摘要
针对红外图像中模糊目标难以分割和识别的情况,提出以小波多尺度滤波分解后的逼近图像作为自适应域值进行图像二值化的方法,由此分割出模糊图像中的目标,提取目标的轮廓边缘。然后,提出一种新的矩计算方法提取目标的不变性特征。实验结果表明,该方法具有很强的抗噪和抗扰性能,能有效提取复杂背景中模糊目标的平移、缩放和旋转不变量,极大提高了运动模糊目标识别的可靠性。
Since it is difficult to segment and recognize the blurred target in IR images, a binarization method based on the approximation resulted from wavelet multi-scale filtering with the adaptive threshold is proposed. The targets in the blurred images can be segmented and the contour edge can be extracted. Then, a new algorithm of moment calculation is used to extract the invariable features. Experimental results show that this approach is very robust and efficient to extract the invariants of displacement, zoom and rotation of target, which can improve the reliability of recognition of arbitrary blurred target.
出处
《光电工程》
EI
CAS
CSCD
北大核心
2007年第4期1-6,共6页
Opto-Electronic Engineering
基金
中国科学院国防科技创新基金项目(CXJJ-65)
关键词
红外图像
模糊目标
特征提取
综合不变矩
Infrared image
Blurred target
Feature extraction
United invariant moment
