摘 要
本研究提出「基於影像處理與機器學習技術之聚丙烯（Polypropylene，PP）共擠壓（co-extrusion）保護膜分層厚度分析系統」，係由保護膜截面取樣裝置、數位顯微鏡、後端數據分析平台所組成。保護膜截面取樣裝置採3D列印技術設計製成，用於裁切取得保護膜截面（cross-section），進而置於數位顯微鏡以拍攝截面影像。後端數據分析平台主要包含分層厚度分析模組、數據可視化模組、與資料庫模組。分層厚度分析模組以本研究所提出之影像處理結合機器學習分析機制，計算出截面影像中保護膜之基材層（base layer）與黏著層（adhesive layer）之厚度與比例等數據。這些數據可進一步透過數據可視化模組呈現，並記錄於資料庫模組中。實際測試結果顯示，所提之分析機制之準確度達92％，可實現自動分析之目的。經由本系統之研製，保護膜產品品質檢測流程得以標準化與資訊化，亦同時縮短檢測作業時間，進而實現產品品管數位轉型之目標。
關鍵詞：保護膜、共擠壓、影像處理、機器學習、數位轉型
Abstract
This study proposes a layered thickness analysis system for polypropylene (PP) co-extrusion protective films, utilizing image processing and machine learning. The system comprises a 3D-printed film cross-section sampling device, a digital microscope, and a back-end data analysis platform. The sampling device is used to obtain the cross-section of the film, which is then imaged by the digital microscope. The back-end data analysis platform consists of a layered thickness analysis module, a data visualization module, and a database module. The proposed image processing with machine learning scheme accurately calculates the thickness and ratio of the film's base and adhesive layers in the cross-section image. These results can be further presented through the data visualization module and recorded in the database module. The experimental results show that the accuracy of the system has reached 92%, achieving automatic analysis. This system can standardize and digitize the quality inspection process for film products, shorten inspection time, and realize the digital transformation of product quality control.
Keywords: Protective film, Co-extrusion, Image processing, Machine learning, Digital transformation