摘要
为了解决传统车辆检测存在的问题,提高车辆检测的准确度,本文提出将区域卷积神经网络算法应用到车辆检测中。该算法利用图像的颜色层次特征,获取潜在的车辆候选区域;建立卷积神经网络结构,使用车辆样本库进行特征训练,提取候选区域特征;选定正负样本进行SVM分类器训练,采用SVM分类器进行最终的候选区域分类,最后得到车辆信息。本文使用的算法能够检测出图像中的车辆,剔除非车辆区域,有效提高车辆检测的准确性,并且具有一定的实时性。
In order to solve the existing problems of vehicle detection and improve the accuracy of vehicle detection, the regional convolution neural network algorithm is applied to vehicle detection. The algorithm uses the color hierar-chy of the image to obtain the potential vehicle candidate area. The convolution neural network structure is estab-lished, and the vehicle sample library is used for feature training to extract the candidate region characteristics. The SVM classifier is selected for the positive and negative samples, and the final candidate region is classified by SVM classifier. Finally, the vehicle information is obtained. The algorithm used in this paper can detect the vehicle in the image, eliminate the non-vehicle area, improve the accuracy of vehicle detection and it has a certain real-time.
作者
封晶
Feng Jing(School of Information Engineering,Jiangxi University of Science and Tecnnology,Jiangxi Ganzhou 341000)
出处
《科技广场》
2017年第3期10-14,共5页
Science Mosaic
关键词
候选区域
车辆检测
卷积神经网络
深度学习
支持向量机
Candidate Area
Vehicle Detection
Convolution Neural Network
Depth Learning
Support Vector Na- chine