CN-121982702-A - Sugarcane germination state monitoring system based on image recognition

CN121982702ACN 121982702 ACN121982702 ACN 121982702ACN-121982702-A

Abstract

The invention discloses a sugarcane germination state monitoring system based on image recognition, which relates to the technical field of agricultural intelligent monitoring and comprises an image acquisition module, an image recognition module, a dynamic growth model construction module, a prediction early warning module and a decision support module; according to the invention, a dynamic growth model is constructed based on the extracted time sequence characteristic data, the germination progress of sugarcane buds is predicted by utilizing a fusion time sequence analysis and machine learning algorithm, and potential growth problems are early warned in time when growth data deviate from a normal curve.

Inventors

LI MING
JING YAN
FANG WEIKUAN
He Panshan
LUO TING
WU XIAOQING
ZHOU HUI
YAN HAIFENG

Assignees

广西壮族自治区农业科学院

Dates

Publication Date: 20260505
Application Date: 20260114

Claims (10)

1. The sugarcane germination state monitoring system based on image recognition is characterized by comprising an image acquisition module, an image recognition module, a dynamic growth model construction module, a prediction early warning module and a decision support module; The image acquisition module is used for acquiring time sequence images in the sugarcane germination process, and comprises a high-resolution camera and a timing shooting device, wherein the timing shooting device is used for controlling the shooting of the high-resolution camera according to a set time interval; the image recognition module is used for recognizing the time sequence image and extracting time sequence characteristic data of sugarcane buds, and adopts a deep learning algorithm based on an improved convolutional neural network; The dynamic growth model construction module is used for constructing a dynamic growth model based on the time sequence characteristic data, adopts a fusion time sequence analysis and machine learning algorithm and comprises a model updating unit; The prediction and early warning module predicts the germination progress of the sugarcane buds by using the dynamic growth model and early warns potential growth problems when growth data deviate from a normal curve; the decision support module provides a planting decision basis based on a prediction early warning result, and comprises a data visualization unit and a decision suggestion generation unit.
2. The sugarcane germination state monitoring system based on image recognition according to claim 1, wherein when a timing shooting device in an image acquisition module determines shooting time intervals, historical growth experimental data of at least 50 groups of different sugarcane varieties are collected, average values and variances of growth rates of sugarcane buds of the varieties are counted, the data are analyzed through multiple linear regression to determine weight coefficients of balance data acquisition timeliness and system energy consumption, and finally, the optimal shooting time intervals are calculated according to the average values and variances of the growth rates of the sugarcane varieties to be monitored currently and in combination with the weight coefficients.
3. The sugarcane germination state monitoring system based on image recognition according to claim 1, wherein when the self-adaptive optical lens of the high-resolution camera in the image acquisition module adjusts the focal length, the distance between the camera and sugarcane buds is measured firstly, then a nonlinear relation model of the distance and the focal length is fitted according to imaging definition experimental data under a large number of different distances, and finally the focal length of the lens is automatically adjusted according to the measured distance by using the model.
4. The sugarcane germination state monitoring system based on image recognition according to claim 1, wherein the image recognition module is based on a deep learning algorithm of an improved convolutional neural network, and the core feature extraction formula is as follows: wherein Is the first Layer number The characteristic map of the image is shown in the figure, Is an activation function; And Weights and offsets for convolution kernels; as a function of the mechanism of attention, Is the weight.
5. The sugarcane germination status monitoring system based on image recognition according to claim 1, wherein the improved convolutional neural network in the image recognition module adopts a data enhancement strategy in the training process, and comprises sugarcane germination image generation based on a generation countermeasure network, wherein the formula is: wherein In order to generate the sugarcane bud image, The generator network is input as random noise , Is the image of the real sugarcane buds, Is the fusion coefficient.
6. The sugarcane germination status monitoring system based on image recognition according to claim 1, wherein the dynamic growth model construction module fuses a time sequence analysis and a machine learning algorithm, and a model training loss function thereof is as follows: wherein For the mean square error of the predicted value and the true value, Predicting distribution for model With historical data distribution Is used for the distribution of KL of the formula (I), Is a model parameter Is used to determine the regularization term of (c), 、、 Is a weight coefficient.
7. The sugarcane germination state monitoring system based on image recognition as claimed in claim 1, wherein the model updating unit in the dynamic growth model construction module adopts an online learning algorithm to update the model in real time, and the update step formula is as follows: wherein Is the first The learning rate of the secondary update; Is the initial learning rate; is the attenuation coefficient; For the number of updates.
8. The sugarcane germination state monitoring system based on image recognition according to claim 1, wherein the deviation calculation of the predictive early warning module is calculated by the following formula: wherein For the degree of deviation of the generated data, Is the first The real-time values of the individual timing characteristics, 、 Is the first The historical mean and standard deviation of the individual timing characteristics, Is the first Weights for the timing characteristics.
9. The sugarcane germination state monitoring system based on image recognition according to claim 1 is characterized in that when a data visualization unit in a decision support module achieves data display, time sequence feature data, a germination progress prediction curve, growth deviation data and the like in a predicted early warning result are classified and sorted, then corresponding visualization forms are selected according to different data types, the time sequence feature data are displayed in a dynamic line graph mode to show a change trend, germination progress prediction is compared with an actual value and is displayed in a double-shaft histogram mode, abnormal growth deviation early warning is conducted by adopting a color gradient thermodynamic diagram to highlight a risk area, finally an interactive control is arranged on a display interface to support a planter to intuitively obtain key information of sugarcane germination states through scaling, screening time intervals and clicking and checking data detail operation, and prediction early warning results and decision suggestions are assisted to be understood.
10. The sugarcane germination state monitoring system based on image recognition according to claim 1, wherein when the decision suggestion generation unit in the decision support module determines suggestion priority, firstly, the decision preference of at least 100 growers is studied to determine time weight coefficients, then the expected occurrence time of potential problems, expected benefits after decision making and decision implementation cost data are collected, finally, risk urgency indexes and input-output ratio indexes are calculated respectively by combining the time weight coefficients, and then the two indexes are fused according to weight to obtain the priority of the decision suggestion.

Description

Sugarcane germination state monitoring system based on image recognition Technical Field The invention relates to the technical field of agricultural intelligent monitoring, in particular to a sugarcane germination state monitoring system based on image recognition. Background Along with the acceleration of the modern agricultural process, accurate agriculture and intelligent monitoring technology become key means for improving the yield and quality of crops, and sugarcane is taken as an important cash crop, and the germination state of the sugarcane directly influences the subsequent growth cycle and the final yield. The traditional sugarcane germination state monitoring method mainly has the defects that firstly, the monitoring result is greatly influenced by subjective factors and possibly has large difference among different observers due to the fact that manual observation and experience judgment are relied on, secondly, the real-time and continuous monitoring of the growth state of the sugarcane is difficult to achieve, the optimal management time is easy to miss, thirdly, the traditional method lacks scientific data analysis and prediction capability, the germination progress and the potential growth problem of the sugarcane cannot be accurately predicted, and therefore an effective planting management strategy is difficult to formulate, and finally, under a large-scale planting scene, the manual monitoring is high in cost and low in efficiency, and the requirements of modern agriculture on efficient and accurate management of the sugarcane cannot be met, so that the traditional technology has difficulty in meeting the requirements of modern planting management. Aiming at the defects of the traditional sugarcane germination state monitoring method, the invention provides a sugarcane germination state monitoring system based on image recognition. Disclosure of Invention The invention aims to make up the defects of the prior art and provides a sugarcane germination state monitoring system based on image recognition, which can collect time sequence images in the sugarcane germination process through a high-resolution camera and a timing shooting device, identify the images by using a deep learning algorithm based on an improved convolutional neural network, extract time sequence characteristic data of sugarcane buds, further, the system predicts the germination progress of the sugarcane buds by constructing a dynamic growth model and utilizing a fusion time sequence analysis and machine learning algorithm, and early warn potential growth problems in time when growth data deviate from a normal curve. The invention provides a sugarcane germination state monitoring system based on image recognition, which comprises an image acquisition module, an image recognition module, a dynamic growth model construction module, a prediction early warning module and a decision support module; The image acquisition module is used for acquiring time sequence images in the sugarcane germination process, and comprises a high-resolution camera and a timing shooting device, wherein the timing shooting device is used for controlling the shooting of the high-resolution camera according to a set time interval; the image recognition module is used for recognizing the time sequence image and extracting time sequence characteristic data of sugarcane buds, and adopts a deep learning algorithm based on an improved convolutional neural network; The dynamic growth model construction module is used for constructing a dynamic growth model based on the time sequence characteristic data, adopts a fusion time sequence analysis and machine learning algorithm and comprises a model updating unit; The prediction and early warning module predicts the germination progress of the sugarcane buds by using the dynamic growth model and early warns potential growth problems when growth data deviate from a normal curve; the decision support module provides a planting decision basis based on a prediction early warning result, and comprises a data visualization unit and a decision suggestion generation unit. Further, when the timing shooting device in the image acquisition module determines a shooting time interval, historical growth experimental data of at least 50 groups of different sugarcane varieties are collected, average values and variances of growth rates of sugarcane buds of the varieties are counted, the data are analyzed through multiple linear regression to determine weight coefficients for balancing data acquisition timeliness and system energy consumption, and finally an optimal shooting time interval is calculated according to the average values and variances of the growth rates of the sugarcane varieties to be monitored currently and the weight coefficients. Further, when the self-adaptive optical lens of the high-resolution camera in the image acquisition module adjusts the focal length, the distance between the camera and the sugarcane