标准双色水尺的图像法水位测量  

Image-based water level measurement with standard bicolor staff gauge

作　　者：张振[1] 周扬 王慧斌[1] 高红民[1] 刘海韵[1] Zhang Zhen;Zhou Yang;Wang Huibin;Gao Hongmin;Liu Haiyun(College of Computer and Information Engineering,Hohai University,Nanjing 211 I00,China)

机构地区：[1]河海大学计算机与信息学院

出　　处：《仪器仪表学报》2018年第9期236-245,共10页Chinese Journal of Scientific Instrument

基　　金：国家重点研发计划(2017YFC0405703);国家自然科学基金青年基金(51709083);中央高校基本科研业务费专项(2017B16914);江苏省自然科学基金青年基金(BK20170891)项目资助

摘　　要：图像法水位测量利用图像处理技术自动检测水位线并识别水尺读数,具有非接触测量、无温漂、结果可追溯、系统造价低等优点。然而受到户外视频监控系统在拍摄视角倾斜、成像分辨率低、光照条件复杂等方面的制约,水尺刻度及字符的检测和识别实际上存在较大困难,导致现有方法的稳定性和适宜性差。针对上述问题,从摄影测量的角度提出了一种解决方案。首先根据标准双色水尺的样式设计模板图像;然后通过人工选取的控制点建立感兴趣区域和模板图像间的透视投影变换关系,将水尺图像配准到正射坐标系下;接下来根据配准图像中设置的采样窗口计算自适应分割阈值将其二值化;最终在二值图像的水平投影曲线中检测水位线,并根据模板图像的物理分辨率将其坐标换算为水位测量值。在不同条件下开展了4组现场试验。结果表明,本方法对于日夜光照下拍摄的低分辨率图像均具有较好的鲁棒性,测量分辨率和精度分别达到1 mm和1 cm,满足水文测验的要求。Image-based water level measurement method automatically detects the water line and recognizes its reading on the staff gauge with image processing technique. It has the advantages such as non-contact measurement, no temperature drift, traceable result, low system cost and etc. However, limited by tilt shooting angle, low imaging resolution and complex illumination in outdoor video surveillance systems, the detection and recognition of scales and characters on the staff gauge are quite difficult in practice, resulting in poor stability and suitability of the existing methods. Aiming at these problems, a solution is proposed from the perspective of photogrammetry. Firstly, a template image is designed according to the style of standard bicolor staff gauge. Secondly, the perspective projection transformation relationship between the region of interest（ROI） and the template image is established through manually selected control points, then the staff gauge image is registrated to orthographic coordinate system. Thirdly, according to the sampling windows set in the registration image the adaptive segmentation threshold is calculated and binarized. Finally, the water line is detected in the horizontal projection curve of the binary image, its coordinates are converted to the water level measurement value according to the physical resolution of the template image. Four sets of field tests were conducted under different conditions. Results show that the method is robust for the low resolution images recorded under both day and night illuminations. The measurement resolution and accuracy achieve 1 mm and 1 cm, respectively, which meets the hydrometry requirements.

关 键 词：机器视觉 水位测量 标准水尺 水位线检测 图像处理 

分 类 号：TH815[机械工程—仪器科学与技术]