摘要
利用飞机在光学遥感图像中的形状特点和灰度变化特点,本文提出了一种新的飞机目标检测方法一圆周频率滤波法。首先,采用圆周频率滤波器对全图进行滤波,在飞机中心点附近像素通过滤波后幅值通常比其他非飞机区域的幅值大,从而通过保留滤波后幅值大的点作为飞机中心的候选点;然后,通过区域生长式聚类确定最终的飞机位置。本文的方法用于8幅1.0-4.0米分辨率的光学遥感图像的飞机目标提取,总共65架飞机,正确检测出59架,5个虚警。实验证明圆周频率滤波法用于遥感图像的飞机目标检测,是一种即简单又有效的方法。
This paper presents a new approach-Circle-Frequency filter （CF-filter） to detect airplanes in panchromatic remotesensing images,in which both the shape and luminance features are used. First, the whole image is filtered by the CF-filter. Since airplane centers appear bright while background appears dimming, the points with higher outputs of CF-filtered ＂image＂ are selected as candidate points of airplane centers. Then,a region-growing clustering method is used to locate airplane centers from the candidates. 8 panchromatic images of 1.0m ～ 4.0m resolution, including 65 airplanes, are tested with this approach. 59 airplanes are detected and 5 times are false alarmed. The experiments demonstrate that the presented approach using CF-filter is effective in airplane detection.
出处
《信号处理》
CSCD
北大核心
2007年第4期539-543,共5页
Journal of Signal Processing
关键词
圆周频率滤波器
目标检测
飞机检测
遥感图像处理
Circle-frequency filter
Target detection
Airplane detection
Remote-sensing image processing.
