运动小目标检测最优图像帧数概率模型  

Optimal frames probability model of small moving target detection

作　　者：郑红[1] 隋强强[1] 陈磊[1] 

机构地区：[1]北京航空航天大学自动化科学与电气工程学院,北京100191

出　　处：《系统工程与电子技术》2011年第1期8-12,共5页Systems Engineering and Electronics

基　　金：国家自然科学基金(60543006)资助课题

摘　　要：针对运动小目标轨迹检测方法存在的最长轨迹拟合范围难以确定的问题,研究了轨迹长度与小目标检测概率分布之间的可描述性和适用性,分析了基于二项式概率分布的轨迹长度优化模型参数与最优轨迹模型帧数之间的关系,连续轨迹中丟失图像帧数与预测误差之间的关系,以及丟帧引起的轨迹拟合长度有效性限制问题,建立了小目标轨迹检测最优帧数概率模型。利用不同干扰下的40组紫外小目标连续视频,进行轨迹建立帧数、轨迹跟踪帧数、连续丟帧有效性测试,验证了文中所建立模型的有效性。Due to the shortage of confirming the length of the small moving targets＇ track detection,a new method is proposed to solve the problem of trace-fitting length,and the relative characteristics of the trace length and small target detection probability distribution are studied.The problem of trace-fitting length limitation caused by allowable continuous losing frames in the process of trace prediction is discussed.The probability model of small targets detection is established.Through the model,the accuracy of the trace prediction is improved,and an optimal solution of the target trace-fitting frame number is obtained.The proposed model is tested by using 40 groups＇ images of different interferential ultraviolet small targets.The result shows that the model is efficient.

关 键 词：运动小目标检测 概率模型 最优帧数模型 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]