摘要
为实现动态背景下运动目标的实时检测,提出一种基于稀疏光流场分割的目标检测算法。通过块匹配法计算相邻两帧图像间的稀疏光流场,利用K均值聚类算法分割光流场获取运动目标候选区域。由于初步提取的运动目标存在目标空洞、边缘不平滑等特性,将光流场分割图和彩色图像分割结果进行融合,检测完整运动目标。对比实验与分析结果表明,该方法对于动态背景下的目标有很好的检测性,能够完整地提取运动目标,检测效率满足实时性检测的需求。
To achieve moving object real-time detection in dynamic scenes,an algorithm based on sparse optical flow field segmentation was proposed.The sparse optical flow field between two adjacent frames was calculated by block matching,and the K-Means cluster algorithm was used to segment the field to obtain the candidate region of moving object.Due to the target candidate existing target hole and rough edge,sparse optical flow field segmentation and color image segmentation were fused to obtain complete moving object.The comparative experiments and analysis illustrate that the proposed method shows good detection performances in terms of moving objects under dynamic scenes.It can extract the whole moving target and the detection efficiency meets the needs of real-time detection.
作者
李果家
李显凯
LI Guo - jia LI Xian - kai(Faculty of Land Resource Engineering, Kunming University of Science and Technology, Kunming 650093,Chin)
出处
《计算机工程与设计》
北大核心
2017年第11期3029-3035,共7页
Computer Engineering and Design
基金
国家自然科学基金地区基金项目(41561082
41161061)
关键词
动态背景
实时检测
稀疏光流场
块匹配
K均值聚类
图像分割
dynamic scenes
real-time detection
sparse optical flow field
block matching
K-Means cluster
image segmentation
