毕业论文移动版

摘要在智能视频监控、自动车辆驾驶等领域，车辆检测算法有着重要而广泛的应用前景。在真实场景中，检测对象往往存在着相互遮挡、检测对象尺寸过小和图像边缘截断等问 题，导致车辆检测效率低、效果也不够好。于此同时，随着大数据的应用和计算机运算能 力的发展，深度学习算法在计算机视觉领域得到了广泛的应用，它有着很强的特征学习能 力，并具有一定的鲁棒性。因此，本文将基于深度学习算法，研究复杂场景下的车辆检测 算法，获得一体化的车辆检测模型，在一个框架下统一完成特征学习、分析、车辆识别和 定位等。整个车辆检测框架可以分为三个部分，1）提取图像中对象 proposal；2）利用深 度卷积神经网络来提取对象 proposal 的特征并进行分析；3）车辆识别和坐标定位。最后 在 Kitti 数据集上进行车辆检测实验，验证了该算法的正确性和有效性。77316

毕业论文关键词    车辆检测   深度学习   车辆识别  车辆定位

毕 业 设 计 说 明 书 外 文 摘 要

Title Car detection algorithm with deep learning

Abstract In the field of intelligent video surveillance and automatic vehicle driving, car detection algorithm has an important and broad application prospect。 In the practical scene, the detection objects are often with the mutual occlusion, small-size, edge truncation and other problems, and then the traditional car detection algorithms lack the initiative and effectiveness。 Meanwhile, with the application of big data and the development of computer computing capability, the deep learning method has been widely applied in the field of computer vision, has a strong ability of the feature learning and a certain robustness。 Therefore this thesis proposes the car detection algorithm under the guide of deep learning, obtains an integrative car detection model, and finishes the feature learning, analysis, car recognition and location。 The whole car detection framework has three parts: 1) extracting the object proposals  from an image; 2) obtaining and analyzing the image features of object proposals by the deep convolutional neural network; 3) achieving the results of car recognition and location。 Finally, the experimental results on the Kitti dataset well demonstrate that the proposed algorithm considerably outperforms other traditional methods。 

Keywords Car Detection  Deep Learning  Car Recognition  Car Location

本科毕业设计说明书 第 I 页

目 录

1 引言1 

1。1 课题研究背景与意义 1

1。2 国内外研究的现状 1

1。3 车辆检测存在的问题 2

2 深度学习介绍3 

2。1 深度学习基本思想3 

2。2 深度学习发展概述3 

2。3 深度学习一些常用模型 4

2。3。1 自动编码器4 

2。3。2 稀疏编码5 

2。3。3 限制波尔兹曼机 5

2。3。4 深度信置网络6 

2。3。5 (责任编辑：qin)