本論文提出了一種檢測移動物體的方法,適用於靜態攝影機。在許多影像處理的應用中,移動物體的檢測屬於基本步驟,目的在於縮小檢測的範圍,讓處理速度大幅提升。無論是車牌辨識、監視系統中之移動偵測、視訊壓縮…等等,都能派上用場。本論文所提出的方法,是基於背景相減技術,使用時間軸平均值法建立簡單背景模型,根據後續檢測結果,排除移動物體像素,使背景模型更加準確。由於時間軸平均值法存在殘影問題,利用時間差異法進行改善。再對物體進行簡易的邊緣修補,解決時間差異法所造成的物體破碎問題,並對物體進行追蹤,取得物體移動軌跡,若長時間在相同位置微動,則判定為會移動的背景物體。本研究在實驗結果顯示,無論是處理速度或是準確率,都能有不錯的表現。
In this paper, we propose a method for detecting a moving object, for stationary camera. In many video processing applications, the detection of a moving object belongs to the basic steps used to reduce the detection range, so that the processing speed is increased significantly. Such as license plate recognition, surveillance system for motion detection, video compression ... and so on, can use this method. The proposed method is based on background subtraction techniques. Selective update using temporal averaging to create a simple timeline background model. According to the above results exclude the pixels of moving objects, making the background model is more accurate. Since the selective update using temporal averaging method has residual shadow problem, use the temporal differencing method improvement. And then the edges of objects were simple fixes to solve broken object problems caused by temporal differencing. Finally, the object tracking to obtain the movement locus of the object, if the jog in the same position for a long time, it is determined that the background move object. The experimental results show that processing speed or accuracy, can have a good performance.