偏振光太阳位置确定人工神经网络。
Polarized light sun position determination artificial neural network.
出版信息
Appl Opt. 2022 Feb 20;61(6):1456-1463. doi: 10.1364/AO.453177.
Our previous work has constructed a polarized light orientation determination (PLOD) artificial neural network. Although a PLOD network can determine the solar azimuth angle, it cannot determine the solar elevation angle. Therefore, this paper proposes an artificial neural network for polarized light solar position determination (PLSPD), which has two branches: the solar azimuth angle determination branch and the solar elevation angle determination branch. Since the solar elevation angle has no cyclic characteristics, and the angle range of the solar elevation angle is different from that of the solar azimuth angle, the solar elevation angle exponential function encoding is redesigned. In addition, compared with the PLOD, the PLSPD deletes a local full connection layer to simplify the network structure. The experimental results show that the PLSPD can determine not only the solar azimuth angle but also the solar elevation angle, and the solar azimuth angle determination accuracy of the PLSPD is higher than that of the PLOD.
我们之前的工作构建了一个偏振光方位确定（PLOD）人工神经网络。虽然 PLOD 网络可以确定太阳方位角，但它不能确定太阳高度角。因此，本文提出了一种用于偏振光太阳位置确定（PLSPD）的人工神经网络，它有两个分支：太阳方位角确定分支和太阳高度角确定分支。由于太阳高度角没有周期性特征，并且太阳高度角的角度范围与太阳方位角的角度范围不同，因此重新设计了太阳高度角的指数函数编码。此外，与 PLOD 相比，PLSPD 删除了一个局部全连接层，以简化网络结构。实验结果表明，PLSPD 不仅可以确定太阳方位角，还可以确定太阳高度角，并且 PLSPD 的太阳方位角确定精度高于 PLOD。
Zotero 插件