基于无人机高光谱和BP神经网络的城市水体污染监测

我国大多数流域存在不同程度的水体污染，城市水体污染防治与监测是一项艰巨而漫长的任务，而传统水质监测和卫星遥感方法在水体面积较大、水流运动不稳定、周边地形复杂的河流或城市湖泊的水质监测表现为适用性差、准确度低。基于无人机的高光谱遥感技术具有覆盖范围广、数据获取快速等特点，对城市水体污染监测具有一定的应用价值。以珠海市城市水域为研究对象、无人机高光谱数据为数据源，利用线性回归模型和BP (Back Propagation) 神经网络模型方法，分别建立了波段组合反射率与水体叶绿素a、氨氮和磷酸盐3种水质指标之间的最优反演模型，并通过实际样品验证了该模型在城市水体中的适用性。该研究结果不仅为大数据驱动的水质分析提供了重要的技术支持，也为无人机技术应用于城市水体污染程度评价和动态监测提供新方法。

Water pollution monitoring based on unmanned aerial vehicle (UAV) hyperspectral and BP neural network

Water pollution is a common issue to most basins in China. Pollution prevention and monitoring of urban water are an arduous and lengthy task. It is a great challenge to conduct water quality monitoring in rivers or urban lakes with large surface areas, unstable water flow, complex surrounding terrain by traditional water quality monitoring and satellite remote sensing for their poor applicability and low accuracy. Unmanned Aerial Vehicle (UAV)-based hyperspectral remote sensing technology has wide coverage and fast data acquisition, and is applicable to urban water pollution monitoring. The present study takes urban waters of Zhuhai City as the research object, and uses UAV hyperspectral data as the data source. By processing the hyperspectral images, a linear model and a Back Propagation (BP) neural network model were established to simulate the optimal mathematical mapping between the combined reflectance of the waveband and the key water quality parameters (chlorophyll a, ammonium, phosphate). The applicability of the models in urban water bodies was demonstrated by actual samples, providing a new method for urban water pollution evaluation and dynamic monitoring. The findings not only provide important technical support for big-data-driven water quality analysis, but also for urban water pollution monitoring using UAV remote sensing technology.