摘要
针对复杂场景条件下运动目标检测方法存在的局限性,提出了一种基于运动检测和静止图像分割相融合的算法。采用相邻帧差法结合建立的假设检验模型进行自适应的运动目标检测;为消除孔径效应和噪声的影响,根据运动目标检测的结果,在当前帧利用区域增长法融合运动分割的结果。试验结果表明,算法能从复杂场景的图像序列中有效地检测和提取出运动目标,并有很强的鲁棒性。
In view of the limits existing in detecting moving targets in complex scenes, a fusion algorithm based on moving targets detection and static image segmentation is proposed. Adaptive moving targets can be detected by using a hypothesis test model and the adjacent frame difference. In order to eliminate aperture effect and noise influence, the moving segmentation results are fused by means of region-growing method in current frame according to the detected results of the moving targets. The test results show that the moving targets can be effectively detected and extracted from the image sequence of the complex scenes. The proposed algorithm is effective with strong robustness.
出处
《光电工程》
CAS
CSCD
北大核心
2004年第B12期36-39,共4页
Opto-Electronic Engineering
基金
北京市南中轴路 BRT 智能公交系统项目
关键词
目标探测
计算机视觉
活动目标
鲁棒性
Target detection
Computer vision
Moving target
Robustness
