摘要
针对现有行人检测方法速度慢、无法满足实时性检测需求的缺点,提出一种基于边缘对称性和改进的等价局部二值模式的行人检测方法 ES-IULBP(Edge Symmetry and Improved Uniform Local Binary Patterns)。该方法首先对输入的图像进行垂直边缘提取并计算对称性,完成行人的初检测,确定行人候选区;然后引入等价局部二值模式,并对其改进,进行行人的纹理特征提取;最后结合线性支持向量机进行行人验证。实验结果表明,与基于梯度方向直方图特征的行人检测方法相比,ES-IULBP检测速度快、准确率高,并具有较强的鲁棒性。
In order to overcome the drawbacks of existing pedestrian detection method that its speed is slow and cannot meet real-time de-tection requirement,we proposed a pedestrian detection method which is based on edge symmetry and improved uniform local binary patterns, namely ES-IULBP.First,the method extracts the vertical edges of the inputting image and computes their symmetry to pre-detect the pedestri-an,and sets the candidate region of pedestrians.Then it introduces the uniform local binary patterns and improves it,and extracts the texture feature of pedestrian,at last it combines the linear support vector machine to validate the pedestrian.Experimental result proves that com-pared with the pedestrian detection methods based on histogram feature in gradient orientations,the ES-IULBP method has fast detection speed and high accuracy,and has stronger robustness as well.
出处
《计算机应用与软件》
CSCD
2015年第10期216-219,247,共5页
Computer Applications and Software
关键词
行人检测
边缘对称
等价局部二值模式
梯度直方图特征
线性支持向量机
Pedestrian detection
Edge symmetry
Uniform local binary patterns （ULBP）
Gradient histogram feature
Linear support vector machine
