基于卡尔曼滤波融合的改进神经网络油菜成熟度预测方法

Improved neural network prediction method of vegetables maturity based on Kalman filtering fusion

油菜在智能植物生长柜中的生长发育过程会发生显著的形态变化。准确掌握油菜成熟度对调节智能植物生长柜环境参数设置、节约资源能源具有重要的意义。本文利用图像分割和边缘检测技术来提取冠层叶面积、株高和根系长度、根系侧面积等形态特征,分别建立神经网络模型并对其特征参数进行训练,实现对蔬菜成熟度的预测。提出基于卡尔曼滤波的成熟度预测信息融合方法,将预测准确性提高到95.5%。

Modality of rape in the plant growth cabinet will change significantly in the process of growth.The maturity of rape is important for governor to adjust the parameter and saving the resource.In this paper,we extract leaf crown projected area,plant height,root length and root side area and other data of external morphological characteristics through image segmentation and edge detection techniques.Then we train the characteristic parameters of aboveground and underground parts to realize the prediction of the vegetable maturity by neural network.Finally,we improve the accuracy of prediction to 95.5% with the use of Kalman filter.