【摘要】 为解决传统的遥感数据融合模型在高异质性地表区域的融合精度不高的问题,本研究在考虑地表温度的前提下构建了一种新的多源遥感数据时空融合模型。首先,为消除反演得到的可见光与红外信息出现不对等的影响,将地表测温空间尺度为综合多地表参数的数据尺度模型。然后利用低尺度低空间分辨率的转换残差模拟地表温度。由于地表温度特征复杂,构建BP神经网络,并根据实际情况调整网络结构与初始参数值。然后融合光学遥感数据与SAR数据,根据高、低空间的分辨率影像计算、反演遥感地表参数,最后将尺度转换模型应用于预测时间相位的参数融合过程中。仿真实验结果表明:上述设计的模型的融合精度要比传统模型略高,证明该模型实现了设计初衷。更多还原

【Abstract】 The fusion accuracy of traditional remote sensing data fusion models is not high in highly heterogeneous surface areas. In this study, a new spatiotemporal fusion model of multi-source remote sensing data was constructed under the premise of considering the land surface temperature. First of all, in order to eliminate the influence of non-equivalence of visible light and infrared information obtained by inversion, the spatial scale of surface temperature measurement should be transformed, so as to obtain the data scale conversion model integrating multiple surface parameters. Then the surface temperature is simulated by using the conversion residuals at low scale and low spatial resolution. Because of the complex characteristics of surface temperature, the BP neural network was constructed, and the network structure and initial parameter values were adjusted according to the actual situation. Then optical remote sensing data and SAR data are fused to calculate and invert remote sensing surface parameters according to high and low spatial resolution images. Finally, the scale transformation model is applied to predict the parameter fusion process of time and phase. The simulation results show that the fusion accuracy of the above model is slightly higher than that of the traditional model, which proves that the model has realized the original intention of the design.更多还原