在視訊監控、機器人視覺的應用，物件偵測與追蹤扮演著重要的角色。本論文致力於提出一個強健且高效率的演算法，以便進一步實現即時動態物件偵測和追蹤的嵌入式系統。 在移動物偵測方面，我們採用漸進法建構背景，在變動的背景下，仍可適應性的更新的背景資訊，以便可靠地擷取出移動物件。接著我們應用PSO演算法進行移動物之追蹤。先建立移動物樣版，並利用粒子群的全域最佳化搜尋特性進行非線性、非固定數目的移動物追蹤。經由單人、多人、及有遮蔽情況的追蹤實驗，我們的方法在準確性與執行效率均較傳統方法為佳。 最後我們將此一演算法移植到嵌入式系統上，結合Pan/Tilt攝影機架設於移動式機器人之頭部進行即時的移動物偵測和追蹤。此一系統在極其有限的記憶體資源和低速的處理器的硬體資源下，以達到即時運算的性能需求。 Visual tracking has been a popular application in computer vision, for example, public area surveillance, and robot vision, etc. This paper presents a robust and efficient algorithm for detecting and tracking moving objects, so that we can achieve an embedded system for real-time detection and tracking of moving objects. For detection, we utilize a progressive and adaptive background generation. Even if the unstable noise, for example, light changed and fluttering leaves, we still extract the foreground objects exactly. Then particle swarm optimization (PSO) is used for tracking as a search strategy. First, build the target model, and through the PSO, we can track moving objects in the nonlinear system. Experiment shows that the proposed method can track the single person, multiple people even when occluded, and is more efficient and accurate than the traditional methods. Eventually, we transplant the algorithm into the embedded system, and the CPU is ARM Cortex-M3. The COMS sensor is placed on a pan/tilt platform in order to track the moving object. Limited on the CPU and memory, the experiments and analysis still show the efficiency.
