應用電腦視覺方法於水果品質檢測之研究

字體大小:A- A A+

單位商業會員附件下載:
論文名稱: 應用電腦視覺方法於水果品質檢測之研究
研究生姓名: 林念德
指導教授姓名: 温志皓
出版年: 2021
學校名稱: 國防大學
系所名稱: 運籌管理學系
關鍵字: 品質檢測;電腦視覺;卷積神經網路;YOLO;Quality inspection;Computer vision;Convolutional neural network;YOLO
摘要: 隨著經濟快速發展、生活水準提升及健康意識抬頭,消費者對於追求更高品質的水果需求不斷增長。然而,從農人口的老化及全球化競爭等現象,世界各國面臨糧食安全與資源永續發展的問題引發高度關注,更影響政府對農業決策的推動方向,同時帶動水果品質評估的技術發展。外部品質對於水果是非常重要的感官屬性,它不僅攸關市場價值、消費者偏好及選擇,更可能反映出內部品質。在過去的幾十年中,新興的電腦視覺系統和影像辨識在食品行業中獲得廣泛使用,隨著資通訊技術與設備日益精良,「深度學習」技術所發展的「卷積神經網路」架構已被證明成為水果品質檢測的強大工具。然而,在臺灣果品市場及國軍日常需求量較大的水果品項當中,如何將水果於提供至消費者面前時,即預先妥善完成品質監控是極為關鍵的問題。本研究成功藉由「YOLOv3」演算法延伸出非破壞性、快速且正確率接近100% 的電腦視覺方法,輔助費時耗力且缺乏客觀性的人工目視檢驗,同時有效提升國軍副食供應站的水果品質分類及檢驗作業效率,對於未來國軍應用於物聯網技術或建置雲端系統之參考價值有極大助益。With the rapid economic development, rising living standards and rising health awareness, consumer demand for higher quality fruits is growing. However, from the phenomenon of aging agricultural population and global competition, the problems of food security and sustainable development of resources in all countries of the world have aroused great concern, which affects the direction of government's agricultural decision-making, and promotes the technological development of fruit quality assessment. External quality is a very important sensory attribute for fruit, which not only concerns market value, consumer preferences and choices, but also reflects internal quality to a certain extent. In the past few decades, emerging computer vision systems and image recognition have been widely used in the food industry, and with the increasing sophistication of communication technologies and equipment, the ""Deep Learning"" technology has proven to be a powerful tool for fruit quality testing. Despite this, in Taiwan's fruit market and the Taiwan Army's daily demand for fruit items, how to provide fruit in front of consumers, that is, to properly complete quality control in advance is a critical issue. The successful extension of non-destructive, fast and correct computer vision method by the ""YOLOv3"" algorithm, auxiliary time-consuming and labor-consuming and lack of objectivity of manual visual inspection, while effectively improving the fruit quality classification and inspection operation efficiency of Taiwan's side food supply station, is of great benefit to the future reference value of Taiwan Army's application to IoT technology or the construction of cloud systems.
資源連結: 前往查看