摘要
针对红外弱小目标检测提出了一种新的算法.算法首先对图像进行均值滤波处理以减少噪声点,然后利用基于多尺度小波核函数的最小二乘向量机对图像进行局部灰度曲面拟合,再通过二阶方向导数算子计算出其特征图像并将连续几帧特征图像融合,最后采用对比度分割方法确认目标位置.仿真实验表明,该方法不仅具有良好的适应性和检测效果,而且具有较强的时效性.
A new method is presented for detecting the infrared weak and small targets.In the method,the mean filtering approach is firstly adopted to reduce the noise points of the image,and then the least-squares support vector machine(LS-SVM)is employed,which is based on multi-scale wavelet kernel function,to fit the local intensity surface of the image.Afterwards,its feature image is worked out by using the second-order directional derivative operators and fuse several consecutive frames of feature image.Finally,the target position is determined by the contrast segmentation method.Simulation results suggest that the proposed method not only demonstrates good adaptation and detection capability,but also shows efficiency.
出处
《红外与激光工程》
EI
CSCD
北大核心
2006年第z4期251-257,共7页
Infrared and Laser Engineering
基金
国家自然基金(60572048)
关键词
红外序列图像
小目标检测
最小二乘向量机
多尺度小波核函数
IR image sequence
Small target detection
Least squares support vector machines
Multi-scale wavelet kernel function
