摘要
针对传统检测跟踪算法无法有效处理迎头点目标的问题,提出了基于红外序列图像相邻两帧间匹配的检测算法。根据迎头点目标的红外序列图像特点,采取点对点标记匹配结果,点对点统计标记次数,阈值门限输出迎头点目标。算法无需假定目标点位置,检测过程不随背景复杂程度和目标点数目变化而变化,尤其解决了高噪声情况下迎头点目标的检测和跟踪问题。算法结构简单、存储量小、运算速度快,根据多次仿真和实拍照片实验,验证了该方法的可行性和有效性。
In order to solve the problem that traditional detection and tracking algorithms can not deal with head-on point targets effectively, an algorithm for head-on point targets detecting and tracking is presented based on infrared image sequences matching. Aiming at the characteristics of head-on point targets in the images sequences, the match result is marked directly in a point-to-point manner, the times of pixel points are counted which have been marked in the match process and an appropriate threshold is selected to show the real point targets. The algorithm merits lie in no need to suppose the doubtful target point in advance, and the tracking process does not change while the background complex degree or the number of point targets change. Especially, this algorithm can solve the problem of point target detection under the high noise environment. According to the simulations and experiments with real images, it has been confirmed that the method is feasible and valid.
出处
《光电工程》
CAS
CSCD
北大核心
2009年第5期1-5,共5页
Opto-Electronic Engineering
基金
海军预研项目(4010605020402)
关键词
图像处理
目标跟踪
目标检测
点目标
红外图像
image processing
target tracking
target detection
point target
infrared image
