為了防止水患發生，造成對人民生命財產安全的危害，監視攝影機被廣泛使用在易淹水地區低窪處、河川等地方，以便於颱風豪雨期間監視該區域水位。近年來，隨著監視攝影機解析度的進步，配合圖像辨識技術的研究，利用監視攝影機自動測量水位的方式亦被廣泛的應用。
本論文提出了一種使用單個相機的自動圖像水位測量系統，利用單一相機傳回的影像來偵測河流的水位。由於一般採用鏡頭大都需要進行鏡頭的失真校正，我們先在實驗室進行相機校正，最主要的目的是為了求得相機的內部參數及失真係數，用以校正因鏡頭失真的圖像。再利用一連串的數位影像處理技術，將圖像中的雜訊干擾消除，以利於現場水尺的影像中辨識出目前水線的候選位置。為了從候選水線決定出正確的水位位置，本系統提出一種水線決策機制，並採用水尺的感興趣區域(region of interest, ROI)影像提供給水線決策機制，決定出正確的水位線。所提水線決策機制將同一個輸入圖像分別同時進行2個處理程序的數位影像處理，並分別產生候選水線以及水線位置參考。從圖像中得到目前水線位置座標後，經由參考水位和控制點的設定，將圖像中像素距離轉換成真實世界的距離，系統即可準確的計算出真實世界的水位。
在系統的實際應用上，由於監視鏡頭需架設於戶外排水渠道旁，為了避免鏡頭被人為破壞和為了得到較大的視野(field of view, FOV)，通常使用鐵桿架設起來，這個情況造成攝影相機光軸與現場水尺平面存在某個非正交角度，因而產生了透射失真，這使得我們無法從水線在像素平面的座標位置變化，得到水線的座標位置變化與真實的水位變化的一個線性關係來計算當前的水位，因此透射失真的現象需改善，避免水位計算結果產生誤差。此外，由於監視鏡頭架設在鐵桿上，容易因人為或風、地震等自然因素產生振動，這使得所拍攝到的圖像也因相機振動而產生偏移，導致圖像中的水線坐標位置亦隨著偏移，造成了水位偵測結果的錯誤。由於前揭實務應用上常遭遇到的問題，使得傳統水尺圖像水位偵測方式之結果存在一定的誤差值。
承上，為了改善前揭實務上遭遇的問題，本論文所提出的系統採用逆透視映射(inverse perspective mapping, IPM)的技術來校正圖像的透射失真，利用IPM技術將拍攝到的影像從一個視平面投影到另外一個視平面的過程，達到攝影機像素平面與真實世界中水尺上水位變化的近似於線性關係。實驗發現圖像經由IPM技術校正後的水位測量結果與真實水位相較，仍然存在著些許的誤差，故我們提出一個距離校正模組(Distance calibration module, DISCAL)來進一步優化水位測量結果。另外，為了防止相機振動產生的圖像偏移干擾水位測量的準確性，我們利用歸一化互相關 (normalized cross-correlation, NCC) 技術和提出一個相機振動校準機制來加以消除其影響。
本論文所提單監視鏡頭自動水位測量系統架設點位於一座抽水站的進流渠道，進流渠道為2米寬、2.2米深的鋼筋混凝土矩形結構。水位變化來自將進流渠道上的水閘門開啟引水流入渠道，觀測水位上升狀況，待水位上升到高點再將水閘門關閉後開啟抽水泵，將水抽排出抽水站，觀測水位下降狀況，實驗結果證實本系統能有效追蹤水位即時的變化。本系統所提水位估算機制可以有效改善傳統水尺圖像水位偵測方式的水位估算系統，在低水位EL-0.45m時，改善誤差從4.82cm降低到1.8cm，改善約63%。另外，採用所提出的距離校正模組後，更可以將誤差值降低到小於0.5cm，相較於傳統水尺圖像水位偵測方式，在低水位EL-0.45m時，誤差改善提高到90%以上。在其他水位的偏差值亦均小於0.5 cm。根據實驗結果證明，本系統可以有效改善透射失真現象所造成的誤差，特別適合局限空間進行圖像水位測量。

In order to prevent the occurrence of floods causing harm to the safety of people's lives and properties, surveillance cameras are widely being used in low-lying places and rivers in flood-prone areas, so that the water level in the area can be monitored during typhoons and heavy rains. In recent years, with the improvement of the resolution of surveillance cameras and the increased research on the image recognition technology, the methods of using surveillance cameras to automatically measure the water level have also been widely studied.
This dissertation proposed an automatic image water level measurement system through a single camera to detect the water level of a river by using the images returned by the single camera. For this purpose, we first calibrated the camera in the laboratory since most of the lenses used would generally require lens distortion correction. The main purpose of this step was to obtain the camera's intrinsic parameters and distortion coefficients to correct the distorted image. A series of basic digital image processing techniques was used to eliminate the noise interference in the image, in order to facilitate the identification of the current candidate position of the waterline in the image of the on-site staff gauge. In order to determine the correct water level position from the candidate waterline, this system proposed a waterline decision mechanism, and used the region of interest (ROI) image of the staff gauge image for the purpose of determining the correct water level. The proposed waterline decision mechanism performed the digital image processing of two different processing programs with the same input image at the same time and generates candidate waterlines and waterline position references respectively. It was demonstrated that after obtaining the coordinates of the current waterline position from the image, the system could accurately calculate the water level elevation in the real world by converting the pixel distance in the image into the real-world distance through the setting of the reference water level and the control points.
In the practical application of the system, the surveillance camera needs to be erected beside the outdoor drainage channel. In order to avoid the camera being damaged by human as well as to obtain a larger field of view (FOV), it was usually erected with iron rods, which causes the camera’s optical axis and the staff gauge plane nonorthogonal. There was a nonorthogonal angle between the optical axis of the camera and the pixel plane of the staff gauge, resulting into a perspective distortion so that the actual length represented by the unit pixel varies with the position of the pixel on the pixel plane, which makes it impossible to obtain the water level from the change of the coordinate position of the waterline on the pixel plane. Since the water level was calculated by the linear relationship between the change of the coordinate position of the waterline on the pixel plane and the water level on the staff gauge in the real world, eliminating the perspective distortion can effectively improve the results of the water level estimation. In addition, since the surveillance camera was mounted on an iron pole, it would be easy to vibrate due to human or natural factors such as wind and earthquake, which could make the captured image also offset due to camera vibration, causing the waterline coordinate position in the image to change. Also, with the offset, there would be a possibility for the deviations in the water level detection results. Thus, due to the problems often encountered in the practical application of the previous disclosure, there are certain deviations in the water level detection results of the traditional image-recognition method.
Continuing from the above, in order to improve the problems encountered in the practical applications, this proposed system used the inverse perspective mapping (IPM) technology to correct the perspective distortion of the image by projecting the captured image from one viewing plane to another viewing plane. After IPM transformation, the actual length represented by the unit pixel will not vary with the position of the pixel on the pixel plane. The experiments revealed that the measurement results of the water level after the image were corrected by IPM technology but however still had a little deviation when compared with the real water level elevation. Therefore, we proposed a distance calibration module to further optimize the water level corrected by IPM technology. In addition, in order to prevent the image shifts caused by camera vibration from interfering with the accuracy of water level measurements, we utilized the normalized cross-correlation (NCC) technique as well as a proposed camera vibration calibration mechanism to counteract its effects.
The proposed automatic water level measurement system was mainly divided into two parts：client and server. The client was installed on the intake channel of the pumping station, and the server was installed indoors. The intake channel of the pumping station is a reinforced concrete rectangular structure with a width of 2 meters and a depth of 2.2 meters. The water level changes came from opening the floodgate on the intake channel to divert the water into the channel, by observing the rising water level. The floodgate was closed and then the pumping pump was turned on to pump the water out of the pumping station when the water level rose to a high point. Then, the falling water level was observed. The results confirmed that the system could effectively track the real-time changes in water level. It was thus highlighted that the water level estimation mechanism proposed in this system could effectively improve the water level estimation system using the traditional image-recognition method. It was found when the low water level was EL-0.45m, the improvement deviation was reduced from 4.82 cm to 1.8 cm, i.e., an improvement of about 63%. In addition, after using the proposed distance calibration module, the deviation value could be reduced to less than 0.5 cm. Interestingly when compared with the traditional image-recognition method, the deviation rectification was improved to more than 90% at the low water level of EL-0.45 m. The deviation values at other water level elevations were also found to be less than 0.5 cm. Thus, based on the experimental results, it was suggested that the system could effectively improve the deviations caused by the perspective distortion phenomenon, and it would be especially suitable for the image water level elevation measurement in the confined space.

[1] T. J. Buchanan and W.P. Somers, "Stage Measurement at Gaging Stations," in Techniques of Water-Resources Investigations of the United States Geological Survey, 3thed., US: USGS, pp. 1-28, 1968.
[2] H. H. Shih and D. L. Porter, "Error Models for Stilling Well-Float Type Tide Gages," in Proc. IEEE-OCEANS, Boston, MA, USA, pp. 16-18, Sep. 1981.
[3] B.Y. Lee and B.Y. Park, "Development of high precision underground water level meter using a buoyant rod load cell technique," KSAFM, vol. 1, no. 1, pp. 36-40, 1999.
[4] S. W. Wang, C. C. Chen, C. M. Wu, and C. M. Huang, "A continuous water-level sensor based on load cell and floating pipe," in Proc. IEEE-ICASI, Chiba, Japan, pp. 151-154, Apr. 2018.
[5] A. Stateczny, "Radar Water Level Sensors for Full Implementation of the River Information Services of Border and Lower Section of the Oder in Poland," in Proc. 17th IEEE Int. Radar Symp., Krakow, Poland, May 2016.
[6] G. Wang, C. Gu, J.A. Rice, T. Inoue, and C. Li, "Highly Accurate Noncontact Water Level Monitoring using Continuous-Wave Doppler Radar," in Proc. IEEE-WiSNet, Austin, TX, USA, pp. 19-21, Jan. 2013.
[7] C. Gu, W. Xu, G. Wang, T. Inoue, J. A. Rice, L. Ran, and C. Li, "Noncontact Large-Scale Displacement Tracking: Doppler Radar for Water Level Gauging," in Proc. IEEE Microw. Wirel. Compon. Lett, vol. 24, no. 12, pp. 899-901, Dec. 2014.
[8] S. Ben-Yaakov, C. Golan, and S. Kessler, "A stand-alone ultrasonic ranging system for hydrological water stage measurements," IEEE Trans. Instrum. Meas., vol. 41, no. 5, pp. 699-702, Oct. 1992.
[9] N. Zachos, C. Kosmatopoulos, and T. Laopoulos, "A wireless network of remote measuring stations: application in water level monitoring," in Proc. IEEE-ISIE, Dubrovnik, Croatia, pp. 153-156, Jul. 1995.
[10] M. Meribout, M. Habli, A. Al-Naamany, and K. Al-Busaidi, "A new ultrasonic-based device for accurate measurement of oil, emulsion, and water levels in oil tanks," in Proc. 21st IEEE Instrum. Meas. Tech. Conf., Como, Italy, vol. 3, pp. 1942- 1947, May 2004.
[11] M. Mousa and C. Claudel, "Poster abstract: Water Level Estimation in Urban Ultrasonic/Passive Infrared Flash Flood Sensor Networks Using Supervised Learning," in Proc. 13th Int. Symp. Inf. Process. Sensor Netw., Berlin, Germany, pp. 277-278, Apr. 2014.
[12] R. M. Shrenika, S. S. Chikmath, A. V. Ravi Kumar, Y. V. Divyashree, and R. K. Swamy, "Non-contact water level monitoring system implemented using LabVIEW and Arduino," in Proc. IEEE Int. Conf. Recent. Adv. Electron. and Commun. Tech., pp. 306-309, 2017.
[13] S. Sachio, A. Noertjahyana, and R. Lim, "IoT based water level control system," in Proc. 3rd Technol. Innov. Manag. and Eng. Sci. Int. Conf., pp. 1-5, 2018.
[14] S. J. Yue, H. Cui, M. Hanafi, S. Mariam Shafie, and N. Abd Salim, "IoT Based Automatic Water Level and Electrical Conductivity Monitoring System," in Proc. 8th IEEE Conf. Syst., Proc. & Ctrl., pp. 95-100, 2020.
[15] M. Iwahashi, S. Udomsiri, Y. Imai, and S. Muramatsu, "Water level detection for functionally layered video coding," in Proc. IEEE Int. Conf. Image Process., San Antonio, TX, USA, pp. 321-324, Oct. 2007.
[16] J. Y. Zheng and S. Sinha, "Line cameras for monitoring and surveillance sensor networks," in Proc. 15th ACM Int. Conf. Multimedia, Augsburg, Germany, pp. 433-442, Sep. 2007.
[17] H. Sakaino, "Camera-Vision-Based Water Level Estimation," IEEE Sen. J., vol. 16, no. 21, pp. 7564-7565, Nov. 2016.
[18] T. W. Feng, J. F. Xiong, J. C. Xiao, J. Q. Liu, and Y. Q. He, "Real-time riverbank line detection for USV system," in Proc. IEEE Int. Conf. Mechatron. and Autom., pp. 2546-2551, 2019.
[19] S. Bin, C. Zhang, L. Li, and J. Wang, "Research on Water Level Measurement Based on Image Recognition for Industrial Boiler," in Proc. 26th IEEE-CCDC, Changsha, China, pp. 1344-1348, Jun. 2014.
[20] T. Hies, P. S. Babu, Y. Wang, R. Duester, H. S. Eikaas, and T. Meng, "Enhanced water-level detection by image processing," in Proc. 10th Int. Conf. on Hydroinformatics, Hamburg, Germany, pp.1-9, Jan. 2012.
[21] J. Kim, Y. Han, and H. Hahn, "Embedded implementation of image-based water-level measurement system," IET Comput. Vis., vol. 5, no. 2, pp. 125-133, Mar. 2011.
[22] J. Pan, Y. Yin, J. Xiong, W. Luo, G. Gui, and H. Sari, "Deep Learning-Based Unmanned Surveillance Systems for Observing Water Levels," IEEE Access, vol. 6, pp. 73561-73571, Nov. 2018.
[23] Z. Zhang, Y. Zhou, H. Liu, and H. Gao, "In-situ water level measurement using NIR-imaging video camera," Flow Meas. Instrum., vol. 67, pp. 95-106, Jun. 2019.
[24] F. Lin, W. Y. Change, L. C. Lee, H. T. Hsiao, W. F. Tsai, and J. S. Lai, "Applications of image recognition for real-time water level and surface velocity," in Proc. IEEE-ISM, Anaheim, CA, USA, pp. 259-262, Dec. 2013.
[25] J. Yu and H. Hahn, "Remote Detection and Monitoring of a Water Level Using Narrow Band Channel," J. Inf. Sci. Eng., vol. 26, pp. 71-82, Jan. 2010.
[26] T. E. Gilmore, F. Birgand, and K. W. Chapman, "Source and magnitude of error in an inexpensive image-based water level measurement system," J. Hydrol., vol. 496, pp. 178-186, Jul. 2013.
[27] Y. T. Lin, Y. C. Lin, and J. Y. Han, "Automatic water-level detection using single-camera images with varied poses," Meas., vol. 127, pp. 167-174, May 2018.
[28] M. Bruinink, A. Chandarr, M. Rudinac, P. J. van Overloop, and P. Jonker, "Portable, automatic water level estimation using mobile phone cameras," in Proc. IAPR Int. Conf. Mach. Vis. Appl., pp. 426-429, May 2015.
[29] C. S. Zhang, G. Feng, Z. Y. Liu, B. Sun, and H. P. Zhang, "Research on water level gauge image identification based on intelligent remote viewing," in Proc. Chin. Ctrl. Decis. Conf., pp. 1561-1566, May 2016.
[30] Z. K. Xu, J. Feng, Z. Z. Zhang, and C. F. Duan, "Water level estimation based on image of staff gauge in smart city," in Proc. IEEE SmartWorld/SCALCOM/UIC/ATC/CBDCom/IOP/SCI, pp. 1341-1345, Oct. 2018.
[31] M. Jodeau, A. Hauet, A. Paquier, J. Le Coz, and G. Dramais, "Application and evaluation of LS-PIV technique for the monitoring of river surface velocities in high flow conditions," Flow Meas. Instrum., vol. 19, pp. 117-127, Apr. 2008.
[32] M. C. Lee, J. M. Leu, H. C. Chan, and W. C. Huang, "The measurement of discharge using a commercial digital video camera in irrigation canals," Flow Meas. Instrum., vol. 21, pp. 150-154, Jan. 2010.
[33] Z. Zhang, Y. Zhou, Y. Li, Y. Ye, and X. Li, "IP camera-based LSPIV system for on-line monitoring of river flow," in Proc. IEEE-ICEMI, Yangzhou, China, pp. 357-363, Oct. 2017.
[34] Z. Zhang, "A flexible new technique for camera calibration," IEEE Trans. Pattern Anal. Mach. Intell., vol. 22, no. 11, pp. 1330-1334, Nov. 2000.
[35] J. Canny, "A Computational Approach to Edge Detection," IEEE Trans. Pattern Anal. Mach. Intell., vol. PAMI-8, no. 6, pp. 679-698, Nov. 1986.
[36] R. O. Duda and P. E. Hart, "Use of the Hough transformation to detect lines and curves in pictures," Comm. ACM, vol. 15, no. 1, pp. 11-15, Jan. 1972.
[37] J. P Lewis, "Fast normalized cross-correlation," Vision Interface., vol. 10, no. 1, pp.120-123, 1995.
[38] F. Zhao, Q. Huang, and W. Gao, "Image Matching by Normalized Cross-Correlation," IEEE-ICASSP, Toulouse, France, vol. 2, pp. 729-732, May 2006.
[39] Y. Raghavender Rao, Nikhil Prathapani, and E. Nagabhooshanam, "APPLICATION OF NORMALIZED CROSS CORRELATION TO IMAGE," IJRET, vol. 3, no. 5, pp.12-16, May 2014.
[40] H. A. Mallot, H. H. Bulthoff, J. J. Little, and S. Bohrer, "Inverse perspective mapping simplifies optical flow computation and obstacle detection," Biol. Cybern., vol. 64, no. 3, pp.177-185, Jan. 1991.
[41] R. Hartley and A. Zisserman, "Multiple View Geometry in Computer Vision," 2nd ed., Cambridge University, pp.153-166, 2000.
[42] J. Jeong and A. Kim, "Adaptive inverse perspective mapping for lane map generation with SLAM, "13th Int. Conf. Ubiquitous Robot & Ambient Intell., pp. 38-41, 2016.
[43] L. C. Kuo and C. C. Tai, "Implementation of Inverse Perspective Mapping for Camera-Vision Water-Level Measurements," in Proc. International Computer Symposium (ICS), Tainan, Taiwan, pp. 348-351, Dec. 2020.
[44] L. C. Kuo and C. C. Tai, “Automatic water-level measurement system for confined-space applications,” RSI, vol. 92, no. 8, Art. no. 085001, Aug. 2021.