基于深度卷积神经网络特征层融合的小尺度行人检测

针对目前行人检测算法计算量过大和对小尺度行人检测精度不高的问题,提出了一种基于深度卷积神经网络特征层融合的小尺度行人检测方法,设计了一种包含9个卷积层的深度神经网络架构.在进行行人检测时,首先,对输入图像进行分块预处理操作,避免损失原始图像的视觉信息;然后,将网络不同层的卷积特征进行融合,提升行人特征的区分能力和表达能力,进而提升行人检测的精度,在保证检测精度的同时有效降低网络的复杂度.在INRIA、Caltech等公共行人数据集上的实验结果表明,所提出的行人检测方法能够有效检测小尺度的行人,且网络架构的参数量更少,检测速度更快,能得到更高精度的行人检测结果.

Small Scale Pedestrian Detection Based on Feature Layers Fusion of Deep Convolutional Neural Network

In order to solve the problems of excessive computation of pedestrian detection algorithm and low detection accuracy of small scale pedestrian detection,a small scale pedestrian detection method based on the fusion of feature layers of deep convolutional neural network (CNN) is proposed,which contains nine convolutional layers.In the process of pedestrian detection,the input image was preprocessed into sub-blocks to avoid the loss of the visual information of the original image.The different convolutional features of CNN were fused to improve the ability of distinguishing and expressing pedestrian characteristics,thereby improving the accuracy of pedestrian detection.While ensuring the detection accuracy,the implementation complexity of the network was reduced effectively.Experimental results on public pedestrian datasets,such as INRIA and Caltech,show that the proposed pedestrian detection method can effectively detect small scale pedestrians,with fewer network architecture parameters,faster detection speed,and higher accuracy of pedestrian detection.