Deep learning method and device for identifying plant number and leaf age in corn seedling stage, and storage medium

Abstract

The invention relates to the technical field of crop image recognition processing, in particular to a deep learning method and device for corn seedling stage plant number and leaf age recognition and a storage medium. Then the plant number and leaf age of corn seedlings are accurately predicted through an improved recognition model YOLOv8n-LP, and then mixed comparison of multiple evaluation indexes and multiple models is adopted to verify that the method is high in precision, high in performance, small in model space proportion and high in adaptability; the large-scale standardized monitoring of the growth condition of the corn in the seedling stage is realized, the farmland management efficiency can be remarkably improved, a scientific basis can be provided for crop production, and agricultural intelligentization and sustainable development are assisted.

Description

Technical Field

[0001] The present invention relates to the technical field of crop image recognition and processing, and particularly to a deep learning method, device, and storage medium for identifying the number of plants and leaf age of maize at the seedling stage. Background Art

[0002] Maize is one of the crops with the largest planting area globally. Although the planting area has decreased in recent years, its consumption has increased significantly over the past decade. With limited arable land resources, increasing maize yield per unit area to meet the growing food demand of the population has become an important goal for global agricultural development.

[0003] In this context, we found that during the entire growth cycle of maize, the seedling stage is a crucial stage determining yield potential, and the emergence rate and leaf development status are important indicators of early growth, having a direct impact on subsequent growth and final yield.

[0004] With the development of precision agriculture, the demand for dynamic monitoring of crop growth is increasing day by day. Especially at the critical seedling stage of maize, it is particularly important to accurately monitor the number of plants and leaf age of the crop. These data are not only the basis for evaluating crop growth status but also important references for optimizing farmland management decisions and predicting crop yields.

[0005] However, at present, the monitoring of maize at the seedling stage still mainly relies on manual surveys. This traditional method is time-consuming, laborious, and inefficient, and is greatly affected by human subjective factors, making it difficult to meet the requirements of rapid and accurate monitoring of large-scale farmland.

[0006] In addition to relying on manual surveys, the application of remote sensing technology has provided new possibilities for agricultural monitoring. For example, collecting farmland information through drones provides a new technical means for crop growth monitoring. However, in complex field environments (such as weed coverage, shadow interference, large differences in soil background, etc.), the monitoring accuracy of existing remote sensing technology is limited and it cannot effectively adapt to complex scenarios. Moreover, general image processing algorithms lack optimization for specific characteristics of maize at the seedling stage, making it difficult to accurately identify the number of maize plants and leaf age. At the same time, the processing speed of existing technologies is slow, making it difficult to meet the needs of real-time monitoring and rapid decision-making for large-scale farmland, especially showing deficiencies in precision agriculture management. Therefore, there are still many limitations in applying remote sensing technology to maize at the seedling stage. For the above reasons, the present invention designs a deep learning method, device, and storage medium for identifying the number of plants and leaf age of maize at the seedling stage, constructs a lightweight and intelligent automatic recognition system, and realizes large-scale standardized monitoring of the growth status of maize at the seedling stage. It can not only significantly improve farmland management efficiency but also provide a scientific basis for crop production, contributing to agricultural intelligence and sustainable development. Summary of the Invention

[0007] The object of the present invention is to overcome the deficiencies of the prior art and provide a deep learning method, device and storage medium for identifying the number of plants and leaf age at the seedling stage of corn. By constructing a lightweight and intelligent automatic identification system, it realizes large-scale standardized monitoring of the growth status of corn at the seedling stage, which can not only significantly improve the efficiency of farmland management, but also provide a scientific basis for crop production, contributing to agricultural intelligence and sustainable development.

[0008] The present invention provides a deep learning method for identifying the number of plants and leaf age at the seedling stage of corn, including the following steps: S1, Acquisition of original images: S1-1, Use near-ground equipment and UAV cameras to obtain data of corn at the seedling stage, and the data are all near-ground images and UAV images in the V2 stage to V5 stage; The V2 stage is the two-leaf age stage, that is, the stage when two corn leaves are fully unfolded; the V5 stage is the five-leaf age stage, that is, the stage when five corn leaves are fully unfolded; S1-2, After stitching the UAV images, crop the stitched images to obtain the plot images of the corn seedling stage to be identified; S2, Training of the recognition model: S2-1, Organize the processed images into two data sets: near-ground NG and UAV UAV; S2-2, Draw bounding boxes around each corn seedling for manual annotation; S2-3, In order to increase the diversity and quantity of the UAV image data training set, perform data augmentation processing on the UAV image data; S2-4, The near-ground NG data set and the UAV UAV data set are independent of each other, and both are divided into training set, validation set and test set according to 8:1:1 to train the model; S3, Recognition model YOLOv8n-LP: The recognition model YOLOv8n-LP adjusts the backbone network, network neck and network head components based on YOLOv8n, and reduces the number of parameters without affecting the accuracy; S4, Counting the number of plants at the corn seedling stage: S4-1, Use the trained model to predict the plot images of the corn seedling stage to be identified, generate pictures with single-plant bounding boxes, and save the corresponding box position information to a txt file; S4-2, Each line in the txt file records the confidence and position information of a box. By counting the number of lines in the txt file, the predicted number of corn seedlings is obtained; For S4-3 and S4-1, the actual number of plants in the picture of the bounding box of a single plant is obtained by labeling the image and generating box information before training, and the statistical counting is carried out using the same method as the prediction result; For S4-4, by comparing the prediction result and the actual number of plants, it is used to evaluate the accuracy of the corn seedling stage counting result; For S5, corn seedling leaf age estimation: Using the near-ground NG platform and UAV dataset, the recognition of corn leaf tips is carried out based on the recognition model YOLOv8n-LP, which is used to estimate the leaf age of single corn plants and plots; For S6, measurement result verification: Using evaluation indicators such as Precision (p), Recall (r), and Average Precision (AP) to evaluate the performance of the recognition model YOLOv8n-LP, and using the coefficient of determination R 2 , Root Mean Square Error RMSE, and Relative Root Mean Square Error rRMSE to evaluate the accuracy of the model recognition result. The specific formulas are as follows: Precision: Formula 1: ; Recall: Formula 2: ; Average Precision: Formula 3: ; Where is the differential unit in the integral formula, is an algebraic symbol; Coefficient of determination R 2 : Formula 4: ; Root Mean Square Error RMSE: Formula 5: Relative Root Mean Square Error rRMSE: Formula 6: ; Where TP represents the number of samples correctly predicted as positive by the model, FP represents the number of samples that are actually negative but mispredicted as positive by the model, FN represents the number of samples that are actually positive but mispredicted as negative by the model, represents the actual value of the i-th observation, represents the predicted value of the i-th observation, represents the average value of the observations.

[0009] Specifically, S3 includes: For S3-1, the backbone network adopts the C2f_DAttention module, which is used to enhance the dynamic attention mechanism in the feature extraction process, thereby improving the ability to capture the details of corn leaves; S3-2. The network neck is a bidirectional feature pyramid network (BiFPN), and a 3×3 convolutional layer is added before each multi-scale feature fusion to reduce the computational cost. S3-3. The network head adopts the EfficientHead module, which uses shared convolutional operations and multi-scale information fusion to reduce the computational load and improve the detection accuracy. S3-4. Layer Adaptive Model Pruning (LAMP) is implemented on the recognition model YOLOv8n-LP to dynamically adjust the pruning rate of each layer, remove unimportant layers, and retain the parameters of key layers, for high-precision and high-efficiency maize plant detection and further reduce the computational requirements.

[0010] S5 specifically includes: S5-1. When counting the seedlings, first generate the cropped images of individual maize plants, and label the leaf tips by the labeling method, thus constructing a new dataset related to the leaf age. S5-2. The new dataset related to the leaf age includes the near-ground NG platform dataset and the unmanned aerial vehicle (UAV) dataset, which are independent of each other, and both are divided into training set, validation set, and test set in the ratio of 8:1:1 to train the model. S5-3. First predict the sum of the leaf ages of all individual maize plants, and then estimate the average leaf age of the plot by dividing the sum of the leaf ages by the predicted number of plants. The true leaf age value of the plot is obtained by dividing the sum of the true leaf ages at the time of labeling by the true number of plants.

[0011] The near-ground device is a digital camera fixed on a high pole.

[0012] The V2 stage to V5 stage is the stage from two fully expanded leaves to five fully expanded leaves.

[0013] The data augmentation process includes brightness adjustment, histogram equalization, white balance, and color setting adjustment.

[0014] An electronic device includes a processor and a memory storing a program. The program includes instructions that, when executed by the processor, cause the processor to execute the deep learning method for identifying the number of maize seedlings and leaf age.

[0015] A non-transitory computer-readable storage medium storing computer instructions for causing a computer to execute the deep learning method for identifying the number of maize seedlings and leaf age.

[0016] Compared with the prior art, the present invention acquires data of maize seedlings at the seedling stage through ground-based equipment and UAV cameras, processes the data, and then accurately predicts the number of maize seedlings and leaf age through the improved recognition model YOLOv8n-LP. Subsequently, a hybrid comparison of multiple evaluation indicators and multiple models is adopted to verify the high precision, high performance, small model space occupancy, and high adaptability of the present invention, realizing large-scale standardized monitoring of the growth status of maize at the seedling stage. It can not only significantly improve the efficiency of farmland management, but also provide a scientific basis for crop production, contributing to the intelligentization and sustainable development of agriculture. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 Schematic diagram of ground-based data acquisition of the present invention.

[0018] Figure 2 Schematic diagram of UAV data acquisition and processing of the present invention.

[0019] Figure 3 Schematic diagram of UAV data augmentation processing of the present invention.

[0020] Figure 4 Schematic diagram of the logic of the recognition model YOLOv8n-LP of the present invention.

[0021] Figure 5 Schematic diagram of the detection result of ground-based image of plant counting of the present invention.

[0022] Figure 6 Schematic diagram of the detection result of UAV image of plant counting of the present invention.

[0023] Figure 7 Schematic diagram of the detection process of leaf age estimation of the present invention.

[0024] Figure 8 Comparison of the evaluation results of various maize seedling detection models of the present invention Figure 1 。

[0025] Figure 9 Comparison of the evaluation results of various maize seedling detection models of the present invention Figure 2 。

[0026] Figure 10 Schematic diagram of the comparison of maize seedling counting of the model of the present invention on different platforms under various conditions: where Figure (a) is the application of the YOLOv8n-L model to ground-based data, Figure (b) is the application of the YOLOv8n-LP model to ground-based data, Figure (c) is the application of the YOLOv8n-L model to UAV data, and Figure (d) is the application of the YOLOv8n-LP model to UAV data.

[0027] Figure 11Confusion matrix for the model of the present invention used for leaf age estimation: where Figure (a) is the YOLOv8n-L model applied to near-ground data, Figure (b) is the YOLOv8n-LP model applied to near-ground data, Figure (c) is the YOLOv8n-L model applied to UAV data, and Figure (d) is the YOLOv8n-LP model applied to UAV data.

[0028] Figure 12 Schematic diagram of the Precision-Recall curve of the present invention. Detailed implementation manners

[0029] The present invention will be further described below in conjunction with the accompanying drawings.

[0030] See Figures 1 to 12 , the present invention provides a deep learning method for identifying the number of plants and leaf age at the seedling stage of corn: I. Acquisition of original images: As Figures 1 - 2 shown, to ensure the accuracy and applicability of the analysis results, a near-ground device (Daheng MER2-302-56U3M / C digital camera fixed on a 4.1-meter-high pole) and a DJI M600-Pro UAV equipped with a Sony α7II digital camera are used to obtain data for corn at the seedling stage.

[0031] Data acquisition is carried out from the V2 stage (two fully expanded leaves) to the V5 stage (five fully expanded leaves). A total of 1918 near-ground images and 744 UAV images are collected. The original UAV images are stitched using Agisoft Metashape software and then cropped using ArcGIS 10.8 software to obtain plot images.

[0032] II. Training of the corn seedling identification model: The processed images are sorted into two datasets: near-ground (NG) and UAV. These two datasets are independent of each other and are divided into training set, validation set, and test set in a ratio of 8:1:1 to train the model.

[0033] The LabelImg tool is used to draw bounding boxes around each corn seedling for manual annotation. The NG dataset contains 1918 images (annotating 74,887 seedlings), and the UAV dataset contains 744 images (annotating 90,176 seedlings). Since the UAV data is relatively small, to increase the diversity and quantity of the training set and enable the model to learn more powerful features, brightness adjustment ( Figure 3 a), histogram equalization ( Figure 3 b), white balance ( Figure 3 c), and color setting adjustment ( Figure 3 d) and other data augmentation processes are performed on it.

[0034] III. Recognition Model YOLOv8n-Light-Pruned (YOLOv8n-LP): YOLOv8n-LP is an optimized variant of YOLOv8n. By adjusting the backbone, neck, and head components, it reduces the number of parameters without sacrificing accuracy.

[0035] In the backbone network, the C2f_DAttention module is adopted, enhancing the dynamic attention mechanism in the feature extraction process and improving the ability to capture details of corn leaves.

[0036] The network neck is replaced with a bidirectional feature pyramid network (BiFPN), and a 3x3 convolutional layer is added before each multi-scale feature fusion to reduce the computational cost.

[0037] In the network head, the EfficientHead module is adopted. By sharing convolutional operations and multi-scale information fusion, it reduces the computational load and improves the detection accuracy.

[0038] In addition, to further reduce the computational requirements, layer adaptive model pruning (LAMP) is implemented on the model. It dynamically adjusts the pruning rate of each layer, removes unimportant layers, and retains the parameters of key layers to ensure high accuracy and efficiency in corn plant detection.

[0039] IV. Counting of Maize Seedling Plants: The trained model is used to predict the images of plots that need to be counted, generating pictures with single-plant bounding boxes, and saving the corresponding box position information to a txt file. Each line records the confidence and position information of a box. By counting the number of lines in the txt file, the predicted number of maize seedlings can be obtained.

[0040] The actual number of plants in each image is obtained by marking the image and generating box information before training, and the same method as the prediction result is used for statistical counting. By comparing the prediction result with the actual number of plants, it can be used to evaluate the accuracy of the maize seedling counting result.

[0041] V. Estimation of Maize Seedling Leaf Age: Using the NG platform and UAV dataset, the proposed model is used to identify the leaf tips of maize, thereby estimating the leaf age of individual maize plants and plots.

[0042] The steps are similar to those of plant counting. When counting seedlings, first generate the cropped images of individual maize plants, and label the leaf tips by the labeling method, thereby constructing a new dataset related to the leaf age.

[0043] Among them, the NG dataset contains 17,712 single-plant maize images, including 15,940 in the training set and 1,772 in the test set; the UAV platform dataset contains 14,244 images, including 12,819 in the training set (25,639 after data augmentation) and 1,425 in the test set. Similarly, these two leaf age datasets are divided into training set, validation set and test set according to 8:1:1, and the YOLOv8n-LP model is used for training. First, the leaf age of a single plant of maize is predicted, and then the average leaf age of the plot is estimated by dividing the leaf age by the predicted number of plants. The true leaf age value of the plot is obtained by dividing the true leaf age at the time of annotation by the true number of plants.

[0044] VI. Verification of measurement results: Evaluation metrics such as Precision, Average Precision, and Recall are used to evaluate the performance of the model, and the coefficient of determination (R2), Root Mean Squared Error (RMSE), and relative Root Mean Squared Error (rRMSE) are used to evaluate the accuracy of the model recognition results.

[0045] Precision refers to the proportion of the number of samples (TP) correctly predicted as positive by the model among all samples predicted as positive (TP+FP). The higher the precision, the higher the accuracy of the model in predicting positive samples. The expression is as follows: Precision: Formula 1: ; Recall refers to the proportion of the number of samples (TP) correctly predicted as positive by the model among all samples actually being positive (TP+FN). The higher the recall, the stronger the ability of the model to correctly identify samples actually being positive. The expression is as follows: Recall: Formula 2: ; Average Precision refers to the average of the precision at different recall thresholds of the model, reflecting the overall performance of the object detection or classification model at different decision thresholds. The higher the AP, the higher the precision the model can maintain at different recall levels. Its calculation expression is as follows: Average Precision: Formula 3: ; Formula 3 is an integral formula, is the differential unit in the integral formula, without special meaning; is an algebraic symbol, and in Formula 3 As a whole, it represents a smoothed form of the Precision-Recall curve (as shown in Figure 12 ), which is transformed into a monotonically decreasing function. Here, r’ can be replaced by any symbol without affecting the calculation result.

[0046] The coefficient of determination refers to the degree of fit between the predicted values and the actual values of the model, measuring the model's ability to explain the data. The closer the R² value is to 1, the better the model can explain the data changes and the better the prediction effect. Among them, represents the actual value of the i-th observation, represents the predicted value of the i-th observation, represents the average value of the observations. Its calculation expression is as follows: Coefficient of determination R 2 : Formula Four: ; The root mean square error refers to the root mean square of the error between the predicted value and the actual value of the model, measuring the magnitude of the model prediction error. The smaller the RMSE, the lower the prediction error of the model and the better the fitting effect. Its calculation expression is as follows: Root mean square error: Formula Five: ; The relative root mean square error refers to the ratio of the root mean square error (RMSE) to the mean of the true values, measuring the relative magnitude of the model prediction error relative to the actual observations. The smaller the rRMSE, the smaller the prediction error of the model relative to the data fluctuations and the better the fitting effect.

[0047] Relative root mean square error: Formula Six: ; VII. Model comparison statistics: 1. Model comparison: YOLOv8n-LP is compared with other classical recognition models. As shown in Figures 8 - 9 , the evaluation results of various corn seedling detection models are shown: Among the existing models, YOLOv8n has the highest precision of 0.945, R² of 0.90, and rRMSE of 8.53%. The proposed YOLOv8n-LP model maintains high performance (p = 0.953, rRMSE = 6.25%), while significantly reducing the model size (1.8MB) and being very efficient in practical applications.

[0048] 2. Corn seedling stage plant counting: The proposed YOLOv8n-L and YOLOv8n-LP models on different observation platforms ( Figure 7), the corn seedling counting performance is excellent under conditions such as leaf age, image resolution, seedling distribution, and planting density. The R², RMSE, and rRMSE indicators of each experiment are summarized in Table 1.

[0049] The results show that the YOLOv8n-L and YOLOv8n-LP models exhibit a consistent performance trend under different factors, and the rRMSE is less than 10%.

[0050] 3. Estimation of corn seedling leaf age: Considering that field crop management is usually carried out at the plot scale, in this embodiment, the accuracy of estimating leaf age in plot images is further evaluated based on single-plant leaf age data. The experimental results show that the two models exhibit good adaptability on different data acquisition platforms. The R², RMSE, and rRMSE indicators of each experiment are summarized in Table 2.

[0051] The results show that YOLOv8n-LP performs well on both the NG and UAV datasets, achieving accurate estimation results, with rRMSE values of 5.73% and 9.24% respectively.

[0052] Confusion matrix ( Figure 11 ) shows that the actual values are highly consistent with the predicted values, and most values are concentrated on the diagonal. This concentration indicates that the model's predictions are basically accurate. Figure 11 a and Figure 11 b represent the performance of the YOLOv8n-L model on the NG and UAV datasets respectively, Figure 11 c and Figure 11 d represent the performance of the YOLOv8n-LP model on the NG and UAV datasets respectively.

[0053] The confusion matrix is used to show the accuracy of the classification model. In this confusion matrix, the horizontal axis is the actual leaf age, the vertical axis is the leaf age predicted by the model, and the numbers are the number of samples belonging to that situation. For example, in (a), the 217 in the upper left corner corresponds to the coordinates of the horizontal and vertical axes both being V2, indicating that there are 217 samples that actually belong to the V2 leaf age and are detected as the V2 leaf age by the model.

[0054] The above is only the preferred implementation manner of the present invention, which is only used to help understand the method and its core idea of the present application. The protection scope of the present invention is not limited to the above embodiments. All technical solutions falling within the idea of the present invention belong to the protection scope of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements should also be regarded as the protection scope of the present invention.

[0055] The present invention overall solves the problems of insufficient accuracy and low efficiency caused by the dependence on manual methods for monitoring maize seedlings in the prior art, as well as the insufficient accuracy due to the inability of traditional remote sensing methods to adapt to complex field environments, the lack of precise identification and slow processing due to the lack of optimization of specific features and the inability to adapt to large-scale monitoring and decision-making. By constructing a lightweight and intelligent automatic identification system, it realizes large-scale standardized monitoring of the growth status of maize seedlings, and through promoting the popularization and application of automation technologies in the field of precision agriculture, the present invention can not only significantly improve the efficiency of farmland management, but also provide a scientific basis for crop production, contributing to agricultural intelligence and sustainable development.

Claims

1. A deep learning method for identifying the number of corn seedlings and leaf age, characterized in that: The following steps are involved: S1, original image acquisition: S1-1, using near-ground equipment and drone cameras to obtain data on corn seedlings, the data are near-ground images and drone images from V2 to V5 stages; The V2 stage is the two-leaf stage, i.e., the stage with two corn leaves fully expanded; the V5 stage is the five-leaf stage, i.e., the stage with five corn leaves fully expanded; S1-2, after stitching the drone images, cropping the stitched images to obtain an image of the plot where corn seedlings need to be identified; S2, training of recognition model: S2-1, organize the processed images into two datasets: near-ground NG and UAV; S2-2, draw a bounding box around each corn seedling and perform manual annotation; S2-3, in order to increase the diversity and quantity of the drone impact data training set, perform data expansion processing on the drone image data; S2-4, the near-ground NG dataset and the UAV dataset are independent of each other, and are both divided into a training set, a validation set, and a test set according to an 8:1:1 ratio to train the model; S3, recognition model YOLOv8n-LP: The recognition model YOLOv8n-LP adjusts the backbone network, network neck and network head components based on YOLOv8n to reduce the number of parameters without affecting the accuracy; S4, corn seedling plant counts: S4-1, using the trained model to predict the plot image that needs to identify the corn seedling stage, generate an image with a single plant boundary box, and save the corresponding box position information in a txt file; S4-2, each line in the txt file records the confidence and position information of a box, and the predicted number of corn seedlings is obtained by counting the number of lines in the txt file; S4-3, the actual number of plants in the image of the single plant boundary box in S4-1 is obtained by marking the image and generating frame information before training, and statistical counting is performed using the same method as the prediction result; S4-4, used to evaluate the accuracy of corn seedling counting results by comparing the predicted results with the actual plant numbers; S5, estimation of leaf age of corn seedlings: Using the near-ground NG platform and the UAV dataset, the corn leaf tip is identified based on the recognition model YOLOv8n-LP to estimate the leaf age of individual corn plants and plots; S6, measurement result verification: The performance of the recognition model YOLOv8n-LP is evaluated using the precision, recall and average precision evaluation indicators, and the determination coefficient R is used to evaluate the performance of the recognition model YOLOv8n-LP. 2 , root mean square error RMSE and relative root mean square error rRMSE are used to evaluate the accuracy of the model recognition results. The specific formulas are as follows: Precision: Formula 1: ; Recall: Formula 2: ; Average Precision: Formula 3: ;in is the differential unit in the integral, is an algebraic symbol; Coefficient of determination R 2 :Formula 4: ; Root mean square error RMSE: Formula 5: Relative root mean square error rRMSE: Formula 6: ; Among them, TP represents the number of samples correctly predicted as positive examples by the model, FP represents the number of samples that are actually negative examples but incorrectly predicted as positive examples by the model, and FN represents the number of samples that are actually positive examples but incorrectly predicted as negative examples by the model. represents the actual value of the ith observation, represents the predicted value of the ith observation, Represents the mean of the observations.

2. The deep learning method for identifying the number of corn seedlings and leaf age according to claim 1, characterized in that: The S3 specifically includes: S3-1, the backbone network adopts the C2f_DAttention module to enhance the dynamic attention mechanism in the feature extraction process, thereby improving the ability to capture the details of corn leaves; S3-2, the neck of the network is a bidirectional feature pyramid network BiFPN, and a 3×3 convolution layer is added before each multi-scale feature fusion to reduce the computational cost; S3-3, the network head adopts the EfficientHead module, which is used to reduce the computational load and improve the detection accuracy through shared convolution operation and multi-scale information fusion; S3-4, layer adaptive model pruning LAMP is implemented on the recognition model YOLOv8n-LP, dynamically adjusting the pruning rate of each layer, removing unimportant layers, while retaining the parameters of key layers, for high-precision and high-efficiency corn plant detection, and for further reducing computing requirements.

3. The deep learning method for identifying the number of corn seedlings and leaf age according to claim 1, characterized in that: The S5 specifically includes: S5-1, when counting seedlings, firstly, a cropped image of a single corn plant is generated, and the leaf tips are annotated by the labeling method, thus constructing a new dataset related to leaf age; S5-2, the new datasets related to leaf age include the near-ground NG platform dataset and the UAV dataset, which are independent of each other and are divided into training set, validation set and test set according to 8:1:1 to train the model; S5-3, first predict the sum of the leaf ages of all individual corn plants, then divide the sum of the leaf ages by the predicted number of plants to estimate the average leaf age of the plot. The actual leaf age value of the plot is obtained by dividing the sum of the actual leaf ages at the time of marking by the actual number of plants.

4. The deep learning method for identifying the number of corn seedlings and leaf age according to claim 1, characterized in that: The near-ground device is a digital camera fixed on a high pole.

5. The deep learning method for identifying the number of corn seedlings and leaf age according to claim 1, characterized in that: The V2 stage to the V5 stage are from the stage of two fully expanded leaves to the stage of five fully expanded leaves.

6. The deep learning method for identifying the number of corn seedlings and leaf age according to claim 1, characterized in that: The data expansion process includes brightness adjustment, histogram equalization, white balance and color setting adjustment.

7. An electronic device, characterized in that: It includes a processor and a memory storing a program, wherein the program includes instructions, and when the instructions are executed by the processor, the processor executes the deep learning method for identifying the number of corn seedlings and leaf age as described in any one of claims 1 to 6.

8. A non-transitory computer-readable storage medium storing computer instructions, characterized in that: The computer instructions are used to enable a computer to execute the deep learning method for identifying the number of corn seedlings and leaf age as described in any one of claims 1 to 6.

Images