應用人工智慧技術作基板缺陷檢測

深度學習（Deep Learning）是一種對資料與圖像進特徵學習，利用建立網路結構與演算法計算，使機器之判斷與思維接近我們預設之目標，幫助作業流程檢測與製程更加快速及便利。本研究以OpenR8軟體為基礎，採用卷積神經網路（Convolutional Neural Network, CNN）為架構，利用Caffe為溝通橋梁，針對PCB電子零件影像進行瑕疵檢驗，透過半監督式學習模式，使機器訓練更有效率與精準，並分析瑕疵種類幫助品檢人員在生產線上進行篩檢，有效降低漏失率（Underkill rate）與誤判率（Overkill rate）也減輕檢視人員在目測時之負擔，提高整體生產流程之效率，後續也可以透過瑕疵品進行製程能力分析找出相關問題並克服與改善。

關鍵字

人工智慧；深度學習；瑕疵檢測；卷積神經網路

並列摘要

Deep Learning is a kind of learning of data and images. It uses network structure to build and algorithm calculations, so that the judgment and thinking of the machine are close to our preset goals, helping the process detection and process to be faster and more convenient. Based on the OpenR8 software, this study uses Convolutional Neural Network (CNN) as the framework, uses Caffe as a communication bridge, and performs flaw detection on PCB electronic parts images. Through semi-supervised learning mode, the machine training is further improved.Efficient and accurate, and analysis of the type of help the quality inspection personnel to screen on the production line, effectively reducing the underkill rate and the overkill rate also reduce the burden on the visual inspection staff, improve the overall production process Efficiency, follow-up can also be used to analyze the process capability to find out the relevant problems and overcome and improve.

並列關鍵字

Artificial intelligence ； Deep learning ； Flaw detection ； Convolutional neural network

國際替代計量

應用人工智慧技術作基板缺陷檢測

全文下載

主題瀏覽