Research And Design Of Plant Growth Stage Monitoring System Based On Convolutional Neural Network

Plant growth monitoring is one of the most important aspects of smart agriculture.However,traditional plant monitoring methods mainly rely on manual work,which not only increases the cost,but also makes the monitoring results subject to subjective influence.With the rapid development of science and technology,agricultural planting is transforming from artificial to automatic.Plant growth condition monitoring including contact type and noncontact,in the process of contact measurement of plant is vulnerable to damage,so the plant growth is proposed based on the non-contact monitoring method in this paper,the condition monitoring system,using convolution neural network and migration study to identify its growth stage and by using image processing techniques to detect the growth condition,the main research contents include:To solve the problem of obtaining the data set of plant growth stage,an image acquisition system of the data set was constructed in this paper.According to the growth characteristics of plants,the data set of growth stage identification was made for the experiment.In this paper,a neural network is proposed based on image generation technology and transfer learning,including image generation network and image recognition network.Image generation network uses generative adversation network DCGAN to increase the number of data sets and reduce the degree of overfitting.The image recognition network adopts transfer learning and improves the transfer learning framework to improve the recognition accuracy.During the recognition model training at the plant growth stage,the SGD-M optimizer has the problem of local optimum oscillation,which will lead to the iteration falling into the local optimum during the training process,and finally,the accuracy is difficult to increase.To solve this problem,based on the internal relationship between curvature radius and gradient descent,two optimizer optimization algorithms,SGD-MS and SGD-MA,are proposed in this paper.Experimental results show that the improved algorithm can effectively solve the above problems.In this paper,based on LoRa transmission technology of Internet of Things and mechanical structure construction method of monitoring mechanism,a plant growth state monitoring system was built,and a non-contact plant monitoring method was designed.Combined with software and hardware technology,nondestructive monitoring of plant growth state was realized.Combined with the above methods,a plant growth stage monitoring system was constructed in this paper,which could realize the functions of real-time monitoring of plant environment,identification and parameter measurement of plant growth stage.It provides a reference for monitoring the growth status of plants with similar phenotypes.It is of great significance to the development of plant research and intelligent agriculture.