Abstract: Pedestrian detection remains an important task in the theory research and practical application of object detection. Designing an efficient describing method to extract the pedestrian features and applying classifier to realize dichotomy is a popular way in the area of pedestrian detection at present. Deep convolutional neural network has made great success on image and audio, which is the important component of deep learning. Artificial designed methods of feature extracting have an imperfect description of pedestrian in the complex background. To solve this problem, a method based on deep convolutional neural network with multi-layers is proposed. This paper analyzes the influence to the detection accuracy of the number of layer, the size of convolution kernel and the dimension of features, which provides an reference to optimize the related parameter. The experiment results show that the proposed method is a feasible way to detect pedestrian and perform an higher accuracy compared with the traditional methods on the self-build dataset.

Key words: pedestrian detection, deep learning, convolutional neural network, feature extracting

摘要： 行人检测一直是目标检测研究与应用中的热点。目前行人检测主要通过设计有效的特征提取方法建立对行人特征的描述，然后利用分类器实现二分类。卷积神经网络作为深度学习的重要组成，在图像、语音等领域得到了成功应用。针对人工设计的特征提取方法难以有效表达复杂环境下行人特征的问题，提出采用多层网络构建深度卷积神经网络实现对行人检测的方法。系统分析了卷积神经网络层数、卷积核大小、特征维数等对识别效果的影响，优化了网络参数。实验结果表明该方法对于行人检测具有很高的识别率，优于传统方法。

关键词: 行人检测, 深度学习, 卷积神经网络, 特征提取