1 中科星图智慧科技有限公司 行业应用事业部，山东 青岛 266000
2 中科星图股份有限公司 行业应用事业部，北京 101300
3 生态环境部卫星环境应用中心 数据中心， 北京 100094
研究采用卫星遥感技术获取高分辨率遥感影像水体样本数据集，基于深度卷积神经网络从高分辨遥感影像中提取水体并进行黑臭水体智能监测，提出了一种改进U-Net的黑臭水体检测网络模型(IWDNet)。基于U-Net结构引入跳跃式多尺度特征融合，结合通道注意力机制、卷积注意力模块、通道与空间注意力机制生成不同多尺度特征融合注意力机制(MFFAM)模块进行对比，并引入空洞卷积扩大网络感受野，最终实现黑臭水体的识别检测。实验证明: 基于跳跃式多尺度融合与CBAM注意力机制的黑臭水体检测网络(MFFCBAM-IWNet)模型有效提升了识别精度，在高分辨遥感影像水体样本数据集上表现最佳，总体精度达98.56%，Kappa系数达0.9784。
神经网络 遥感影像 黑臭水体检测 neural network remote sensing image impaired water detection
Author Affiliations
Abstract
National Key Laboratory of Tunable Laser Technology, Harbin Institute of Technology, Harbin 150001, China
We report a monolithic Tm:YLF micro laser in this Letter. In order to improve the relaxation oscillation of the laser, both ends of the crystal are coated, making the Tm:YLF crystal itself a resonant cavity. The micro laser is pumped by a 792 nm laser diode operated in the continuous wave (CW) mode. We obtain maximum output powers of 7.78 and 10.4 W at the total incident power of 43.6 W with focus lenses of 37.5 and 40 mm, respectively, corresponding to the slope efficiencies of 25.6% and 40.0% and the optical–optical conversion efficiencies of 17.8% and 23.8%. It is clear that the amplitude of the relaxation oscillation is smaller and the beam quality is better with the focus length of 37.5 mm; however, the laser with the focus length of 40 mm produces a higher output power and a more stable wavelength centering at 1878.44 nm.
140.3480 Lasers, diode-pumped 140.3425 Laser stabilization 140.3430 Laser theory Chinese Optics Letters
2016, 14(6): 061401
