CAJ | 학술논문

研究在动态背景下对视频图像序列进行运动目标的检测。主要包括三个步骤，分别是运动估计，运动补偿和目标检测。在运动估计中采用的主要是基于特征点匹配算法。这种算法与传统的块匹配算法最大的好处在于它的数据量少，计算简单迅速而且图像的匹配可靠性更高。最后用计算机视觉库OpenCV进行实现。
연구재동태배경하대시빈도상서렬진행운동목표적검측。주요포괄삼개보취，분별시운동고계，운동보상화목표검측。재운동고계중채용적주요시기우특정점필배산법。저충산법여전통적괴필배산법최대적호처재우타적수거량소，계산간단신속이차도상적필배가고성경고。최후용계산궤시각고OpenCV진행실현。
The moving object detection in dynamic background in the video image sequence is studied. The method includes three steps, they are motion estimation, motion compensation and object detection. At the motion estimation, an algorithm based on the feature points matching is taken. The advantages of this algorithm is that it needs fewer data and indicates faster calculating speed compared to the block matching algorithm and the matching of the video image sequence is more reliable. At last the algorithm is implemented using an computer vision library named OpenCV.