采用小波变换和数学形态学的小目标检测

鉴于常用检测方法不能准确稳定地检测出复杂背景中小目标,结合小波变换和数学形态学,提出了一种小目标检测新方法.首先对图像进行单尺度小波变换,提取高频分量系数;其次,利用阈值算法将各个高频分量系数图像转化为二值图像后对其进行多结构元素形态学滤波,滤波结果与原二值图像相减后在差值图像上得到可能的小目标.将3个方向的高频系数的检测结果相关联获得单帧检测结果;最后将多个单帧检测结果进行流水线检测,得到最终的检测结果.仿真结果表明该方法能够准确稳定地检测出信噪比(SNR)大于2的弱小目标.

Small target detection based on wavelet transform and mathematical morphology

Since it is difficult to detect the small target stably and accurately in complex background with the traditional methods,a new small target detecting method is proposed based on wavelet transform and mathematical morphology.Firstly,the image is decomposed using single-scale wavelet transform and the high frequency components are extracted.Secondly,threshold algorithm is used for converting each high frequency component image to binary image.The binary images are filtered by mathematical morphology method using multi-structure elements.The possible small targets are detected in the differential images obtained by subtracting the previous binary images and the filtering results.Detection results in three directions are associated to get the single-frame detection result.Finally,pipeline detecting scheme is used to detect the small target by using several single-frame detection results for obtaining the final detection result.Simulation results show that the proposed algorithm can detect accurately the small target with Signal to Noise Ratio(SNR) bigger than 2.