基于动态方差卡尔曼滤波的高洪期河流流量监测研究

Research on river fl ow monitoring during high fl ood period based on dynamic variance Kalman fi ltering

摘要:

为解决传统高洪期河流流量监测方法精度低、抗干扰能力弱等问题，通过高清摄像头非接触式采集河流水面图像，利用 Sobel 算法提取水位线，采用光流法分析水面像素的时空运动特征，将光流矢量转换为实际流速，实现局部流速估计。以流量和水位为状态变量，区域流速和水位为观测值，构建状态空间模型，并利用动态噪声协方差在线更新机制，通过融合多源观测数据与卡尔曼滤波迭代方法，对河流流量进行估计。实验结果表明：采用所提方法进行高洪期河流流量监测时，漂移累积率稳定在 1.5% 以下，能够实现从水位线提取、局部流速估计到全局流量反演的全流程优化。动态方差卡尔曼滤波算法的应用可有效提升流量监测的精度与稳定性，为高洪期河流流量监测提供新的有效途径。

Abstract:

To address the problems of low accuracy and weak anti-interference ability of traditional river flow monitoring methods during the high fl ood period，high-defi nition cameras were used to non-contact collect river surface images. The Sobel algorithm was applied to extract water level lines，and the optical flow method was employed to analyze the spatiotemporal motion characteristics of water surface pixels. Optical fl ow vectors were converted into actual fl ow velocities to achieve local velocity estimation. A state-space model was constructed with fl ow rate and water level as state variables，and regional fl ow velocity and water level as observations，incorporating a dynamic noise covariance online update mechanism. River flow was estimated by integrating multi-source observation data with iterative Kalman fi ltering operations. Experimental results showed that when the proposed method was applied to river fl ow monitoring during high fl ood period，the drift accumulation rate remained stable below 1.5%，enabling full-process optimization from water level extraction and local velocity estimation to global flow inversion. The application of the dynamic variance Kalman filtering algorithm effectively enhances the accuracy and stability of fl ow monitoring，providing a new and eff ective approach for river fl ow monitoring during high fl ood period.