基于传统图像处理算法和YOLOv4的水位识别方法研究

Water level recognition method based on traditional image processing algorithm and YOLOv4

摘要:

水位监测是水利建设的重点问题,为及时掌握水情、预防洪涝灾害,提出了一种智能图像水位识别系统解决方案。对多种情况下的水尺图片利用传统图像算法进行图像预处理后,使用基于YOLOv4的深度学习水位识别算法,对采集的图像进行训练,实现水位自动识别。实验结果表明,基于YOLOv4的深度学习水位识别算法能够有效的通过水尺图像读取当前的水位,算法误差仅在1~2cm左右,符合工程水位监测误差要求。

Abstract:

Water level monitoring is a key issue in water conservancy construction. In order to grasp the water regime in time and prevent flood disasters, an intelligent image water level recognition system solution is proposed After using traditional image algorithms for image preprocessing of water gauge images in various situations, the deep learning water level recognition algorithm based on YOLOv4 is used to train the collected images to achieve automatic water level recognition.. Experimental results show that the deep learning water level recognition algorithm based on YOLOv4 can effectively read the current water level through the water scale image, and the algorithm error is only about 1~2cm, which meets the requirements of engineering water level monitoring error.