基于改进Adaboost算法的视频车辆轮廓检测算法研究

发布时间：2018-06-18 07:27

本文选题：视频车辆轮廓 + adaboost算法　；参考：《中原工学院》2017年硕士论文

【摘要】：视频车辆轮廓检测技术作为智能交通中的关键技术,在日常生活中有着广泛的应用前景。而智能交通对于视频车辆轮廓检测技术也有着实时性和准确性等严苛的要求。而在检测时,视频图像中复杂的背景以及各式各样的干扰,是目前视频车辆轮廓检测技术面临的问题。而随着国内外学者的不断努力,各式各样的检测算法层出不穷。Adaboost算法是近些年来比较流行的机器学习算法,在人脸识别领域有着出色的表现。同样,adaboost算法可以应用于视频车辆轮廓检测领域。本文根据经典的adaboost算法,在其基础上做出了一些改进,研究工作如下:(1)介绍视频车辆轮廓检测技术研究的背景意义,了解其中面临的问题。(2)介绍分析几种现有的视频车辆轮廓检测算法,了解其原理,并总结出存在的问题。(3)详细了解adaboost算法,理解其原理和实现过程,依次介绍haar特征、积分图以及分类器的训练及选取过程。(4)提出一种改进的adaboost算法进行视频车辆轮廓检测。首先,针对算法学习过程中,haar特征计算量过于庞大且耗时的现象,提出了对训练样本进行裁剪,去除样本边缘像素,有效减少特征数量,从而降低了计算量。(5)Adaboost算法检测在对视频图像进行检测时,滑动子窗口会在待检图像上依次滑过,图像中无关信息都需要被检测一遍,相当耗时。提出了使用光流法来获取视频图像中的运动区域作为感兴趣区域,在感兴趣区域中使用canny算子进行边缘检测,通过边缘能量筛选感兴趣区域,排除非感兴趣区域。最终使用adaboost算法对感兴趣区域进行检测,减小了检测区域降低了检测时间。(6)通过设置阈值来对检测结果进行筛选,提升准确率。结合算法的优点和不足对未来的发展进行展望。
[Abstract]:As the key technology of intelligent transportation, video vehicle contour detection technology has a wide application prospect in daily life. Intelligent Transportation has strict requirements such as real-time and accuracy for video vehicle contour detection technology. In detection, the complex background and various kinds of interference in video images are the current problems of video vehicle contour detection technology. With the continuous efforts of scholars at home and abroad, a variety of detection algorithm. Adaboost algorithm is a popular machine learning algorithm in recent years, and has a good performance in the field of face recognition. The adaboost algorithm can also be used in the field of video vehicle contour detection. Based on the classical adaboost algorithm, some improvements are made in this paper. The research work is as follows: 1) the background significance of the video vehicle contour detection technology is introduced. (2) introduce and analyze several existing video vehicle contour detection algorithms, understand their principles, and summarize the existing problems. (3) understand the adaboost algorithm in detail, understand its principle and implementation process, and then introduce the characteristics of haar in turn. An improved adaboost algorithm is proposed for video vehicle contour detection. First of all, aiming at the phenomenon that the computation of haar feature is too large and time-consuming in the learning process of the algorithm, the training sample is clipped to remove the edge pixels of the sample, and the number of features is reduced effectively. Therefore, the computational complexity of the algorithm is reduced. When detecting the video image, the sliding subwindow will slip through the image to be checked, and the irrelevant information in the image needs to be detected once, which is time-consuming. In this paper, an optical flow method is proposed to obtain the moving region of the video image as the region of interest, and the canny operator is used to detect the edge of the region of interest. The region of interest is filtered by edge energy, and the region of non-interest is excluded. Finally, the adaboost algorithm is used to detect the region of interest, which reduces the detection time and reduces the detection time. Combining the advantages and disadvantages of the algorithm, the future development is prospected.
【学位授予单位】：中原工学院
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U495;TP391.41

【参考文献】