基于机器学习的级联行人检测方法研究

Research on Cascaded Pedestrian Detection Technology Based on Machine Learning

针对传统的HOG+SVM行人检测模型易受行人所处位置、影像大小和背景干扰等因素影响的问题,提出了一种基于选择性搜索和机器学习的级联行人检测算法。综合运用密集搜索、分割搜索和随机采样搜索方法,改进选择性搜索产生的候选窗口,可以减少大量非行人窗口的数量。将HOG特征和Alex Net模型相结合,可有效提高行人检测的效率,降低误检率和漏检率。

The traditional HOG + SVM pedestrian detection model is easily affected by the location,image size and background interference of pedestrians,so a cascaded pedestrian detection algorithm based on selective search and machine learning is proposed. The number of non-pedestrian windows can be reduced by using the methods of intensive search,segmentation search and random sampling search to improve the candidate windows generated by selective search. The combination of HOG features and AlexNet model can effectively improve the efficiency of pedestrian detection and reduce the false detection rate and missed detection rate.