摘要
运动目标检测与识别研究是计算机视觉中最重要的研究领域之一。在目标检测中存在场景动态变化、光照变化等同题,在目标跟踪中则存在目标的遮挡、重叠及目标关联等问题。提出一种有效的运动目标检测方法,较好地解决以上问题。首先利用背景差分方法建立背景模型,再对背景模型进行实时更新,以适应视频本身和光线的变化,最后使用形态学方法消除噪声和运动阴影带来的影响。并对检测到的目标应用区域跟踪技术,引入2个参数,实现跟踪匹配,很好地处理了目标之间的相互遮挡问题。实验结果表明,该方法快速有效,能够满足实时的需要。
Detection and tracking of moving objects is one the most important parts of computer vision and video processing.The difficulties,such as how to deal with dynamic changing scenes,moving objects' overlap and occlusion on detection and recognition effectively are perplexing people.An effective method was proposed to solve these problems.First,an effective detection model of moving object was built,which addressed the problems aforementioned.And then,the background model was updated realtimely in order to accommodate the changes of illumination and others.After that,morphological operation was used to reduce the negative impact of disturbance and noise.Regional tracking technology was applied to detected targets and two parameters were introduced to process the overlapping between many objects.Experimental results showed that the model has a better performance in effectiveness and realtime aspects.
出处
《河北北方学院学报(自然科学版)》
2015年第6期11-15,共5页
Journal of Hebei North University:Natural Science Edition
关键词
视频信息处理
背景差分
目标跟踪
目标检测
video information processing
background modeling
object tracking
object detection