Abstract: Pedestrian detection is an important research direction of object detection. A problem that the pedestrian detection algorithm misses detection in the case of complex scenes and objects that are too small. Based on the Faster R-CNN detection algorithm, a deep network pedestrian detection based on shallow feature fusion guidance is proposed. Through the HOG feature, the improved LBP features and deep network features are combined to obtain accurate pedestrian characteristics, and a series of experiments are carried out on the internationally widely used pedestrian dataset. The results show that the proposed improved method has better performance in detecting accuracy and rate.

Key words: pedestrian detection, convolutional neural network, feature fusion

摘要： 行人检测是目标检测中的一个重要研究方向。针对行人检测算法在复杂场景和目标太小情况下漏检的问题，在Faster R-CNN检测算法的基础上，提出一种基于浅层特征融合引导的深层网络行人检测。通过HOG特征、改进的LBP特征与深度网络特征融合获得准确的行人特征，在国际上广泛使用的行人数据集上进行一系列实验。结果表明，所提出的改进方法在检测准确率和速率方面都有所提高。

关键词: 行人检测, 卷积神经网路, 特征融合