Method and system for detecting rice lodging based on unmanned aerial vehicle image recognition

Abstract

The application discloses a method and system for detecting rice lodging based on unmanned aerial vehicle image recognition, and particularly relates to the field of agricultural technology; by setting a high-resolution camera on the unmanned aerial vehicle and combining a convolutional neural network model to analyze captured rice field image data, comprehensive and accurate monitoring of the rice field can be realized, the growth state and lodging condition of the rice can be identified in a timely manner, the agricultural machinery operation route and field management time are adjusted according to the detection report, the lodging area is preferentially processed by the agricultural machinery, the efficiency and accuracy of field management are improved, the overall growth quality of the rice is improved, yield loss is reduced, resource allocation is optimized, and the intelligence and efficiency of agricultural production are significantly improved.

Description

Technical Field

[0001] This invention relates to the field of agricultural technology, specifically to a method and system for detecting rice lodging based on image recognition from unmanned aerial vehicles (UAVs). Background Technology

[0002] UAV-based image recognition for rice lodging detection refers to using UAVs equipped with high-resolution cameras to take aerial photographs of rice paddies, acquiring detailed image data. Image recognition algorithms are then used to automatically analyze these images, identifying and assessing the area and extent of rice lodging. This provides accurate and timely data support for agricultural production, improving management efficiency and production outcomes. However, existing rice lodging detection technologies suffer from limited monitoring range and insufficient data accuracy, resulting in high detection costs and low efficiency. This makes it difficult to quickly and accurately detect rice lodging in large-scale planting areas, thus impacting agricultural production efficiency and profitability. Summary of the Invention

[0003] The purpose of this invention is to provide a method and system for detecting rice lodging based on UAV image recognition, so as to overcome the shortcomings of the prior art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a method for detecting rice lodging based on UAV image recognition, comprising the following steps:

[0005] S1: A high-resolution camera is set on the drone to capture the number of rice paddies in the area in real time and divide them into several monitoring areas. The drone also acquires image data of each monitoring area from different angles and at different times in real time according to the preset drone aerial photography path, so as to ensure comprehensive monitoring of the rice paddies.

[0006] S2: After preprocessing the image data of rice fields captured by the high-resolution camera of the drone, the color and morphological features related to rice in the image data are extracted through a convolutional neural network model;

[0007] S3: Analyze the parameter changes of the color and morphological characteristics of rice in the image data, and evaluate the rice growth status in each monitoring area based on the analysis results;

[0008] S4: Based on the assessment results of rice growth status in each monitoring area, the rice growth status is divided into normal growth status and lodging status, and the rice field monitoring area corresponding to the lodging status is divided into lodging monitoring area.

[0009] S5: Summarize and statistically analyze the lodged areas and generate corresponding test reports. Adjust the agricultural machinery operation routes and field management time according to the test reports to improve the testing efficiency of rice.

[0010] In a preferred embodiment, color and morphological features related to rice are extracted from image data using a convolutional neural network model, specifically as follows:

[0011] Images of rice paddies captured by a drone's high-resolution camera;

[0012] Use Gaussian filtering to remove noise and normalize the image;

[0013] Rice regions were extracted by segmenting using a color threshold.

[0014] Multi-layer convolutional operations extract color and morphological features. Each convolutional layer scans the image through a convolutional kernel to extract local features. The specific calculation expression is as follows: In the formula, fL(k)(x,y) is the feature map output by the k-th convolutional kernel in the L-th layer, and W ij (k) These are the convolution kernel weights, f(L-1) is the input feature map of the previous layer, and b (k) σ is the bias term, and σ is the activation function;

[0015] Max pooling is used to reduce the dimensionality of the feature map;

[0016] Expand the pooled feature map and extract global features through a fully connected layer;

[0017] The Softmax function is used to output the probability of the classification result, predicting the color and morphological features related to rice in the image.

[0018] In a preferred embodiment, the parameter variations of the color and morphological features of rice in the image data are analyzed, and a saturation deviation index of rice color is generated based on the color feature variations of rice in the image data. The method for obtaining the saturation deviation index is as follows:

[0019] The color space of the original image is converted from RGB to HSV to extract the saturation channel. The number of pixels for each saturation value in the saturation channel S is counted to obtain a histogram. The cumulative distribution function of the saturation channel histogram is calculated and then normalized. The specific calculation expression is as follows: In the formula, S eq (x, y) is the saturation value after equalization, M and N are the number of rows and columns of the image, respectively, MN is the total number of pixels, and L is the maximum level of saturation value; CDF(S(x, y)) represents the total number of pixels with a saturation value of S(x, y);

[0020] The calculated equalized saturation values ​​are analyzed, and the saturation deviation index is calculated. The specific calculation expression is as follows: In the formula, SD eq MeanS is the saturation deviation index. eq (x, y) represents the mean of the saturation values.

[0021] In a preferred embodiment, the changes in the parameters of the color and morphological features of rice in the image data are analyzed, and the abnormal morphological increase ratio of rice is generated based on the changes in the morphological feature parameters of rice in the image data. The method for obtaining the abnormal morphological increase ratio is as follows:

[0022] Real-time acquisition of morphological feature data at t time points to form a time series {X} t The morphological feature values ​​are normalized, and the changing trend of the morphological features is analyzed using an autoregressive integral moving average model. t =c+(φ1X) t-1 +Φ2X t-2 +...+Φ p X t-p +θ1ec t-1 +θ2ect-2+...+θpect-p, where Xt is the morphological feature value at time t, c is a constant term, Φ is an autoregressive coefficient, ec is a white noise term, and θ is a moving average coefficient;

[0023] An autoregressive integral moving average model is used to predict morphological characteristic values ​​over a future period, and a corresponding dataset {EX} is established. t+1 EX t+2 ..., EX t+n The real-time acquired rice morphological feature data is compared with the predicted values ​​of rice morphological feature data, and the residual is calculated. The residual of the rice morphological feature data is compared with the pre-set residual reference threshold. The time point when the residual of the rice morphological feature data is greater than or equal to the residual reference threshold is taken as the abnormal time point. The ratio between the number of abnormal morphologies and the total number of morphologies at each abnormal time point is calculated, that is, the abnormal morphology ratio at the abnormal time point. The rate of change of the abnormal morphology ratio between two adjacent time points is calculated, that is, the abnormal morphology increase ratio is calculated.

[0024] In a preferred embodiment, the saturation deviation index and the abnormal morphology increase ratio are normalized, and the evaluation coefficient of rice growth status in each monitoring area is calculated using the normalized saturation deviation index and the abnormal morphology increase ratio.

[0025] In a preferred embodiment, the evaluation coefficient of rice growth status in each monitoring area is compared with a standard threshold. If the evaluation coefficient of rice growth status in the monitoring area is greater than or equal to the standard threshold, the quality of rice growth status in the monitoring area is good, and the rice growth status is classified as normal growth status, and the corresponding monitoring area is classified as normal monitoring area. If the evaluation coefficient of rice growth status in the monitoring area is less than the standard threshold, the quality of rice growth status in the monitoring area is poor, and the rice growth status is classified as lodging status, and the corresponding monitoring area is classified as lodging monitoring area.

[0026] In a preferred embodiment, in S5, the lodged areas are summarized and counted, and the total number of all lodged monitoring areas is marked as D. e ; Generate a detection report recording the location Li and lodging condition Ci of each lodging monitoring area, report = {Li, Ci | D} e , e = 1, 2, ..., n}; n is a positive integer greater than 0. Adjust the operation route of the agricultural machinery according to the detection report, and prioritize the lodging monitoring area. New route = optimization function (original route, report), where the optimization function adjusts the original operation route according to the location information of the lodging monitoring area.

[0027] Increase attention to and maintenance of lodging monitoring areas, and adjust field management time. The adjusted management time formula is: Tnew=Told+ΔT×De; where Tnew is the new field management time, Told is the original field management time, ΔT is the added management time factor, and De is the indicator variable for whether it is a lodging monitoring area.

[0028] The present invention also provides a system for detecting rice lodging based on UAV image recognition, including an image data acquisition module, a feature extraction module, a growth status assessment module, a monitoring area division module, and a management module;

[0029] Image data acquisition module: A high-resolution camera is set on the drone to capture the number of rice paddies in the area in real time and divide them into several monitoring areas. The camera also acquires image data of each monitoring area from different angles and at different times in real time according to the preset drone aerial photography path, ensuring comprehensive monitoring of the rice paddies.

[0030] Feature extraction module: After preprocessing the image data of rice fields captured by the high-resolution camera of the drone, the color and morphological features related to rice in the image data are extracted through a convolutional neural network model;

[0031] Growth status assessment module: Analyzes the parameter changes of color and morphological characteristics of rice in image data, and assesses the growth status of rice in each monitoring area based on the analysis results;

[0032] Monitoring area division module: Based on the assessment results of rice growth status in each monitoring area, the rice growth status is divided into normal growth status and lodging status, and the monitoring area of ​​rice field corresponding to the lodging status is divided into lodging monitoring area.

[0033] Management module: It summarizes and statistically analyzes lodging areas and generates corresponding test reports. Based on the test reports, it adjusts the operation routes of agricultural machinery and the time of field management to improve the testing efficiency of rice.

[0034] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0035] 1. This invention utilizes a drone equipped with a high-resolution camera to conduct real-time aerial photography, acquiring detailed image data of rice paddies. A convolutional neural network model is then used to extract the color and morphological characteristics of the rice, analyze their changes, and assess the growth status of rice in each monitored area. Based on the assessment results, the rice growth status is divided into normal growth and lodging states. Lodging areas are then summarized and statistically analyzed to generate a detailed monitoring report.

[0036] 2. This invention, through detection reports, adjusts the operating routes of agricultural machinery to prioritize lodging monitoring areas and optimizes field management time, increasing attention and maintenance of lodging areas. This improves the accuracy and efficiency of rice lodging detection, reduces detection costs, ensures comprehensive and accurate monitoring of rice in large-scale planting areas, and enhances agricultural production management efficiency and yield. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0038] Figure 1 This is a flowchart of the method of the present invention.

[0039] Figure 2 This is a system module diagram of the present invention. Detailed Implementation

[0040] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, 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.

[0041] Example 1

[0042] Please see Figure 1 As shown in the figure, the method for detecting rice lodging based on UAV image recognition in this embodiment includes the following steps:

[0043] S1: A high-resolution camera is set on the drone to capture the number of rice paddies in the area in real time and divide them into several monitoring areas. The drone also acquires image data of each monitoring area from different angles and at different times in real time according to the preset drone aerial photography path, so as to ensure comprehensive monitoring of the rice paddies.

[0044] S2: After preprocessing the image data of rice fields captured by the high-resolution camera of the drone, the color and morphological features related to rice in the image data are extracted through a convolutional neural network model;

[0045] S3: Analyze the parameter changes of the color and morphological characteristics of rice in the image data, and evaluate the rice growth status in each monitoring area based on the analysis results;

[0046] S4: Based on the assessment results of rice growth status in each monitoring area, the rice growth status is divided into normal growth status and lodging status, and the rice field monitoring area corresponding to the lodging status is divided into lodging monitoring area.

[0047] S5: Summarize and statistically analyze the lodged areas and generate corresponding test reports. Adjust the agricultural machinery operation routes and field management time according to the test reports to improve the testing efficiency of rice.

[0048] In S1, a high-resolution camera is mounted on a drone to capture the number of rice paddies in the area in real time and divide them into several monitoring zones. Based on a preset drone aerial photography path, image data of each monitoring zone is acquired in real time from different angles and at different times, ensuring comprehensive monitoring of the rice paddies. Specifically:

[0049] High-resolution cameras mounted on drones can capture detailed features of rice paddies, such as plant height, color, and shape. These details provide a precise data foundation for subsequent image recognition and lodging detection.

[0050] To achieve efficient monitoring, drones will divide the rice paddies into several monitoring zones. This ensures that each zone receives comprehensive and detailed monitoring, avoiding any omissions. Each monitoring zone can be divided according to the actual conditions of the field, ensuring comprehensive coverage without duplication.

[0051] The drone's aerial photography path is pre-set, ensuring that the drone flies along the optimal route to cover all monitored areas. The pre-set aerial photography path takes into account the best flight altitude, speed, and angle to acquire high-quality image data.

[0052] As the drone flies along a pre-set aerial photography path, it captures image data of each monitored area in real time. High-resolution cameras can capture images of each area from different angles and at different times, ensuring that image information of the rice paddies can be obtained under different lighting conditions and growth stages.

[0053] By acquiring image data from different angles and time periods in real time, drones can comprehensively monitor rice paddies. They can not only capture the growth status of the rice but also promptly detect problems such as lodging, pests, and diseases.

[0054] The acquired image data is transmitted to a ground control center or cloud server for processing and analysis. Image recognition algorithms can automatically identify and assess the growth status and lodging conditions of rice. The analysis results can generate detailed monitoring reports, which are provided to farmers and agricultural managers to help them make accurate agricultural management decisions.

[0055] In this invention, by setting up a high-resolution camera on a drone and capturing image data of the monitoring area in real time from different angles and at different times according to a preset aerial photography path, comprehensive monitoring of rice paddies is achieved. This enables efficient and accurate acquisition of detailed information about rice paddies, providing strong data support for agricultural production and improving management efficiency and production results.

[0056] S2: After preprocessing the image data of rice fields captured by the high-resolution camera of the drone, the color and morphological features related to rice in the image data are extracted by a convolutional neural network model.

[0057] The image data of rice fields captured by the high-resolution camera of the drone are preprocessed by using Gaussian filtering or median filtering to remove noise from the images.

[0058] Gaussian filtering: Smooths an image by using Gaussian blur to reduce noise interference. Its principle is to convolve a Gaussian kernel with the image to suppress high-frequency noise.

[0059] Median filtering: Replacing the current pixel value with the median of neighboring pixels can effectively remove salt-and-pepper noise.

[0060] Image pixel values ​​are normalized to the range of [0,1] to eliminate the influence of differences in brightness between different images.

[0061] The image is segmented using color thresholding or region growing algorithms to extract the rice-growing regions of interest.

[0062] Color thresholding: Based on color features, a threshold is set to separate the target region from the background.

[0063] Region growing: Starting from the seed point, adjacent pixels are gradually assigned to the same segmentation region based on similarity criteria.

[0064] The color and morphological features related to rice in image data are extracted using a convolutional neural network model, specifically:

[0065] Multiple convolutional layers are used to extract low-level features (such as edges, textures, etc.) and high-level features (such as color distribution, morphological features, etc.) of an image.

[0066] Convolution operation: Each convolutional layer scans the image using a convolution kernel to extract local features. The specific calculation expression is as follows: In the formula, fL(k)(x,y) is the feature map output by the k-th convolutional kernel in the L-th layer, and W ij (k) These are the convolution kernel weights, f(L-1) is the input feature map of the previous layer, and b (k) σ is the bias term, and σ is the activation function;

[0067] Pooling layers are used to reduce the dimensionality of feature maps, thereby reducing computational complexity while preserving important features;

[0068] The maximum value within the pooling window is taken as the output of that window, thus preserving the significance of the features;

[0069] The pooled feature map is unfolded into a one-dimensional vector, and global features are extracted through a fully connected layer.

[0070] Each node is connected to all nodes in the previous layer, and features are extracted comprehensively.

[0071] The output layer can be used to classify (e.g., identify different degrees of lodging) or regress (e.g., predict the lodging area).

[0072] The Softmax function is used to output the probability of the classification result, predicting the color and morphological features related to rice in the image data.

[0073] In this invention, through preprocessing steps and a convolutional neural network model, color and morphological features can be effectively extracted from rice paddy images captured by a high-resolution UAV camera. These features provide a reliable data foundation for subsequent rice lodging detection and analysis.

[0074] S3: Analyze the parameter changes of the color and morphological characteristics of rice in the image data, and evaluate the rice growth status in each monitoring area based on the analysis results.

[0075] The changes in the parameters of color and morphological features of rice in image data are analyzed. Based on the changes in the color features of rice in the image data, a saturation deviation index of rice color is generated. The method for obtaining the saturation deviation index is as follows:

[0076] The color space of the original image is converted from RGB to HSV to extract the saturation channel. The number of pixels for each saturation value in the saturation channel S is counted to obtain a histogram. The cumulative distribution function of the saturation channel histogram is calculated and then normalized. The specific calculation expression is as follows: In the formula, S eq (x, y) is the saturation value after equalization, M and N are the number of rows and columns of the image, respectively, MN is the total number of pixels, and L is the maximum number of saturation levels (usually 256); CDF(S(x, y)) represents the total number of pixels with a saturation value of S(x, y);

[0077] The calculated equalized saturation values ​​are analyzed, and the saturation deviation index is calculated. The specific calculation expression is as follows: In the formula, SD eq MeanS is the saturation deviation index. eq (x, y) represents the mean of the saturation values.

[0078] A larger saturation deviation index indicates a greater variation in the color saturation of rice within the monitored area. This may signify uneven rice growth, with plants exhibiting significant differences in health status. Rice in some areas may be well-nourished and growing healthily, while rice in others may be affected by pests, diseases, water shortages, or other environmental factors, leading to reduced saturation or abnormal color.

[0079] A large saturation deviation index may also indicate uneven nutrition or fertilization in certain areas of the paddy field. In this case, some areas of rice may have higher color saturation, appearing dark green, while other areas may have lower saturation, appearing light green or yellow, showing differences in plant growth.

[0080] In summary, a larger saturation deviation index indicates a more uneven rice growth status in the monitored area, potentially indicating various growth problems. This situation requires further investigation and analysis to determine the specific causes and to implement corresponding agricultural management measures to improve the overall growth condition of the rice paddies.

[0081] This study analyzes the parameter variations of color and morphological features of rice in image data. Based on these variations, it generates the abnormal morphological increase ratio, which can be used to determine the morphological changes of rice within a monitoring area, particularly the growth trend of abnormal morphologies. The abnormal morphological increase ratio is generated by analyzing the morphological features of rice in image data and calculating the trends of these features over time.

[0082] The method for obtaining the abnormal morphology increment ratio is as follows:

[0083] Real-time acquisition of morphological feature data at t time points to form a time series {X} t The morphological feature values ​​are normalized, and the changing trend of the morphological features is analyzed using an autoregressive integral moving average model. t =c+(φ1X) t-1 +φ2X t-2 +...+φ p X t-p +θ1ec t-1 +θ2ect-2+...+θpect-p, where Xt is the morphological feature value at time t, c is a constant term, φ is an autoregressive coefficient, ec is a white noise term, and θ is a moving average coefficient;

[0084] An autoregressive integral moving average model is used to predict morphological characteristic values ​​over a future period, and a corresponding dataset {EX} is established. t+1 EX t+2 ..., EX t+n The real-time acquired rice morphological feature data is compared with the predicted values ​​of rice morphological feature data, and the residual is calculated. The residual of the rice morphological feature data is compared with the pre-set residual reference threshold. The time point when the residual of the rice morphological feature data is greater than or equal to the residual reference threshold is taken as the abnormal time point. The ratio between the number of abnormal morphologies and the total number of morphologies at each abnormal time point is calculated, that is, the abnormal morphology ratio at the abnormal time point. The rate of change of the abnormal morphology ratio between two adjacent time points is calculated, that is, the abnormal morphology increase ratio is calculated.

[0085] The increasing proportion of abnormal morphologies reflects the changes in rice morphological characteristics over a period of time. By calculating the increasing proportion of abnormal morphologies within a monitoring area, the growth status and health level of rice in that area can be assessed.

[0086] A high proportion of abnormal morphological changes indicates significant changes in the morphological characteristics of rice plants within the monitoring area, with an increase in abnormalities. This usually suggests poor rice growth, potentially influenced by multiple factors such as pests and diseases, nutrient deficiencies, over-irrigation, or other environmental stresses. In such cases, agricultural management measures should be implemented as soon as possible, such as fertilization, pest and disease control, or adjustments to irrigation plans, to improve the rice's growing environment and health.

[0087] When the proportion of abnormal morphological changes is low or remains stable, it indicates that the morphological characteristics of rice plants in the monitored area show little change and few abnormalities. This suggests that the rice is growing well, the plants are healthy and stable, and the growing environment is suitable. At this time, current agricultural management measures can be maintained, and the rice growth status should be monitored regularly to ensure its healthy growth.

[0088] The saturation deviation index and the increase ratio of abnormal morphology were normalized, and the evaluation coefficient of rice growth status in each monitoring area was calculated using the normalized saturation deviation index and the increase ratio of abnormal morphology.

[0089] For example, the present invention can use the following formula to calculate the evaluation coefficient of rice growth status in each monitoring area, the calculation expression being: In the formula, AS is the evaluation coefficient of rice growth status in each monitoring area, and SD... eq Rk is the saturation deviation index. e denoted as the abnormal morphology increasing ratio, a1 and a2 are the saturation deviation index and the proportional coefficient of the abnormal morphology increasing ratio, and a2>a1>0.

[0090] S4: Based on the assessment results of rice growth status in each monitoring area, the rice growth status is divided into normal growth status and lodging status, and the rice field monitoring area corresponding to the lodging status is divided into lodging monitoring area.

[0091] The evaluation coefficients of rice growth status in each monitoring area are compared with standard thresholds. If the evaluation coefficient of rice growth status in a monitoring area is greater than or equal to the standard threshold, it indicates that the rice growth status in the monitoring area is of good quality, and the rice growth status is classified as normal growth status, and the corresponding monitoring area is classified as normal monitoring area. If the evaluation coefficient of rice growth status in a monitoring area is less than the standard threshold, it indicates that the rice growth status in the monitoring area is of poor quality, and the rice growth status is classified as lodging status, and the corresponding monitoring area is classified as lodging monitoring area.

[0092] S5: Summarize and statistically analyze the lodged areas and generate corresponding test reports. Adjust the agricultural machinery operation routes and field management time according to the test reports to improve the testing efficiency of rice.

[0093] The lodging areas were summarized and statistically analyzed, and the total number of all lodging monitoring areas was marked as D. e ; Generate a detection report recording the location Li and lodging condition Ci of each lodging monitoring area, report = {Li, Ci | D} e , e = 1, 2, ..., n}; n is a positive integer greater than 0. Adjust the operation route of the agricultural machinery according to the detection report, and give priority to the lodging monitoring area. New route = optimization function (original route, report), where the optimization function adjusts the original operation route according to the location information of the lodging monitoring area, so that the agricultural machinery gives priority to these areas.

[0094] Increase attention to and maintenance of lodging monitoring areas, and adjust field management time. The adjusted management time formula is: Tnew=Told+ΔT×De; where Tnew is the new field management time, Told is the original field management time, ΔT is the added management time factor, and De is the indicator variable for whether it is a lodging monitoring area.

[0095] In this embodiment, a high-resolution camera is mounted on a drone to capture the number of rice paddies in the area in real time and divide them into several monitoring areas. The drone acquires image data of each monitoring area from different angles and at different times according to a preset aerial photography path, ensuring comprehensive monitoring of the rice paddies. After preprocessing the image data of the rice paddies captured by the drone's high-resolution camera, a convolutional neural network model is used to extract color and morphological features related to rice from the image data. The parameter changes of the color and morphological features of rice in the image data are analyzed, and the growth status of rice in each monitoring area is evaluated based on the analysis results. Based on the evaluation results of the rice growth status in each monitoring area, the rice growth status is divided into normal growth status and lodging status, and the monitoring areas of rice paddies corresponding to the lodging status are designated as lodging monitoring areas. The lodging areas are summarized and statistically analyzed, and corresponding detection reports are generated. Based on the detection reports, the agricultural machinery operation routes and field management times are adjusted to improve the efficiency of rice detection.

[0096] Example 2

[0097] Please see Figure 2 As shown in the figure, the system for detecting rice lodging based on UAV image recognition in this embodiment includes an image data acquisition module, a feature extraction module, a growth status assessment module, a monitoring area division module, and a management module.

[0098] Image data acquisition module: A high-resolution camera is set on the drone to capture the number of rice paddies in the area in real time and divide them into several monitoring areas. The camera also acquires image data of each monitoring area from different angles and at different times in real time according to the preset drone aerial photography path, ensuring comprehensive monitoring of the rice paddies.

[0099] Feature extraction module: After preprocessing the image data of rice fields captured by the high-resolution camera of the drone, the color and morphological features related to rice in the image data are extracted through a convolutional neural network model;

[0100] Growth status assessment module: Analyzes the parameter changes of color and morphological characteristics of rice in image data, and assesses the growth status of rice in each monitoring area based on the analysis results;

[0101] Monitoring area division module: Based on the assessment results of rice growth status in each monitoring area, the rice growth status is divided into normal growth status and lodging status, and the monitoring area of ​​rice field corresponding to the lodging status is divided into lodging monitoring area.

[0102] Management module: It summarizes and statistically analyzes lodging areas and generates corresponding test reports. Based on the test reports, it adjusts the operation routes of agricultural machinery and the time of field management to improve the testing efficiency of rice.

[0103] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0104] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0105] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0106] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0107] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0108] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for detecting rice lodging based on UAV image recognition, characterized in that: Includes the following steps; S1: A high-resolution camera is set on the drone to capture the number of rice paddies in the area in real time and divide them into several monitoring areas. The drone also acquires image data of each monitoring area from different angles and at different times in real time according to the preset drone aerial photography path, so as to ensure comprehensive monitoring of the rice paddies. S2: After preprocessing the image data of rice fields captured by the high-resolution camera of the drone, the color and morphological features related to rice in the image data are extracted through a convolutional neural network model; S3: Analyze the parameter changes of the color and morphological characteristics of rice in the image data, and evaluate the rice growth status in each monitoring area based on the analysis results; S4: Based on the assessment results of rice growth status in each monitoring area, the rice growth status is divided into normal growth status and lodging status, and the rice field monitoring area corresponding to the lodging status is divided into lodging monitoring area. S5: Summarize and statistically analyze the lodged areas and generate corresponding test reports. Adjust the agricultural machinery operation routes and field management time based on the test reports to improve the testing efficiency of rice. The color and morphological features related to rice in image data are extracted using a convolutional neural network model, specifically: Images of rice paddies captured by a drone's high-resolution camera; Use Gaussian filtering to remove noise and normalize the image; Rice regions were extracted by segmenting using a color threshold. Multi-layer convolutional operations extract color and morphological features. Each convolutional layer scans the image through a convolutional kernel to extract local features. The specific calculation expression is as follows: In the formula, It is the feature map output by the k-th convolutional kernel in the L-th layer. These are the convolution kernel weights. It is the input feature map of the previous layer. It is a bias term. It is an activation function; Max pooling is used to reduce the dimensionality of the feature map; Expand the pooled feature map and extract global features through a fully connected layer; The Softmax function is used to output the probability of the classification result, predicting the color and morphological features related to rice in the image. The changes in the parameters of color and morphological features of rice in image data are analyzed. Based on the changes in the color features of rice in the image data, a saturation deviation index of rice color is generated. The method for obtaining the saturation deviation index is as follows: The color space of the original image is converted from RGB to HSV to extract the saturation channel. The number of pixels for each saturation value in the saturation channel S is counted to obtain a histogram. The cumulative distribution function of the saturation channel histogram is calculated and then normalized. The specific calculation expression is as follows: In the formula, is the saturation value after equalization, M and N are the number of rows and columns of the image, MN is the total number of pixels, and L is the maximum level of saturation value; The saturation value represents The total number of pixels; The calculated equalized saturation values ​​are analyzed, and the saturation deviation index is calculated. The specific calculation expression is as follows: In the formula, The saturation deviation index. This represents the average saturation value.

2. The method for detecting rice lodging based on UAV image recognition according to claim 1, characterized in that: The changes in the color and morphological features of rice in image data are analyzed. Based on these changes, the percentage increase in abnormal morphological features of rice is generated. The method for obtaining this percentage increase is as follows: Real-time acquisition of morphological feature data at t time points to form a time series { The morphological feature values ​​were normalized, and an autoregressive integral moving average model was used to analyze the changing trends of the morphological features. ,in, Here, is the morphological characteristic value at time t, c is a constant term, and ϕ is the autoregressive coefficient. This is the white noise term, and θ is the moving average coefficient. We use an autoregressive integral moving average model to predict morphological characteristic values ​​over a future period and establish a corresponding dataset. The real-time acquired rice morphological feature data is compared with the predicted values ​​of rice morphological feature data, and the residual is calculated. The residual of the rice morphological feature data is compared with the pre-set residual reference threshold. The time point when the residual of the rice morphological feature data is greater than or equal to the residual reference threshold is taken as the abnormal time point. The ratio between the number of abnormal morphologies and the total number of morphologies at each abnormal time point is calculated, that is, the abnormal morphology ratio at the abnormal time point. The rate of change of the abnormal morphology ratio between two adjacent time points is calculated, that is, the abnormal morphology increase ratio is calculated.

3. The method for detecting rice lodging based on UAV image recognition according to claim 2, characterized in that: The saturation deviation index and the increase ratio of abnormal morphology were normalized, and the evaluation coefficient of rice growth status in each monitoring area was calculated using the normalized saturation deviation index and the increase ratio of abnormal morphology.

4. The method for detecting rice lodging based on UAV image recognition according to claim 3, characterized in that: The evaluation coefficients of rice growth status in each monitoring area are compared with standard thresholds. If the evaluation coefficient of rice growth status in a monitoring area is greater than or equal to the standard threshold, the quality of rice growth status in the monitoring area is good, and the rice growth status is classified as normal growth status, and the corresponding monitoring area is classified as normal monitoring area. If the evaluation coefficient of rice growth status in a monitoring area is less than the standard threshold, the quality of rice growth status in the monitoring area is poor, and the rice growth status is classified as lodging status, and the corresponding monitoring area is classified as lodging monitoring area.

5. The method for detecting rice lodging based on UAV image recognition according to claim 4, characterized in that: In S5, the lodging areas are summarized and statistically analyzed, and the total number of each lodging monitoring area is marked as... ; Generate a detection report recording the location Li and lodging condition Ci of each lodging monitoring area. = n is a positive integer greater than 0. Adjust the agricultural machinery's operating route based on the inspection report, prioritizing the treatment of lodging monitoring areas. The optimization function adjusts the original work route based on the location information of the lodging monitoring area. Increase attention and maintenance of lodging monitoring areas, and adjust field management time. The adjusted management time formula is as follows: Where Tnew is the new field management time, and Told is the original field management time. It is an added management time factor.

6. A system for detecting rice lodging based on UAV image recognition, used to implement the method for detecting rice lodging based on UAV image recognition as described in any one of claims 1-5, characterized in that: It includes an image data acquisition module, a feature extraction module, a growth status assessment module, a monitoring area division module, and a management module; Image data acquisition module: A high-resolution camera is set on the drone to capture the number of rice paddies in the area in real time and divide them into several monitoring areas. The camera also acquires image data of each monitoring area from different angles and at different times in real time according to the preset drone aerial photography path, ensuring comprehensive monitoring of the rice paddies. Feature extraction module: After preprocessing the image data of rice fields captured by the high-resolution camera of the drone, the color and morphological features related to rice in the image data are extracted through a convolutional neural network model; Growth status assessment module: Analyzes the parameter changes of color and morphological characteristics of rice in image data, and assesses the growth status of rice in each monitoring area based on the analysis results; Monitoring area division module: Based on the assessment results of rice growth status in each monitoring area, the rice growth status is divided into normal growth status and lodging status, and the monitoring area of ​​rice field corresponding to the lodging status is divided into lodging monitoring area. Management module: It summarizes and statistically analyzes lodging areas and generates corresponding test reports. Based on the test reports, it adjusts the operation routes of agricultural machinery and the time of field management to improve the testing efficiency of rice.

Images