基于局部纹理分析的小目标检测方法

Dim targets detection method based on local texture analysis

在多分辨率分析的基础上,提出了一种基于局部纹理分析的小目标检测新方法。利用目标与背景的局部纹理特征差异,在局部距离像上完成小目标的检测;而传统的检测方法利用目标点亮度值与背景亮度值不同进行目标与背景的分离,因此本方法更能体现目标与背景的差异,在强杂波背景中实现高效的检测。对实测数据的仿真实验表明,该方法能有效地检测占极少(1-3 个)像素的小目标;同时,在小波分解阶段分别采用 Mallat 算法和 à trous 算法时,该方法都能在 0.65秒内完成单帧检测,具有良好的实时性能。

A dim targets detection method based on local texture analysis is proposed on the basis of Multi-Resolution Analysis (MRA). The detection of dim targets is carried out by using local MRA character vector difference of target and background, while the traditional methods always use the gray value difference of target and background of infrared image to realize dim target detection. So this method can reflect the difference between targets and background better and high-efficiency detection in the strong clutter background can be performed. The simulation experiments for practical measured data show that the method can effectively detect dim targets with 1-3 pixels. When Mallat algorithm and a trous algorithm is applied in wavelet decomposition phase, respectively, single-frame detection can be completed by this method in 0.65 s. This represents it with good timeliness.