摘要
为了提高不同光线环境下阴影检测的准确度和稳定性,提出了一种自适应的阴影检测算法。设计了一种阴影检测器,利用候选前景中像素YUV分量变化比率判别阴影像素,其检测阈值由阈值估计器得到。阈值估计器利用全局纹理和抽样推断的方法统计计算出当前光线环境下所需的阈值。整个阴影检测过程不需要人工干预,适应于各种复杂动态的场景。对代表不同光线条件的标准测试视频的检测实验表明,本文算法能够自适应地检测得到各目标阴影区域,具有较好的稳定性和实时性,综合检测指标达到94%以上。
In order to improve the accuracy and stability of the shadow detection under di{ferent lighting conditions,an adaptive shadow detection algorithm is proposed. A shadow detector is designed to identify shadow pixds which use the change ratios of YUV components as conditions. The detection thresholds are calculated by a threshold estimator. Under different lighting conditions, the estimator uses global tex- ture and sampling deduction methods to statistically calculate the needed thresholds. The whole detecting process requires no manual intervention and can adapt to any dynamic and complex scene. Experiments on standard test videos show that this algorithm has good stability and real-time performance and can a- daptively detect the shadow area of videos under different lighting conditions; The average comprehensive index of the proposed algorithm can reach more than 94%.
出处
《光电子．激光》
EI
CAS
CSCD
北大核心
2012年第11期2174-2179,共6页
Journal of Optoelectronics·Laser
关键词
阴影检测
YUV色彩空间
边缘检测
全局纹理
抽样推断
自适应阈值
shadow detection
YUV color space
edge detectiom global texture； sampling deduction
a daptive threshold
