摘要
行人检测是计算机视觉中的关键技术之一,在智能交通领域有大量实际应用,如何在提高行人检测精度的同时提高检测速度一直是研究的热点.首先采用基于高斯混合模型的背景建模方法分离出运动目标,将原始视频序列转换为二值图片,得到大量固定大小的训练样本;然后提取样本图片的HOG特征,通过SVM训练得到分类器;接着用固定大小的滑动窗口检测行人,并提出了一种滑动窗口优化算法来筛选检测结果;进而用前景像素密度估算方法调整检测结果,输出最终统计人数,最后实验表明方法的有效性.
Pedestrian detection is one of the key technologies in computer vision. It has applied in intelligent transportation field widely. How to improve the detection precision as well as the detection speed is a hot research topic. Background modeling method based on Gauss mixture model is used in this paper to separate the moving target from the background. Then the original video sequence can be converted into binary image for training. Then the HOG feature in the sample images are extracted through the Support Vector Machines. A size-fixed sliding window is used to detect pedestrians. Here an optimization algorithm using sliding window is proposed to screen the test results. Then the foreground pixel density estimation method is used to adjust the detection result. Finally the experimental results show the proposed method is effective.
出处
《浙江工业大学学报》
CAS
北大核心
2015年第2期212-216,共5页
Journal of Zhejiang University of Technology
基金
国家自然科学基金资助项目(61104095)
关键词
行人检测
高斯混合模型
HOG
背景建模
pedestrian detection
Gaussian mixture model
HOG
background modeling
