基于数据融合的日光温室传感器布设

针对北方日光温室内环境监测传感器的布设问题,设计了基于神经网络、分批估计理论与自适应加权平均融合算法的日光温室传感器布设方案.利用11个监测点采集到的温室内西红柿生长环境温度和湿度数据,在运用BP神经网络进行缺失值补全的基础上,结合分批估计理论和自适应加权平均融合算法进行多传感器数据融合.通过对比融合值与原始数据的相对误差,选择最佳传感器数据,以此为基础确定最优传感器布设区域.结果表明,相对于算数平均融合与自适应加权平均融合,基于分批估计的自适应加权平均融合方法可以更合理地反应多传感器数据特征.

Layout of sensors in sunlight greenhouse based on data fusion

Aiming at the problem of environmental monitoring sensor layout in north sunlight greenhouse, a layout scheme for sensors based on the neural network, batch estimation theory and adaptive weighted average fusion algorithm was designed. With the temperature and humidity data for the tomato growing environment collected from 11 monitoring points in the greenhouse and based on the complement of missing values with the BP neural network, the multi-sensor data fusion was performed in combination with the patch estimation theory and adaptive weighted average fusion algorithm. Through comparing the relative error between the fusion value and original data, the optimal sensor data were selected, and thus the optimal sensor layout area was determined. The results show that compared with the arithmetic mean fusion and adaptive weighted average fusion, the adaptive weighted average fusion method based on the batch estimation can reflect the characteristics of multi-sensor data more reasonably.