文献类型：会议论文

英文题名：Pedestrian Detection Method Based on Faster R-CNN

作者：Hui Zhang[1]; Yu Du[2]; Shurong Ning[1]; Yonghua Zhang[1]; Shuo Yang[1]; Chen Du[3]

机构：[1] Smart City Coll., Beijing Union Univ., Beijing, China; [2] Coll. of Robot., Beijing Union Univ., Beijing, China; [3] Beijing Key Lab. of Inf. Service Eng., Beijing Union Univ., Beijing, China

第一机构：北京联合大学继续教育学院

会议日期：15-18 Dec. 2017

会议地点：Hong Kong, China

语种：英文

外文关键词：convolution - feature extraction - feedforward neural nets - image classification - learning (artificial intelligence) - object detection - object recognition - pattern clustering - pedestrians - traffic engineering computing

摘要：Pedestrian detection based on computer vision is an important branch of object recognition, which is applied to intelligent monitoring, intelligent driving, robot and so on. At present, many pedestrian detection methods are proposed. However, because of the complexity of the background, pedestrian posture diversity and pedestrian occlusions, pedestrian detection is still a challenge which calls for precise algorithms. In this paper, the fast Region-based Convolutional Neural Network (Faster R-CNN) is used. Firstly, image features were extracted by CNN. After that, we built up a Region Proposal Network to extract regions that might contain pedestrians combined with K-means cluster analysis. And the region is identified and classified by detection network. Finally, the method was tested in the INRIA data set. The results show that the method of pedestrian detection based on Faster R-CNN, which achieves the accuracy of 92.7%, performs better, compared with other algorithms.