CAJ | 학술논문

结合高速公路交通视频的特点，提出了监控中运动车辆对象提取的方法。该方法利用混合背景差分法提取运动目标，并对目标进行形态学处理，最后结合后退 N 帧、半监督学习的阈值选取的思想提取车辆对象特征最显著的视频帧。实验结果表明，该方法有效地改善了传统的混合高斯背景差分提取运动对象过程中出现的孔洞、噪声和车辆对象非完整性及特征不显著的问题，进而优化了车辆对象的提取流程。
결합고속공로교통시빈적특점，제출료감공중운동차량대상제취적방법。해방법이용혼합배경차분법제취운동목표，병대목표진행형태학처리，최후결합후퇴 N 정、반감독학습적역치선취적사상제취차량대상특정최현저적시빈정。실험결과표명，해방법유효지개선료전통적혼합고사배경차분제취운동대상과정중출현적공동、조성화차량대상비완정성급특정불현저적문제，진이우화료차량대상적제취류정。
In this paper , combined with the characteristics of highway traffic video , it puts forward the method of moving vehi-cle object extraction in monitoring . The method using the mixed background difference method to extract moving objects , and using morphological processing , back N frames and semi supervised to extract cars in the video . The experimental results show that the method improves the problems of holes , noise and non-integrity and nonsignificant characteristics of vehicle in the process of mov-ing object extraction in traditional mixed Gauss backgound differential , this method can effectively improve the extraction process of vehicle object .