基于发布订阅结构的计算机视觉跟踪研究

Computer Vision Tracking Studying Based on Publish/Subscribe Structure

文中报道了使用发布订阅中间件搭建的一个跟踪系统,通过单摄像头采集图像,经过分析处理,计算出运动物体的相对位置,同时控制摄像头转动,对运动物体进行实时追踪.整个系统主要包括4个方面,以发布/订阅为中心结构连接3个子系统、目标获取、目标跟踪、采集摄像头控制.文中为卡尔曼滤波定义了新的向量,预测目标运动趋势;根据物体的运动信息自行设计了摄像头控制程序.算法有充分的理论依据和实验验证.

This paper presents a tracking system using Publish/Subscribe middleware （Pub/ Sub）. The image data is collected by monocular camera, and analyzed to calculate the position of moving object. The real-time tracking of the moving object is performed by controlling the rotation of camera. The entire system is composed mainly of four parts, i.e. Using Pub/Sub to join three subsystems, target acquisition, tracing the moving object and controlling camera. The improvement of this system include （1） Give a definition of a new vector for Kalman filter; （2） Predict moving tendency; （3） Redesign a program for camera controlling according to the vector of moving information. The arithmetic exployed in this paper is reliable and application.