摘要
针对多目标的检测,本文提出一种采用多源图像分形特征的特征级融合检测方法。首先对多目标检测的特点进行了分析,对分形理论进行了介绍,然后详细介绍了该融合检测算法的思路和原理。该算法首先由红外图像阈值分割出部分目标;然后利用分维数图的统计特征可以增强分形维数的奇异性,在可见光图像的分维数图中搜索与已检测出的目标区域具有相近分形统计特征的区域,进行标记;再根据"距离相似度准则"进行目标的聚类识别,排除背景干扰,最终检测出全部目标。实验结果表明该融合检测算法能有效地进行多目标的检测与识别。
As to the detection of multi-target, a feature-level fusion detection scheme which uses fractal feature of multi-source images is proposed. Firstly, the character of multi-target detection is analyzed, the fractal theory is introduced, and then the theory of fusion detection scheme is introduced in detail. Some targets are segmented in infrared image. Afterwards, using the statistic character can build up irregularity of fractal dimension, and areas are searched and signed in the visible light fractal dimension image, which has close fractal statistic feature to the target area detected. At last, the whole targets are successfully detected after clustering and recognition based on the rule of distance similarity degree. The experimental results show that the fusion detection scheme can detect and recognize target effectively
出处
《光电工程》
CAS
CSCD
北大核心
2009年第12期11-15,共5页
Opto-Electronic Engineering
基金
电子科技大学青年科技基金
关键词
多目标检测
图像融合
分形特征
统计特征
multi-target detection
image fusion
fractal feature
statistical feature
