摘要: 将小波母函数嵌入人工神经网络的神经元形成紧致型小波神经网络,将此种网络用于遥感测温的数据拟合中,提升了纯粹的BP神经网络的拟合精度。结合红沿河核电站无人机红外测温试验,对其采集的一组温度数据采用小波神经网络进行拟合。对实验数据进行了统计分析,结果表明,小波神经网络能保证拟合误差在很小的范围之内,并且优于BP神经网络。本文中对于遥感测温数据的拟合误差控制在了0.4℃以内,可以满足测量要求。
Abstract:
The wavelet mother function is embedded in the neurons of artificial neural network to form the compact wavelet neural network. This kind of network is applied to the data fitting of remote sensing temperature measurement, which could improve the fitting accuracy of the pure BP neu- ral network. Based on the infrared temperature measurement experiment at Hongyanhe nuclear plant using unmanned aerial vehicle, the data fitting is conducted for a group of obtained tempera- ture data. The statistical analysis is performed for the experimental data, and the results show that wavelet neural network could ensure the fitting errors in a small range, which is better than the BP neural network. In this paper, the fitting error of remote sensing temperature measurement is controlled at 0.4˚C which can meet the measurement requirements.