一种基于卷积神经网络的车辆检测方法

作者:叶运生; 刊名:农业装备与车辆工程 上传者:常允艳

【摘要】介绍了一种基于卷积神经网络的车辆识别方法。该方法首先对车道线进行边缘检测,采用车道线模型进行匹配,从而确定道路感兴趣区域。然后采集道路视频,对其中的车辆目标进行标注,制作车辆数据集,再设计一种卷积神经网络,利用车辆数据集训练检测器,使检测器适应于车辆二分类识别的任务。最后在道路感兴趣区域中检测车辆。相较于传统的车辆识别方法,该方法具有较好的准确性与鲁棒性,在复杂行驶环境下的识别效果令人满意。

全文阅读

第 57 卷 第 2 期 Vol. 57 No. 2 2019 年 2 月 February 2019 农业装备与车辆工程 AGRICULTURAL EQUIPMENT & VEHICLE ENGINEERING doi:10.3969/j.issn.1673-3142.2019.02.011 一种基于卷积神经网络的车辆检测方法 叶运生 (230009 安徽省 合肥市 合肥工业大学 汽车与交通工程学院) [ 摘要 ]介绍了一种基于卷积神经网络的车辆识别方法。该方法首先对车道线进行边缘检测,采用车道线模型进行匹配,从而确定道路感兴趣区域。然后采集道路视频,对其中的车辆目标进行标注,制作车辆数据集,再设计一种卷积神经网络,利用车辆数据集训练检测器,使检测器适应于车辆二分类识别的任务。最后在道路感兴趣区域中检测车辆。相较于传统的车辆识别方法,该方法具有较好的准确性与鲁棒性,在复杂行驶环境下的识别效果令人满意。 [ 关键词 ] 卷积神经网络;车辆识别;鲁棒性 [ 中图分类号 ] TP391 [ 文献标识码 ] A [ 文章编号 ] 1673-3142(2019)02-0044-05 Deep Convolutional Neural Networks for Vehicle Detection Ye Yunsheng (School of Automotive and Traffic Engineering, Hefei University of Technology, Hefei City, Anhui Province 230009, China) [Abstract] This paper introduces a method of vehicle identification based on convolution neural network. Firstly, the method detects the edge and uses the lane line model to match, so as to determine the area of interest on the road. Secondly, the road video is collected to label the vehicle targets and make the vehicle data set, and then design a convolution neural network. The vehicle data set is used to train the detector so that the detector can adapt to the task of vehicle identification. Finally, the vehicle is detected in the area of interest of the road. Compared with the traditional vehicle identification method, this method has better accuracy and robustness, and has a satisfactory recognition effect under complicated driving conditions. [Key words] convolutional neural network; vehicle detecti

参考文献

引证文献

问答

我要提问