建立水色影像智慧化分析技術 應用於水庫水質辨識

Build Intelligent Analysis Technology For Water Color Image For Reservoir Water Quality Monitoring

黃佑祥(You-Xiang Huang)

指導教授 ： 施武陽

中原大學/工學院/環境工程學系/碩士(2022年)

https://doi.org/10.6840/cycu202201613

摘要

近年來於環境監測的相關技術越來越成熟，在水體監測的方面更是有許多不同類型的監測方式，從水底到水面甚至遠至太空以外的衛星影像，都有與水質相關的監測項目，相比於傳統接觸式的採樣方式已有很大的進步。根據文獻中在遙測方面有許多水體反射光譜對於水質的研究中證明兩者之間具有相關性，而在水色模擬中也能準確地還原與現場水色相同的顏色，因此在水色的呈現對於水質數據的相關性也具有一定的研究意義。
本研究採取三座水庫分別為石門水庫、鯉魚潭水庫及湖山水庫之入水口與取水口的水樣及影像數據，於現場透過影像感測器取得水色的影像，並且採取水樣進行現場以及實驗室的量測。影像整理後進行影像的處理，將影像數據以及水質數據進行相關性的分析，探討其之間的關聯性。透過影像處理技術，將水面波紋的部分濾除，可以有效提升影像數據與水質間的相關性。在斯皮爾曼相關性分析的結果中，表層水的G/R比值與透明度、濁度及葉綠素-a之相關程度分別為0.87、-0.76及-0.77，而在中層水中影項數據與水質數據相關程度皆低於0.7，因此在水色的呈現上以表層水的水質為主要影像。在各個水庫中根據不同水庫的不同特性，其影像數據與水質數據間也有不同項目具有顯著之相關程度，對於未來探討水色及水質間的相關性有很大的幫助。
未來建議增加研究的時間長度，除不同季節外還能有豐水期與枯水期的水質數據，並且在分析部分將加入統計檢定、敏感度分析和其他相關性分析等，另外就目前影像處理的獲取及處理的流程能夠再加強，提升水色影像的準確性，對於分析的準確度也能提升，若將應用於水質的監測中，周圍環境光的影響將成為所面臨的挑戰。

關鍵字

水色 ； 水質 ； 影像處理

並列摘要

Recently, technologies related to environmental monitoring have become more and more mature, and many types of monitoring methods have emerged in the field of water monitoring. It is a great improvement over traditional contact sampling methods. According to the literature, there are many water reflectance spectra in the field of telemetry. Research on water quality has proved that there is a correlation between the two, and the watercolor simulation can accurately restore the same color as the field watercolor. Therefore, the presentation of watercolor is very important for water quality data. This correlation also has certain research significance.
In this study, water samples and image data were collected from the inlets and intakes of three reservoirs: Shimen, Liyutan, and Hushan. The watercolor images are acquired on-site through the image sensor, and water samples are collected for on-site experimentation. Room measurement. After image processing, correlation analysis was performed on data image and water quality data, and the correlation between them was discussed. The ripples on the water surface can be filtered through image processing technology, which can effectively improve the correlation between image data and water quality. The results of Spearman correlation analysis showed that the correlations between the surface water G/R ratio and transparency, turbidity and chlorophyll-a were 0.87, -0.76 and -0.77, respectively, while the correlations between the shadow item data and the water quality data of the reclaimed water were all the same. It is lower than 0.7, so the surface water quality is the main image of watercolor. In each reservoir, according to the different characteristics of reservoirs, data image and water quality data also have different items, and the degree of correlation is significant, which is of great help for future research on the correlation between watercolor and water quality.
In the future, it is recommended to increase the length of the study. In addition, correlation analysis included statistical verification and sensitivity analysis for different seasons, and water quality data from wet and dry periods were added to the analysis part. In addition, the current image processing acquisition and processing process can be further enhanced to improve the accuracy of watercolor images, and the accuracy of analysis can also be improved. The effect of ambient light can be a challenge if applied to water quality monitoring.

並列關鍵字

water color ； water quality ； image processing

參考文獻

[1] Shenglei Wang,Junsheng Li,Wenzhi Zhang,Chang Cao,Fangfang Zhang,Qian Shen,Xianfeng Zhang and Bing Zhang (2021)

[2] Shoji Tominaga, Shogo Nishi, Ryo Ohtera (2021)

[3] Zhiqiang Chen (2007)

[4] MatiasBonanseaaMaría, ClaudiaRodriguezb, LucioPinottic, SusanaFerrero (2014)

[5] M. Priya, Ashwini K., Janice Vedha, D. Diviya (2019)