影像處理技術應用於透地雷達資料之研究

Abstract

近年來，透地雷達的非破壞性探測技術，被廣泛的應用於偵測地下管線。由於台灣多為非均勻地質層，造成透地雷達影像中的非金屬管線影像更容易受到影響而誤判。而低對比度的透地雷達影像，使得專家辨識管線變的更為困難。增強弱化及低對比度的管線影像而不受整體影像中的雜訊影響為主要研究方向。本論文嘗試使用數學形態學裡的形態交離轉換法及運算拋物線及建立拫物線模型對管線影像偵測並做區域性的影像增強，使得拋物線能有效與背景區分開來。觀察影像特徵後，設計幾組適用的形態學結構元素做為偵測管線影像，再運用拋物方程式產生模型，以達成區域性的影像增強。實驗結果顯示本論文提出的方法能有效提升管線解析度。

Recently, technology of un-destructive exploring of GPR(Ground Penetrating Radar) is extensively applied in underground pipes detection. Since un-uniform of geology layer in Taiwan, it is easier to have erroneous judgment in nonmetal pipeline images made by GPR. And low contrast and weak GRP images make experts encounter more difficulties while identifying pipes. Meanwhile, low contrast GPR images make experts even harder to determine the pipelines. The main aim of this research is to enhance the weakened and low contrast pipeline images while avoiding the unity of the images from affected by noise. This paper is trying to apply morphological hit-or-miss transform to establish the parabola model for pipeline image detection, and to locally enhance the image. So that, the parabola can be separated from the background more effectively. After observing the image features, some suitable morphological structuring elements are designed to detect the pipeline images. Then parabola equation is used to produce the parabola model, to attain local image enhancement. Experiment results show that the approaches proposed in this thesis can increase the pipeline resolution effectively.