透過您的圖書館登入
IP:18.191.240.243
  • 學位論文

機械視覺配合雙CCD三維量測判釋土石軌跡與粒徑分佈之研究

The Study of Machine Vision with Two CCD Applied on the Track of the Debris Flow and the Particle Size Distribution

指導教授 : 張守陽 陳彥璋

摘要


本研究結合「機械視覺」與「雙CCD三維量測」理論,計算影像點之空間三維座標,針對不同情境之土石影像,求算「特定物運移軌跡」及「粒徑分佈」等土石特性。 本研究首先藉由雷射點標示出雙CCD左右影像控制點,利用控制點之三維資訊,計算出雙CCD內外部參數;進行左右影像裡尋找共同點,依據「立體視覺法」計算待測點之三維座標。繼而針對多顆礫石進行判釋,先於左右影像相同範圍,以特定物形心位置利用「象限法」及「向量法」找出左右影像內之相同特定物,得以求算多顆粒之三維座標資訊。多顆礫石得以判識後,配合機械視覺推求各礫石粒徑大小,並進行角度校正及比例尺校正,繪製粒徑分佈圖,並分析各代表粒徑。最後於判釋礫石流動時之三維運移軌跡上,利用流動方向、最小臨近法與特定物大小,尋找前後張影像相同之礫石,並求得表面礫石之三維軌跡座標。 經由室內實驗結果得知, CCD在距離目標物1公尺進行量測,其X、Y及Z軸座標值之均方根誤差各約為0.11、0.22及0.16公分;傾斜拍攝多顆土石,經角度與比例尺校正後,求算之粒徑誤差由8.56%減少為2.63%,誤差明顯改善。礫石運移軌跡之室內實驗發現,「雙CCD三維量測」系統所計算之最大距離誤差約為-4.89%,誤差範圍尚屬合理。希冀本實驗方法可提供未來土石流特性之監測參考。

並列摘要


The study is to calculate stereo coordinates of sediments by using the stereo and machine vision techniques of non-contacted type, and then the characteristics of sediments’ trajectories and grain-size distributions are evaluated based on various sediment pictures of different scenarios. Laser spots easily to be found in two CCD cameras are used to aid the stereo measures of targets or multiple rocks. The trajectories and grain sizes are also analyzed. The stereo positions of the laser spots are measured based on stereo vision techniques using two CCD cameras. The known coordinates of control points are used to calibrate the parameters of two CCD cameras. The measured point correspondences between pairs of points from the left and the right images are established. The 3D coordinates of the measured point are reconstructed by using stereo vision techniques. In the case of multiple rocks, the point correspondences are first searched in the right and the left images to find the centroids of the specific rocks using a quadrant method and a vector method. The grain-size distributions of the specific rocks will be analyzed. Using the information of flow direction, grain size, 3D coordinates of specific rocks, the trajectories of the specific rocks can be estimated based on the nearest neighbor approach. According to the laboratory experiments, the stereo vision techniques using two CCD cameras are feasible to estimate the grain-size distributions and trajectories of the specific rocks. The measure errors of 1 m distance are about 0.11cm, 0.22cm and 0.16cm at X, Y and Z axes respectively. The maximum trajectory error is about -4.89%. The system will be improved to detect debris flows and to prevent sediment related disasters.

參考文獻


[32] 劉榮斌,機械視覺與攝影測量判讀土石流特性參數之初步研究,碩士論文,國立台北科技大學環境規劃與管理研究所,台北,2006。
[4] 尹孝元、黃清哲、連惠邦、李秉乾、周天穎、王晉倫,「自動化土石流觀測系統之發展及應用」,中華水土保持學報,第三十七卷,第二期,2006,第91-110頁。
[17] 張守陽、林志平、游勝任,「影像處理應用於土砂災害判識之研究」,台南:電子計算計於土木水利工程應用研討會論文集,2005,第13-18頁。
[8] 吳榮峰,大尺度質點影像量測法之應用,碩士論文,國立成功大學水利及海洋工程研究所,台南,2003。
[10] 李明靜,河川表面流速與流量非接觸式量測方法之發展及應用,博士論文,國立成功大學水利及海洋工程研究所,台南,2003。

延伸閱讀