頁籤選單縮合
題 名 | 倒傳遞類神經網路應用於BGA錫球瑕疵檢測=Application of Back-Propagation Neural Network for Image Inspection of BGA Shape Defects |
---|---|
作 者 | 陳詩豐; 丁冠文; | 書刊名 | 龍華科技大學學報 |
卷 期 | 29 2010.06[民99.06] |
頁 次 | 頁1-8 |
分類號 | 446.8405 |
關鍵詞 | 倒傳遞類神經網路; BGA錫球瑕疵檢測; 二值化; Back-propagation neural network; Inspection of BGA shape defects; Binarization; |
語 文 | 中文(Chinese) |
中文摘要 | 本研究主要是應用倒傳遞類神經網路,來辨別錫球外形瑕疵以及分類,提升外形檢測的準確率,同時開發二維BGA錫球檢測系統。並以Visual Basic配合Matlab影像處理做為開發工具。檢測系統能在BGA錫球有偏移及外形瑕疵時,能夠正確辨別出錫球外型與位置並以此分類錫球之瑕疵,檢測瑕疵項目包含:球偏移、多球、球過大、球過小、球變形、球相連以及缺球。 在自動檢測系統的研究方法與處理程序方面,首先影像經由二值化,取得錫球的像素、座標位置等資訊。藉由條件法則判斷錫球外形及位置瑕疵,以及使用倒傳遞類神經網路來辨識及分類外形及位置瑕疵之錫球。實驗結果證明系統BGA缺陷辨識方法所採用外形及位置二個倒傳遞類神經網路來作為BGA錫球檢測外形與位置瑕疵,尤其是對一般最難辨識的球變形與球相連的缺陷,達到自動檢測及高效率辨識能力的需求。 |
英文摘要 | This study is mainly utilizing Back-Propagation Neural Network technology to identify the shape of defective tin ball and to promote the accuracy of the inspection. by developing two dimensional BGA optical inspecting system, incorporate with Matlab which developed by Visual Basic as developing tool of function database for image process. Inspecting system is capable of detecting the shape of tin ball and its location precisely and classifying its quality. Inspecting items include, ball offset, ball presence, over size, under size, ball deformation, ball bridging and ball absence. For the study and processing procedure developments of the automatic inspecting system, image is processed by preprocess based on its grey value to identify the edge accuracy in order to acquire the coordinate of the tin ball precisely and BGA board bias angle relevant information. Based on the predefined criteria, defines the dimension and location defects of the tin ball and utilizes Back-Propagation Neural Network technology identifying and classifying the shape defects of the tin ball. Experiments within this study prove that the designed two Back-Propagation Neural networks for both shape and location inspections can correctly identify and classify the shape defects of the tin ball which also can achieve and contribute the requirements for the automatic inspection and high efficiency of identification capabilities. |
本系統中英文摘要資訊取自各篇刊載內容。