摘要
利用平均灰度绝对差最大测度（ＡＧＡＤＭ）估计目标背景的不相似性，提出了一种红外小目标检测算法，分析了预处理前后图像信噪比增益与目标大小的关系，进而建立了检测概率和信噪比增益的关系，理论上证明了该算法具有优越的性能。接着，在分析了ＭＰＰ并行计算机体系结构的基础上，提出了两种基于ＭＰＰ的小目标检测并行快速实现算法。理论分析表明，基于ＭＰＰ的ＡＧＡＤＭ小目标检测并行快速实现算法运算时间复杂性为Ｏ（Ｌ２ｉ／Ｎ）。
In this paper, an algorithm for small infrared target detection using the average gray absolute difference maximum (AGADM) that can compute dissimilarity between an object and a scene is presented, then the signal\|to\|noise gain between original images and their preprocessed results as a function of an object size is analyzed and the theoretical relation between detection probability and the signal\|to\|noise gain is also established, which has shown the performance enhancement of this algorithm theoretically. After analyzing the structure of a MPP computer, two implementations for AGADM\|based small target detection on a parallel MPP computer are presented. Theoretical analysis has shown that the computational complexity for MPP\|based small target detection is O(L\+2\-i/N).\=\;
出处
《红外与激光工程》
EI
CSCD
1999年第3期13-18,共6页
Infrared and Laser Engineering
关键词
目标检测
自动目标识别
红外技术
应用
Small target\ \ Target detection\ \ Automatic target recognition
