Abstract:
Mountainous regions are prone to dammed lake disasters due to their rough topography, scant vegetation, and high summer rainfall. By measuring water level variation, monitoring systems can detect dammed lake events when mudslides block rivers or boost water level. Therefore, an automatic monitoring alarm method based on a hybrid segmentation algorithm is proposed. The algorithm uses the k-means clustering algorithm to segment the picture scene in the RGB color space and the region growing algorithm on the image green channel to select the river target from the segmented scene. The pixel water level variation is used to trigger an alarm for the dammed lake event after the water level has been retrieved. In the Yarlung Tsangpo River basin of the Tibet Autonomous Region of China, the proposed automatic lake monitoring system was installed. We pick up data from April to November 2021, during which the river experienced low, high, and low water levels. Unlike conventional region growing algorithms, the algorithm does not rely on engineering knowledge to pick seed point parameters. Using our method, the accuracy rate is 89.29% and the miss rate is 11.76%, which is 29.12% higher and 17.65% lower than the traditional region growing algorithm, respectively. The monitoring results indicate that the proposed method is a highly adaptable and accurate unmanned dammed lake monitoring system.
摘要:
山区地形粗糙,极易发生堰塞湖灾害,很少的植被,夏季降雨量高。通过测量水位变化,当泥石流阻塞河流或提高水位时,监测系统可以检测堰塞湖事件。因此,提出了一种基于混合分割算法的自动监控报警方法。该算法使用k-means聚类算法在RGB颜色空间中分割图片场景,在图像绿色通道上使用区域生长算法从分割的场景中选择河流目标。像素水位变化被用于在已经检索到水位之后触发针对堰塞湖事件的警报。在中国西藏自治区雅鲁藏布江流域,拟议的自动湖泊监测系统已安装。我们收集了2021年4月至11月的数据,在此期间河流经历了低谷,高,低水位。与传统的区域生长算法不同,该算法不依赖于工程知识来拾取种子点参数。使用我们的方法,准确率为89.29%,漏检率为11.76%,比传统的区域生长算法高29.12%,低17.65%,分别。监测结果表明,该方法是一种适应性强、准确度高的无人堰塞湖监测系统。