CAJ | 학술논문

针对夜间交通监控面临的亮度不足、能见度不佳与光线反射干扰等，造成视频车辆检测器不能全天候准确检测的问题，提出一种基于车灯亮度、颜色、形状、位置、纹理等运动信息的夜间车辆检测系统。先找出可能为车灯的亮点，然后再判断属于同一车的可能性，并进行车灯配对，最后针对路面上碰触虚拟检测线的一对亮点进行计数。实验结果表明，本方法的平均准确率在96％以上，具有检测率高、鲁棒性强的特点，能够满足城市道路及高速公路的实时检测需求。
침대야간교통감공면림적량도불족、능견도불가여광선반사간우등，조성시빈차량검측기불능전천후준학검측적문제，제출일충기우차등량도、안색、형상、위치、문리등운동신식적야간차량검측계통。선조출가능위차등적량점，연후재판단속우동일차적가능성，병진행차등배대，최후침대로면상팽촉허의검측선적일대량점진행계수。실험결과표명，본방법적평균준학솔재96％이상，구유검측솔고、로봉성강적특점，능구만족성시도로급고속공로적실시검측수구。
Aiming at the problem of the video vehicle detector in the night traffic monitoring that cannot accurately detect all‐weather which caused by short of the brightness , poor visibility and light reflection interference . A kind of nighttime vehicle detection system based on light brightness , color , shape , location , texture movement information is proposed . Firstly , to identify possible light spot , and judge the possibility of belonging to the same car , and then do the light matching , finally count the pair of light spots touching the virtual detection line on the road . The experiment results show that the accuracy of this method in nighttime vehicle detection is above 96% , and this method is with the characteristics of high detection rate , strong robustness , and can meet the demand of real‐time detection of city road and highway .