摘要
针对视频环境下行人检测多数采用窗口滑动方法识别慢、不能快速找到行人窗口的缺点,提出了一种基于组合算法的行人目标识别方法,利用高斯混合模型方法提取视频中的运动前景,划定一个泛目标窗口,再使用HOG-l bp联合特征训练的分类器对泛目标窗口进行分类,得到分类结果,对行人目标进行标记。经实验验证:该方法相对于当前行人检测方法,检测速度和正确率都取得了很好的效果。
Aiming at shortcomings of pedestrian detecting methodes mostly use window slipping, and the method can't quickly find suspected pedestrian windows in video environment, propose a method to detect pedestrian based on combination algorithms. Firstly sports foreground in video is extracted by Gaussian mixture model (GMM) method, delimit a suspected pedestrian windows. Then, use classifier based on hog-lbp combined feature training to classify generic target window, and get classification, label pedestrian target result. Through verification of experiment,the method get a good resul: on detecting speed and accuary compared with current pedestrian detection method.
出处
《传感器与微系统》
CSCD
2017年第7期150-153,共4页
Transducer and Microsystem Technologies