一种基于图像的全自动河道水位估计方法
An Image-Based Automatic River Level Measurement Method

DOI: 10.12677/CSA.2021.119225, PP. 2199-2205

Keywords: 图像，水位估计，Faster R-CNN网络，霍夫变换
Image, Water Level Estimation, Faster R-CNN Net, Hough Transform

Full-Text   Cite this paper   Add to My Lib

Abstract:

针对水尺图像人工读数不及时、耗费人力的问题，本文提出了一种基于图像的全自动河道水位估计方法。首先使用Faster R-CNN网络检测水尺区域图像，然后对水尺图像进行倾斜调正，最后使用图像处理技术获得水位信息。该方法通过霍夫变换对水尺图像进行直线检测，根据检测到的直线来实现水尺图像的自动调正，有效解决了由于摄像机拍摄角度不同导致的图像中刻度尺倾斜程度不同的问题；提出一种根据先验知识修正刻度尺图像的思路，根据已知的刻度尺相关信息来自查通过像素级计算得到的刻度尺矩阵是否合理，有效解决了刻度尺图像可能存在的刻度尺磨损、拍摄反光等问题，大大提高了刻度尺读数的可靠性。实验结果表明，本文所提出的河道水位估计方法可以在多种拍摄条件下准确地获得河道的水位信息。
An image-based automatic river level measurement method is proposed in this paper, in order to solve the problem of poor real-time performance and labor-intensive cause of manual measurement. First, the Faster R-CNN network is used to detect the water level gauge image. Secondly, adjust the tilt of the water level gauge image. Finally, the water level information is obtained through image processing technology. In particular, because of the different camera shooting angles, the tilt degree of the water level gauge image is different. In order to solve the problem, an automatic adjustment of the water level gauge image by straight line detection is used in this method. Furthermore, an idea of correcting the water level gauge image based on prior knowledge was proposed. According to the known appearance information of the water level gauge, check whether the matrix calculated with pixel level is reasonable. This method can effectively solve the problem of unclear images caused by the scale wear and the reflection of the shooting, which greatly improves the measurement reliability. Experimental results show that the river water level measurement method proposed in this paper can accurately obtain river water level information under a variety of shooting conditions.

References

[1]	安全, 范瑞琪. 常用水位传感器的比较和选择[J]. 水利信息化, 2014(3): 52-54.
[2]	陈贺璋, 庄静. 基于图像分割的舰船吃水深度检测方法研究[J]. 舰船科学技术, 2017, 39(16): 13-15.
[3]	刘晓杰, 张旻, 罗印升, 薛波. 图像处理技术在船舶水尺刻度线检测中的应用[J]. 舰船科学技术, 2017, 39(16): 10-12.
[4]	冉鑫, 彭将辉. 基于图像处理的船舶水尺刻度识别[J]. 上海海事大学学报, 2012, 33(2): 6-9.
[5]	江帆, 陈德为. 基于图像识别技术的船舶水尺刻度线检测算法的研究[J]. 数字技术与应用, 2013(1): 114-117.
[6]	罗婧, 施朝健, 冉鑫. 一种视频图像船舶吃水线自动检测方法[J]. 船海工程, 2012, 41(1): 30-32.
[7]	朱学海, 张帅, 张东星, 张阿平, 罗陨飞. 基于机器视觉与深度学习的船舶水尺智能识别技术研究与应用[J]. 检验检疫学刊, 2019, 29(2): 101-104.
[8]	Ren, S.Q., et al. (2017) Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks. IEEE Trans-actions on Pattern Analysis and Machine Intelligence, 39, 1137-1149. https://doi.org/10.1109/TPAMI.2016.2577031
[9]	Ballard, D.H. (1981) Generalizing the Hough Transform to Detect Arbitrary Shapes. Pattern Recognition, 13, 111-122. https://doi.org/10.1016/0031-3203(81)90009-1
[10]	Rosenthal, D.L., et al. (1979) A Simple Method of Producing a Monolayer of Cervical Cells for Digital Image Processing. Analytical and Quantitative Cytology, 1, 84-88.

Full-Text