基于图像匹配的库水位变动识别研究

Research on reservoir water level change identification based on image matching

针对库水位传统量测方法中水尺易锈蚀倾斜,精度低且成本高的问题,该文提出一种不依赖水尺,基于图像匹配的库水位变动识别方法。首先对监控相机拍摄的库坝上游面光学图像进行畸变消除和透视变换,消除相机误差。进一步选取包含水面的感兴趣区域(region of interest,ROI)图像进行自适应二值化分割、形态学处理等前处理操作,将水面和库坝特征分离,突出水位线位置。最后,利用归一化互相关匹配算法(normalised cross correlation,NCC)对水位变动前后的两幅图像进行匹配计算,识别水位线变动距离。通过室内试验与现场测试验证上述方法实用性。结果表明:基于图像匹配的库水位动态识别方法可准确识别水位变动,相对误差约为5%,此算法鲁棒性较高。该研究可为库水位自动化、低成本监测提供一种新思路。

The traditional measurement methods of reservoir water level are limited by corrosion and tilting,low frequency and high cost.To address such problems,this paper proposes a method to identify the changes of reservoir water level based on image matching without relying on the water level ruler.Firstly,the optical image of the upstream surface of the dam taken by the monitoring camera is distortion-removed and perspective-transformed to eliminate camera errors.Subsequently,the image of the region of interest(ROI)containing the water surface is selected for pre-processing operations such as adaptive binarization and morphological processing to separate the water surface and reservoir dam features and highlight the water level line location.Finally,the normalised cross correlation(NCC)matching algorithm is used to match the two images before and after the water level change to identify the distance of water level change.The practicality of the above proposed method is verified through indoor experiments and field tests.The results show that the dynamic identification method based on image matching can accurately identify water level changes with a relative error of about 5%,and the algorithm has high robustness.This study can provide a new method for automated and low-cost monitoring of reservoir water levels.