行人偵測在許多應用中扮演著重要的角色，尤其是自動車輛的安全駕駛輔助系統。這是一個富有挑戰性的題目，因為人體的動作有很大的變化性。本論文提出了一個行人偵測的駕駛輔助系統；即是在移動的車子上執行行人偵測。我們只用一支攝影機即完成整個行人偵測的過程，且準確率高、執行速度快。系統的架構包含了行人偵測模組和追蹤模組。在行人偵測模組中，我們整合了AdaBoost (Adaptive-Boosting)和樣板比對法。首先，我們利用AdaBoost方法訓練我們的行人偵測分類器，在影片中利用此分類器挑選看起來像人的區域。第二，針對這些區域來比對行人的形狀驗證是否出現行人。在追蹤模組中，我們採用CAMSHIFT (Continuous Adaptive Mean Shift) 追蹤演算法來追蹤行人。 實驗結果顯示此系統能夠接近即時的處理並且在複雜背景下擁有高度的準確率。測試的影片解析度為320 x 240每秒30個畫面，行車平均速度每小時40至50公里。由實驗可證實本論文所提出行人偵測的方法是可行且有效的。 Pedestrian detection plays an important role in several applications, especially in driving assistance system or autonomous vehicle. However, it is difficult to detect pedestrians because of the variations of walking style and the movement of human body over time even for the same person. In this thesis, we propose a pedestrian detection system for driving assistance system. The ultimate goal of the system is detecting pedestrian from a moving vehicle. In the proposed system, we only use a single video camera in our practice yet we can achieve fast and accurate detection results. The system consists of two main modules including pedestrian detection module and pedestrian tracking module. AdaBoost (Adaptive-Boosting) method and template matching technique are integrated in the pedestrian detection module. Firstly, we use AdaBoost to train the pedestrian detector and select pedestrian candidates over the image sequences. Secondly, we apply template matching technique to verify pedestrian appearances. In the pedestrian tracking module, we adopt CAMSHIFT (Continuously Adaptive Mean Shift) algorithm to track the mis-detected pedestrian. The experimental results show that the system is fast and near real time. Moreover, high accuracy detection rate can be accomplished even against cluttering backgrounds. In our experiments, the test set video sequences are taken from the real scene with 320 x 240 image resolutions and the vehicle speed is 40-50 km/h in average. The results reveal that the proposed method is feasible and efficient in detecting pedestrian.