影像處理應用於雨滴偵測系統

Abstract

台灣的地理位置四周環海，四季分明，雖然面積只有 35,759.5 km^2，但累積年雨量卻非常驚人，台灣地理位置位在熱帶氣旋活耀的地區，氣象局平均每年也會發佈 4-5 個颱風警報，颱風可以說是台灣主要的天然災害之大宗。龐大的雨量就是造成災害的主要元兇，本論文與財團法人國家實驗研究院儀器科技研究中心合作，計畫內容是災害預警儀器技術開發之計畫，主要目的是開發出使用攝影機之儀器，運用影像處理的技術完成雨滴偵測與分析。

本論文主要分為兩個部分，其一為雨滴的偵測，觀察到雨滴的外型類似橢圓，所以運用帕斯卡定理判斷出影像中雨滴的部分，也就是橢圓形狀當作目標，達到擷取雨滴位置資訊。第二部分為雨滴的追蹤，從雨滴影像拍攝的快門時間，反映到位置的移動，造成殘影的產生，運用軌跡計算預測出雨滴在於前後張影像的落點，使之資訊相聯結，達到雨滴追蹤的目的。將兩部分結合，完成雨滴影像擷取與分析之系統，並加強討論其需改善的部分，使整個系統更加齊全，增加其實際投入於災防計畫的可行性。

Taiwan is an island surrounding with ocean, and its climate has a distinct feature of monsoons. Though the area of Taiwan is merely 35,759.5 km^2, the mean annual cumulative rainfall is actually astonishing. Moreover, Taiwan is located in the region where tropical cyclone generates actively. The meteorological bureau also issues average 4 to 5 typhoon alerts every year. Thus, typhoon could be considered as the main reason of natural disaster. From above, we could conjecture that the huge rainfall is a major cause of the disaster. This paper is cooperating with NARLabs Instrument Technology Research Center(ITRC). The main purpose of plan is disaster preventing and controlling. The methodology of plan is to develop a new instrument of raindrop detection and analysis by using image processing technique.

This paper was divided into two parts. One is to detect raindrop and study the elliptical similarity due to its outlook; therefore, we introduce the Pascal Theorem in our paper to analyse the raindrop's shape. The other is to track the path of raindrop. Using haulage shadow of raindrop is very helpful to calculate and predict a raindrop position and its associated information at each time step. By combing the above two parts, our newly developed instrument can complete the measurement of the raindrop for the weather forecast. Finally, we conclude the study with some comments of raindrop detection instrument for improving the disaster prevention.