Method for monitoring and consistency evaluation of sida hermaphrodita in the field during flowering period

Abstract

This invention provides a method for monitoring the flowering period and assessing the consistency of milk thistle in field settings. High-resolution remote sensing images are acquired by using a drone to cruise and photograph the target area. A deep learning-pre-trained milk thistle flower recognition model is then used to accurately extract milk thistle flower information from the remote sensing images. Based on the recognition results, a flowering period display map is generated, enabling visualized monitoring of the dynamic flowering period of milk thistle within the study area. Simultaneously, a flowering period consistency evaluation index system is constructed to quantitatively assess the synchronicity of flowering periods among different plots or varieties. This method provides scientific and objective decision support for milk thistle variety breeding, adaptability evaluation, and yield prediction.

Description

Technical Field

[0001] This invention relates to the field of milk thistle flowering period monitoring, and in particular to a method for monitoring and evaluating the consistency of milk thistle flowering period in the field. Background Technology

[0002] Milk thistle (Silybum marianum (L.) Gaertn.), also known as water pheasant, milk thistle, rat tendon, and milk thistle seed, is an annual or biennial herbaceous plant belonging to the genus Silybum in the family Asteraceae. Milk thistle has functions such as clearing heat and detoxifying, protecting the liver, promoting bile secretion, and improving brain function, and occupies an important position in the domestic and international herbal medicine market.

[0003] The differences in flowering phenotypes among different milk thistle varieties can directly affect seed maturity and yield stability. Monitoring the entire flowering phenotype process is of great significance for the breeding of superior varieties and large-scale yield estimation. Traditional milk thistle flowering monitoring methods rely on manual observation, which is not only time-consuming and labor-intensive with low quantitative levels, but also has low monitoring efficiency and accuracy when facing large-scale planting areas. Summary of the Invention

[0004] This application provides a method for monitoring and assessing the consistency of milk thistle flowering time in open fields. This method can improve the efficiency and accuracy of milk thistle flowering time monitoring. The technical solution is as follows:

[0005] A method for monitoring the flowering period and assessing the consistency of milk thistle in open fields, comprising:

[0006] Drones were used to inspect the study area and acquire remote sensing images of the area.

[0007] The milk thistle flower in the remote sensing image was identified using a pre-trained milk thistle flower recognition model, and the milk thistle flower recognition results were obtained.

[0008] The flowering period of the milk thistle is determined based on the milk thistle flower identification results, and a flowering period display image is generated;

[0009] Based on a pre-set evaluation index system for flowering period consistency, the flowering period synchronicity among different plots or varieties within the study area is assessed.

[0010] In this embodiment, a drone is used to conduct cruise photography of the target area to acquire high-resolution remote sensing images. Combined with a pre-trained milk thistle flower recognition model, milk thistle flower information is accurately extracted from the remote sensing images. Based on the recognition results, a spatiotemporal distribution map of flowering period is generated, enabling visualized monitoring of the dynamic flowering period of milk thistle within the study area. Simultaneously, a flowering period consistency evaluation index system is constructed to quantitatively assess the synchronicity of flowering periods among different plots or varieties. This method provides scientific and objective decision support for milk thistle variety breeding, adaptability evaluation, and yield prediction.

[0011] To better understand and implement this invention, the following detailed description is provided in conjunction with the accompanying drawings. Attached Figure Description

[0012] Figure 1 This is a flowchart of a method for monitoring the flowering period and evaluating the consistency of milk thistle in a field, according to one embodiment of the present invention.

[0013] Figure 2 This is a flowchart of a method for monitoring the flowering period and evaluating the consistency of milk thistle in a field, as described in another embodiment of the present invention.

[0014] Figure 3 This is a flowchart of a method for monitoring the flowering period and evaluating the consistency of milk thistle in a field, as described in another embodiment of the present invention.

[0015] Figure 4 This is a schematic diagram illustrating the annotation of milk thistle flowers in a remote sensing image according to one embodiment of the present invention;

[0016] Figure 5 This is a graph showing the change in the loss function value of the model in one embodiment of the present invention;

[0017] Figure 6 This is a graph showing the variation of the average crossover ratio in one embodiment of the present invention;

[0018] Figure 7 This is a flowchart of a method for monitoring the flowering period and evaluating the consistency of milk thistle in a field, as described in another embodiment of the present invention.

[0019] Figure 8 This is a schematic diagram illustrating the identification results of milk thistle flowers at the initial flowering stage and the full flowering stage in one embodiment of the present invention;

[0020] Figure 9 This is a flower density diagram of milk thistle flowers during the initial flowering stage and the full flowering stage in one embodiment of the present invention;

[0021] Figure 10 This is a histogram of flower density for milk thistle varieties 0-5 during their initial flowering stage, as shown in one embodiment of the present invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0023] It should be understood that the described embodiments are merely some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of the embodiments of this application.

[0024] The terminology used in the embodiments of this application is for the purpose of describing particular embodiments only and is not intended to limit the embodiments of this application. The singular forms “a,” “the,” and “the” used in the embodiments of this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0025] In the following description, when referring to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims. In the description of this application, it should be understood that the terms "first," "second," "third," etc., are used only to distinguish similar objects and are not necessarily used to describe a specific order or sequence, nor should they be construed as indicating or implying relative importance. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.

[0026] Furthermore, in the description of this application, unless otherwise stated, "several" refers to two or more. "And / or" describes the correspondence between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. The character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0027] Please see Figure 1 This application provides a method for monitoring the flowering period and assessing the consistency of milk thistle in the field, including the following steps:

[0028] S110: Use drones to inspect the study area and obtain drone remote sensing images of the study area;

[0029] Drones can be equipped with multispectral, hyperspectral, thermal infrared, and LiDAR sensors. Drones with centimeter-level spatial resolution can be selected. By adjusting the flight parameters of the drone, drone remote sensing images at different times can be collected to construct a multi-time series (daily / weekly) sea thistle image database.

[0030] In this embodiment, the drone can be a DJI Mavic 3M drone, which has a centimeter-level RTK navigation and positioning system, which can improve the positioning accuracy of image data. The drone also carries a multispectral sensor, which can simultaneously acquire information from different wavelengths. Please refer to Table 1, which is a parameter table for the DJI Mavic 3M drone.

[0031] Table 1. Specifications of the DJI Mavic 3M drone

[0032] Aircraft parameters Parameter value Camera parameters Parameter value Maximum takeoff altitude 6000m Image sensor 4 / 3CMOS Longest flight time 45min focal length 24mm Maximum range 32km Maximum photo size 5280×3956 Maximum horizontal flight speed 15m / s Photo format JPEG Hovering accuracy Vertical ±0.1m Effective pixels 20 million

[0033] Optionally, drones can be used periodically to inspect the study area and collect remote sensing images according to a set inspection frequency. The inspection frequency can be determined based on the terrain and weather conditions of the study area. For example, under favorable weather conditions such as good light intensity and low wind speed, dynamic inspections can be conducted at intervals of 3 to 5 days.

[0034] "Good light intensity" means the light intensity is greater than the set light intensity threshold, and "low wind speed" means the wind speed is less than the set wind speed threshold. The wind speed threshold can be 8 m / s.

[0035] The inspection route of the drone can be set based on the planting status of milk thistle in the study area. The inspection route can be generated using existing inspection route generation algorithms, or it can be pre-input into the drone by the user. In this embodiment, the drone's flight mission is planned using a drone ground station and transmitted to the drone to inspect the flowering status of milk thistle in the study area. The remote sensing images collected by the drone can be stored in a memory card or other storage medium, or transmitted back to a designated receiving device in real time via wireless communication technology.

[0036] During inspections, the drone's flight altitude was fixed at 12 meters, and the overlap rates for the drone's heading and lateral directions were set to 80% and 70%, respectively. Simultaneously, an RTK centimeter-level positioning system was activated to improve spatial positioning accuracy. This ensured that the remote sensing images acquired by the drone could generate a spatially continuous orthophoto dataset, guaranteeing the integrity of the image stitching.

[0037] Milk thistle has distinctive morphological characteristics. Using drones, high-frequency monitoring of the flowering dynamics of milk thistle in the study area can be achieved. This not only allows for the acquisition of vegetation canopy images with sub-centimeter spatial resolution, but also enables the complete recording of the flowering sequence characteristics of different varieties through periodic aerial photography. This effectively overcomes the interference of terrain differences, climate fluctuations, and human management factors, and improves the accuracy of monitoring.

[0038] S120: Use a pre-trained milk thistle flower recognition model to identify milk thistle flowers in the remote sensing image and obtain milk thistle flower recognition results;

[0039] The milk thistle flower recognition model is used to identify milk thistle flowers in remote sensing images and obtain the location of the milk thistle flowers in the image.

[0040] Specifically, milk thistle has a single flower and a single fruit. The plant grows to a height of 1-2m, with a capitulum inflorescence and tubular flowers that are reddish-purple. It has a distinct spherical structure and a large diameter, and its spectral and structural features are obvious. The milk thistle flower recognition model can learn the morphology and features of milk thistle flowers in labeled data, thereby achieving accurate segmentation of flower regions in images.

[0041] Milk thistle flower recognition models can be built based on the U-net architecture or existing machine learning algorithms.

[0042] Specifically, the milk thistle flower recognition model learns the morphology and features of milk thistle flowers in labeled data to accurately segment the flower region in the image, thereby identifying the flowering status of milk thistle.

[0043] By utilizing the milk thistle flower recognition model, the flowering situation in a large-scale milk thistle planting study area can be analyzed automatically and quickly, improving the efficiency of milk thistle flowering analysis in the study area.

[0044] S130: Determine the flowering period of the milk thistle based on the milk thistle flower identification result, and generate a flowering period display image;

[0045] Specifically, the flowering period of each milk thistle variety in the research area can be statistically analyzed based on the milk thistle flower identification results, and a flowering period display map can be generated.

[0046] Flowering period display map is used for visual monitoring of the flowering period dynamics of milk thistle within the study area.

[0047] S140: Based on a preset evaluation index system for flowering period consistency, the flowering period synchronicity among different plots or varieties within the study area is evaluated.

[0048] The evaluation index system for flowering period consistency can include evaluation indicators such as the coefficient of variation and Moran's index, which can be used to evaluate consistency.

[0049] By quantitatively assessing the synchronicity of flowering periods among different plots or varieties, we can identify the temporal and spatial distribution of flowering periods among various plots or varieties, providing scientific and objective decision support for milk thistle variety breeding, adaptability evaluation, and yield prediction.

[0050] In this embodiment, a drone is used to conduct cruise photography of the target area to acquire high-resolution remote sensing images. Combined with a pre-trained milk thistle flower recognition model, milk thistle flower information is accurately extracted from the remote sensing images. Based on the recognition results, a spatiotemporal distribution map of flowering period is generated, enabling visualized monitoring of the dynamic flowering period of milk thistle within the study area. Simultaneously, a flowering period consistency evaluation index system is constructed to quantitatively assess the synchronicity of flowering periods among different plots or varieties. This method provides scientific and objective decision support for milk thistle variety breeding, adaptability evaluation, and yield prediction.

[0051] In step S110, the drone can be controlled to collect remote sensing images according to the environmental and weather conditions of the study area.

[0052] Optionally, the remote sensing imagery may include remote sensing images of milk thistle from multiple periods. The periods can be determined based on the flowering time of the milk thistle.

[0053] In the embodiments of this application, multiple periods may include the initial flowering period, the peak flowering period, and the end of the flowering period of milk thistle. The key period can be determined based on the historical flowering time of this milk thistle variety.

[0054] In step S120, the milk thistle flower recognition model can be a pre-trained model.

[0055] Specifically, a pre-collected sample set can be used to pre-train a milk thistle flower recognition model. The sample set may include remote sensing images labeled with the distribution areas of milk thistle flowers.

[0056] like Figure 2 As shown, in one embodiment, before identifying milk thistle flowers in the remote sensing image using a pre-trained milk thistle flower recognition model, the following steps are included:

[0057] S210: Construct the milk thistle flower recognition model based on the U-net architecture;

[0058] U-Net is a convolutional neural network architecture specifically designed for image segmentation tasks. The U-Net architecture consists of an encoder (downsampling) and a decoder (upsampling), with features fused through skip connections. The encoder comprises multiple repeating convolutional blocks and pooling layers, used to extract multi-level features from the image. The decoder progressively restores resolution through transposed convolutions or upsampling layers, and, combined with the skip connection features from the encoder, restores the spatial resolution of the image to generate the segmentation result.

[0059] The milk thistle flower recognition model in this application employs a symmetrical encoder-decoder structure. The encoder extracts image features through multi-layer convolution and pooling operations, progressively capturing the red-purple spectral characteristics of the milk thistle flower and the morphological structure of its capitulum. The decoder, through upsampling and skip connections, fuses multi-scale features to recover the flower's detailed information. An adaptive mechanism strengthens the weights of relevant color channels while suppressing interference from the green background, effectively resolving the boundary blurring problem between the flower and the green leaf background. The milk thistle flower recognition model combines shallow local details (such as flower edges) with deep global semantics (such as the inflorescence outline) using skip connections, achieving pixel-level segmentation of milk thistle inflorescences in complex backgrounds while maintaining high-resolution feature transfer, thus enabling accurate analysis of the flower's morphology.

[0060] Using a milk thistle flower identification model to identify milk thistle flowers in remote sensing images can automatically and quickly analyze the flowering status of milk thistle flowers in large-area planting areas, improving monitoring efficiency.

[0061] S220: Based on samples in the sample database, the milk thistle flower recognition model is pre-trained.

[0062] Optionally, the sample database may include multiple remote sensing images of milk thistle flowers collected in advance by the user. These milk thistle flower remote sensing images can be field-acquired images or images acquired via the network.

[0063] Specifically, when pre-training the milk thistle flower recognition model, the accuracy of the milk thistle flower recognition model can be verified by combining the results of the quadrat survey, so that the parameters of the model can be adjusted to meet the preset accuracy requirements.

[0064] like Figure 3 As shown, in another embodiment, the sample database can be remote sensing imagery collected by a drone. Before pre-training the milk thistle flower recognition model based on samples from the sample database, the following steps are included:

[0065] S310: Mark the areas where milk thistle flowers are located in the remote sensing image and generate a tag file;

[0066] Specifically, such as Figure 4 As shown, you can select the area where the milk thistle flower is located using a polygonal selection tool to generate a label file.

[0067] S320: Construct a sample database based on the tag file and the remote sensing image.

[0068] The sample database can contain spatiotemporally aligned image-label pairs. The samples in the sample database not only meet the input size requirements of the milk thistle flower recognition model, but also enhance the model's generalization ability through sample diversity, thereby improving the accuracy of training and recognition of the milk thistle flower recognition model.

[0069] In one embodiment, the milk thistle flower recognition model is pre-trained based on samples in a sample database, including:

[0070] The sample database is divided into multiple subsets based on the K-fold cross-validation method, and the multiple subsets are used in turn to train and validate the milk thistle flower recognition model.

[0071] K-fold cross-validation divides the dataset into k non-overlapping subsets of roughly equal size and performs k iterations. In each iteration, a different subset is selected as the test set, and the remaining k-1 subsets are combined as the training set. The model is trained on the training set in each iteration and then evaluated on the corresponding test set. The average of the k evaluations is used as an estimate of the model's performance. By alternating between training and validation, the stability and generalization ability of the model under different data combinations are evaluated, avoiding overfitting caused by specific sample distributions.

[0072] During model training, convolution is first used for downsampling, then layer by layer of features are extracted. These features are then upsampled, resulting in an image where each pixel corresponds to its category. After one hundred iterations, the optimal training model is obtained.

[0073] During training, the loss function can be used to visualize the model's training progress. By analyzing the loss function, the model's state can be determined, whether overfitting or underfitting exists, and the model parameters can be adjusted accordingly to ensure the model meets the target accuracy requirements.

[0074] like Figure 5 As shown, this is a graph showing the change of the model's loss function values ​​on the training and validation sets as the number of training epochs increases. The red solid line represents the loss function value (trainloss) on the training set, the green solid line represents the loss function value (valloss) on the validation set, and the green dashed line and brown dashed line are the training loss and the smoothed validation loss curves, respectively, used to observe the overall trend.

[0075] like Figure 6As shown, it is a graph showing the change of the model's average crossover ratio (mIoU) as the number of training iterations increases. The average crossover ratio is the average crossover ratio of each class in the dataset. The crossover ratio refers to the ratio of the intersection and union of the true label and the predicted value of that class. The average crossover ratio can reflect the model's performance in each class, thus providing a more comprehensive and intuitive measure of the model's overall performance.

[0076] Optionally, other existing performance evaluation metrics can also be used to determine the model's performance. In this embodiment, precision can be used to measure the model's prediction accuracy, using the formula: Precision = TP / (TP + FP). Wherein, TP (True Positive): the number of samples that are both true and predicted as positive; FP (False Positive): the number of samples that are true and predicted as positive.

[0077] Alternatively, recall can be used to measure the model's positive sample recognition rate. Recall refers to the number of positive samples the model identifies out of all positive samples. The formula is: Recall = TP / (TP + FN), where TN (True Negative): the number of samples that are true negatives and predicted as negative, and FN (False Negative): the number of samples that are true positives but predicted as negative. The F1-score (F1) is used to determine the model's overall performance; F1 is the harmonic mean of precision and recall. The formula is: F1 = 2 × (Precision × Recall) / (Precision + Recall). Intersection over Union (IoU) is the most commonly used metric in segmentation tasks, used to verify the degree of overlap between the predicted and labeled regions.

[0078] Experiments showed that the pre-trained milk thistle flower recognition model achieved a segmentation accuracy of over 90%, effectively improving the efficiency and accuracy of milk thistle flower recognition.

[0079] Due to factors such as ambient light, flight attitude, or cloud cover, some remote sensing images of drones may suffer from blurriness, abnormal exposure, or insufficient overlap. Therefore, after obtaining the remote sensing images, preprocessing is necessary to improve the accuracy of milk thistle flower identification.

[0080] like Figure 7 As shown, specifically, before constructing the sample database based on the tag file and the remote sensing image, the process further includes:

[0081] S410: Remove remote sensing images whose image quality is lower than the target image quality threshold;

[0082] Specifically, the image quality of remote sensing images can be determined based on existing image quality detection algorithms such as sharpness assessment algorithms and edge detection algorithms. By removing remote sensing images whose image quality is lower than the target image quality threshold, effective data with clear texture and uniform color can be retained, thereby improving the accuracy of milk thistle flower identification.

[0083] S420: Perform image stitching on the remote sensing images of each period to generate an orthophoto map;

[0084] An orthophoto map is a planar map created from orthophotos, featuring a kilometer grid, inset borders, and annotations.

[0085] Alternatively, existing 3D reconstruction algorithms such as Structure from Motion (SFM) can be used to generate orthophoto maps.

[0086] S430: Using the first phase of remote sensing imagery as a standard, perform histogram matching on the remaining phases of remote sensing imagery;

[0087] Histogram matching is an image enhancement method that adjusts the pixel value distribution of two images to make their histograms as similar as possible. By performing histogram matching on remote sensing images from multiple different periods, the radiometric differences in images caused by external factors are reduced, ensuring the transferability of the subsequent milk thistle flower recognition model and the consistency of image analysis across different periods.

[0088] Specifically, histogram matching is used to match feature points and align spatial coordinates, eliminating perspective distortion and geometric errors, and generating a standardized base map covering the entire study area, providing a spatial benchmark for subsequent time-series analysis. Secondly, because images from different periods may exhibit inconsistent brightness and contrast distributions due to weather changes, sensor parameter fluctuations, or differences in vegetation growth stages, direct overlay analysis can easily introduce interference. Therefore, the first image was selected as the reference benchmark, and histogram matching was performed on images from other periods. This significantly reduces the interference caused by external factors, ensuring the comparability of spectral characteristics between images from different periods, and providing a reliable data foundation for subsequent milk thistle flower detection and growth monitoring.

[0089] Histogram matching reduced the differences in image radiation caused by different lighting conditions on different dates, and constructed a high spatiotemporal density observation matrix for flowering period consistency analysis.

[0090] S440: Crops the remote sensing images from each period to the same size.

[0091] By cropping each remote sensing image to the same size, the recognition efficiency of subsequent milk thistle flower identification models is improved. In this embodiment, the remote sensing image can be cropped into samples of size 256*256.

[0092] Optionally, before constructing the sample database based on the tag file and the remote sensing image, the method further includes:

[0093] Invalid samples are removed from the remote sensing images; the invalid samples include remote sensing images with blurred areas, background interference, or incomplete annotations.

[0094] Background interference can refer to an image containing too much bare soil, shadows, or edge areas. It can be determined based on whether the proportion of bare soil, shadows, or edge areas in the image exceeds a preset target proportion.

[0095] Invalid samples can be removed through manual verification or by using existing image detection algorithms to automate the screening process, thereby improving the data quality of the sample database.

[0096] The sample database constructed in this application embodiment has the characteristics of high quality and spatiotemporal alignment. It can not only retain the key flowering period growth information of thistle, but also reduce data redundancy, providing efficient and standardized input for model training and recognition. The sample diversity increases the generalization ability of the thistle flower recognition model.

[0097] In one embodiment, generating a flowering period display image includes:

[0098] Based on kernel density analysis, flower density maps of milk thistle flowers at various stages were generated.

[0099] Kernel density analysis is a nonparametric method for estimating the probability density function of a random variable. Based on the data itself, it estimates and analyzes the distribution of data by placing a kernel function around the data points and performing smoothing. In this embodiment, kernel density analysis can be used to generate flower density maps of different milk thistle varieties based on an adaptive bandwidth strategy. The flower density map can reflect the spatial distribution characteristics and dynamic changes in flowering period of milk thistle varieties.

[0100] Please see Figure 8-9 In this embodiment of the application, based on the kernel density analysis method, flower density maps of milk thistle flowers at the initial flowering stage (August 2nd) and the full bloom stage (August 13th) of the study area are generated. Figure 8 (a) and (b) show the identification results of milk thistle flowers at the initial flowering stage and full bloom stage in the study area, respectively. Figure 9 (a) and (b) are flower density diagrams of milk thistle flowers at the initial flowering stage and the peak flowering stage, respectively.

[0101] In one embodiment, the method further includes:

[0102] The study area was divided into several grids, and the flowering density of each variety of milk thistle in each grid during the same period was statistically analyzed to generate a flower density histogram; the flower density histogram included the flowering density of each variety of milk thistle.

[0103] Please see Figure 10 These are flower density histograms for milk thistle varieties 0-5 during their initial flowering stage. The flower density in the flower density histogram represents the number of milk thistle flowers per square meter.

[0104] As shown in the figure, it can be seen that the milk thistle of variety 0 has the earliest flowering period, while the milk thistle of variety 4 has the latest flowering period.

[0105] The flower density histogram is used to display the flowering density of different milk thistle varieties at the same time, which makes it convenient for users to monitor the flowering period of different milk thistle varieties and provides important technical support for the planting management and breeding of milk thistle.

[0106] Optionally, the flowering characteristics of milk thistle in each grid, such as the number of flowers, flowering density, and flowering duration, can be statistically analyzed. The coefficient of variation can be used to evaluate the flowering consistency of milk thistle flowers in each grid, thereby quantitatively analyzing the flowering synchronicity and spatial distribution patterns of different milk thistle varieties.

[0107] The coefficient of variation (CV) is the ratio of the standard deviation to the mean. It directly reflects the degree of dispersion of a set of data relative to its mean. A larger CV indicates greater relative dispersion of the data, and vice versa.

[0108] In this embodiment, the coefficient of variation can effectively reflect the spatial consistency of the flowering period of a particular variety. The smaller the value, the closer the number of flowers in each grid and the more concentrated the flowering period; the larger the value, the more uneven the distribution and the poorer the consistency of the flowering period. Therefore, the spatial distribution characteristics are characterized by statistically analyzing the coefficient of variation of milk thistle flowers for each variety.

[0109] In one embodiment, the method further includes:

[0110] Obtain the Moran Index as follows;

[0111]

[0112] Where I is the Moran index, n is the number of grids, xi and xj are the flower densities of grids i and j, respectively, and x - ΣiΣj is the average flower density, wij is an element in the spatial weight matrix, representing the spatial relationship between units i and j (e.g., adjacent is 1, non-adjacent is 0); W is the sum of all spatial weights, i.e., ΣiΣj wij.

[0113] Moran's I value is typically between [-1, 1]:

[0114] I>0 indicates positive spatial autocorrelation, meaning that similar values ​​cluster together, indicating that the flowering period is spatially consistent;

[0115] I<0 indicates negative spatial autocorrelation, meaning that high and low values ​​are interspersed, indicating a large difference in flowering period.

[0116] I≈0 indicates no spatial autocorrelation, meaning the flowering period is randomly distributed.

[0117] Moran's I is a classic spatial autocorrelation statistical indicator widely used to analyze the clustering or dispersion of geospatial variables. In this application, Moran's I is used to assess whether flowering period changes within different grids exhibit spatial clustering characteristics, determine whether a particular variety blooms earlier or later than other regions, and detect whether flowering time changes are continuous in geographic space. This application utilizes high-resolution remote sensing images of the study area acquired by UAVs to achieve high-frequency monitoring of milk thistle flowering period dynamics. A deep learning-pre-trained milk thistle flower recognition model is used to identify milk thistle flowers in the UAV remote sensing images, thereby providing a consistent assessment of the flowering period of different milk thistle varieties in the study area. This provides key technical support for applications such as milk thistle variety breeding and yield estimation. Compared to traditional manual sampling and monitoring, this application can effectively improve monitoring efficiency, avoid data bias caused by subjective judgment, and improve the accuracy of milk thistle flowering period monitoring.

[0118] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0119] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for monitoring the flowering period and evaluating the consistency of milk thistle in open fields, characterized in that, include: Drones were used to inspect the study area and acquire remote sensing images of the area. The pre-trained milk thistle flower recognition model is used to identify milk thistle flowers in the remote sensing image and obtain milk thistle flower recognition results. The pre-trained milk thistle flower recognition model is built based on the U-net architecture. The flowering period of the milk thistle is determined based on the milk thistle flower identification results, and a flowering period display image is generated; Based on a pre-set evaluation index system for flowering period consistency, the flowering period synchronicity among different plots or varieties within the study area is evaluated. The evaluation includes: quantitatively analyzing the flowering synchronicity and spatial distribution patterns of different milk thistle varieties, and / or determining whether the planting area corresponding to a certain milk thistle variety flowers flowers earlier or later than the planting area of ​​other milk thistle varieties, and detecting whether the changes in flowering time are continuous in geographical space. The study area was divided into 1m×1m grids, and the flowering density of each variety of milk thistle in each grid during the same period was counted to generate a flower density histogram. The flower density histogram included the flowering density of each variety of milk thistle. Flower density histograms are used to display the flowering density of different milk thistle varieties at the same time, enabling management and breeding of various milk thistle varieties.

2. The method for monitoring and evaluating the flowering period of milk thistle in the field according to claim 1, characterized in that, Before identifying milk thistle flowers in the remotely sensed image using a pre-trained milk thistle flower recognition model, the following steps are included: The milk thistle flower recognition model is constructed based on the U-net architecture; The milk thistle flower recognition model is pre-trained based on samples in the sample database.

3. The method for monitoring and evaluating the flowering period of milk thistle in the field according to claim 2, characterized in that, Before pre-training the milk thistle flower recognition model based on samples in the sample database, the following steps are included: The areas where milk thistle flowers are located in the remote sensing image are marked, and a tag file is generated; A sample database is constructed based on the tag files and the remote sensing images.

4. The method for monitoring the flowering period and evaluating the consistency of milk thistle in the field according to claim 3, characterized in that, The remote sensing images include remote sensing images from multiple periods, wherein the multiple periods are determined based on the flowering time of milk thistle, and the multiple periods include the initial flowering period, the full flowering period, and the final flowering period of milk thistle; before constructing the sample database based on the tag file and the remote sensing images, the following steps are also included: Remove remote sensing images whose image quality is below the target image quality threshold; The remote sensing images from each period are stitched together to generate an orthophoto map; Histogram matching was performed on the remaining remote sensing images using the first phase of remote sensing images as the standard. The remote sensing images from each period are cropped to the same size.

5. The method for monitoring and evaluating the flowering period of milk thistle in the field according to claim 3, characterized in that, Based on samples in the sample database, the milk thistle flower recognition model is pre-trained, including: The sample database is divided into multiple subsets based on the K-fold cross-validation method, and the multiple subsets are used in turn to train and validate the milk thistle flower recognition model.

6. The method for monitoring the flowering period and evaluating the consistency of milk thistle in the field according to claim 1, characterized in that, Generate a display image showing the flowering period, including: Based on kernel density analysis, flower density maps of milk thistle flowers at various stages were generated.

7. The method for monitoring and evaluating the flowering period of milk thistle in the field according to claim 1, characterized in that, The synchronicity of flowering time among different plots or varieties within the study area was evaluated, including: Count the number of milk thistle flowers in each 1m×1m grid, and calculate the standard deviation and mean. The coefficient of variation for each variety of milk thistle flower was obtained as follows: CV = (σ / μ) × 100% in, CV Let σ represent the coefficient of variation, σ represent the standard deviation, and μ represent the mean.

8. The method for monitoring and evaluating the flowering period of milk thistle in open fields according to claim 1, characterized in that, The synchronicity of flowering time among different plots or varieties within the study area was evaluated, including: Obtain the Moran Index as follows; I = (n / W) × (Σi Σj wij (xi - x (xj - x) )) / Σi (xi - x )² in, I Here, n is the Moran index, xi and xj are the flower densities at grids i and j, respectively, and x is the number of grids. Σi Σj wij is the average flower density, wij is an element in the spatial weight matrix, which is 1 when grids i and j are adjacent and 0 when they are not adjacent; W is the sum of all spatial weights, i.e. Σi Σj wij; I>0 indicates positive spatial autocorrelation, meaning that similar values ​​cluster together, indicating that the flowering period is spatially consistent; I<0 indicates negative spatial autocorrelation, meaning that high and low values ​​are interspersed, indicating a large difference in flowering period. I ≈ 0 indicates no spatial autocorrelation, meaning that the flowering period is randomly distributed.

Images