Sugarcane germination state monitoring system based on image recognition

Abstract

The invention discloses a sugarcane germination state monitoring system based on image recognition, and relates to the technical field of agricultural intelligent monitoring, the system comprises the following components: an image acquisition module, an image recognition module, a dynamic growth model construction module, a prediction and early warning module and a decision support module; according to the method, a dynamic growth model is constructed based on extracted time sequence characteristic data, the germination progress of sugarcane buds is predicted by using a fusion time sequence analysis and machine learning algorithm, and potential growth problems are warned in time when growth data deviates from a normal curve, and in addition, a decision support module is used for generating a decision suggestion through a data visualization unit and a decision suggestion generation unit. Visual data display and scientific decision basis are provided for farmers, the farmers are helped to deal with abnormal growth in time, a planting management strategy is optimized, and precise agriculture and scientific planting are achieved.

Description

Technical Field

[0001] This invention relates to the field of agricultural intelligent monitoring technology, specifically to a sugarcane germination status monitoring system based on image recognition. Background Technology

[0002] With the acceleration of agricultural modernization, precision agriculture and intelligent monitoring technology have become key means to improve crop yield and quality. As an important economic crop, the germination status of sugarcane directly affects the subsequent growth cycle and final yield.

[0003] Traditional methods for monitoring sugarcane germination status have several shortcomings: First, they rely on manual observation and experience, leading to significant subjective influences on monitoring results and substantial differences between observers. Second, manual monitoring struggles to provide real-time, continuous monitoring of sugarcane growth, easily missing optimal management opportunities. Third, traditional methods lack scientific data analysis and predictive capabilities, failing to accurately predict germination progress and potential growth problems, thus hindering the development of effective planting and management strategies. Finally, in large-scale planting scenarios, manual monitoring is costly and inefficient, failing to meet the demands of modern agriculture for efficient and precise management. Therefore, traditional technologies are no longer sufficient to meet the needs of modern sugarcane planting and management.

[0004] In view of the shortcomings of traditional methods for monitoring the germination status of sugarcane, this invention proposes a sugarcane germination status monitoring system based on image recognition, which is of particular importance. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a sugarcane germination status monitoring system based on image recognition. It can acquire time-series images of the sugarcane germination process through a high-resolution camera and a timed shooting device, and use a deep learning algorithm based on an improved convolutional neural network to recognize the images and extract the time-series feature data of the sugarcane buds. Then, the system constructs a dynamic growth model and uses a fusion of time series analysis and machine learning algorithms to predict the germination progress of the sugarcane buds, and provides timely warnings of potential growth problems when the growth data deviates from the normal curve.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a sugarcane germination status monitoring system based on image recognition, the system comprising the following components: an image acquisition module, an image recognition module, a dynamic growth model construction module, a prediction and early warning module, and a decision support module;

[0007] The image acquisition module: acquires time-series images of sugarcane sprouting process. The image acquisition module includes a high-resolution camera and a timed shooting device. The timed shooting device is used to control the high-resolution camera to shoot at set time intervals.

[0008] The image recognition module: identifies the time-series image and extracts the time-series feature data of sugarcane buds. The image recognition module adopts a deep learning algorithm based on an improved convolutional neural network.

[0009] The dynamic growth model construction module: constructs a dynamic growth model based on the time series feature data. The module adopts a fusion of time series analysis and machine learning algorithms and includes a model update unit.

[0010] The prediction and early warning module: uses the dynamic growth model to predict the germination progress of sugarcane buds, and provides early warning of potential growth problems when the growth data deviates from the normal curve;

[0011] The decision support module provides planting decision-making basis based on the prediction and early warning results. The module includes a data visualization unit and a decision suggestion generation unit.

[0012] Furthermore, when determining the shooting time interval in the image acquisition module, the timing shooting device first collects at least 50 sets of historical growth experiment data of different sugarcane varieties, and statistically analyzes the average and variance of the sugarcane bud growth rate of each variety; then, it uses multiple linear regression analysis to determine the weighting coefficients that balance the timeliness of data acquisition and system energy consumption; finally, it calculates the optimal shooting time interval based on the average and variance of the growth rate of the sugarcane variety to be monitored, combined with the weighting coefficients.

[0013] Furthermore, when the adaptive optics lens of the high-resolution camera in the image acquisition module adjusts its focal length, it first measures the distance between the camera and the sugarcane sprout; then, based on a large amount of experimental data on imaging clarity at different distances, it fits a nonlinear relationship model between distance and focal length; finally, based on the measured distance, it uses this model to automatically adjust the lens focal length to ensure that clear images of sugarcane sprouts can be acquired within a distance range of 1-5m.

[0014] Furthermore, the deep learning algorithm based on an improved convolutional neural network in the image recognition module has the following core feature extraction formula: ,in For the first Layer Each feature map For the activation function, a modified Swish function is used. ,in The trainable parameter was determined by gradient descent training on a dataset containing 2000 sugarcane sprout images, and its value was 0.3. and The convolution kernel weights and biases were fine-tuned through transfer learning using 800 labeled sugarcane sprout images, based on an ImageNet pre-trained model. This is an attention mechanism function used to highlight features of the sugarcane bud region. The weight is 0.5, which is determined by analyzing the contribution of the attention mechanism in the sugarcane sprout recognition task.

[0015] Furthermore, the improved convolutional neural network in the image recognition module employs data augmentation strategies during training, including sugarcane sprout image generation based on generative adversarial networks, as shown in the following formula: ,in For the generated sugarcane bud image, This is a generator network with random noise as input. , Images of real sugarcane buds. The fusion coefficient is 0.3, which is determined by evaluating the similarity between the generated image and the real image. This ensures that the generated image retains both real features and diversity, thereby expanding the training dataset and improving the model's generalization ability.

[0016] Furthermore, the dynamic growth model construction module integrates time series analysis and machine learning algorithms, and its model training loss function is: ,in The mean square error between the predicted and actual values. Predicting distribution for the model Distribution of historical data KL divergence, For model parameters The regularization term, , , The weighting coefficients are 0.6, 0.3, and 0.1, respectively, and were determined by grid search on 100 sets of simulated sugarcane sprout growth data to achieve a balance between prediction accuracy and model generalization ability.

[0017] Furthermore, the model update unit in the dynamic growth model construction module uses an online learning algorithm to update the model in real time, and its update step size formula is: ,in For the first The learning rate is updated next time; The initial learning rate is 0.01. This is the attenuation coefficient, with a value of 0.001; The number of updates is determined by testing the convergence speed of the model on real-time data, so as to enable the model to quickly adapt to new changes in sugarcane bud growth while maintaining stability.

[0018] Furthermore, the deviation of the prediction and early warning module is calculated using the following formula: ,in To represent the deviation of growth data, For the first Real-time values ​​of a time-series feature , For the first The historical mean and standard deviation of each time-series characteristic were obtained through statistical analysis of three years of sugarcane sprout growth data. For the first The weights of each time-series feature were determined using the analytic hierarchy process (AHP). For features such as daily bud growth increment and bud morphology change rate, five agricultural experts were invited to conduct pairwise comparisons and scoring. After constructing a judgment matrix, the weights were calculated, with values ​​ranging from [0.3, 0.7]. Greater than the threshold At that time, it issues an early warning of potential growth problems. The value was set to 1.5, determined through retrospective analysis of 200 abnormal growth cases.

[0019] Furthermore, when the data visualization unit in the decision support module displays data, it first categorizes and organizes the time-series characteristic data, germination progress prediction curves, and growth deviation data in the prediction and early warning results. Then, it selects the corresponding visualization form according to different data types. The time-series characteristic data uses a dynamic line chart to show the trend of change, the comparison between the predicted and actual germination progress is presented using a dual-axis bar chart, and the abnormal growth deviation early warning uses a color gradient heatmap to highlight the risk area. Finally, interactive controls are set on the display interface, allowing growers to intuitively obtain key information on the sugarcane germination status by zooming, filtering time intervals, and clicking to view data details, thus assisting in understanding the prediction and early warning results and decision suggestions.

[0020] Furthermore, when determining the priority of recommendations in the decision support module, the decision recommendation generation unit first surveys the decision preferences of at least 100 growers to determine the time weight coefficient; then it collects data on the expected occurrence time of potential problems, the expected benefits after taking a decision, and the cost of implementing the decision; finally, it combines the time weight coefficient to calculate the risk urgency index and the input-output ratio index, and then merges the two indicators according to their weights to obtain the priority of the decision recommendations, thereby providing growers with a scientific decision ranking.

[0021] Compared with existing technologies, this image recognition-based sugarcane germination status monitoring system has the following advantages:

[0022] I. This system constructs a dynamic growth model based on extracted time-series feature data, uses integrated time series analysis and machine learning algorithms to predict the germination progress of sugarcane buds, and provides timely warnings of potential growth problems when growth data deviates from the normal curve. In addition, the decision support module provides growers with intuitive data displays and scientific decision-making basis through data visualization and decision suggestion generation units, helping growers to respond to growth anomalies in a timely manner, optimize planting management strategies, and achieve precision agriculture and scientific planting.

[0023] Second, this system uses a high-resolution camera and a timed shooting device to collect time-series images of sugarcane sprouting. It then uses a deep learning algorithm based on an improved convolutional neural network to identify the images and extract the time-series feature data of the sugarcane sprouts. This automated monitoring method not only improves the accuracy of data acquisition, but also ensures the timeliness of data acquisition and the balance of system energy consumption by automatically adjusting the shooting time interval according to the growth rate of the sugarcane variety through the timed shooting device, thereby significantly improving the monitoring efficiency.

[0024] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0026] Figure 1 This is a flowchart of the operation of a sugarcane germination status monitoring system based on image recognition.

[0027] Figure 2 This is an overall flowchart of a sugarcane germination status monitoring system based on image recognition. Detailed Implementation

[0028] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0029] Example 1

[0030] Large-scale sugarcane plantations require monitoring 10 different sugarcane varieties. To ensure timely capture of bud growth changes during subsequent image acquisition while minimizing system energy consumption, at least 50 sets of historical growth experimental data for each variety were first collected. By statistically analyzing the average and variance of bud growth rates for each variety, the differences in growth patterns among different varieties were accurately determined. Subsequently, multiple linear regression analysis was used to determine the weighting coefficients balancing data acquisition timeliness and system energy consumption. Determining these coefficients effectively avoids energy waste due to frequent equipment starts caused by excessively short acquisition intervals, and also prevents missing critical growth nodes due to excessively long intervals. This, combined with the current data to be monitored, allows for a more comprehensive understanding of the differences in bud growth patterns among different varieties. The average and variance of the growth rate of each variety were used to calculate the optimal shooting interval for each variety. Early-maturing varieties grow faster, so the shooting interval was set to 4 hours to track their rapid growth process in real time. Mid- and late-maturing varieties grow relatively slowly, so the shooting interval was set to 6 hours to reduce energy consumption while ensuring monitoring effectiveness. At the same time, one high-resolution camera with an adaptive optics lens was deployed every 5 acres in the planting base. The camera was installed 1.5 meters above the sugarcane bud growth area. This density and height ensured that each camera covered the area without blind spots and avoided monitoring blind areas, laying the foundation for subsequent comprehensive collection of time-series images of sugarcane buds.

[0031] The timed imaging device controls the high-resolution camera to start shooting according to the pre-calculated optimal time interval. This operation ensures that image data can be acquired during the critical period of sugarcane bud growth, providing continuous and complete time-series image support for subsequent analysis. Before shooting, the camera automatically measures the actual distance to the sugarcane bud. Based on the nonlinear relationship model between distance and focal length fitted by a large number of experimental data on imaging clarity at different distances, the lens focal length is automatically adjusted. This process can effectively solve the imaging blur problem caused by changes in the distance between the camera and the sugarcane bud, ensuring that the time-series images captured each time are clear and distinguishable. This ensures that the subsequent image recognition module can accurately extract the features of the sugarcane bud. After shooting is completed at the set interval each day, the image acquisition module stores all time-series images according to variety and shooting time for easy retrieval and analysis later. At the same time, the images are synchronously transmitted to the image recognition module to ensure that the image data can enter the next processing stage in a timely manner, avoiding data accumulation and delaying the analysis process.

[0032] The image recognition module employs a deep learning algorithm based on an improved convolutional neural network to perform frame-by-frame recognition on the received time-series images. The formula is as follows: ,in For the first Layer Each feature map For activation functions; and The convolution kernel weights and biases; For attention mechanism functions, Using weights, this algorithm possesses strong feature extraction capabilities, accurately distinguishing sugarcane buds from the background and avoiding recognition errors caused by background interference. The algorithm highlights key areas of the sugarcane bud through an attention mechanism function, allowing the model to focus more on the growth characteristics of the bud and reduce the influence of irrelevant areas on feature extraction. Then, combining convolutional kernel weights and biases, it extracts features from the image, generating temporal feature data for each sugarcane bud at different time points, including indicators such as bud length, bud thickness, and leaf unfolding degree. These indicators comprehensively reflect the growth status of the sugarcane bud, providing core data support for the subsequent construction of a dynamic growth model. During the recognition process, the algorithm employs a data augmentation strategy based on generative adversarial networks for sugarcane bud image generation, fusing the generated sugarcane bud image with the real sugarcane bud image. The formula is: ,in For the generated sugarcane bud image, This is a generator network with random noise as input. , Images of real sugarcane buds. As the fusion coefficient, this strategy can effectively expand the training sample size, especially by supplementing images of sugarcane buds in rare growth states, thereby improving the model's accuracy in recognizing sugarcane buds in different growth states and avoiding the problem of poor model generalization ability due to insufficient samples.

[0033] After receiving time-series feature data, the dynamic growth model construction module uses an algorithm that integrates time series analysis and machine learning to construct dynamic growth models for various sugarcane varieties. The model training loss function is as follows: ,in The mean square error between the predicted and actual values. Predicting distribution for the model Distribution of historical data KL divergence, For model parameters The regularization term, , , Using weighted coefficients, time series analysis can capture the temporal correlation of sugarcane sprout growth, while machine learning algorithms can uncover complex nonlinear relationships between feature data. The fusion of the two can make the model more closely match the actual growth pattern of sugarcane sprouts and improve the model's fitting accuracy to the growth trend.

[0034] During model training, the core parameters are mean squared error, KL divergence, and model parameter regularization. These are combined with corresponding weight coefficients to calculate the loss function. Mean squared error measures the deviation between the model's predicted values ​​and the true values. KL divergence ensures that the model's predicted distribution remains consistent with historical data distribution, preventing model bias. Regularization prevents overfitting. These three parameters work together to optimize model parameters and improve prediction accuracy. The model update unit is updated in real-time using an online learning algorithm. Based on the initial learning rate, decay coefficient, and number of updates, the learning rate for each update is adjusted using the learning rate update step size formula: ,in For the first The learning rate is updated next time; The initial learning rate; The attenuation coefficient; The formula allows the learning rate to decay reasonably with each update, ensuring that the model converges quickly in the early stages of training and optimizes stably in the later stages. After collecting new time-series feature data every 24 hours, the model is automatically updated, enabling the model to adapt to environmental changes and growth stage transitions during the sugarcane bud growth process in a timely manner, ensuring that the model always has a high predictive ability.

[0035] The prediction and early warning module calls the constructed dynamic growth model to predict the germination progress of sugarcane buds for each variety, generating a germination progress prediction curve. Planting managers can use this curve to intuitively understand the growth trend of sugarcane buds over a future period and make planting plans in advance. Simultaneously, it calculates the deviation of growth data. By statistically analyzing the deviation between the real-time value and the historical mean of each time-series feature, and combining the historical standard deviation and feature weights, a comprehensive deviation value is obtained using the following formula: ,in To represent the deviation of growth data, For the first Real-time values ​​of a time-series feature , For the first Historical mean and standard deviation of each time-series feature For the first The weight of each time-series feature is quantified to reflect the degree of difference between the current growth state of sugarcane buds and the normal growth state, providing an objective basis for early warning. When the deviation value exceeds the preset threshold, the system automatically issues an early warning of potential growth problems, clearly marking the abnormal time-series features and the corresponding degree of deviation, helping managers to quickly locate the root cause of the problem. At the same time, the system pushes early warning information through the sound and light alarm device in the base monitoring center and the manager's mobile APP to ensure that managers can be informed of abnormal situations in a timely manner and avoid losses due to delayed handling.

[0036] The data visualization unit of the decision support module categorizes and organizes time-series characteristic data, germination progress prediction curves, and growth deviation data. This logical categorization makes the data presentation more logical and easier for managers to understand. Time-series characteristic data uses dynamic line charts to display trends, clearly showing the changes in various sugarcane bud characteristics over time and helping managers identify subtle changes during growth. The comparison between predicted and actual germination progress values ​​is presented using a dual-axis bar chart, providing a direct comparison of the differences between predictions and actual values, and offering a reference for evaluating model accuracy and adjusting planting strategies. The abnormal growth deviation warning uses a color gradient heatmap to highlight risk areas, quickly differentiating risk levels through color intensity, allowing managers to prioritize high-risk areas. The display interface includes interactive controls, allowing managers to zoom, filter time intervals, and click to view data details. To acquire key information, improve the flexibility and efficiency of data queries, and assist managers in deeply understanding forecast and early warning results, the decision-making suggestion generation unit first surveys 100 experienced growers to determine the time weighting coefficient. This coefficient makes the decision-making suggestions more in line with actual planting and management habits, improving the operability of the suggestions. Then, it collects data on the expected occurrence time of potential problems, the expected benefits of the decision, and the implementation costs, providing a basis for quantitatively evaluating the value of the decision. Finally, it calculates the risk urgency index and the input-output ratio index and integrates them according to weight to determine the priority of decision-making suggestions. For the growth problems warned of, it outputs specific suggestions such as irrigation adjustment, fertilization optimization, and pest and disease control, marking the implementation priority and expected effects. This helps managers rationally allocate their energy among multiple pending tasks, prioritize the execution of high-priority, high-return decisions, and improve planting management efficiency and economic benefits.

[0037] Example 2

[0038] Three specialty sugarcane varieties were tested on a small family farm. First, 50 sets of historical growth experimental data for each of the three varieties were collected. The average and variance of the sugarcane bud growth rate for each variety were calculated. Through multiple linear regression analysis, the weight coefficients for balancing the timeliness of data collection and system energy consumption were determined. Combined with the average and variance of the growth rate of the current test varieties, the optimal shooting time interval was calculated and uniformly set to 5 hours. Two high-resolution cameras with adaptive optics lenses were set up at both ends of the test area of ​​the farm to ensure coverage of all the test sugarcane.

[0039] The timed shooting device controls a high-resolution camera to shoot at a set time interval of 5 hours. Before shooting, the camera automatically measures the distance to the sugarcane buds. Based on a non-linear relationship model between distance and focal length fitted by a large amount of experimental data on image clarity at different distances, the lens focal length is automatically adjusted to ensure that the time sequence images are clear. After shooting, the image acquisition module stores the time sequence images locally and transmits them to the local server of the family farm via wireless network, and pushes them to the grower's computer terminal simultaneously.

[0040] The image recognition module employs a deep learning algorithm based on an improved convolutional neural network to recognize time-series images. The algorithm focuses on the sugarcane bud region through an attention mechanism function and extracts time-series feature data of the sugarcane bud by combining convolutional kernel weights and biases, including key indicators such as bud height, bud color, and number of buds. During the recognition process, a data augmentation strategy based on generative adversarial networks for sugarcane bud image generation is adopted to fuse generated images with real images, thereby improving the stability and accuracy of the model recognition and ensuring that the growth differences of sugarcane buds of different varieties can be accurately captured.

[0041] After receiving time-series feature data, the dynamic growth model construction module uses an algorithm that integrates time series analysis and machine learning to build a unique dynamic growth model for each trial variety. During model training, the loss function is calculated by combining mean squared error, KL divergence, and model parameter regularization terms with corresponding weight coefficients to optimize model parameters. The model update unit uses an online learning algorithm, which adjusts the update step size according to the initial learning rate, decay coefficient, and update number using the learning rate update step size formula. Every 48 hours of new time-series feature data is collected, the model is automatically updated in real time to adapt to changes in sugarcane bud growth.

[0042] The prediction and early warning module uses a dynamic growth model to predict the germination progress of sugarcane buds of various varieties and generates an intuitive germination progress prediction curve. At the same time, it calculates the deviation of growth data through a deviation calculation formula, compares the real-time value of each time-series feature with the historical mean and historical standard deviation, and obtains the comprehensive deviation by combining feature weights. When the deviation exceeds the preset threshold, the system sends early warning information through the grower's computer terminal and mobile APP, clearly informing them of abnormal characteristics and possible growth problems, such as water shortage and insufficient nutrients.

[0043] The data visualization unit of the decision support module categorizes and organizes various types of data. Time-series characteristic data is displayed using dynamic line charts to show trends. Predicted and actual germination progress values ​​are compared using biaxial bar charts. Abnormal growth deviation areas are highlighted using color gradient heatmaps. The display interface features simple and easy-to-understand interactive controls. Growers can quickly obtain key information by zooming in and out of images, filtering time ranges, and clicking to view details. The decision suggestion generation unit first surveys the decision preferences of 100 similar family farm growers to determine the time weight coefficient. Then, it collects data on the expected occurrence time of potential problems, expected benefits of decisions, and implementation costs. It calculates risk urgency indicators and input-output ratio indicators and integrates them according to weight to determine the priority of decision suggestions. It then outputs targeted decision suggestions to growers, such as adjusting watering frequency and supplementing specific fertilizers, and marks the implementation steps and expected effects of the suggestions.

[0044] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A sugarcane germination status monitoring system based on image recognition, characterized in that, The system comprises the following components: an image acquisition module, an image recognition module, a dynamic growth model construction module, a prediction and early warning module, and a decision support module; The image acquisition module: acquires time-series images of sugarcane sprouting process. The image acquisition module includes a high-resolution camera and a timed shooting device. The timed shooting device is used to control the high-resolution camera to shoot at set time intervals. The image recognition module: identifies the time-series image and extracts the time-series feature data of sugarcane buds. The image recognition module adopts a deep learning algorithm based on an improved convolutional neural network. The dynamic growth model construction module: constructs a dynamic growth model based on the time series feature data. The module adopts a fusion of time series analysis and machine learning algorithms and includes a model update unit. The prediction and early warning module: uses the dynamic growth model to predict the germination progress of sugarcane buds, and provides early warning of potential growth problems when the growth data deviates from the normal curve; The decision support module provides planting decision-making basis based on the prediction and early warning results. The module includes a data visualization unit and a decision suggestion generation unit.

2. The sugarcane germination status monitoring system based on image recognition according to claim 1, characterized in that, When determining the shooting time interval in the image acquisition module, the timed shooting device first collects at least 50 sets of historical growth experimental data of different sugarcane varieties, and calculates the average and variance of the sugarcane bud growth rate of each variety; then, it uses multiple linear regression analysis to determine the weighting coefficients that balance the timeliness of data acquisition and system energy consumption; finally, it calculates the optimal shooting time interval based on the average and variance of the growth rate of the sugarcane variety to be monitored, combined with the weighting coefficients.

3. The sugarcane germination status monitoring system based on image recognition according to claim 1, characterized in that, When adjusting the focal length of the adaptive optics lens of the high-resolution camera in the image acquisition module, the distance between the camera and the sugarcane bud is first measured; then, based on a large amount of experimental data on imaging clarity at different distances, a nonlinear relationship model between distance and focal length is fitted; finally, based on the measured distance, the lens focal length is automatically adjusted using this model.

4. The sugarcane germination status monitoring system based on image recognition according to claim 1, characterized in that, The deep learning algorithm based on an improved convolutional neural network in the image recognition module has the following core feature extraction formula: ,in For the first Layer Each feature map For activation functions; and The convolution kernel weights and biases; For attention mechanism functions, As weight.

5. The sugarcane germination status monitoring system based on image recognition according to claim 1, characterized in that, The improved convolutional neural network in the image recognition module employs data augmentation strategies during training, including sugarcane sprout image generation based on generative adversarial networks, with the following formula: ,in For the generated sugarcane bud image, This is a generator network with random noise as input. , Images of real sugarcane buds. This is the fusion coefficient.

6. The sugarcane germination status monitoring system based on image recognition according to claim 1, characterized in that, The dynamic growth model construction module integrates time series analysis and machine learning algorithms, and its model training loss function is: ,in The mean square error between the predicted and actual values. Predicting distribution for the model Distribution of historical data KL divergence, For model parameters The regularization term, , , These are the weighting coefficients.

7. The sugarcane germination status monitoring system based on image recognition according to claim 1, characterized in that, The model update unit in the dynamic growth model construction module uses an online learning algorithm to update the model in real time, and its update step size formula is as follows: ,in For the first The learning rate is updated next time; The initial learning rate; The attenuation coefficient; This represents the number of updates.

8. The sugarcane germination status monitoring system based on image recognition according to claim 1, characterized in that, The deviation of the prediction and early warning module is calculated using the following formula: ,in To represent the deviation of growth data, For the first Real-time values ​​of a time-series feature , For the first Historical mean and standard deviation of each time-series feature For the first Weights of each time-series feature.

9. A sugarcane germination status monitoring system based on image recognition according to claim 1, characterized in that, When the data visualization unit in the decision support module displays data, it first categorizes and organizes the time-series characteristic data, germination progress prediction curves, and growth deviation data in the prediction and early warning results. Then, it selects the corresponding visualization form according to different data types. The time-series characteristic data uses a dynamic line chart to show the trend of change, the comparison between the germination progress prediction and the actual value is presented using a dual-axis bar chart, and the abnormal growth deviation early warning uses a color gradient heatmap to highlight the risk area. Finally, interactive controls are set in the display interface, allowing growers to intuitively obtain key information on the sugarcane germination status by zooming, filtering time intervals, and clicking to view data details, thus assisting in understanding the prediction and early warning results and decision suggestions.

10. A sugarcane germination status monitoring system based on image recognition according to claim 1, characterized in that, When determining the priority of recommendations in the decision support module, the decision recommendation generation unit first surveys the decision preferences of at least 100 growers to determine the time weight coefficient; then it collects data on the expected occurrence time of potential problems, the expected benefits after taking a decision, and the cost of implementing the decision; finally, it calculates the risk urgency index and the input-output ratio index by combining the time weight coefficient, and then merges the two indicators according to their weights to obtain the priority of the decision recommendations.

Images