透過您的圖書館登入
IP:18.226.177.223
  • 學位論文

色彩模型於機器視覺檢測的驗證

Evaluation of Color Models for Color Classification and Segmentation

指導教授 : 蔡篤銘 博士
若您是本文的作者,可授權文章由華藝線上圖書館中協助推廣。

摘要


本研究的主要目的是探討色彩分類的技術,其結果可應用於工業界 的彩色分類與量測。黑白影像中所包含的資訊只有亮度,現今的產品大 多都包含了豐富的色彩,色彩所包含的資訊有色相(Hue)、亮度(Inten- sity) 及飽和度(Saturation)這三種。若是利用黑白影像處理彩色物體 時,要將整個色彩空間投影到一條黑白灰階直線,會造成多種顏色都對 映到同一灰階(gray level)中,而不易分辨。 本研究利用14種不同色彩模型(color models)中的42個色彩指標( color features) 進行色彩的分類,以貝氏分類法(Bayes classifier) 及逐步指標搜尋(stepwise search) 法找尋一個適合工業檢測的指標組 合,使其應用於工業環境之色彩檢測上,能在不同亮度的環境下,對於 色相及飽和度這兩種色彩特徵,在色彩分類能達到最佳的辨識能力。最 後利用所選擇之最佳指標組合進行色彩分割並採用眾數平滑化( modal smoothing )法去除誤判之雜點,修正分割後的影像。在本研究中將以 PANTONE 色票來驗證研究方法,並經由彩色瓷磚、彩色紡織品、彩色印 刷品及印刷電路板驗證實驗方法,由實驗結果得知,在最佳的指標組合 模式下,可得到良好的色彩分割結果。

並列摘要


Recent work has shown that color machine vision has great potential when applied to automated classification and inspection in industry. In contrast to traditional monochrome images that measure only intensity information, color images allows to be measured in hue, saturation and intensity, and, therefore, obtain a richer information. In this research, we study the use of the Bayes classifier for color classification and segmentation. The best color descriptors used in the Bayes classifier for color classification are selected among 31 distinct color features from 14 existing color models using a bottom-up stepwise search technique. Both constant lighting and varying lighting are con- sidered, and their corresponding color descriptor sets for the best dis- crimination of individual hues and saturations are determined. In color segmentation, the Bayes classification with the best selected color fea- tures is applied to assign the color in a pixel-by-pixel basis. A modal smoothing procedure follows to remove noise in the segmented image. In experiments, 20 hues, each with 7 saturations, selected from the PANTONE color formula guide are used for performance evaluation. Experimental results show that the Bayes classifier with the selected color features and the modal smoothing have generated good color segmentation.

參考文獻


【28】 黃正霖、李芳繁、林慶福,"應用彩色機器視覺技術選別蔬菜種子之研究",農業機械學報1994年12月﹐1-11頁。
【 1】 Billmeyer, F. W. and Saltzman, M., Principles of Color Technology, 2nd Edition, Wiley-Interscience Publication, New York, 1981.
【 2】 Dhawan, A. P. and Sicsu, A., "Segmentation of Images of Skin Lesions Using Color and Information of Surface Pigmentation", Computerized Medical Imaging and Graphics, Vol, 16, pp.163-177, 1992.
【 3】 Gonzalez, R. C. and Woods, R. E., Digital Image Processing, Addison-Wesley, New York, 1992.
【 4】 Jain, A. K., Fundamentals of Digital Image Processing, Prentice-Hall, New Jersey, 1989.

被引用紀錄


陳佩鈴(1998)。邏輯迴歸樹應用於印刷電路板之瑕疵分類〔碩士論文,元智大學〕。華藝線上圖書館。https://www.airitilibrary.com/Article/Detail?DocID=U0009-0112200611365372
葉佳榮(1998)。彩色雙攝影機之物件追蹤〔碩士論文,元智大學〕。華藝線上圖書館。https://www.airitilibrary.com/Article/Detail?DocID=U0009-0112200611354375

延伸閱讀