复杂背景下红外图像弱小点目标自动识别仿真

传统识别方法受到低信噪比、低对比度、缺乏弱小点目标的形状及纹理信息等因素影响,尤其在复杂背景下,弱小点目标自动识别准确率较低,针对此问题,提出一种基于BEMD(二维经验模态分解算法)的红外图像弱小点目标自动识别方法,根据待识别图像的频谱特性,并结合分频段处理方式。对比了不同滤波器的性能,并建立了图像滤波器组,采用滤波器组将弱小点目标图像分解到不同子频域中;对子频段图像进行罗宾逊滤波处理,提取弱小点目标。采用多层经验模态分解算法对原始弱小点目标图像输入函数分解为二维本征模态函数,通过微分计算来获取原始图像与背景区域之间的差,分割出弱小点目标区域。通过局部逆熵分割弱小点目标区域的高频信息来获取各个模态函数的弱小点目标识别结果。实验结果表明,所提方法能够高效且准确地提取出弱小点目标,更好地抑制复杂背景。

Automatic Recognition of Infrared Image Weak Point Targets in Complex Background

Traditional recognition method is influenced by low SNR, low contrast, lack of shape and texture information of weak point target. Especially in complex background, the accuracy of automatic recognition for weak point targets is low. Therefore, a method to automatically recognize the dim point targets in infrared image based on BEMD(Bidimensional Empirical Mode Decomposition) was proposed. Based on spectral characteristics of the image to be identified and the frequency division method, this method compared the performance of different filters and the established the image filter bank. And then, our method used the filter bank to decompose the dim point target image into different sub-frequency domains. In addition, we applied Robinson filtering into sub-band image and thus to extract the dim point target. Furthermore, we used multi-layer empirical mode decomposition algorithm is used to decompose the input function of original dim point target image into a two-dimensional intrinsic mode function. Through the differential calculation, we obtained the difference between the original image and the background region, so that the region of dim point target could be segmented. Finally, we segmented the high-frequency information of dim point target region by local inverse entropy, so as to find the target recognition result of each modal function. Simulation results prove that the proposed method can extract the dim point target efficiently and accurately. Meanwhile, this method can suppress the complex background better.