近年來近視人口急遽的上升，且年紀介於16到18歲的學生近視率也達到80%左右，在考量到未來高度近視的人數也逐漸攀升，導致為來可能有視覺障礙的人口也會逐漸增加，對於這種狀況也會使照顧人員的數量不足。 本論文提出一種利用樹莓派作為邊緣裝置的盲人輔助系統，主要提升視覺障礙人口生活的自主性，減緩照護所需的人力資源。此系統運用了深度學習技術，結合優化後的YOLOv5-Lite模型，能夠提供導航、物品辨識和環境感知等功能的裝置，並利用MQTT等相關通訊協議，能夠及時操作多項裝置，而為了兼顧隱私和資料安全，系統採用邊緣運算核心，這不僅保護了個人隱私，也降低所需成本。 ;In recent years, the number of people with myopia has risen sharply, with the myopia rate among students aged 16 to 18 reaching approximately 80%. As the number of individuals with high myopia continues to grow, this trend may lead to an increase in the population with visual impairments in the future, potentially resulting in a shortage of caregiving personnel. This paper proposes a blind assistance system utilizing Raspberry Pi as an edge device, aimed at enhancing the independence of people with visual impairments and reducing the demand for human caregiving resources. The system employs machine learning techniques, incorporating an improved YOLOv5-Lite model to deliver functions such as navigation, object recognition, and environmental awareness. By using communication protocols like MQTT, the system is capable of operating multiple devices in real time. To address privacy and data security concerns, the system is designed with an edge computing core, which not only protects personal information but also reduces overall costs.
