小波理論於智慧型影像處理在鋼構橋梁表面□蝕面積檢測之應用

Abstract

近年來，由於電腦技術的發展、成本花費降低與電腦處理器運算能力的不斷提昇讓影像處理(image processing)技術具有實用性，讓我們可以利用影像處理的過程，將複雜的影像加以分析。因此，本研究之目的在於開發一種智慧型橋梁表面□蝕面積評估判斷模式，其結合小波轉換(wavelet transform)、影像處理與MATLAB 程式語言，對於鋼構橋梁表面塗裝□蝕狀況進行辨認與量測。在系統中利用了小波轉換的基底有較短的持續時間(time duration)、具有多重解析(multi-resolution analysis)的特性和時頻兩域局部特徵的能力，來處理數位影像。 本系統先使用影像檔案之標準差來判斷R、G、B三個獨立的影像平面的差異性，選取差異性大的二維數位訊號影像平面來做小波轉換，二維小波轉換產生四塊矩陣，一塊代表低解析部的部分，三塊代表細節的部分。低解析部的部分含有影像中最多的能量，亦即包含了最多的資訊和影像的輪廓，因此對低解析部的係數進行處理，使得重建後的影像能達到求取□蝕面積的目的，並對□蝕面積再進行相同的處理，分出深淺不同的區域面積，再分別乘以折減係數求取折減後的□蝕面積，並結合規範進行□蝕面積的評估與補救方式，對於規範中不確定的地帶採取模糊(fuzzy)判斷。希望能對□蝕影像的評估與判斷提供具有客觀性、一致性且量化的程序。

In the recent years, the advancement in computer technology allowed image processing to become more practical and cost-efficient. Following the continuous advancement in algorithm ability of computer processors, we are able to analyze the complicated images with image processing. Therefore, this study developed an intellectual bridge surface coating evaluation judgment model which combines wavelet transform, image processing and MATLAB programming language to identify and measure the corrosion situation of the surface coating on steel bridges. The system used the wavelet transform to gain shorter time duration, multi-resolution analysis feature, and time and frequency domain to process images. The system first uses standard deviation to determine the discrepancies in image plane of R, G, B, and selects 2D digital signal image with large discrepancy for wavelet transform. The 2D wavelet transform produces 4 matrixes, one represents the low resolution part, and the other three represent the details. The low resolution part contains the most part of energy in the image, also the most information and image outline. Processing the parameters of the low resolution part can obtain the corrosion area from the reconstructed image, and process the corrosion area in the same way to divide it into areas of different shades. Then, multiple it with discounted parameters to obtain the discounted corrosion area, and combines with the norm to evaluate and formulate remedy for the corrosion area. Fuzzy concept is adopted for ambiguous area. This study expects to provide objective, consistent, and quantified procedure for evaluation and judgment of corrosion images.