CAJ | 학술논문

影响基于视频检测的车型分类系统准确率的一个主要因素是采集的车辆外型参教的准确性.针对这种情况,提出了基于多源数据融合的方法提取车辆的外型参数,并使用SVM(支持向量机)对车辆进行分类.实验结果表明,多源数据融合的方法能够有效控制在采集过程中产生的噪音干扰和镜头畸变引起的误差,提高车型参数的准确性.使用支持向量机分类能够克服神经网络中无法避免的局部极值问题.该方法能够提高车型分类准确率,实时性强,适用于实时车型分类系统.
영향기우시빈검측적차형분류계통준학솔적일개주요인소시채집적차량외형삼교적준학성.침대저충정황,제출료기우다원수거융합적방법제취차량적외형삼수,병사용SVM(지지향량궤)대차량진행분류.실험결과표명,다원수거융합적방법능구유효공제재채집과정중산생적조음간우화경두기변인기적오차,제고차형삼수적준학성.사용지지향량궤분류능구극복신경망락중무법피면적국부겁치문제.해방법능구제고차형분류준학솔,실시성강,괄용우실시차형분류계통.
Research shows that a major factor that influences vehicle classification system based on video accuracy rate is the accurate of the vehicle parameters.According to this situation,this paper presents a method based on multi-source data fusion to detract vehicles' shape parameters and uses Support Vector Machine(SVM) to class vehicles.The experimental results indicate that multi-source data fusion method can control noise disturbance in collecting procedure and the errors caused by camera lens distortion effectively and increase the inaccuracy of the vehicle shape parameters.Using support machine to class vehicles can overcome the problem of local extreme values unavoidably in neural network.The suggested method can increase the accuracy of vehicles classification and has a strong timeliness,be suitable for the real-time vehicle classification system.