城市河道水资源是重要的生态资源, 近年来城市工业的不断发展, 导致河道水污染问题日益突出。 治理水污染的前提条件是及时掌握河道水质状况(即水质类型), 传统采样化验的监测方法精度高, 但较费时费力, 该研究提出一种基于光谱二阶微分波动指数的水质类型快速识别方法, 可实现城市河道水质类型的快速排摸。 首先使用光谱差分计算得到光谱二阶微分曲线, 对曲线进行平滑处理消除噪声和其他干扰; 之后使用滑动窗口提取曲线上的局部极大值和极小值点, 并设置最小距离阈值逐步去除虚假极值点, 再使用三次样条插值法得到包含光谱二阶微分曲线的双包络线, 最后利用上下包络线计算得到光谱二阶微分的波动指数曲线。 通过对各类样本的波动曲线分析后发现, 在720～740, 750～770以及820～840 nm处各类水体的波动指数差异较大, 随后统计了各类水体光谱在这三波段内的平均波动指数的均值、 标准差等统计特征, 发现平均波动指数与水质级别具有正向相关关系, 水质级别越高水质状况越差, 平均波动指数也越大。 为验证光谱二阶微分波动指数可用于城市河道水质的快速识别, 将各类水体光谱样本随机划分为训练集和测试集, 结合LSSVM(least-squares support vector machine)构建水质类型识别模型, 将上述三个特征波段的平均波动指数作为特征输入, 经测试, 各类样本的平均识别准确率达80.65%, 误差不超过一类的识别精度超过95%。 基于光谱二阶微分波动指数的河道水质高光谱识别方法识别精度较高, 可作为城市河道水质类型快速排摸的一种辅助技术手段。

Urban river water resources are important ecological resources. In recent years, the continuous development of urban industry has led to the increasingly prominent problem of river water pollution. The traditional sampling and testing methods have high precision, but it is time-consuming and laborious. This study proposes a rapid identification of water quality types based on the second-order differential fluctuation index of the spectrum. The method can realize rapid displacement of water quality types in urban rivers. The method first uses the spectral difference calculation to obtain the second-order differential curve of the spectrum, and smoothes the curve to eliminate noise and other disturbances; then uses the sliding window to extract the local maximum and minimum values on the curve, and sets the minimum distance threshold to gradually remove The extreme false point is obtained by using the cubic spline interpolation method to obtain the double envelope of the second-order differential curve of the spectrum. Finally, the fluctuation index curve of the second-order differential of the spectrum is calculated by using the upper and lower envelopes. After analyzing the fluctuation curves of various samples, it is found that the fluctuation indexes of various water bodies are different at 720~740, 750~770 and 820~840 nm, and then the average fluctuations of various water bodies in these three bands are counted. Statistical characteristics such as the mean value and standard deviation of the index show that the average fluctuation index has a positive correlation with the water quality level. The higher the water quality level, the worse the water quality condition and the larger the average fluctuation index. In order to verify the second-order differential fluctuation index of the spectrum, it can be used for the rapid identification of urban river water quality. The water body spectral samples are randomly divided into training sets and test sets, and LSSVM is used to construct the water quality type, identification model. The average fluctuation index is used as the character input. After testing, the average recognition accuracy of each type of sample is 80.65%, and the recognition accuracy of no more than one type exceeds 95%. The high-spectral identification method of river water quality based on spectral second-order differential fluctuation index proposed by this study has high recognition accuracy and can be used as an auxiliary technical means for rapid detection of urban river water quality types.