在狹窄且雜亂的環境中，微型無人機的自主避障是一項極具挑戰性的任務。
由於微型無人機所搭載的嵌入式晶片功耗極低，因此難以負擔複雜的計算需求。
為解決此限制，本研究提出一種基於 Lukas-Kanade Pyramid 光流演算法的避障解
決方案，此演算法具備非常低的運算複雜度，特別適合在資源有限的微型無人機
嵌入式晶片上運行。
本研究所採用的微型無人機搭載之晶片架構，不僅適合低功耗的即時處理，
且內建多核心同步處理資源，研究中即充分利用此特性以優化光流演算法的執行
效率，大幅提升運算速度與性能。本避障系統透過前視攝像頭即時捕捉環境影像，
利用光流計算即時估測物體的運動狀態，並將運算得到的數值即時轉換成飛行控
制命令，使無人機能夠迅速並自主地避開障礙物。實驗結果顯示，所提出的基於
光流的避障系統於實際飛行測試中表現穩定且可靠，能有效應對複雜環境下的導
航挑戰，證實此方案在微型無人機自主避障領域中的實用性與效能。

In narrow and cluttered environments, autonomous obstacle avoidance for micro drones presents a significant challenge. Due to the extremely low power consumption of the embedded chips used in micro drones, complex computational tasks are challenging to perform. To address this limitation, this research proposes an obstacle avoidance solution based on the Lukas-Kanade Pyramid optical flow algorithm, which features very low computational complexity, making it especially suitable for execution on resource-constrained embedded chips in micro drones.
The micro drone's embedded chip architecture adopted in this study not only supports low-power real-time processing but also includes built-in multi-core synchronous processing capabilities. The study leverages this feature to optimize the execution efficiency of the optical flow algorithm, significantly enhancing computational speed and performance. The system captures environmental imagery in real-time using a forward-facing camera and employs optical flow calculations to estimate object motion instantly. These computational results are immediately converted into flight control commands, enabling the drone to quickly and autonomously avoid obstacles. Experimental results demonstrate that the proposed optical flow-based obstacle avoidance system performs stably and reliably in real-world flight tests, effectively managing navigation challenges in complex environments, thereby validating the practicality and effectiveness of this solution for autonomous obstacle avoidance in micro drones.