摘要
提出一种静止摄像机条件下车辆和行人的支持向量机(SVM)识别方法。首先根据背景差分法对监控视频中的运动目标进行检测,提取出运动目标的基本轮廓,然后利用数学形态学方法对目标进一步检测处理。用星形向量表示法对运动目标提取8个特征,以及高度、宽度和高宽比作为另外3个特征,通过构造SVM分类器,实现了基于SVM的图像分类和识别。实验结果表明,该方法能够在视频监控中快速准确地对运动的车辆和行人进行检测和分类,平均识别率达到96.97%。
In this paper, an approach based on SVM to recognize the moving vehicles and pedestrian with a static camera is presented. Firstly, the moving object in the video can be detected by background subtraction, and their basic outline is extracted. Then, the target by mathematical morphology is given. Eight features of the moving object with center radiation are picked up, and height, width, aspect ratio as the other three characteristics. Image classification and recognition based on SVM are achieved by constructing the SVM classifier. The experimental results show that the pedestrians and vehicles are detected and classified accurately in the surveillance video, and the average recognition rate is 96.97%.
出处
《电视技术》
北大核心
2011年第15期1-3,15,共4页
Video Engineering
关键词
支持向量机
目标检测
背景差分
特征提取
目标识别
SVM
object detection
background subtraction
feature extraction
object recognition
