攝影機檢定應用於土石流現場

Abstract

近年來攝影機作為監測設備已廣泛應用於土石流現地，但獲得的資料有限，造成無法提供足夠的資訊去做預警，本研究設想提供現場攝影機的檢定結果，建立現場世界座標與影像中像素座標的轉換關係，等到利用影像技術分析出土石流水面線時，可以提供轉換座標的流程去計算出土石流的流深。而本研究是針對在現場尺度約為100公尺範圍的大尺度情況下所設計的流程，利用在現場測量的方式，沿著河道方向設立斷面並檢定攝影機的內參數，根據理論一台攝影機只會有一組內參數，但因為現場尺度較大，造成每個斷面的測量點在影像深度的差異會較大，及測量時本身的測量誤差，會造成每個斷面在檢定參數上會有誤差，而本研究的目的是為了要反轉換出現場的座標，故為了讓轉換結果更準確便選擇此方法，在建立出各斷面的檢定參數之後，即可轉換影像中在斷面內水面線的世界座標，首先取得每個測量斷面平均 的世界座標及像素座標，並在影像中將其依序連線，並求出線上每個點對應的，再利用此點去計算與各斷面的直線距離，決定將未知點分配到垂直距離最短的斷面，利用該斷面的檢定參數及相對應的 帶入式子去反轉換最後的世界座標。之後利用數值實驗及現場測量來驗證本研究所設計的流程並列出轉換結果的誤差，最後將流程實際套用到事件發生後利用影像技術分析出的水面線，去計算出土石流水面線的實際流深。

In recent years, cameras have been widely used as monitoring equipment in the field of debrisflow, but the data obtained are limited that cannot provide sufficient information for early warning.This study aims to provide on-site camera calibration results and establish a conversion relationship between on-site world coordinates and pixel coordinates in the image, so that when the water surface line of the debrisflow is analyzed by imaging technology, the conversion process of coordinates can be provided to calculate the flow depth of the debrisflow. In this study, the process is designed for a large scale of about 100 meters in the field, using the field measurement method to set up a section along the river direction and calibrate the internal parameters of the camera. According to the theory, a camera will only have one set of internal parameters, but because of the large scale of the site, the difference in image depth of each measurement point in each section will be larger, and the measurement error of the measurement itself will cause errors in the calibration parameters of each section, and the purpose of this study is to inverse convert the coordinates of the site, so in order to make the conversion results more accurate, this method is chosen. After establishing the calibration parameters of each section, we can convert the world coordinates of the water surface lines in the section in the image, First,get the average world coordinates and pixel coordinates of each measured section, and connect them in the image in order, and find the corresponding distance of each point on the line, and then use this point to calculate the straight line distance from each section, and decide to assign the unknown point to the section with the shortest vertical distance, and use the calibration parameters of the section and the corresponding equation is brought into the equation to inverse convert the final world coordinates.Afterwards, numerical experiments and field measurements are used to verify the process designed by this study and to list the errors of the conversion results.