摘要
提出了一种基于深度神经网络的车辆特征识别方法,通过车辆特征智能检测识别实现交通智能监控和管理.采用三维区域轮廓扫描方法进行车辆图像采集和几何形状判断,对采集的车辆图像进行边缘轮廓检测和信息增强处理,突出车辆的类别属性特征点,在仿射不变区域对车辆角点分布信息进行直方图均衡化处理,实现车辆像素特征点的提取.对提取的像素特征点采用深度神经网络进行分类训练,实现车辆特征的智能识别.选取大量交通视频图像进行实验,仿真结果表明采用该方法进行车辆特征识别的成功率较高,输出车辆特征点正确的像素总数较多,对目标车辆的准确检测定位性能较好.
An identification method of vehicle features depth based on neural network is proposed,it can realize the vehicle intelligent identification for intelligent traffic monitoring and management. By using 3D contour scanning method for vehicle image collectionand geometry of judgment,and the vehicle image edge detection and contour information enhancement,the categorical feature pointshighlight the vehicles. In the affine invariant region of the vehicle corner distribution information histogram is equalized,and the pixelfeature extraction is realized. The vehicle classification training of pixel feature extraction using deep neural network,intelligent recognition of vehicle features is realized. Through a large number of traffic video images,the simulation results show that using the methodthe success rate of vehicle feature recognition is high,and the total number of pixels of the vehicle is more accurate,which shows thatthe detection and localization of the target vehicle is high.
出处
《河南工程学院学报(自然科学版)》
2017年第4期44-48,共5页
Journal of Henan University of Engineering:Natural Science Edition
关键词
深度神经网络
车辆
特征识别
分类
像素
deep neural network
vehicle
feature recognition
classification
pixel