【摘要】 针对空中远距离红外小目标检测的实际问题,提出了一种基于非抽样轮廓波变换的检测算法。首先利用非抽样轮廓波变换的优良性质,通过分析噪声系数、背景边缘系数和目标系数在尺度间的不同特性,计算各个信号在尺度间的相关系数并归一化。接下来,按照自适应阈值法抑制噪声和背景边缘系数,然后通过反变换得到抑制背景增强目标的图像。最后,结合目标面积信息选择适当阈值,对重构图像进行分割,生成单帧检测结果并进一步利用帧间目标位置的相关性完成小目标检测过程。试验结果表明,提出的算法能够准确地检测目标。相对于通常的小目标检测算法,本算法在背景抑制方面具有一定的优势,能够获得相对较高的信噪比。更多还原

【Abstract】 In order to solve the practical problem of the infrared small target’s detecting in the sky,a detecting algorithm based on the NSCT(nonsubsampled contourlet transform) was proposed.Firstly,taking advantage of the excellent property of the transform,this algorithm analyzed the different property of the NSCT coefficients of noise,background edge and signal,then computed the normalized correlation coefficients between scale for the transform coefficients.Subsequently,according to the adaptive method of threshold,the coefficients of background edge and noise were suppressed and then the image which includes the enhanced target was acquired by the inverse transform.In the end,taking the target area into account,the reconstructed image was partitioned and the final detecting result was acquired by considering the position correlation of the target between frames.Experimental results show that the method given by this paper can detect small target accurately.Compared with some traditional method,it has certain advantage in background suppressing and can acquire high SNR(signal noise ratio) value.更多还原