利用近距离目标红外图像的机动检测问题研究

机动检测问题是目标跟踪(尤其是对近距离目标跟踪)过程中的一类关键性问题.传统的机动检测方法一般将目标看做点目标来处理,而在能得到目标红外图像的情况下,依据红外图像跟踪可以得到更快、更准确的机动检测性能,本文提出了一种针对近距离目标红外图像的依据类散度的小波包变换、初始聚类后叠代逼近以及能量最强直线段提取的图像处理方法,该方法在较低信噪比情况下依然能够保持较好的目标中心的跟踪精度,并进一步改善机动检测的性能.

Maneuver Detection Using Short Range Target Infrared Image

Maneuver detection is critical especially for tracking of short range target.Former methods deal with target as a point,however,it is proved that the use of information from an image sensor can improve the quality and reliability of maneuver detection.In this paper,we introduce an infrared image processing method with the application of wavelets,in which multi scale basis selection is based on class separability got from a recurrence process from initial classification to final classification.We also introduce a method of line extraction to estimate three centers of target.This algorithm keeps relatively fine performance even in low SNR environment.