货船船舱内运输的各种车辆,需要对其进行有效的监控及识别跟踪定位,以使车辆间运行路径互不干扰,保证调度效率.该文针对船舱这种特殊条件下的车辆图像识别和跟踪系统进行设计,采用广角镜头结合多摄像机分布实现低高度大范围的拍摄,采用样条函数对失真图像进行校正,采用Harris算法结合拉伸变换进行图像拼接,获得整个船舱的完整图像.提出采用基于主视觉词(principal visual word,PVW)的牌照识别算法对坦克和装甲车等牌照进行识别,克服迷彩等复杂背景的影响.根据船舱的实际尺寸,设计16:1的实验系统,通过在车辆上设置标识点对车辆的定位精度进行验证,最终获得±0.5 m的车辆定位精度.%As for various vehicles transported in the cabin of cargo ship, effective monitoring and identification tracking shall be carried out to ensure there is no mutual interfere between vehicles and guarantee dispatching efficiency. Thus in the article, design was carried out for vehicle image recognition and tracing system according to the special conditions in cabin, and wide-angle lens combining with multi-camera distribution was adopted to achieve low-altitude and wide-shooting. Spline function was adopted to correct the distorted image and Harris algorithm combined with stretching transformation was introduced for image stitching, so as to obtain the entire image of the whole cabin. Meanwhile, license recognition algorithm based on principal visual word (PVW) was proposed to recognize the license of tanks and armored vehicles to overcome the influence of complicated background such as camouflage. A 16:1 test system was designed according to the real dimension of the cabin and the positioning accuracy of vehicle reached ±0.5 m finally by setting marks on the vehicle to verify the positioning precision.
展开▼
