基于支持向量机车辆检测的算法

发布时间：2018-05-12 17:30

本文选题：车辆检测 + 支持向量机　；参考：《江西师范大学》2014年硕士论文

【摘要】：随着社会的不断发展和科技的不断进步，未来的人造系统和产品的特征之一就是智能化，，就是用计算机代替人类某些方面的活动。在智能交通领域，进行各种智能化交通行为分析的第一步就是车辆检测。因此车辆检测的准确度和速度对智能交通系统的影响非常大。本文基于目标检测技术中支持向量机的方法对车辆检测方法进行研究，对目标主流的检测算法进行了实验对比，并从检测速度的角度对传统的算法进行了改进。主要内容如下：（1）研究了统计学习理论以及支持向量机的思想、实现算法和应用前景；（2）分析了图像特征提取中HOG特征提取以及增强性HOG特征提取技术；（3）在Visual Studio2008编程环境下进行了基于OPENCV和基本HOG特征的实验和基于OPENCV和改进HOG特征的实验从准确度和检测速度两方面对算法性能进行评估；（4）采用非线性支持向量机训练的模型和线性支持向量机的检测思路，对传统算法进行改进，实验结果表明，改进的算法不仅提高了检测准确度，而且降低了检测消耗时间；（5）结合传统支持向量机的检测思路和DPM（可变形部件模型）的检测方法，提出仅仅采用DPM中的根滤波器来对车辆进行检测的算法，实验结果表明，该方法虽然准确率不如传统的支持向量机检测，但是仍然能够准确地检测出车辆，同时检测耗时大大减少，可以适用于一些对准确度要求不是很高，而对检测速度要求较高的场合。
[Abstract]:With the development of society and the progress of science and technology, one of the characteristics of future artificial systems and products is intelligence, that is, the use of computers to replace some aspects of human activities. In the field of intelligent transportation, vehicle detection is the first step to analyze various intelligent traffic behaviors. Therefore, the accuracy and speed of vehicle detection have great influence on its. In this paper, based on the support vector machine (SVM) method of target detection, the vehicle detection method is studied, and the main detection algorithm of target is compared experimentally, and the traditional algorithm is improved from the point of view of detection speed. The main contents are as follows: 1) the theory of statistical learning and the idea of support vector machine (SVM) are studied. Secondly, the techniques of HOG feature extraction and enhanced HOG feature extraction in image feature extraction are analyzed. 3) experiments based on OPENCV and basic HOG features and experiments based on OPENCV and improved HOG features are carried out in Visual Studio2008 programming environment. The performance of the algorithm is evaluated in terms of accuracy and detection speed. The traditional algorithm is improved by using the training model of nonlinear support vector machine and the train of thought of linear support vector machine. The experimental results show that the improved algorithm not only improves the detection accuracy but also reduces the detection time. 5) combined with the traditional SVM and DPM (deformable component Model) detection method, a vehicle detection algorithm using only the root filter in DPM is proposed. The experimental results show that, Although the accuracy of this method is not as good as that of traditional SVM detection, it can detect vehicles accurately, and the detection time is greatly reduced, so it can be applied to some cases where the accuracy requirement is not very high. On the other hand, the testing speed is high.
【学位授予单位】：江西师范大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U495

【参考文献】