摘要
针对 ATR 系统小目标识别实时性的需求,提出了一种快速的小目标检测算法,该方法在分析形态滤波的基本运算即膨胀、腐蚀后发现:形态滤波计算量大、实时性差的原因在于膨胀和腐蚀的计算过程中存在比较运算的冗余,即当新像素进入结构元中时,需要重新比较所有像素的灰度值,而提出的形态滤波算法则仅需把新像素的灰度值与前一次的结果值相比较而得到该次运算的结果,这样就大大减少了比较次数而提高了算法的速度.最后将算法移植到以 TSM320C6203为核心的硬件平台上实现了目标检测的实时处理.实验表明:该快速算法较一般算法的计算时间减少70%.
To satisfy the requirement of small target real-time detection in ATR system,a fast algorithm for small targets detection is proposed.After analyzing the basic operation of morphological filtering,viz.erosion and dilation,the large amount of computation redundancy is found in the previous operation results in great complexity and pool real-time ability.Particularly,when a new pixel enters a structuring element,all pixels in the element will be compared with the new pixel.The morphological filtering algorithm proposed in this paper is able to make only one comparison between the new pixel and the result obtained in the latest comparison before.This greatly decreases comparison times and improves computation efficiency.In the end,the algorithm is implemented in the TSM320C6203 platform to detect real-time targets.Series of experiments show that the computation time is reduced by 70%.
出处
《红外与激光工程》
EI
CSCD
北大核心
2006年第z4期272-276,共5页
Infrared and Laser Engineering
关键词
小目标
结构元
形态滤波
实时检测
Small target
Structuring element
Morphological filtering
Track detection
