain the enhanced plot boundary feature. The multi-source data features included: remote sensing feature, DEM feature and elevation difference edge feature, and then input the enhanced plot boundary feature and the remote sensing feature into the multi-task plot recognition layer and the output layer to obtain the plot recognition area; Solution 2 is to only use remote sensing images, and then directly input the remote sensing features into the multi-task plot recognition layer and the output layer to obtain the corresponding recognized plot area; compare the recognized plot areas of different solutions with the target plot area to obtain the proportion of the recognized plot areas of each group of data within a reasonable range under different solutions. The effectiveness test results of the plot boundary edge enhancement method are shown in Table 1.

[0027] Table 1. Effectiveness test results of the plot boundary edge enhancement method

[0028] As can be seen from Table 1, adopting the solution of combining multi-source data features and the plot boundary edge enhancement layer proposed by the present invention can effectively improve the accuracy of crop planting plot area recognition, which is beneficial to assisting crop feature recognition and non-crop feature recognition, thereby improving the accuracy and reliability of crop planting plot monitoring.

[0029] By combining multi-source data combination and multi-task crop planting plot recognition model, using the differences between remote sensing data and elevation data to achieve edge enhancement of plot boundaries, and at the same time using the output features of the plot boundary edge enhancement layer and the non-crop feature mask extraction layer to assist the multi-objective recognition of the multi-task plot recognition layer, thereby effectively improving the accuracy and reliability of crop planting plot monitoring.

[0030] Obtain the health status of the crop planting plot according to crop characteristics, non-crop characteristics and plot characteristics; Further, the calculation process of the health status of the crop planting plot includes: Obtain the crop index, non-crop index and plot index according to crop characteristics, non-crop characteristics and plot characteristics respectively; Further, crop characteristics include: crop coverage, photosynthetic intensity and crop height; non-crop characteristics include: pest quantity, weed coverage and building area; Further, the crop index can be expressed as: ; ; where is the crop index; is the crop coverage weight value; is the crop coverage; is the crop coverage standard value; is the photosynthetic intensity weight value; is the photosynthetic intensity; is the photosynthetic intensity standard value; is the crop height weight value; is the crop height; is the crop height standard value; Further, the crop coverage weight value, photosynthetic intensity weight value, and crop height weight value are set to 0.35, 0.3, and 0.35 respectively; the crop coverage standard value, photosynthetic intensity standard value, and crop height standard value are greatly affected by seasons, so they need to be flexibly adjusted according to the corresponding seasons; Further, the non-crop index can be expressed as: ; ; wherein, is the non-crop index; is the exponential function; is the pest weight; is the pest quantity; is the weed coverage weight; is the weed coverage; is the weed coverage threshold; is the building weight; is the building area; is the collection area; the pest weight, weed coverage weight, and building weight are set to 0.35, 0.3, and 0.35 respectively; Further, the plot index can be expressed as: ; wherein, is the plot index; is the value of the i-th plot index; is the threshold of the i-th plot index; is the weight of the i-th plot index, each of which is set to 0.2; the plot indexes include: soil humidity, organic matter content, pH value, salt content, and heavy metal content.

[0031] The crop index, non-crop index, and plot index are weighted and summed to obtain the health degree of the crop planting plot; wherein, the calculation formula of the health degree of the crop planting plot is: ; ; wherein, is the health degree of the crop planting plot; is the crop weight factor; is the crop index; is the non-crop weight factor; is the non-crop index; is the plot weight factor; is the plot index.

[0032] Furthermore, the crop weight factor, non-crop weight factor, and plot weight factor are set to 0.4, 0.3, and 0.3 respectively; the setting of the weight factor is affected by the regional environment, so it needs to be adjusted according to the actual situation.

[0033] To illustrate the health of the crop planting plots proposed by the present invention, three groups of identification data from different planting plots are randomly selected for testing, denoted as Sample One, Sample Two, and Sample Three respectively; according to the calculation results of the crop index, non-crop index, and plot index, the health of the crop planting plots in different plots is weighted; the test results of the health of the crop planting plots are shown in Table 2.

[0034] Table 2. Test Results of the Health of Crop Planting Plots

[0035] By weighting different types of features output by the multi-task crop planting plot identification model, the health of the planting plot is comprehensively measured from crops, non-crops, and the plot itself, providing a basis for subsequent yield prediction of the crop planting plot, thereby effectively improving the accuracy and reliability of crop planting plot monitoring.

[0036] Compare the health of the crop planting plot with a preset threshold. If it exceeds the preset threshold, an early warning message is output; otherwise, the remote sensing image and the health of the crop planting plot are input into the crop planting plot yield prediction model for processing to obtain the plot yield prediction information.

[0037] Furthermore, the process of obtaining the plot yield prediction information includes: Using historical remote sensing images, historical health, and historical yields to train the crop planting plot yield prediction model to obtain the historical plot yield impact coefficient; Input the remote sensing image, the health of the crop planting plot, and the historical plot yield impact coefficient into the trained crop planting plot yield prediction model for parameter update to obtain the plot yield impact coefficient; Combine the plot yield impact coefficient with the current plot yield data to obtain the plot yield prediction information.

[0038] In this embodiment, the crop planting plot yield prediction model uses a dual-branch network; among them, the remote sensing image branch takes the remote sensing image as input and uses the EfficientNet network to output high-level semantic features; the health branch takes the health feature of the crop planting plot as input and uses the lightweight MLP network to extract the temporal change features of health; then, the scSE attention module is used to dynamically weight and fuse the dual-branch features; finally, the fused dual-branch features are decoded to obtain the yield information.

[0039] Taking the historical plot yield impact coefficient obtained by training the crop planting plot yield prediction model as prior knowledge, and combining real-time remote sensing images and the health of crop planting plots to obtain plot yield prediction information; the plot yield prediction information can be used to further adjust the monitoring of crop planting plots, thereby improving the accuracy and reliability of the monitoring of crop planting plots.

[0040] Embodiment 2 The present invention proposes a crop planting plot monitoring system based on remote sensing images. The structure of the system is as Figure 3 shown, including: a data acquisition module, a data preprocessing module, a plot identification module, a plot health module, a plot yield prediction module, and an output module; The data acquisition module is used to obtain remote sensing images, DEM data, and LiDAR point cloud data of crop planting plots; The data preprocessing module is used to preprocess the remote sensing images, DEM data, and LiDAR point cloud data to obtain preprocessed remote sensing images, preprocessed DEM data, and preprocessed LiDAR point cloud data; The plot identification module is used to input the preprocessed remote sensing images, preprocessed DEM data, and preprocessed LiDAR point cloud data into a multi-task crop planting plot identification model to obtain enhanced plot boundary features and non-crop feature mask features; combining the enhanced plot boundary features, non-crop feature mask features, and the multi-task crop planting plot identification model to obtain crop features, non-crop features, and plot features; Further, the processing process of the plot identification module includes: Inputting the preprocessed remote sensing images, preprocessed DEM data, and elevation difference edges into the input layer and feature extraction layer of the multi-task crop planting plot identification model to obtain remote sensing features, DEM features, and elevation difference edge features; Inputting the remote sensing features, DEM features, and elevation difference edge features into the plot boundary edge enhancement layer of the multi-task crop planting plot identification model to obtain enhanced plot boundary features; inputting the remote sensing features into the non-crop feature mask extraction layer to obtain non-crop feature mask features; Inputting the remote sensing features, enhanced plot boundary features, and non-crop feature mask features into the multi-task plot identification layer and output layer of the multi-task crop planting plot identification model to obtain crop features, non-crop features, and plot features.

[0041] The plot health module is used to obtain the health of crop planting plots according to the crop features, non-crop features, and plot features; The plot yield prediction module is used to input the remote sensing images and the health of crop planting plots into the crop planting plot yield prediction model for processing to obtain plot yield prediction information; Further, the process by which the plot yield prediction module obtains the plot yield prediction information includes: Training a crop planting plot yield prediction model using historical remote sensing images, historical health levels, and historical yields to obtain historical plot yield impact coefficients; Inputting the remote sensing image, the health level of the crop planting plot, and the historical plot yield impact coefficients into the trained crop planting plot yield prediction model for parameter update to obtain plot yield impact coefficients; Combining the plot yield impact coefficients with the current plot yield data to obtain plot yield prediction information.

[0042] To illustrate the plot yield prediction method proposed by the present invention, three groups of data from different planting plots are selected for testing, denoted as Test 1, Test 2, and Test 3 respectively; among them, the data includes: remote sensing images, the health levels of crop planting plots, and historical plot yield impact coefficients obtained through training of the crop planting plot yield prediction model; using the data to update the model to obtain the plot yield impact coefficients of each group; combining the method for obtaining plot yield prediction information to obtain the plot yield prediction information of different plots; the plot yield prediction results are shown in Table 3.

[0043] Table 3. Plot Yield Prediction Results

[0044] The output module is used to compare the health level of the crop planting plot with a preset threshold. If it exceeds the preset threshold, a warning message is output; otherwise, the plot yield prediction information is output.

[0045] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A crop planting plot monitoring method based on remote sensing images, characterized in that: include: Obtain remote sensing images, DEM data and LiDAR point cloud data of crop planting plots; Preprocessing the remote sensing image, the DEM data and the LiDAR point cloud data to obtain a preprocessed remote sensing image, preprocessed DEM data and elevation difference edges; Constructing a multi-task crop planting plot recognition model, inputting the pre-processed remote sensing image, the pre-processed DEM data and the elevation difference edge into the multi-task crop planting plot recognition model, and obtaining enhanced plot boundary features and non-crop object mask features; Combining the enhanced plot boundary features, the non-crop feature mask features and the multi-task crop planting plot recognition model, crop features, non-crop features and plot features are identified; Obtaining the health of the crop-planting plot according to the crop characteristics, the non-crop characteristics and the plot characteristics; Comparing the health of the crop planting plot with a preset threshold, and outputting a warning message if the preset threshold is exceeded; Otherwise, the remote sensing image and the health of the crop planting plot are input into the crop planting plot yield prediction model for processing to obtain the plot yield prediction information.

2. A crop planting plot monitoring method based on remote sensing images according to claim 1, characterized in that: The process of preprocessing the remote sensing image, the DEM data and the LiDAR point cloud data includes: correcting, enhancing and normalizing the remote sensing image to obtain the preprocessed remote sensing image; filtering the DEM data to obtain the preprocessed DEM data; filtering and normalizing the LiDAR point cloud data to obtain preprocessed LiDAR point cloud data; and extracting edges from an elevation difference map obtained by processing the preprocessed LiDAR point cloud data to obtain the elevation difference edges.

3. The crop planting plot monitoring method based on remote sensing images according to claim 1 is characterized in that: The multi-task crop planting plot recognition model includes: an input layer, a feature extraction layer, a plot boundary edge enhancement layer, a non-crop object mask extraction layer, a multi-task plot recognition layer and an output layer.

4. The crop planting plot monitoring method based on remote sensing images according to claim 1 is characterized in that: The process of acquiring the enhanced land parcel boundary feature, the non-crop feature, the crop feature, the non-crop feature and the land parcel feature includes: Inputting the preprocessed remote sensing image, the preprocessed DEM data and the elevation difference edge into the input layer and feature extraction layer of the multi-task crop planting plot recognition model to obtain remote sensing features, DEM features and elevation difference edge features; Inputting the remote sensing features, the DEM features and the elevation difference edge features into the plot boundary edge enhancement layer of the multi-task crop planting plot recognition model to obtain the enhanced plot boundary features; inputting the remote sensing features into the non-crop object mask extraction layer to obtain the non-crop object mask features; The remote sensing features, the enhanced plot boundary features and the non-crop object mask features are input into the multi-task plot recognition layer and the output layer of the multi-task crop planting plot recognition model to obtain the crop features, the non-crop features and the plot features.

5. The crop planting plot monitoring method based on remote sensing images according to claim 1 is characterized in that: The calculation process of the health of the crop planting plot includes: Obtaining a crop index, a non-crop index and a plot index respectively according to the crop characteristics, the non-crop characteristics and the plot characteristics; The crop index, the non-crop index and the plot index are weighted and summed to obtain the health of the crop-planting plot; wherein the calculation formula for the health of the crop-planting plot is: ； in, the health of the plot for planting said crop; is the crop weight factor; is the crop index; is the non-crop weight factor; is the non-crop index; is the land parcel weight factor; is the plot index.

6. The crop planting plot monitoring method based on remote sensing images according to claim 1 is characterized in that: The process of obtaining the yield prediction information of the plot includes: Using historical remote sensing images, historical health and historical yields to train the crop planting plot yield prediction model, and obtain the historical plot yield impact coefficient; The remote sensing image, the health of the crop planting plot and the historical plot yield impact coefficient are input into the trained crop planting plot yield prediction model to update parameters and obtain the plot yield prediction information.

7. A crop planting plot monitoring system based on remote sensing images, characterized in that: include: Data collection module, data preprocessing module, plot identification module, plot health module, plot yield prediction module and output module; The data acquisition module is used to obtain remote sensing images, DEM data and LiDAR point cloud data of crop planting plots; The data preprocessing module is used to preprocess the remote sensing image, the DEM data and the LiDAR point cloud data to obtain preprocessed remote sensing image, preprocessed DEM data and preprocessed LiDAR point cloud data; The plot recognition module is used to input the pre-processed remote sensing image, the pre-processed DEM data and the pre-processed LiDAR point cloud data into the multi-task crop planting plot recognition model to obtain enhanced plot boundary features and non-crop object mask features; Combining the enhanced plot boundary features, the non-crop feature mask features and the multi-task crop planting plot recognition model to obtain crop features, non-crop features and plot features; The plot health module is used to obtain the health of the crop planting plot according to the crop characteristics, the non-crop characteristics and the plot characteristics; The plot yield prediction module is used to input the remote sensing image and the health of the crop planting plot into the crop planting plot yield prediction model for processing to obtain plot yield prediction information; The output module is used to compare the health of the crop planting plot with a preset threshold, and output warning information if it exceeds the preset threshold; Otherwise, output the yield prediction information of the plot.

Images