基于深度学习的车辆检测系统设计
2022-07-21黎雄陈雨聪
黎雄 陈雨聪









摘 要:针对目前智能交通系统实时道路车辆检测中存在的检测效率不高等问题,设计了一种基于深度学习的道路车辆检测算法。首先在搭建好的平台下,通过卷积神经网络对采集的车辆图像数据集進行训练,得到训练后的模型;其次,对该模型内部的层结构进行可视化;最后,通过调节各网络参数及层结构对该模型进行优化。训练的模型通过实验测试,分别对图片和视频进行检测,图像识别准确率高,检测速度快,跟踪精度高,可应用于实时交通系统的检测。
关键词:深度学习;卷积神经网络;车辆识别;OpenCV
中图分类号:TP183 文献标识码:A
Design of Vehicle Detection System based on Deep Learning
LI Xiong1, CHEN Yucong2
(1.Xi'an Institute of Electromechanical Information Technology, Xi'an 710065, China;
2.School of Computing, Xianyang Normal University, Xianyang 712000, China)
42340658@qq.com; 943858467@qq.com
Abstract: Aiming at the low detection efficiency of real-time road vehicle detection in intelligent transportations system, this paper proposes to design a road vehicle detection algorithm based on deep learning. Firstly, based on the built platform, convolutional neural network is used to train the collected vehicle image dataset to obtain the trained model. Secondly, the layer structure inside the model is visualized. Finally, the model is optimized by adjusting all of the network parameters and layer structure. The trained model has been tested by experiments to detect pictures and videos respectively. Experiment results show that the proposed model can be applied to the detection of real-time traffic systems with high image recognition accuracy, fast detection speed, and high tracking accuracy.
Keywords: deep learning; convolutional neural network; vehicle recognition; OpenCV
1 引言(Introduction)
随着大数据时代的到来,越来越多的数据需要处理。而人们对信息处理的要求也越来越严格,既要达到高准确率,又要达到较高的运行速率。卷积神经网络是为了识别二维图像而设计的多层感知器,在处理图像方面具有良好的容错能力与并行的处理以及学习能力。在深度学习图像识别应用上,卷积神经网络具有较高的准确率和处理速度,所以可以作为理论依据来研究动态车辆识别。对深度学习动态车辆的研究,主要是通过摄像头实时监控车辆的位置,继而来判断车辆的类别,这一领域的研究对今后智能辅助领域的数据处理发展有着至关重要的作用[1-2]。与此同时,也面临着艰巨的挑战,如车辆检测方面包括视频序列中背景的更新、复杂环境下背景图片的遮挡和运动车辆在道路中停止时对跟踪的影响,以及复杂环境下车辆检测的实时性要求等[3]。……
