A method and system for obtaining high-throughput of crop canopy coverage by unmanned aerial vehicle remote sensing

Abstract

The present application belongs to the field of low-altitude unmanned aerial vehicle remote sensing and precision agriculture technology, and discloses a crop canopy coverage high-throughput acquisition method and system for unmanned aerial vehicle remote sensing. The method quantitatively describes the visual feature pattern between the crop canopy and the soil background from the high-resolution RGB image; the support vector machine method is used to automatically mine the association mapping between the category semantics and the visual feature pattern, to obtain the preliminary extraction result of the crop canopy and the soil background, and to generate the single-time-phase crop canopy mask image; the dynamic weed removal method of the RGB image time sequence is used to optimize all single-time-phase crop canopy mask images of the crop growth, to respectively count the peanut canopy pixel cluster number and the soil pixel cluster number in each plot in each plot, and to complete the accurate estimation of the crop canopy coverage. The present application provides a basic technical means for the fields of agricultural precision management, crop high-throughput phenotype monitoring and digital breeding, etc.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of low-altitude unmanned aerial vehicle remote sensing and precision agriculture, and discloses a crop canopy coverage high-throughput acquisition method and system for unmanned aerial vehicle remote sensing. BACKGROUND

[0002] The research on crop phenotypes is usually realized by analyzing the parameter characteristics thereof, such as canopy coverage, leaf area index, plant height and the like. A large number of studies have proved that the crop canopy coverage has a significant correlation with the remaining phenotypic parameters. Therefore, the accurate extraction of the crop canopy coverage is very important for the analysis of crop phenotypic characteristics and the accurate identification of germplasm resources. The crop canopy coverage (Fractional Vegetation Cover, FVC) refers to the ratio of the vertical projection area of the crop canopy in each plot area to the total area of the soil, that is, the ratio of vegetation to soil. In the field of agronomy, the crop canopy coverage is an important parameter for describing the distribution of crops on the ground, and can reflect the light interception capacity of crop leaves, judge the growth and health status of crops and predict the biological yield.

[0003] At present, the estimation methods of the crop canopy coverage mainly include the traditional ground measurement method and the remote sensing monitoring method. The commonly used ground measurement method is the grid method, which has high accuracy but low efficiency and is difficult to meet the large-scale crop trait investigation. In recent years, the unmanned aerial vehicle remote sensing technology has been increasingly concerned due to its high precision, low application cost and strong timeliness, and a large number of scholars have carried out many studies on the extraction of the crop canopy coverage using the unmanned aerial vehicle remote sensing technology. For the remote sensing monitoring method, the separation of vegetation and soil is the first and most important step for the accurate extraction of the crop canopy coverage, that is, the image is divided into two categories of vegetation pixels and non-vegetation pixels, and the proportion of the vegetation pixels is directly calculated after classification. In the image with low spatial resolution such as satellite, a single pixel may contain vegetation and non-vegetation information (mixed pixel), and it is impossible to accurately divide the pixel into vegetation or non-vegetation. Therefore, the pixel bisection model is generally used to obtain the proportion of the vegetation component in the pixel by solving the proportion of the vegetation or non-vegetation component in the mixed pixel.

[0004] Different from the satellite images with low spatial resolution, the images obtained by low-altitude unmanned aerial vehicle have high spatial resolution, which can reach 1mm / pixel. Therefore, the canopy coverage can be estimated by directly counting the vegetation pixels and non-vegetation pixels. In the field of agronomy, the segmentation of real farmland scenes is challenged by the unstructured background, light changes and weak contrast of crop background in the complex background of the field, and has become a hot research topic. Therefore, agricultural researchers have conducted a lot of research on crop canopy segmentation. At present, the two main methods are threshold segmentation and machine learning method. Although the threshold-based method is simple and intuitive, it is easy to understand and implement, but this method needs to dynamically determine the segmentation threshold. In the field of machine learning, the crop canopy and the background such as soil often have different visual patterns and presentation rules. Machine learning quantitatively describes the specific visual patterns between each other, establishes a statistical model to automatically mine the correlation mapping between the category semantics and the visual patterns, and is widely used in land cover classification and target recognition in the fields of computer vision and precision agriculture. Therefore, more and more people pay attention to using machine learning method for plant-soil segmentation.

[0005] Through the above analysis, the problems and defects of the prior art are that the prior art uses threshold segmentation, machine learning or deep learning method to realize the estimation of crop coverage, but these methods do not consider the influence of weeds on crop coverage estimation (especially in the seedling stage) in the farmland scene, resulting in an overestimated result. And in the prior art, the influence of weeds on crop canopy coverage estimation is removed from the high-resolution RGB image of the unmanned aerial vehicle, and the precision of high-throughput acquisition of crop canopy coverage is low. SUMMARY

[0006] In order to overcome the problems in the related art, the present application discloses an unmanned aerial vehicle remote sensing crop canopy coverage high-throughput acquisition method and system, specifically relates to a crop canopy coverage high-throughput acquisition technology using low-altitude unmanned aerial vehicle remote sensing.

[0007] The technical solution is as follows: the unmanned aerial vehicle remote sensing crop canopy coverage high-throughput acquisition method comprises the following steps:

[0008] S1, quantitatively describing the visual feature pattern between the crop canopy and the soil background from the high-resolution RGB image;

[0009] S2, using the method of support vector machine to automatically mine the correlation mapping between the category semantics and the visual feature pattern, obtaining the preliminary extraction result of the crop canopy and the soil background, generating a single-time-phase crop canopy mask map, marking the crop canopy area as 1 and the soil background area as 0;

[0010] S3, using the dynamic weed removal method of the RGB image time sequence, optimizing the crop canopy mask diagram of all single-phase crops growing, determining the peanut canopy area and weed area, and marking and removing the weeds in the entire growth period of the crop growing;

[0011] S4, based on the optimized results, respectively counting the number of peanut canopy pixel clusters and soil pixel clusters in each plot in each plot, and completing the accurate estimation of crop canopy coverage.

[0012] In step S1, the crop is peanut, and two color indexes, super green index ExG and vegetation extraction color index CIVE, are used to separate the visual feature pattern between peanut canopy and soil background in the quantitative description of the visual feature pattern between crop canopy and soil background from high-resolution RGB images.

[0013] Further, the super green index ExG is used to calculate the green degree of the crop, and the calculation formula is:

[0014] ExG = 2g - r - b (1)

[0015] In the formula, ExG is the super green index, g is the green band of the image, r is the red band of the image, and b is the blue band of the image.

[0016] The calculation formula of r, g, and b is:

[0017]

[0018] In the formula, R, G, and B are the channel values of the red, green, and blue channels, respectively.

[0019] The vegetation extraction color index CIVE is used to separate green vegetation from soil background, and the calculation formula is:

[0020] CIVE = 0.441R - 0.811G + 0.385B + 18.78745 (3)

[0021] In the formula, CIVE is the vegetation extraction color index.

[0022] In step S2, the method of support vector machine is used to automatically mine the association mapping between category semantics and visual feature pattern, including:

[0023] The SVM classifier model is built using the scikit-learn library in the Python environment, and the SVC function in the scikit-learn library is used to train the training sample data, and the calling form of the function is:

[0024] svm_model = SVC('kernel') (4)

[0025] wherein svm_model is a defined support vector machine training model, SVC() represents a support vector machine classification called, and 'kernel' is a kernel function of the model;

[0026] A support vector machine (SVM) model is trained, using the RGB values of the pixels as feature inputs and the classification labels as output targets; the SVM model is used to replace the threshold method, and the RGB values of each pixel are input into the SVM model to output the classification label; the established SVM classifier model is used to classify the RGB images of each growth period of the peanuts obtained, to obtain the preliminary extraction results of the peanut canopy and soil background, and to obtain a mask image containing only the peanut images of the processed period.

[0027] In step S3, a dynamic weed removal method using the RGB image time series is used, including:

[0028] (1) Cumulative time phase mask generation: multiply all the mask data of the growth periods pixel by pixel, and find the intersection of the crop canopy area to obtain the cumulative time phase mask;

[0029] (2) DBSCAN clustering analysis is performed in combination with the cumulative time phase mask to determine whether each pixel cluster exists continuously in the entire growth period of the crop.

[0030] In step (1), the cumulative time phase mask is obtained, and the expression is:

[0031]

[0032] wherein M is the cumulative time phase mask, T is the sum of the time points, M t is a binary mask generated at time point t, is a pixel point (i, j) identified as a vegetation area at time point t, is a pixel point (i, j) not identified as a vegetation area at time point t, i and j are the row and column of the pixel position.

[0033] In step (2), DBSCAN clustering analysis is performed in combination with the cumulative time phase mask, including:

[0034] The connected canopy pixels are clustered using the DBSCAN clustering algorithm, and the extracted single-time-phase peanut canopy mask is divided into different pixel clusters; in combination with the cumulative time phase mask, it is determined whether each pixel cluster exists continuously in the entire growth period of the peanut; if the current pixel cluster intersects with the cumulative time phase mask, the pixel cluster is a peanut canopy area; otherwise, it is determined that the pixel cluster is a weed area, and the weeds are removed; the process is repeated to mark and remove the weeds in the entire growth period of the peanut.

[0035] In step S4, the accurate estimation of crop canopy coverage is completed, and the expression is:

[0036]

[0037] In the formula, FVC (i,j) is the canopy coverage of the plot region of the i-th row and the j-th column, M (i,j) is the number of crop canopy pixel clusters of the plot region of the i-th row and the j-th column, N (i,j) is the number of soil pixel clusters of the plot region of the i-th row and the j-th column.

[0038] Another object of the present application is to provide a UAV remote sensing crop canopy coverage high-throughput acquisition system, which implements the UAV remote sensing crop canopy coverage high-throughput acquisition method, and the system comprises:

[0039] A visual feature pattern acquisition module is configured to quantitatively describe the visual feature pattern between the crop canopy and the soil background from the high-resolution RGB image.

[0040] A single-time-phase crop canopy mask map generation module is configured to automatically mine the association mapping between the category semantics and the visual feature pattern by using the support vector machine method, obtain a preliminary extraction result of the crop canopy and the soil background, and generate a single-time-phase crop canopy mask map, in which the crop canopy region is marked as 1 and the soil background region is marked as 0.

[0041] An RGB image time sequence dynamic weed removal module is configured to use the dynamic weed removal method of the RGB image time sequence to optimize all single-time-phase crop canopy mask maps of crop growth, determine the peanut canopy region and the weed region, and mark and remove the weeds in the entire growth period of the crop.

[0042] A crop canopy coverage estimation module is configured to, based on the optimized result, respectively count the number of peanut canopy pixel clusters and the number of soil pixel clusters in each plot region in each plot, and complete the accurate estimation of the crop canopy coverage.

[0043] Further, the UAV remote sensing crop canopy coverage high-throughput acquisition system is carried on a computer readable storage medium, the computer readable storage medium stores a computer program, and the computer program can realize the functions in the UAV remote sensing crop canopy coverage high-throughput acquisition system when executed by a processor.

[0044] In combination with all the above technical solutions, the present application has the beneficial effects that the high-resolution RGB image obtained by using the UAV low-altitude remote sensing technology makes it simple and easy to estimate the crop coverage by separating the crop canopy and the soil and other non-vegetation backgrounds.

[0045] The application utilizes the high spatial resolution image acquisition capability of low-altitude unmanned aerial vehicle remote sensing with high efficiency and low cost, combines the advantages of computer vision in feature extraction, realizes information fusion in time dimension through time series images of peanut canopy mask, significantly improves the robustness of peanut field weed removal, realizes reliable separation of plant soil, and further improves the precision and quality of crop canopy coverage estimation. The crop canopy coverage high-throughput acquisition technology designed based on low-altitude unmanned aerial vehicle remote sensing improves the automation and intelligent level of crop canopy coverage estimation, and provides a basic technical means for agricultural precision management, crop high-throughput phenotype monitoring and digital breeding and other fields.

[0046] The present application can realize efficient crop canopy coverage acquisition: the present application can quickly acquire crop canopy coverage data in a high-throughput and automated manner by combining unmanned aerial vehicle remote sensing with high-resolution RGB images, which significantly improves efficiency and accuracy compared to traditional manual measurement or low-resolution image analysis. The present application can realize agricultural production optimization: accurately monitor crop growth to provide data support for precision agriculture, optimize farmland management decisions (such as fertilization, irrigation and weeding), and improve crop yield. The present application can realize weed removal cost reduction: through an automated dynamic weed removal method, the cost of manual weeding is greatly reduced, and the risk of crop yield reduction due to weed competition is reduced. The present application can realize dynamic weed removal: a dynamic weed removal method based on time series is proposed, which combines DBSCAN clustering and cumulative time phase mask analysis to achieve accurate differentiation of crops and weeds. This innovative method is significantly better than traditional single-time static analysis methods. The present application can realize multi-index combined visual feature pattern extraction: the present application uses the excess green index (ExG) and the vegetation extraction color index (CIVE) to model the characteristics of crop canopy and soil background from multiple angles based on high-resolution RGB images, improving the accuracy of identification. The present application can realize high-throughput precision monitoring: traditional canopy coverage monitoring methods rely on manual investigation or low-resolution images. The present application realizes efficient and automated acquisition of crop canopy coverage through high-resolution unmanned aerial vehicle images and machine learning algorithms. The present application can realize accurate differentiation of crop canopy and weed area: the high similarity of weeds and crops in visual features, as well as their dynamic changes in the growth cycle, make it difficult for traditional methods to accurately identify them. The present application combines time series analysis and density clustering algorithm to dynamically remove weeds for the first time. The present application can realize the low-efficiency bottleneck of rapid coverage acquisition: traditional canopy coverage monitoring is limited by low-resolution image accuracy and manual analysis efficiency, making it difficult to be applied efficiently in large-scale farmland. The present application significantly improves processing speed and applicability. The present application can realize robustness in growth period monitoring: previous technologies rely heavily on single-time data and are easily affected by local errors. The dynamic method of cumulative time phase mask generation in the present application significantly enhances the stability and accuracy of data analysis.

[0047] Traditional methods tend to analyze crop canopy coverage through single-phase image analysis, without considering the dynamics of crop growth. The time series dynamic analysis method proposed by the present application effectively overcomes this limitation. Low utilization of high-resolution images: previous studies have relied on low-resolution remote sensing images for large-area monitoring, considering that high-resolution images are difficult to apply efficiently due to large data volume and complex processing. The present application optimizes the algorithm and automates the processing flow, fully utilizing the potential of high-resolution images. Fixed thinking of weed identification: traditional methods usually distinguish weeds by fixed threshold or static feature extraction, the present application breaks through this fixed thinking by combining cluster analysis and time sequence mask map, dynamically adjusting the analysis model to adapt to the characteristics of different growth stages. BRIEF DESCRIPTION OF DRAWINGS

[0048] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and serve to explain the principles of the present disclosure, together with the description;

[0049] Figure 1 is a flow chart of the crop canopy coverage high-throughput acquisition method of unmanned aerial vehicle remote sensing provided by the embodiments of the present application;

[0050] Figure 2 is a principle diagram of the cumulative time phase mask map generation provided by the embodiments of the present application;

[0051] Figure 3 is a schematic diagram of the crop canopy coverage high-throughput acquisition system of unmanned aerial vehicle remote sensing provided by the embodiments of the present application, wherein (a) is an image map of an untreated plot, (b) is a plot image map manually labeled with weed areas, (c) is a plot image map labeled with weed areas by using the present patent, (d) is a binary plot image map without removing weeds, (e) is a binary plot image map after manually removing weeds, (f) is a binary plot image map after removing weeds by using the present patent method;

[0052] In the figure: 1, visual feature mode acquisition module; 2, single-phase crop canopy mask map generation module; 3, dynamic weed removal module of RGB image time series; 4, crop canopy coverage estimation module. DETAILED DESCRIPTION

[0053] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below in combination with the drawings. In the following description, a lot of specific details are set forth in order to fully understand the present application. However, the present application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without departing from the scope of the present application, therefore the present application is not limited by the specific embodiments disclosed below.

[0054] The crop canopy coverage high-throughput acquisition method provided by the unmanned aerial vehicle remote sensing of the embodiment of the application comprises the following steps:

[0055] Embodiment 1, as shown in the figure, the crop canopy coverage high-throughput acquisition method provided by the unmanned aerial vehicle remote sensing of the embodiment of the application comprises the following steps: Figure 1

[0056] S1, quantitatively describe the visual feature mode between the crop canopy and the soil background from the high-resolution RGB image;

[0057] S2, automatically mine the association mapping between the category semantics and the visual feature mode by using the method of the support vector machine, obtain the preliminary extraction result of the crop canopy and the soil background, generate the single-time-phase crop canopy mask image, mark the crop canopy area as 1, and mark the soil background area as 0;

[0058] S3, the dynamic weed removal method of the RGB image time sequence is used to optimize all single-time-phase crop canopy mask images of crop growth, determine the peanut canopy area and the weed area, and mark and remove the weeds in the whole growth period of the crop growth;

[0059] S4, based on the optimized result, the number of peanut canopy pixel clusters and the number of soil pixel clusters in each plot area in each plot are counted respectively, and the accurate estimation of the crop canopy coverage is completed.

[0060] Embodiment 2, the peanut is taken as the research object, the high-resolution RGB image sequence is acquired by using the unmanned aerial vehicle low-altitude remote sensing technology, the canopy area of the peanut in different growth periods is extracted, and a crop canopy coverage high-throughput acquisition method (low-altitude unmanned aerial vehicle remote sensing crop canopy coverage high-throughput acquisition method) is designed. Specifically, it comprises:

[0061] Step 1, data set source.

[0062] ​The present application uses the collected unmanned aerial vehicle image data of peanuts at different growth stages. The collection and production of peanut unmanned aerial vehicle image data are carried out in a peanut test field in a certain county, and the label file is manually annotated in labelme. A certain county belongs to a typical warm temperate semi-humid continental climate, with distinct seasons, rainy and hot seasons, and an average annual temperature of about 14 DEG C. The average altitude of a certain county is 85 meters, the average annual sunshine hours are 2759.1 hours, the average annual precipitation is 901.4 mm, the average annual frost-free period is 199 days, and it is suitable for the growth of peanuts. Data collection started on May 31, 2023, and ended on August 9, 2023, covering the emergence stage, seedling stage, flowering and needle stage, and pod stage of peanuts. The test uses DJI 3 unmanned aerial vehicle to carry visible light camera, and data collection is carried out at noon on a sunny day. Use DJI GS Pro platform for flight planning and flight display, set the lateral overlap to 80%, and the flight height to 20 m. In the DJI ZhiTu software, the flight strip splicing and radiation correction are carried out to generate digital orthophoto map, and ArcGIS 10.4 software is used to divide the plots and batch extract the peanut image at each stage.

[0063] Step 2, image preprocessing.

[0064] Affected by the complex background of the test soil, the image contains a large amount of noise interference. In order to remove image noise pollution and improve the stability of the characteristics, the present application uses Gaussian low-pass filter to enhance the gray images of the R, G and B channels of the image respectively. Through Gaussian low-pass filter, the image can be smoothed, the high-frequency noise in the soil background can be removed, and the influence of small pixel changes on the result can be reduced in the subsequent vegetation index calculation. Finally, the three channels of the processed gray images are merged to obtain the denoising result.

[0065] Step 3, visual feature pattern extraction.

[0066] Color feature is the most basic visual feature in image retrieval, and is also the most widely used basic image attribute in computer vision. This is because compared with other visual features, color feature has less dependence on the size, direction and viewing angle of the image itself, and has better robustness. Color index is calculated based on RGB image linearly or nonlinearly, and has been widely used in crop image semantic segmentation. Reasonable use of color index can enhance the visual contrast of foreground and background, reduce the influence of light conditions, and improve the accuracy of crop segmentation. In the present application, two color indexes, super green index (ExG) and vegetation extraction color index (CIVE), are used to separate the visual feature pattern of peanut canopy and soil background.

[0067] Specifically includes:

[0068] Step 301, super green index.

[0069] Excess Green index (ExG) emphasizes the green degree of crops, and many previous studies have proved that the crop contour can be clearly outlined by using this index. The calculation of the index needs to normalize the color values of the three channels first, and the specific calculation formula is as follows:

[0070] ExG = 2g-r-b (1)

[0071] In the formula, ExG is the excess green index, g is the green band of the image, r is the red band of the image, and b is the blue band of the image.

[0072] The calculation formula of r, g and b is as follows:

[0073]

[0074] In the formula, R, G and B are respectively the channel values of the red, green and blue channels.

[0075] The excess green index formula of the application is simple to calculate, only depends on the linear combination of green (G), red (R) and blue (B) channel values, and has high calculation efficiency. The distinguishing effect of the vegetation area with significant color characteristics is good, and it is especially suitable for scenes with obvious green. By emphasizing the green channel and reducing the weight of red and blue, the vegetation is significantly higher than the background in the index value. It can be directly applied to various optical image data (RGB image), without additional band or complex preprocessing. For images with low resolution and obvious color distribution, ExG can better extract vegetation, which greatly reduces the requirements for equipment.

[0076] Step 302, vegetation extraction color index.

[0077] Color Index of Vegetation Extraction (CIVE) is proposed on the basis of soybean and sugar beet field research, and it is found that CIVE has good effect on separating green vegetation from outdoor soil background. The specific calculation formula is as follows:

[0078] Color Index of Vegetation Extraction (CIVE) is used to separate green vegetation from soil background, and the calculation formula is as follows:

[0079] CIVE = 0.441R-0.811G+0.385B+18.78745 (3)

[0080] In the formula, CIVE is the color index of vegetation extraction.

[0081] In the present application, CIVE gives different weights to R, G and B channels compared to ExG, and more carefully distinguishes between vegetation and non-vegetation areas. It has better vegetation extraction effect in complex background, can improve accuracy in various scenes, and has strong background suppression ability. Under the conditions of light change and high scene complexity, CIVE is relatively more stable and less affected by light conditions.

[0082] Step 4, soil separation model method.

[0083] The method of machine learning can automatically learn the effective features in the data, improve the accuracy and robustness of classification, and the trained machine learning model usually has strong generalization ability and can achieve good classification effect on unseen data, making it more reliable in practical application. Therefore, the present application uses a classic machine learning model, support vector machine, to establish the association mapping between the class semantics and the visual feature pattern. The present application uses the scikit-learn (sklearn) library in the Python environment to build an SVM classifier model (soil separation model), and uses the SVC function in the library to train the training sample data. The calling form of the function is:

[0084] svm_model = SVC ('kernel') (4)

[0085] In the formula, svm_model is the defined support vector machine training model, SVC() represents the support vector machine classification in the support vector machine, and 'kernel' is the kernel function of the model.

[0086] First, a support vector machine (SVM) model is trained, using the RGB values of the pixels as feature input and the classification label (such as canopy and non-canopy) as output target. The SVM model replaces the threshold method, and the RGB value of each pixel point is input into the SVM model to output the classification label. Finally, the established soil separation model is used to classify the RGB images of each growth period of peanuts obtained, to obtain the preliminary extraction results of peanut canopy and soil background, and to generate a single-time-phase peanut canopy mask image, wherein the peanut canopy area is marked as 1 and the soil background area is marked as 0.

[0087] Step 5, weed removal using time series images.

[0088] Peanut fields are prone to weed growth before the ridges are closed (especially at the seedling stage). Due to the fact that the visual feature patterns of weeds and peanut canopies are very similar, conventional soil-segmentation models have difficulty in effectively distinguishing weeds from peanut canopies, especially in complex farmland scenes, and this problem is even more pronounced. To overcome this challenge, the present application designs a dynamic weed removal method using a time series of RGB images based on the fact that weeds do not exist at all growth stages of crops. Specifically, the method comprises the following steps:

[0089] Step 501, as shown in the cumulative phase mask map generation, all the mask data of the growth period is multiplied pixel by pixel, that is, the intersection of the peanut canopy area is obtained, and the cumulative phase mask map is obtained according to formula (5). Figure 2

[0090]

[0091] In the formula, M is the cumulative phase mask map Mask, T is the sum of time points, M t is a binary mask map generated at time point t, is the pixel point (i, j) identified as a vegetation area at time point t, is the pixel point (i, j) not identified as a vegetation area at time point t, i, j is the row and column of the pixel position.

[0092] The mask map can retain the area that is displayed as a peanut plant at all growth stages, thereby providing a reference basis for effectively removing the interference (such as weeds and other noise) that exists in some phases but is not persistent.

[0093] DBSCAN clustering analysis.

[0094] For a single-phase peanut canopy mask map, the present application uses the DBSCAN clustering algorithm to cluster the connected canopy pixels. DBSCAN is a density-based clustering algorithm that divides data points into core points, boundary points, and noise points. Core points represent points with enough (usually at least) points within a certain radius. Boundary points represent points with fewer points within a radius than core points require, but at least belong to one cluster. Noise points are points that are neither core points nor boundary points. It identifies the attributes of points by the distance between points, and then divides the extracted single-phase peanut canopy mask map into different pixel clusters. Then, in combination with the cumulative phase mask map, it is determined whether each pixel cluster persists throughout the entire growth period of the peanut. If the current pixel cluster intersects with the cumulative phase mask map, the pixel cluster is a peanut canopy area; otherwise, it is determined that the pixel cluster is a weed area, thereby achieving weed removal. Repeat the process to mark and remove weeds throughout the entire growth period of the peanut.

[0095] ​Step 6, canopy coverage estimation.

[0096] Based on the soil separation model and the time series image optimization result obtained by the dynamic weed removal method using the RGB image time series, the number of peanut canopy pixel clusters and the number of soil pixel clusters in each plot area are counted respectively, and the canopy coverage of the peanut is calculated according to formula (6).

[0097]

[0098] In the formula, FVC (i,j) is the canopy coverage of the plot area in the i-th row and the j-th column, M (i,j) is the number of crop canopy pixel clusters of the plot area in the i-th row and the j-th column, N (i,j) is the number of soil pixel clusters of the plot area in the i-th row and the j-th column.

[0099] As shown in Figure 3 , the embodiment of the present application provides a high-throughput acquisition system for crop canopy coverage by unmanned aerial vehicle remote sensing, comprising:

[0100] A visual feature pattern acquisition module 1 is used to quantitatively describe the visual feature pattern between the crop canopy and the soil background from the high-resolution RGB image.

[0101] A single-phase crop canopy mask map generation module 2 is used to automatically mine the association mapping between the category semantics and the visual feature pattern by using the support vector machine method, to obtain the preliminary extraction result of the crop canopy and the soil background, to generate a single-phase crop canopy mask map, and to mark the crop canopy area as 1 and the soil background area as 0.

[0102] A dynamic weed removal module 3 of the RGB image time series is used to optimize all single-phase crop canopy mask maps of crop growth by using the dynamic weed removal method of the RGB image time series, to determine the peanut canopy area and the weed area, and to mark and remove the weeds in the entire growth period of crop growth.

[0103] A crop canopy coverage estimation module 4 is used to count the number of peanut canopy pixel clusters and the number of soil pixel clusters in each plot area in each plot based on the optimized result, to complete the accurate estimation of the crop canopy coverage.

[0104] To further illustrate the related effects of the embodiments of the present application, the following experiments are performed: as Figure 3As shown in (a) to (f) in the figure, the vegetation coverage of the binary plot image without removing weeds is calculated, and the plot vegetation coverage is 40.8227%. The binary plot vegetation coverage calculation result after manually marking the weed area is 36.5236%. The vegetation coverage of the binary plot image obtained by using the method of the present application is 36.5555%. It proves that the present application can accurately estimate the vegetation coverage and has a positive effect on removing weeds and identifying peanut areas.

[0105] The above is only a preferred specific embodiment of the present application, but the protection scope of the present application is not limited thereto, and any modification, equivalent replacement and improvement within the technical range disclosed by the present application and within the spirit and principle of the present application should be covered within the protection scope of the present application.

Claims

1. A high-throughput method for obtaining crop canopy cover from unmanned aerial vehicle (UAV) remote sensing, characterized in that, The method includes the following steps: S1, quantitatively describes the visual feature patterns between crop canopy and soil background from high-resolution RGB imagery; S2, using the support vector machine method to automatically mine the association mapping between category semantics and visual feature patterns, obtain the preliminary extraction results of crop canopy and soil background, generate a single-temporal crop canopy mask map, with the crop canopy area marked as 1 and the soil background area marked as 0; S3 utilizes the dynamic weed removal method based on RGB image time series to optimize the crop canopy mask map for all single-phase crop growth, identify the peanut canopy area and weed area, and mark and remove weeds throughout the entire growth period of the crop. S4. Based on the optimized results, the number of peanut canopy pixel clusters and soil pixel clusters in each plot are counted to complete the accurate estimation of crop canopy coverage. In step S3, a dynamic weed removal method based on RGB image time series is used, including: (1) Generate cumulative phase mask image by multiplying the mask data of all growth periods pixel by pixel and finding the intersection of the crop canopy area to obtain the cumulative phase mask image; (2) Combine the cumulative phase mask image and perform DBSCAN cluster analysis to determine whether each pixel cluster persists throughout the entire growth period of the crop; In step S4, the accurate estimation of crop canopy coverage is completed, expressed as follows: （6） In the formula, For the first line, number The canopy coverage of the listed plots of land. For the first line, number The number of crop canopy pixel clusters in the plots listed. For the first line, number The number of soil pixel clusters in the listed plots.

2. The high-throughput method for obtaining crop canopy coverage by UAV remote sensing according to claim 1, characterized in that, In step S1, the crop is peanut. In the quantitative description of the visual feature patterns between the crop canopy and the soil background from the high-resolution RGB image, two color indices, the Super Green Index (ExG) and the Vegetation Extraction Color Index (CIVE), are used to separate the visual feature patterns of the peanut canopy from the soil background.

3. The high-throughput method for obtaining crop canopy coverage by UAV remote sensing according to claim 2, characterized in that, The Super Green Index (ExG) is used to calculate the degree of greenness of crops. The calculation formula is as follows: （1） In the formula, It is a super green index. The green band of the image, The red band of the image, The blue band of the image; The calculation formula is: （2） In the formula, These are the channel values ​​for the red, green, and blue channels, respectively. The vegetation extraction color index (CIVE) is used to separate green vegetation from the soil background. The calculation formula is as follows: （3） In the formula, Color index was extracted from the vegetation.

4. The high-throughput method for obtaining crop canopy coverage by UAV remote sensing according to claim 1, characterized in that, In step S2, the association mapping between category semantics and visual feature patterns is automatically mined using the support vector machine method, including: We build an SVM classifier model using the scikit-learn library in the Python environment, and train it on the training sample data using the SVC function in the scikit-learn library. The function call format is as follows: （4） In the formula, To train a support vector machine model as defined, This indicates that the called function is Support Vector Machine (SVM) classification. This is the kernel function of the model; Train a Support Vector Machine (SVM) model, using the RGB values ​​of pixels as feature inputs and classification labels as output targets; replace the thresholding method with the SVM model, inputting the RGB values ​​of each pixel into the SVM model and outputting classification labels; use the established SVM classifier model to classify the acquired RGB images of peanuts at each growth stage, obtaining preliminary extraction results of peanut canopy and soil backgrounds, and obtaining a mask map containing only peanut images of the processed period.

5. The method for high-throughput acquisition of crop canopy coverage by UAV remote sensing according to claim 1, characterized in that, In step (1), the cumulative phase mask is obtained, and its expression is: （5） In the formula, For the cumulative phase mask, The sum of time points. For at a certain point in time The generated binary mask image, For pixels At the point of time Identified as a vegetated area For pixels At the point of time Not identified as a vegetated area The row and column where the pixel is located.

6. The method for high-throughput acquisition of crop canopy coverage by UAV remote sensing according to claim 1, characterized in that, In step (2), DBSCAN cluster analysis is performed using the cumulative phase mask image, including: The DBSCAN clustering algorithm is used to cluster connected canopy pixels, dividing the extracted single-phase peanut canopy mask into different pixel clusters. Combined with the cumulative phase mask, it is determined whether each pixel cluster persists throughout the entire peanut growth period. If the current pixel cluster intersects with the cumulative phase mask, then the pixel cluster is a peanut canopy region; otherwise, the pixel cluster is determined to be a weed region, and weeds are removed. This process is repeated to mark and remove weeds throughout the entire peanut growth period.

7. A high-throughput system for acquiring crop canopy cover using unmanned aerial vehicle (UAV) remote sensing, characterized in that, The system implements the high-throughput method for obtaining crop canopy cover by UAV remote sensing as described in any one of claims 1-6, and the system includes: The visual feature pattern acquisition module is used to quantitatively describe the visual feature patterns between crop canopy and soil background from high-resolution RGB images; The single-temporal crop canopy mask generation module is used to automatically mine the association mapping between category semantics and visual feature patterns using the support vector machine method, obtain the preliminary extraction results of crop canopy and soil background, and generate a single-temporal crop canopy mask, with the crop canopy area marked as 1 and the soil background area marked as 0. The dynamic weed removal module for RGB image time series is used to optimize the crop canopy mask map of all single time phases of crop growth using the dynamic weed removal method of RGB image time series, identify the peanut canopy area and the weed area, and mark and remove weeds throughout the entire growth period of the crop. The crop canopy coverage estimation module, based on the optimized results, counts the number of peanut canopy pixel clusters and soil pixel clusters in each plot, thereby completing an accurate estimation of crop canopy coverage.

8. The high-throughput crop canopy coverage acquisition system based on UAV remote sensing according to claim 7, characterized in that, The high-throughput crop canopy coverage acquisition system based on UAV remote sensing is mounted on a computer-readable storage medium. The computer-readable storage medium stores a computer program, which, when executed by a processor, can realize the functions of the aforementioned high-throughput crop canopy coverage acquisition system based on UAV remote sensing.

Images