基于卷积神经网络的小目标交通标志检测算法

A Convolutional Neural Network-Based Method for Small Traffic Sign Detection

摘要:

PVANet（performance vs accuracy network）卷积神经网络用于小目标检测的检测能力较弱.针对这一瓶颈问题,采用对PVANet网络的浅层特征提取层、深层特征提取层和HyperNet层（多层特征信息融合层）进行改进的措施,提出了一种适用于小目标物体检测的改进PVANet卷积神经网络模型,并在TT100K（Tsinghua-Tencent 100K）数据集上进行了交通标志检测算法验证实验.结果表明,所构建的卷积神经网络具有优秀的小目标物体检测能力,相应的交通标志检测算法可以实现较高的准确率.

Abstract:

In order to solve the small target detection convolutional neural network algorithm, the PVANet convolutional neural network structure was improved to conduct the experiments of traffic sign detection on the TT100K traffic sign data set. The shallow feature extraction, deep feature extraction, and HyperNet multilayer feature fusion modules were improved. The experimental results show that the improved PVANet convolutional neural network has an excellent detection ability for small target objects, and the corresponding traffic sign detection algorithm can achieve a higher mAP (mean average precision).