A farmland construction quality intelligent sampling method, system, device and storage medium

Abstract

This invention discloses an intelligent sampling inspection method, system, equipment, and storage medium for farmland construction quality, relating to the field of agricultural monitoring technology. The method includes: acquiring remote sensing environmental data of a target area; the remote sensing environmental data includes multi-temporal remote sensing images and ground environmental data; inputting the remote sensing environmental data into a high-standard farmland time-series collaborative sampling model to determine sampling prediction results; the prediction results include a risk heat map and an uncertainty quantification map; the high-standard farmland time-series collaborative sampling model is constructed based on a variogram model, a time-series triggering mechanism, and Monte Carlo techniques, used to dynamically optimize the sampling strategy; and performing uncertainty analysis based on the sampling prediction results to generate the final sampling area. This invention enables precise monitoring and dynamic optimization control of the entire area of ​​high-standard farmland construction quality, providing technical support for the long-term guarantee of farmland quality in black soil regions.

Description

Technical Field

[0001] This invention relates to the field of agricultural monitoring technology, and in particular to an intelligent sampling inspection method, system, equipment and storage medium for farmland construction quality. Background Technology

[0002] The quality of high-standard farmland construction is a key factor in ensuring food security and improving agricultural disaster prevention and mitigation capabilities, and is of great significance to achieving agricultural modernization and sustainable development. However, the current quality inspection system still mainly relies on traditional methods (such as manual visual inspection, tape measure, core drilling, etc.). These methods are inefficient, costly, and difficult to achieve rapid and comprehensive monitoring of construction quality, thus failing to meet the needs of large-scale, high-precision testing.

[0003] With the rapid development of multi-source remote sensing and deep learning technologies, these advanced technologies have been initially applied in large-scale, rapid sampling inspections of high-standard farmland. However, existing technologies are mostly limited to single data sources or single construction content (such as farmland planning, irrigation facilities, and shelterbelt construction), failing to comprehensively cover multiple dimensions of factors such as farmland infrastructure, crop growth status, soil quality, meteorological conditions, and disaster impacts, resulting in insufficient representativeness and comprehensiveness of the sampling results. Furthermore, existing methods lack the ability to dynamically monitor the entire lifecycle of high-standard farmland construction (such as the construction and maintenance periods) and key phenological periods for crops (such as pre-sowing land preparation and post-harvest soil remediation), making it difficult to achieve temporal synergistic optimization of construction quality. Simultaneously, existing technologies have limited ability to resolve spatial heterogeneity, especially in complex terrain or areas with uneven data distribution, where prediction accuracy and reliability urgently need improvement. These limitations severely restrict the scientific assessment and effective supervision of the construction quality of high-standard farmland. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent sampling inspection method, system, equipment, and storage medium for farmland construction quality, which can realize precise monitoring and dynamic optimization and control of high-standard farmland construction quality across the entire area, and provide technical support for the long-term guarantee of farmland quality in black soil areas.

[0005] To achieve the above objectives, the present invention provides the following solution:

[0006] A method for intelligent spot-checking of farmland construction quality includes:

[0007] Acquire remote sensing environmental data of the target area; the remote sensing environmental data includes multi-temporal remote sensing images and ground environment data;

[0008] The remote sensing environmental data is input into the high-standard farmland time-series collaborative sampling model to determine the sampling prediction results; the prediction results include risk heat maps and uncertainty quantification maps; the high-standard farmland time-series collaborative sampling model is constructed based on the variogram model, time-series triggering mechanism and Monte Carlo technique, and is used to dynamically optimize the sampling strategy;

[0009] Uncertainty analysis is performed based on the sampling prediction results to generate the final sampling area.

[0010] Optionally, the process of acquiring remote sensing environmental data of the target area includes:

[0011] The system acquires remote sensing imagery data, ground environment data, manual assessment data of farmland construction quality in the corresponding imagery areas, and corresponding geographic coordinate data for the target area at different construction stages. The remote sensing imagery data includes multi-temporal high-resolution satellite imagery, radar imagery, and thermal infrared imagery. The ground environment data includes infrastructure construction data, crop growth data, meteorological data, and disaster data.

[0012] The remote sensing image data is preprocessed to obtain preliminary correction data; the preprocessing includes radiometric correction, geometric correction, image denoising, and image cropping.

[0013] Using GIS technology, the ground environment data, manual assessment data, and corresponding geographic coordinate data are mapped to the spatial coordinate system of the preliminary correction data for spatial registration, and time alignment is performed using linear interpolation to obtain the final remote sensing environment data.

[0014] Optionally, the high-standard farmland time-series collaborative sampling model specifically includes: an input layer, a network layer, and an output layer;

[0015] The input layer is used to input remote sensing environmental data, denoted as (B,T,W,H,N), where B represents the batch size, T represents the time step, W and H represent the width and height of the image, respectively, and N represents the number of channels of the image.

[0016] The network layer is used for: First, extracting multi-temporal remote sensing image features through the MobileNetV2 network, outputting a shape of (B,T,H',W',C1'), where H' and W' are the height and width of the feature map, and C1' is the number of feature channels; simultaneously, extracting features of facilities such as ditches and roads through the U-Net network, outputting a shape of (B,T,H,W,C2); next, upsampling the output of the MobileNetV2 network to (B,T,H,W,C1), and expanding the output channel number of the U-Net network through a 1x1 convolution to make C1 = C2; subsequently, concatenating the two features along the channel dimension to obtain a shape of ( The features (B, T, H, W, C1+C1) are compressed into (B, T, H, W, C3) using 1x1 convolution. Global average pooling is then applied to the compressed features to obtain a feature vector of shape (B, T, C3). Finally, meteorological time-series data (B, T, C4) and disaster time-series data (B, T, C5) are used as auxiliary features and concatenated with the image features to obtain a fused feature of shape (B, T, C3+C4+C5). These features are then input into an LSTM network for temporal modeling, outputting a feature representation of shape (B, T, hidden_size). A Dropout layer is added after the LSTM to evaluate the model's uncertainty.

[0017] The output layer is used to: process the data through the fully connected layer and output the sampling prediction results.

[0018] Optionally, the MobileNetV2 network specifically includes: a first convolutional layer, a bottleneck module, a second convolutional layer, and a global average pooling layer connected in sequence; wherein, the bottleneck module includes 7 bottleneck layers; the bottleneck layer is composed of 1x1 convolution, 3x3 depthwise separable convolution, and 1x1 convolution.

[0019] Optionally, the U-Net network specifically includes: an encoder, a bottleneck layer, and a decoder connected in sequence; wherein, the encoder consists of multiple convolutional blocks, each convolutional block including two 3x3 convolutional layers, a ReLU activation function, and a 2x2 max pooling layer; the bottleneck layer includes two 3x3 convolutional layers and a ReLU activation function; the decoder consists of multiple upsampling blocks, each upsampling block including a 2x2 upconvolutional layer, a skip connection layer, two 3x3 convolutional layers, and a ReLU activation function.

[0020] This invention also provides an intelligent sampling inspection system for farmland construction quality, comprising:

[0021] The data acquisition unit is used to acquire remote sensing environmental data of the target area; the remote sensing environmental data includes multi-temporal remote sensing images and ground environment data.

[0022] The model prediction unit is used to input the remote sensing environmental data into the high-standard farmland time-series collaborative sampling model to determine the sampling prediction results; the prediction results include a risk heat map and an uncertainty quantification map; the high-standard farmland time-series collaborative sampling model is constructed based on the variogram model, the time-series triggering mechanism and the Monte Carlo technique, and is used to dynamically optimize the sampling strategy;

[0023] The sampling determination unit is used to perform uncertainty analysis based on the sampling prediction results and generate the final sampling area.

[0024] The present invention also provides an electronic device, including a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to perform the above-described intelligent sampling inspection method for farmland construction quality.

[0025] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the intelligent sampling inspection method for farmland construction quality as described above.

[0026] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0027] This invention discloses an intelligent sampling inspection method, system, equipment, and storage medium for farmland construction quality. The method includes acquiring remote sensing environmental data of the target area; the remote sensing environmental data includes multi-temporal remote sensing images and ground environmental data; inputting the remote sensing environmental data into a high-standard farmland time-series collaborative sampling model to determine the sampling prediction results; the prediction results include a risk heat map and an uncertainty quantification map; the high-standard farmland time-series collaborative sampling model is constructed based on a variogram model, a time-series triggering mechanism, and Monte Carlo techniques to dynamically optimize the sampling strategy; and uncertainty analysis is performed based on the sampling prediction results to generate the final sampling area. This invention enables precise monitoring and dynamic optimization control of the entire area of ​​high-standard farmland construction quality, providing technical support for the long-term guarantee of farmland quality in black soil regions. Attached Figure Description

[0028] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 This is a schematic diagram of the time-series collaborative sampling model for high-standard farmland in this embodiment;

[0030] Figure 2This is a schematic diagram of the MobileNetV2 network structure in this embodiment;

[0031] Figure 3 This is a schematic diagram of the U-Net network structure in this embodiment. Detailed Implementation

[0032] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0033] The purpose of this invention is to provide an intelligent sampling inspection method, system, equipment, and storage medium for farmland construction quality, which can realize precise monitoring and dynamic optimization and control of high-standard farmland construction quality across the entire area, and provide technical support for the long-term guarantee of farmland quality in black soil areas.

[0034] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0035] like Figure 1 As shown, this invention provides an intelligent sampling inspection method for farmland construction quality, comprising:

[0036] Step 100: Acquire remote sensing environmental data of the target area; the remote sensing environmental data includes multi-temporal remote sensing images and ground environment data.

[0037] Step 200: Input the remote sensing environmental data into the high-standard farmland time-series collaborative sampling model to determine the sampling prediction results; the prediction results include a risk heat map and an uncertainty quantification map; the high-standard farmland time-series collaborative sampling model is constructed based on the variogram model, time-series triggering mechanism and Monte Carlo technique, and is used to dynamically optimize the sampling strategy.

[0038] Step 300: Based on the sampling prediction results, perform uncertainty analysis to generate the final sampling area.

[0039] As a specific implementation method, the specific calculation process of each of the above steps is as follows.

[0040] 1. Data Acquisition and Preprocessing

[0041] Acquire multi-temporal, high-resolution satellite, radar, and thermal infrared imagery of high-standard farmland at different construction stages; acquire farmland construction data (such as the spatial distribution of infrastructure like ditches, roads, and irrigation facilities); crop growth data (such as crop type and growth cycle); meteorological data (such as daily average temperature, rainfall, and humidity); disaster data (such as the timing and impact range of floods and droughts); and acquire manual assessment data of the construction quality of high-standard farmland in the corresponding image areas, along with their corresponding geographic coordinates.

[0042] The remote sensing images are preprocessed, including radiometric correction, geometric correction, image denoising, and image cropping. Then, spatial registration of the ground environment data and remote sensing images is performed. The acquired ground environment data and manually assessed data are mapped to the spatial coordinate system of the preprocessed remote sensing images using a Geographic Information System (GIS). Finally, temporal registration of the ground data and remote sensing images is performed. Linear interpolation is used to align the multi-time ground data with the multi-temporal remote sensing images.

[0043] 2. Import the multi-temporal remote sensing images and ground data processed in step 1 into the high-standard farmland temporal collaborative sampling model constructed in this invention, such as... Figure 1 As shown, Monte Carlo (Dropout) technology is introduced into the model training process, randomly dropping some neurons during training to assess the model's uncertainty. The specific process includes:

[0044] Input layer: The model input includes multi-temporal remote sensing images and labeled data such as ditches and channels, with shapes of (B,T,W,H,N), where B represents the batch size, T represents the time step, W and H represent the width and height of the image, respectively, and N represents the number of channels of the image.

[0045] Network Structure: First, MobileNetV2 extracts multi-temporal remote sensing image features, outputting a shape of (B,T,H',W',C1'), where H' and W' are the height and width of the feature map, and C1' is the number of feature channels. Simultaneously, U-Net extracts features of facilities such as ditches and roads, outputting a shape of (B,T,H,W,C2). Next, the output of MobileNetV2 is upsampled and adjusted to (B,T,H,W,C1), and the number of channels in the output of U-Net is expanded by a 1x1 convolution to make the number of channels in both networks the same (i.e., C1 = C2). Subsequently, the two features are concatenated along the channel dimension to obtain a feature vector of shape (B,T,H,W,C1+C1), which is then compressed into (B,T,H,W,C3) by a 1x1 convolution. Global average pooling (GAP) is then performed on the compressed features to obtain a feature vector of shape (B,T,C3). Furthermore, meteorological time-series data (shape (B,T,C4)) and disaster time-series data (shape (B,T,C5)) are used as auxiliary features and concatenated with image features to obtain a fused feature with shape (B,T,C3+C4+C5). This feature is then input into an LSTM for time-series modeling, outputting a shape of (B,T,hidden_size). A Dropout layer is added after the LSTM to evaluate the model's uncertainty.

[0046] Output Layer: After processing by the fully connected layer, the model finally outputs the prediction results, including a risk heatmap and an uncertainty quantification map. The risk heatmap is used to identify potential risk areas in farmland construction, while the uncertainty quantification map uses Monte Carlo Dropout technology to evaluate the reliability of the model's predictions.

[0047] like Figure 2 As shown, the MobileNetV2 network includes: the initial convolutional layer Conv2d is a 3×3 convolutional layer with stride=2, used for initial feature extraction; the bottleneck layers (BottleneckBlocks) consist of 1x1 convolutions, 3x3 depthwise separable convolutions (Depthwise Conv), and 1x1 convolutions; Global Average Pooling (GAP) is used to perform global average pooling on the feature maps of each channel to compress the spatial dimension; the last Conv layer is used to adjust the number of output channels and output the result.

[0048] like Figure 3As shown, the U-Net network consists of: an encoder composed of multiple convolutional blocks, each containing two 3x3 convolutional layers (Conv2d), a ReLU activation function, and 2x2 max pooling; a bottleneck layer located between the encoder and decoder, containing two 3x3 convolutional layers (Conv2d) and a ReLU activation function; a decoder composed of multiple upsampling blocks, each containing a 2x2 upconvolutional layer (UpConv), a skip connection (Copy and Crop), two 3x3 convolutional layers (Conv2d), and a ReLU activation function; and a final Conv layer used to map the number of channels to the number of target classes.

[0049] 3. Spatial Correction of Predicted Values. A variogram model is constructed using Kriging interpolation to fully explore the autocorrelation and distribution patterns of the spatial data, providing optimal unbiased estimation for unknown points. The variogram construction formula is as follows:

[0050]

[0051] Where c0 represents microscale variation or measurement error, c represents the variance caused by spatial autocorrelation, and a represents the range of effects of spatial autocorrelation. This method generates a continuous spatial distribution map by analyzing the spatial structure characteristics of measured data and quantifies the uncertainty of interpolation results. It effectively solves the prediction bias problem caused by complex terrain or uneven data distribution in the time series collaborative sampling model of high-standard farmland, and significantly improves the accuracy and reliability of prediction results.

[0052] 4. Embedded Time-Series Triggering Mechanism. A time-series rule base is designed, and a rule engine dynamically triggers adjustments to the sampling strategy. Time-series triggering mechanisms (such as focusing on areas with complex terrain during construction; focusing on areas with abnormal crop growth during maintenance; and focusing on fertilization and irrigation areas during certain phenological periods) are embedded into the time-series collaborative sampling model. This mechanism automatically adjusts the sampling frequency and spatial centroid distribution of the model's input data by monitoring changes in time nodes and farmland construction stages in real time, enabling the model to adapt to the needs of different stages. The optimized input data from the time-series triggering mechanism directly affects the model's prediction results (such as risk heatmaps and uncertainty quantification maps). The prediction result display module uses these results as input to generate visualization results and the final sampled areas.

[0053] 5. Prediction Results Display. After completing the prediction, the model directly outputs a risk heatmap and an uncertainty quantification map. Simultaneously, combining the prediction results with uncertainty analysis, it automatically generates the final sampling area. Through integrated interactive visualization tools, users can dynamically adjust the threshold weighting of uncertainty quantification according to the needs of different periods, flexibly optimizing the division of sampling areas.

[0054] Therefore, this solution has the following beneficial effects:

[0055] This method integrates multi-temporal remote sensing images and ground environment data, combines a time-series triggering mechanism and Monte Carlo Dropout technology to dynamically optimize the sampling strategy, and automatically generates sampling areas based on risk heat maps and uncertainty quantification maps, thereby achieving intelligent and precise sampling of large-scale high-standard farmland.

[0056] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0057] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for intelligent sampling inspection of farmland construction quality, characterized in that, include: Acquire remote sensing environmental data of the target area; the remote sensing environmental data includes multi-temporal remote sensing images and ground environment data; The remote sensing environmental data is input into the high-standard farmland time-series collaborative sampling model to determine the sampling prediction results; the prediction results include risk heat maps and uncertainty quantification maps; the high-standard farmland time-series collaborative sampling model is constructed based on the variogram model, time-series triggering mechanism and Monte Carlo technique, and is used to dynamically optimize the sampling strategy; Uncertainty analysis is performed based on the sampling prediction results to generate the final sampling area; The high-standard farmland time-series collaborative sampling model specifically includes: an input layer, a network layer, and an output layer; The input layer is used to input remote sensing environmental data, denoted as (B, T, W, H, N), where B represents the batch size, T represents the time step, W and H represent the width and height of the image, respectively, and N represents the number of channels of the image. The network layer is used for the following steps: First, it extracts multi-temporal remote sensing image features using the MobileNetV2 network, outputting a shape of (B, T, H', W', C1'), where H' and W' are the height and width of the feature map, and C1' is the number of feature channels. Simultaneously, it extracts features of ditches, roads, and irrigation facilities using the U-Net network, outputting a shape of (B, T, H, W, C2). Next, it upsamples the output of the MobileNetV2 network, adjusting it to (B, T, H, W, C1), and expands the number of output channels of the U-Net network using a 1x1 convolution, making C1 = C2. Then, it concatenates the two features along the channel dimension to obtain a feature vector of shape (B, T, H, W, C1 + C1), and compresses it to (B, T, H, W, C3) using a 1x1 convolution. It then performs global average pooling on the compressed feature to obtain a feature vector of shape (B, T, C3). Finally, it integrates the meteorological time-series data (B, T, C4). Disaster time series data (B, T, C5) are used as auxiliary features and concatenated with image features to obtain a fused feature with shape (B, T, C3 + C4 + C5). The features are then input into an LSTM network for time series modeling, and the output feature representation has shape (B, T, hidden_size). A Dropout layer is added after the LSTM to evaluate the uncertainty of the model. The output layer is used to: process the data through the fully connected layer and output the sampling prediction results.

2. The intelligent sampling inspection method for farmland construction quality according to claim 1, characterized in that, The specific process of acquiring remote sensing environmental data of the target area includes: The system acquires remote sensing imagery data, ground environment data, manual assessment data of farmland construction quality in the corresponding imagery areas, and corresponding geographic coordinate data for the target area at different construction stages. The remote sensing imagery data includes multi-temporal high-resolution satellite imagery, radar imagery, and thermal infrared imagery. The ground environment data includes infrastructure construction data, crop growth data, meteorological data, and disaster data. The remote sensing image data is preprocessed to obtain preliminary correction data; the preprocessing includes radiometric correction, geometric correction, image denoising, and image cropping. Using GIS technology, the ground environment data, manual assessment data, and corresponding geographic coordinate data are mapped to the spatial coordinate system of the preliminary correction data for spatial registration, and time alignment is performed using linear interpolation to obtain the final remote sensing environment data.

3. The intelligent sampling inspection method for farmland construction quality according to claim 1, characterized in that, The MobileNetV2 network specifically includes: a first convolutional layer, a bottleneck module, a second convolutional layer, and a global average pooling layer connected in sequence; wherein, the bottleneck module includes 7 bottleneck layers; the bottleneck layer is composed of 1x1 convolution, 3x3 depthwise separable convolution and 1x1 convolution.

4. The intelligent sampling inspection method for farmland construction quality according to claim 1, characterized in that, The U-Net network specifically includes: an encoder, a bottleneck layer, and a decoder connected in sequence; wherein, the encoder consists of multiple convolutional blocks, each of which includes two 3x3 convolutional layers, a ReLU activation function, and a 2x2 max pooling layer; the bottleneck layer includes two 3x3 convolutional layers and a ReLU activation function; the decoder consists of multiple upsampling blocks, each of which includes a 2x2 upconvolutional layer, a skip connection layer, two 3x3 convolutional layers, and a ReLU activation function.

5. An intelligent sampling inspection system for farmland construction quality, using the method described in any one of claims 1-4, characterized in that, include: The data acquisition unit is used to acquire remote sensing environmental data of the target area; the remote sensing environmental data includes multi-temporal remote sensing images and ground environment data. The model prediction unit is used to input the remote sensing environmental data into the high-standard farmland time-series collaborative sampling model to determine the sampling prediction results; the prediction results include a risk heat map and an uncertainty quantification map; the high-standard farmland time-series collaborative sampling model is constructed based on the variogram model, the time-series triggering mechanism and the Monte Carlo technique, and is used to dynamically optimize the sampling strategy; The sampling determination unit is used to perform uncertainty analysis based on the sampling prediction results and generate the final sampling area.

6. An electronic device, characterized in that, The device includes a memory and a processor, wherein the memory stores a computer program and the processor runs the computer program to enable the electronic device to perform the intelligent sampling inspection method for farmland construction quality according to any one of claims 1-4.

7. A computer-readable storage medium, characterized in that, It stores a computer program, which, when executed by a processor, implements the intelligent sampling inspection method for farmland construction quality as described in any one of claims 1-4.

Images