An agricultural post-disaster verification and yield recovery evaluation method, system, device and medium based on unmanned aerial vehicle images
By using UAV image processing technology, the problems of low efficiency, poor accuracy, and inaccurate information in agricultural post-disaster verification and yield recovery assessment have been solved. It has achieved high-precision disaster identification and yield change trend analysis, supporting rapid and refined agricultural post-disaster assessment.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHINA TOWER CO LTD
- Filing Date
- 2026-01-13
- Publication Date
- 2026-06-12
Smart Images
Figure CN122198302A_ABST
Abstract
Description
Technical Field
[0001] This application belongs to the field of agricultural digital management technology, and specifically relates to a method, system, equipment and medium for agricultural post-disaster verification and yield recovery assessment based on UAV images. Background Technology
[0002] Agriculture is a fundamental industry of the national economy and is significantly affected by natural disasters (such as floods, droughts, pests, and fires). After a disaster occurs, timely and accurate information on the affected areas, the extent of the damage, and the potential yield losses is of great importance for formulating remedial measures, agricultural insurance claims, and helping farmers resume production. The existing technology has the following drawbacks: 1. Manual survey and assessment: Manual methods are inefficient and costly, making it difficult to quickly cover large-scale agricultural areas and resulting in a lag in understanding the dynamics of the disaster; information is highly subjective: there is a lack of unified measurement standards, and manual scoring is prone to bias; data update frequency is low: it is impossible to achieve full-process tracking from post-disaster to recovery. 2. Problems with image registration difficulties and poor disaster location accuracy; 3. Problems with crude disaster severity classification and weak disaster type identification capabilities; 4. The lack of a dynamic production recovery modeling mechanism.
[0003] Therefore, it is a feasible and preferred approach to make up for the shortcomings of existing solutions in terms of practicality, precision, and automation by using an agricultural post-disaster verification and yield assessment process that has the ability to compare and identify images before and after disasters, classify disaster levels, and model yield recovery trends. Summary of the Invention
[0004] To address the aforementioned issues, this application provides a method, system, equipment, and medium for agricultural post-disaster verification and yield recovery assessment based on UAV imagery. This method and system, through an agricultural post-disaster verification and yield assessment process capable of comparing and recognizing pre- and post-disaster images, classifying disaster severity levels, and modeling yield recovery trends, compensates for the shortcomings of existing solutions in terms of practicality, precision, and automation.
[0005] This application provides a method for agricultural post-disaster verification and yield recovery assessment based on UAV imagery, the method comprising, Acquire initial UAV image data corresponding to the pre-disaster and post-disaster time points of the target time, and perform standardization preprocessing and image registration operations on the initial UAV image data to obtain target UAV image data; Image differencing and threshold segmentation are performed on the target UAV image data to construct a disaster change mask map, and actual disaster information is extracted based on the disaster change mask map; Key characteristic indicators of the target farmland area before and after the disaster are extracted. A damage assessment characteristic table is constructed based on the key characteristic indicators. The target farmland area is classified into disaster levels based on the damage assessment characteristic table to obtain the disaster level classification results. A yield change trend model is constructed based on the changes in historical image sequences of fields, and the impact of the current disaster on the yield is assessed based on the yield change trend model to obtain the yield change trend assessment result. The obtained actual disaster information, disaster level classification results, and yield change trend assessment results are used to conduct damage verification and crop yield recovery potential analysis, and the target farmland area is managed accordingly based on the verification and analysis results.
[0006] Furthermore, Acquire initial UAV image data corresponding to the pre-disaster and post-disaster time points at the target time, and perform standardization preprocessing and image registration operations on the initial UAV image data to obtain target UAV image data, specifically including: The drone, equipped with a multispectral camera, takes low-altitude photos of the target farmland area at the target time point to obtain initial drone image data containing the three primary color channels and the near-infrared band. The initial UAV image data is subjected to grayscale normalization and color correction to obtain intermediate UAV image data; A multi-scale matching algorithm based on local feature operators is used to spatially align the intermediate UAV image data at the pixel level through an affine transformation matrix to obtain the target UAV image data.
[0007] Furthermore, Image differencing and threshold segmentation are performed on the target UAV image data to construct a disaster change mask map, and actual disaster information is extracted based on the disaster change mask map, specifically including: A standardized difference map is determined based on the target UAV image data; The standardized difference map is binarized and segmented using the maximum inter-class variance criterion automatic thresholding method to generate the disaster change mask map; Based on the proportion of affected pixels and ground resolution in the disaster change mask image, calculate the actual affected area ratio and spatial distribution range. The actual disaster information includes: the proportion of the actual disaster-affected area and the spatial distribution range.
[0008] Furthermore, Key characteristic indicators of the target farmland area before and after the disaster are extracted. A damage assessment feature table is constructed based on these key characteristic indicators. The target farmland area is then classified into disaster severity levels based on this damage assessment feature table, resulting in a disaster severity level classification. Specifically, this includes: Based on the disaster change mask map, the target farmland area is divided into independent field areas, and the key feature indicators of each independent field area before and after the disaster are extracted. Based on the damage assessment feature table and the set feature change range and threshold or training model results, the target farmland area is divided into mild, moderate and severe levels.
[0009] Furthermore, A yield change trend model is constructed based on the changes in historical image sequences of fields, and the impact of the current disaster on the yield is assessed based on the yield change trend model to obtain the yield change trend assessment results, specifically including: Historical image sequences of each independent field area from the sowing period to the post-disaster recovery period were collected, and the mean vegetation index at key time points was extracted to construct a time series. Based on the time series, the yield change trend model is constructed by first-order linear regression to fit the pre-disaster growth trend slope apre and the post-disaster recovery trend slope apost respectively, and the perturbation slope difference Δa=apost-apre is calculated. Based on the disaster severity classification results, the production trend is determined according to the Δa value, and the production change trend assessment result is obtained.
[0010] Furthermore, The obtained actual disaster information, disaster severity classification results, and yield change trend assessment results are used to conduct damage verification and crop yield recovery potential analysis. Based on the verification and analysis results, corresponding management is implemented for the target farmland area, specifically including: Establish a task lifecycle management mechanism to record the entire process status log from image preprocessing to result archiving; Anomaly detection is performed on the disaster change mask map, the disaster level classification results, and the production change trend assessment results; The platform outputs damage verification reports and production recovery potential analysis results through visualization, and triggers the review process for abnormal data.
[0011] Secondly, based on the same inventive concept, this application provides an agricultural post-disaster verification and yield recovery assessment system based on UAV imagery. The system includes: The image acquisition and preprocessing module acquires the initial UAV image data corresponding to the pre-disaster and post-disaster time points of the target time, and performs standardization preprocessing and image registration operations on the initial UAV image data to obtain the target UAV image data. The disaster change detection module performs image difference and threshold segmentation on the target UAV image data to construct a disaster change mask map, and extracts actual disaster information based on the disaster change mask map; The disaster severity assessment module extracts key characteristic indicators of the target farmland area before and after the disaster, constructs a damage assessment feature table based on the key characteristic indicators, and classifies the target farmland area into disaster severity levels based on the damage assessment feature table to obtain the disaster severity level classification results. The yield trend modeling module constructs a yield change trend model based on the changes in historical image sequences of fields, and assesses the impact of the current disaster on the yield based on the yield change trend model to obtain the yield change trend assessment result. The comprehensive analysis and decision support module performs damage verification and crop yield recovery potential analysis on the obtained actual disaster information, disaster level classification results, and yield change trend assessment results, and manages the target farmland area accordingly based on the verification and analysis results.
[0012] Thirdly, this application also provides an electronic device, including at least one processor and at least one memory electrically connected; The memory is electrically connected to the processor, wherein the memory stores instructions that can be executed by at least one of the processors, the instructions being executed by at least one of the processors to enable at least one of the processors to perform any of the above-described methods for agricultural post-disaster verification and yield recovery assessment based on UAV imagery.
[0013] Fourthly, this application also provides a computer storage medium, wherein a computer program is stored within the computer-readable storage medium; When the computer program is executed by the processor, it implements any of the above-described methods for agricultural post-disaster verification and yield recovery assessment based on UAV imagery.
[0014] Fifthly, this application also provides a computer program product, which is stored in at least one storage medium; The computer program product includes several instructions to cause at least one electronic device to execute any of the above-described methods for agricultural post-disaster verification and yield recovery assessment based on UAV imagery.
[0015] Compared with the prior art, this application has the following advantages: 1. By using a multi-scale feature matching algorithm, spatial overlap and comparative analysis of images under different flight times and angles are achieved, providing a high-precision base map guarantee for disaster mask construction and feature extraction; 2. A disaster mask map generation method based on saliency enhancement and regional consistency optimization is proposed, which can accurately identify disaster-stricken areas in farmland and adapt to multiple types of disaster scenarios; 3. A disaster severity rating system based on multi-dimensional feature indicators was constructed. Through rule mapping, mild, moderate and severe disasters were classified, enabling structured and interpretable risk output of post-disaster images. 4. Introduce the "pre-disaster slope-post-disaster slope" breakpoint slope difference modeling method, and combine it with historical crop growth patterns to accurately depict the yield change trend of disaster-affected crops (stable, declining, or drastic decline), and realize the directional quantification of yield conditions.
[0016] Other features and advantages of this application will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the application. The objectives and other advantages of this application may be realized and obtained by means of the structures pointed out in the description, claims and drawings. Attached Figure Description
[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0018] Figure 1 A flowchart illustrating the method for agricultural post-disaster verification and yield recovery assessment based on UAV imagery, according to an embodiment of this application, is shown. Detailed Implementation
[0019] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0020] Figure 1 This paper illustrates a method for agricultural post-disaster verification and yield recovery assessment based on unmanned aerial vehicle (UAV) imagery, according to an embodiment of this application. For example... Figure 1 As shown in the figure, the method for agricultural post-disaster verification and yield recovery assessment based on UAV images in this application includes the following steps: S1, acquire the initial UAV image data corresponding to the pre-disaster and post-disaster time point of the target, and perform standardization preprocessing and image registration operations on the initial UAV image data to obtain the target UAV image data; S2, perform image difference and threshold segmentation on the target UAV image data to construct a disaster change mask map, and extract actual disaster information based on the disaster change mask map; S3, extract key feature indicators of the target farmland area before and after the disaster, construct a damage assessment feature table based on the key feature indicators, and classify the target farmland area into disaster levels based on the damage assessment feature table to obtain the disaster level classification results. S4. Construct a yield change trend model based on the changes in historical image sequences of the field, and assess the impact of the current disaster on the yield based on the yield change trend model to obtain the yield change trend assessment result. S5. The obtained actual disaster information, disaster level classification results and yield change trend assessment results are used to conduct damage verification and crop yield recovery potential analysis, and the target farmland area is managed accordingly based on the verification results and analysis results.
[0021] In the embodiments of this application, step S1 specifically includes: S11, using a drone equipped with a multispectral camera to take low-altitude photos of the target farmland area at the target time point, and acquire the initial drone image data containing the three primary color channels and the near-infrared band; S12, perform grayscale normalization and color correction on the initial UAV image data to obtain intermediate UAV image data; S13, using a multi-scale matching algorithm based on local feature operators, the intermediate UAV image data is spatially aligned at the pixel level through an affine transformation matrix to obtain the target UAV image data.
[0022] In the specific implementation process, step S1 aims to construct image pairs that can be used for pre-disaster to post-disaster change detection, providing accurate pixel-level correspondences for subsequent disaster area extraction and yield trend modeling, specifically including: 1) Image data acquisition and scene constraints: In an agricultural scenario, a target farmland area Ω was selected. A drone equipped with a multispectral RGB (Red-Green-Blue) camera was used to perform low-altitude aerial photography before the disaster (time t0) and after the disaster (time t1). The image before the disaster was recorded as follows: Post-disaster images are ,in: (x,y) represents the image space coordinates; c∈{R,G,B,NIR} represents the channel, and if it is a multispectral image, it includes the near-infrared (NIR) band. In the specific implementation process, in order to improve the accuracy of subsequent feature comparison, the following consistency must be ensured during the image acquisition process: The flight altitude and angle remain stable (within ±2m and ±3°). Image overlap ≥ 70% to ensure consistent coverage area; Acquire pre-disaster and post-disaster images within a short period (1-3 days) when crop growth does not change drastically; 2) Image normalization preprocessing: To eliminate interference from non-disaster-related factors such as differences in lighting and camera characteristics, image normalization processing is required, including grayscale normalization and linear normalization of the pixel values of each image I.
[0023] in, The raw pixel value for channel c indicates the location before / after the disaster. band The original pixel value at that location; , These are the mean and standard deviation of channel c in the image, respectively. If the image is in RGB format, Retinex (retinal cortex theory) or white balance methods can be used to correct color imbalance in order to ensure the consistency of color space distribution in the images before and after the disaster, and further suppress the impact of lighting differences. 3) Image spatial registration and weak alignment algorithm: During the implementation process, due to slight differences in perspective and displacement in the actual aerial images, image registration is required to reconcile the pre-disaster images. Images of the aftermath of the disaster Establish pixel-level correspondences in space; 1. Feature point extraction and matching: Keypoint sets from pre-disaster and post-disaster images are extracted using local feature operators with strong invariance.
[0024] Among them, each key point i is the index, indicating the first... One key point; This indicates that the image was taken before the disaster. The extracted key point set, where j is the index. This indicates that these are images from after the disaster. The extracted keypoint set includes location, orientation, and descriptors; Hamming distance is used to match the descriptors, constructing a set of corresponding point pairs.
[0025] in, This is the first image before the disaster. Visual feature vectors of key points This is the first post-disaster image. Visual feature vectors of key points Represents the Hamming distance, with a set threshold. Only when the Hamming distance is less than Only if it is considered a successful match.
[0026] By fitting the matching point pairs, the two-dimensional affine transformation matrix A from the pre-disaster image to the post-disaster image is estimated, where, , , , These form linear transformation submatrices, representing geometric transformations such as rotation, scaling, and shearing. Let x and y be the translation amounts:
[0027] Make:
[0028] After performing an affine transformation, the image is then interpolated using bilinear interpolation. Resampling was performed to obtain images similar to those after the disaster. Aligned version .
[0029] In the embodiments of this application, step S2 specifically includes: S21, Determine a standardized difference map based on the target UAV image data; S22, The standardized difference map is binarized and segmented using the maximum inter-class variance criterion automatic thresholding method to generate the disaster change mask map; S23, Calculate the actual affected area ratio and spatial distribution range based on the proportion of affected pixels in the disaster change mask image and the ground resolution; The actual disaster information includes: the proportion of the actual disaster-affected area and the spatial distribution range.
[0030] In the specific implementation process, step S2 aims to use the pre-disaster images that have undergone spatial registration and preprocessing in step S1 as a basis. Images of the aftermath of the disaster A differential image reflecting the disaster-affected area is constructed, and a binary disaster change mask map is generated through saliency enhancement and automated thresholding methods. This allows for the identification of disaster-affected areas in farmland and the quantitative extraction of their area and spatial distribution. Specifically, this includes: 1) Image difference construction: Considering that disaster-stricken farmland areas will undergo significant changes in spectral reflectance, color structure, and texture features, the original image difference function is defined as follows: Using an RGB image, the difference calculation applies to all three channels:
[0031] Where (x,y) represents the pixel position. For each band Calculate the difference between pixel values before and after the disaster separately; if using multispectral images, near-infrared (NIR) or NDVI (Normalized Difference Vegetation Index Difference) differencing should be prioritized. Furthermore, to enhance the contrast of the affected area, a standardized difference map is defined.
[0032] Where, μ D σ D These represent the mean and standard deviation of the difference plot D(x,y), respectively. The image includes the near-infrared band, allowing for further calculation of the difference in NDVI before and after the disaster.
[0033]
[0034] in, The values are near-infrared. The values are for the red band. The difference in NDVI before and after the disaster is represented by ε, which is a small positive number to avoid division by zero. Significant decreases in NDVI usually correspond to agricultural disasters such as floods, diseases, and lodging. 2) Threshold segmentation generates a disaster situation change mask map: For standardized difference plots An image adaptive segmentation algorithm is used to generate a disaster situation change mask map M(x,y). Common methods include: The Otsu's between-class variance criterion automatic thresholding method automatically solves for the optimal segmentation threshold using the Otsu's between-class variance criterion.
[0035] Where L is the number of gray levels in the image; These represent the proportions of pixels on either side of the threshold t; The formula for generating the change map is: (The formula is not provided in the original text.)
[0036] Where M(x,y) is a binary disaster mask image; 3) Extraction of the disaster area's area and spatial dimensions: For the obtained binary disaster mask image The area percentage of pixels with a value of 1 (i.e., the disaster-stricken area) can be calculated and used as a quantitative indicator to participate in the determination of the disaster level and the estimation of production loss. Let the total number of pixels in the entire image be N, where, The corrected disaster mask image has the following number of affected pixels:
[0037] The proportion of the affected area is:
[0038] If we combine the known ground resolution of the image (Unit area / pixel) can be used to further estimate the actual affected area. :
[0039] Where, r g In addition to the ground resolution of the image, each field area can be numbered, and the mask image is divided into independent field units based on the spatial boundary extraction algorithm for subsequent disaster level classification and yield recovery modeling.
[0040] In the embodiments of this application, step S3 specifically includes: S31, Based on the disaster change mask map, the target farmland area is divided into independent field areas, and the key feature indicators of each independent field area before and after the disaster are extracted. S32, based on the damage assessment feature table and the set feature change range and threshold or training model results, the target farmland area is divided into mild, moderate and severe levels.
[0041] In the specific implementation process, step S3 aims to utilize the disaster change mask map extracted in step S2. Within each farmland area (which can be extracted from plot vector boundaries or connected components), multiple salient features (vegetation index, color statistics, texture changes, etc.) are extracted from pre- and post-disaster images to construct a structured disaster assessment feature table. Then, based on the magnitude and threshold of feature changes, or the results of training models, the disaster area is divided into three categories: mild, moderate, and severe, to support subsequent applications such as agricultural insurance and precision intervention. Specifically, this includes: 1) Field area division and numbering: The entire mask image Based on connectivity analysis or by combining existing field boundaries, a set of spatially non-overlapping, uniquely numbered farmland regions is obtained:
[0042] Each Ti is a set of pixel coordinates, representing the i-th farmland area. In the specific implementation process, if the access of the plot layer is supported, the intersection of the disaster areas within the plot can also be obtained by overlaying the vector map and the mask. 2) Feature index design and extraction: For each field Ti, in the pre-disaster image... Post-disaster images Extract the following types of image features: Vegetation index features: The field-level mean is defined as:
[0043] Define the extent of vegetation degradation:
[0044] in, The more negative the value, the more severe the degradation of vegetation cover after the disaster; Color statistical features: Extract the mean and standard deviation of each channel in the RGB channels as color features:
[0045] And the magnitude of the change:
[0046] in, For the range of color change, The color characteristics of the three-color channel before and after the disaster are as follows: the rise of the red channel corresponds to vegetation fading or bare land, and the rise of the blue channel is often accompanied by water reflection. The color characteristics are significantly distinguishable between floods and droughts. 3) Construct a disaster assessment characteristic table: The characteristics of all fields are integrated into a structured table, defined as:
[0047] Based on the above characteristics, the disaster severity level is defined as follows: Slight damage: Vegetation is slightly degraded, but its texture is basically intact; Moderate damage: Vegetation has been significantly degraded, and some of its texture has been destroyed; Severe damage: NDVI significantly decreased, colors appear bare / waterlogged, and textures are chaotic.
[0048] The threshold parameters are set as follows:
[0049] Color and texture are used as auxiliary conditions to cover different types of disasters.
[0050] In the embodiments of this application, step S4 specifically includes: S41, Collect the historical image sequence of each independent field area from the sowing period to the post-disaster recovery period, and extract the mean vegetation index at key time points to construct a time series; S42, Based on the time series, construct the yield change trend model through first-order linear regression to fit the pre-disaster growth trend slope apre and the post-disaster recovery trend slope apost respectively, and calculate the perturbation slope difference Δa=apost-apre; S43, Combining the disaster level classification results, the production trend is determined based on the Δa value to obtain the production change trend assessment result.
[0051] In the specific implementation process, step S4, based on the historical image sequence of each field area, extracts vegetation indices at key time points and models their dynamic change trends at different stages (sowing period, growing period, post-disaster period). Through differential slope modeling, trend identification, and comparison with disaster disturbances, it achieves directional judgments such as "stable, declining, or sharply declining" of crop yield trends, providing a forward-looking basis for agricultural situation analysis and insurance claims. Specifically, it includes: 1) Image time series and key time period selection: The research period is defined as a continuous timeline within the crop life cycle, including key time points such as sowing, growth period, pre-disaster, and post-disaster. The corresponding UAV imagery data is as follows:
[0052] in This represents the image captured at time t for the i-th field; one frame is captured every 7–10 days; for each image at any time, the mean NDVI of the field region Ti is extracted.
[0053] Constructing the complete time series:
[0054] 2) Modeling of pre-disaster growth slope and post-disaster slope: Let the time period before the disaster be t1. tkt_1 \sim t_kt1 tk, after the disaster it becomes tk+1 tmt_{k+1} \sim t_mtk+1 tm, respectively fitted with first-order linear regression models:
[0055]
[0056] in Pre-disaster growth trend slope; Post-disaster recovery trend slope; and These are the intercepts of the trend lines before and after the disaster, respectively. Fitting residuals; The disaster disturbance index (difference in disturbance slope) can be defined as:
[0057] Used to reflect changes in the direction of production trends.
[0058] 3) Trend classification rule setting: Combining the field disaster level and trend disturbance slope in step S3 Establish the following classifications for directional output changes:
[0059] To accurately characterize the ecological recovery capacity, crop growth resilience, and potential production recovery trend of farmland after disasters, a recovery index (RI) system is introduced. Based on the recovery trajectory of image features, a quantifiable, gradeable, and comparable recovery scoring model is constructed by comparing the changes between the pre-disaster baseline and key post-disaster time points. Basic Recovery Index (Single-Dimensional NDVI) The basic definition follows the form previously proposed:
[0060] in, The NDVI value is the value at the point where the disaster occurred (or the first frame after the disaster). The NDVI value of the latest frame (or the last moment of post-disaster tracking); To avoid dividing by zero for tiny positive numbers; If RI(i)>0, it indicates a certain degree of recovery; if RI(i)≈0, it indicates stagnation or complete destruction; if RI(i)<0, it indicates continuous deterioration.
[0061] In the embodiments of this application, step S5 specifically includes: S51, establish a task lifecycle management mechanism to record the entire process status log from image preprocessing to result archiving; S52, perform anomaly detection on the disaster change mask map, the disaster level classification results and the production change trend assessment results; S53 outputs damage verification reports and production recovery potential analysis results through a visualization platform, and triggers the review process for abnormal data.
[0062] In the specific implementation process, it includes: 1) Full lifecycle recording and status management of tasks: Each disaster assessment task is treated as a complete work unit, automatically assigned a unique task number, and managed throughout its entire process from "scheduling initialization" to "output completion." Task status includes, but is not limited to: pending execution, image preprocessing in progress, differential calculation in progress, disaster level extraction in progress, output trend modeling in progress, results archiving in progress, and task completion. The execution logs for each stage are recorded in a structured manner, including timestamps, execution nodes, data summaries, output result paths, etc., for later backtracking and reproduction; 2) Module-level processing node health monitoring: In the core algorithms such as image difference processing, NDVI extraction, disaster segmentation, and recovery index modeling, an embedded operation status monitoring mechanism is implemented, including but not limited to: Image quality check (whether the image is too dark, out of focus, or skewed); Data frame missing detection (whether the historical time sequence is continuous); Model output outlier check (e.g., a significant abnormal increase in NDVI after a disaster). Algorithm runtime threshold check (runtime timeout alarm), etc.; Once an anomaly is detected, the task will be automatically paused or switched to a backup processing node, and an alarm notification will be sent to the management console. 3) Abnormal data and result perception mechanism: In the process of production trend assessment and recovery index calculation, it supports the automatic identification of abnormal fluctuations or logical conflicts in the results, such as: The disaster level is "severe" but the recovery index is abnormally high, indicating "high recovery". The time series trend of the image shows abrupt changes or discontinuities in the data; NDVI fluctuations exceeding the theoretical upper limit for crop growth, etc. The above-mentioned anomalies will trigger the result marking mechanism, and the relevant tasks will be pushed to the review channel, where they can be visually reviewed and annotated on the platform interface.
[0063] Based on the same inventive concept, this application also provides an agricultural post-disaster verification and yield recovery assessment system based on UAV images, corresponding to the above method; The system includes: The image acquisition and preprocessing module acquires the initial UAV image data corresponding to the pre-disaster and post-disaster time points of the target, and performs standardization preprocessing and image registration operations on the initial UAV image data to obtain the target UAV image data. The disaster change detection module performs image difference and threshold segmentation on the target UAV image data to construct a disaster change mask map, and extracts actual disaster information based on the disaster change mask map; The disaster severity assessment module extracts key characteristic indicators of the target farmland area before and after the disaster, constructs a damage assessment feature table based on the key characteristic indicators, and classifies the target farmland area into disaster severity levels based on the damage assessment feature table to obtain the disaster severity level classification results. The yield trend modeling module constructs a yield change trend model based on the changes in historical image sequences of fields, and assesses the impact of the current disaster on the yield based on the yield change trend model to obtain the yield change trend assessment result. The comprehensive analysis and decision support module performs damage verification and crop yield recovery potential analysis on the obtained actual disaster information, disaster level classification results, and yield change trend assessment results, and manages the target farmland area accordingly based on the verification and analysis results.
[0064] Based on the same inventive concept, this application also provides an electronic device. The electronic device of this application includes at least one processor and at least one memory electrically connected to the processor. The memory is electrically connected to the processor, and the memory stores instructions executable by the at least one processor. These instructions are executed by the at least one processor to enable the at least one processor to perform the agricultural post-disaster verification and yield recovery assessment method based on UAV imagery as described above.
[0065] It should be noted that the electrical connections between the various units mentioned above do not necessarily represent the connections between lines. Any indirect connection method can be applied to the embodiments of this application as long as it achieves the purpose of this application.
[0066] Based on the same inventive concept, this application also provides a computer storage medium storing a computer program, which, when executed by a processor, implements the above-described method for agricultural post-disaster verification and yield recovery assessment based on UAV images.
[0067] Based on the same inventive concept, this application also provides a computer program product, which is stored in at least one storage medium; the computer program product includes several instructions to cause at least one computer device to execute the above-described method for agricultural post-disaster verification and yield recovery assessment based on UAV images.
[0068] Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.
Claims
1. A method for post-disaster agricultural verification and yield recovery assessment based on UAV imagery, characterized in that, The method includes: Acquire initial UAV image data corresponding to the pre-disaster and post-disaster time points of the target time, and perform standardization preprocessing and image registration operations on the initial UAV image data to obtain target UAV image data; Image differencing and threshold segmentation are performed on the target UAV image data to construct a disaster change mask map, and actual disaster information is extracted based on the disaster change mask map; Key characteristic indicators of the target farmland area before and after the disaster are extracted. A damage assessment characteristic table is constructed based on the key characteristic indicators. The target farmland area is classified into disaster levels based on the damage assessment characteristic table to obtain the disaster level classification results. A yield change trend model is constructed based on the changes in historical image sequences of fields, and the impact of the current disaster on the yield is assessed based on the yield change trend model to obtain the yield change trend assessment result. The obtained actual disaster information, disaster level classification results, and yield change trend assessment results are used to conduct damage verification and crop yield recovery potential analysis, and the target farmland area is managed accordingly based on the verification and analysis results.
2. The method according to claim 1, characterized in that, Acquire initial UAV image data corresponding to the pre-disaster and post-disaster time points at the target time, and perform standardization preprocessing and image registration operations on the initial UAV image data to obtain target UAV image data, specifically including: The drone, equipped with a multispectral camera, takes low-altitude photos of the target farmland area at the target time point to obtain initial drone image data containing the three primary color channels and the near-infrared band. The initial UAV image data is subjected to grayscale normalization and color correction to obtain intermediate UAV image data; A multi-scale matching algorithm based on local feature operators is used to spatially align the intermediate UAV image data at the pixel level through an affine transformation matrix to obtain the target UAV image data.
3. The method according to claim 1, characterized in that, Image differencing and threshold segmentation are performed on the target UAV image data to construct a disaster change mask map, and actual disaster information is extracted based on the disaster change mask map, specifically including: A standardized difference map is determined based on the target UAV image data; The standardized difference map is binarized and segmented using the maximum inter-class variance criterion automatic thresholding method to generate the disaster change mask map; Based on the proportion of affected pixels and ground resolution in the disaster change mask image, calculate the actual affected area ratio and spatial distribution range. The actual disaster information includes: the proportion of the actual disaster-affected area and the spatial distribution range.
4. The method according to claim 1, characterized in that, Key characteristic indicators of the target farmland area before and after the disaster are extracted. A damage assessment feature table is constructed based on these key characteristic indicators. The target farmland area is then classified into disaster severity levels based on this damage assessment feature table, resulting in a disaster severity level classification. Specifically, this includes: Based on the disaster change mask map, the target farmland area is divided into independent field areas, and the key feature indicators of each independent field area before and after the disaster are extracted. Based on the damage assessment feature table and the set feature change range and threshold or training model results, the target farmland area is divided into mild, moderate and severe levels.
5. The method according to claim 1, characterized in that, A yield change trend model is constructed based on the changes in historical image sequences of fields, and the impact of the current disaster on the yield is assessed based on the yield change trend model to obtain the yield change trend assessment results, specifically including: Historical image sequences of each independent field area from the sowing period to the post-disaster recovery period were collected, and the mean vegetation index at key time points was extracted to construct a time series. Based on the time series, the yield change trend model is constructed by first-order linear regression to fit the pre-disaster growth trend slope apre and the post-disaster recovery trend slope apost respectively, and the perturbation slope difference Δa=apost-apre is calculated. Based on the disaster severity classification results, the production trend is determined according to the Δa value, and the production change trend assessment result is obtained.
6. The method according to claim 1, characterized in that, The obtained actual disaster information, disaster severity classification results, and yield change trend assessment results are used to conduct damage verification and crop yield recovery potential analysis. Based on the verification and analysis results, corresponding management is implemented for the target farmland area, specifically including: Establish a task lifecycle management mechanism to record the entire process status log from image preprocessing to result archiving; Anomaly detection is performed on the disaster change mask map, the disaster level classification results, and the production change trend assessment results; The platform outputs damage verification reports and production recovery potential analysis results through visualization, and triggers the review process for abnormal data.
7. A system for post-disaster agricultural verification and yield recovery assessment based on UAV imagery, characterized in that, The system includes: The image acquisition and preprocessing module acquires the initial UAV image data corresponding to the pre-disaster and post-disaster time points of the target, and performs standardization preprocessing and image registration operations on the initial UAV image data to obtain the target UAV image data. The disaster change detection module performs image difference and threshold segmentation on the target UAV image data to construct a disaster change mask map, and extracts actual disaster information based on the disaster change mask map; The disaster severity assessment module extracts key characteristic indicators of the target farmland area before and after the disaster, constructs a damage assessment feature table based on the key characteristic indicators, and classifies the target farmland area into disaster severity levels based on the damage assessment feature table to obtain the disaster severity level classification results. The yield trend modeling module constructs a yield change trend model based on the changes in historical image sequences of fields, and assesses the impact of the current disaster on the yield based on the yield change trend model to obtain the yield change trend assessment result. The comprehensive analysis and decision support module performs damage verification and crop yield recovery potential analysis on the obtained actual disaster information, disaster level classification results, and yield change trend assessment results, and manages the target farmland area accordingly based on the verification and analysis results.
8. An electronic device, characterized in that, Includes at least one processor and at least one memory electrically connected; The memory is electrically connected to the processor, wherein the memory stores instructions executable by at least one of the processors, the instructions being executed by at least one of the processors to enable at least one of the processors to perform the method for agricultural post-disaster verification and yield recovery assessment based on UAV imagery as described in any one of claims 1-6.
9. A computer storage medium, characterized in that, The computer-readable storage medium stores a computer program. When the computer program is executed by the processor, it implements the method for agricultural post-disaster verification and yield recovery assessment based on UAV images as described in any one of claims 1-6.
10. A computer program product, characterized in that, The computer program product is stored in at least one storage medium; The computer program product includes several instructions to cause at least one electronic device to execute the method for agricultural post-disaster verification and yield recovery assessment based on UAV imagery as described in any one of claims 1-6.