【作者】 李国强；

【导师】 沈艳；

【作者基本信息】 哈尔滨工程大学 ， 应用数学， 2011， 硕士

【摘要】 海杂波信号具有很强的相关性,特别是体现在海尖峰效应当中,这种相关性造成了常规的基于海杂波功率谱的频域滤波方法对海杂波的抑制效果不理想。为了有效地抑制海杂波,本文回避了在常规方法中出现的海杂波相关性问题,提出了一种从图像处理的角度来抑制海杂波的方法。该方法先利用图像复原技术对由航海雷达得到的海浪图像进行图像复原,去除海浪图像中的噪声,得到海杂波图像,之后利用图像分割技术判断海杂波图像中是否含有目标,如果有目标存在,将其分离出来。图像复原部分提出了基于神经网络和数学形态学的图像复原技术,既继承了神经网络的拟合性质和收敛性质,又应用一种新型的形态学变形虫结构元素克服了传统滤波方法的缺点,在保持图像边缘信息的前提下,有效地平滑了图像中的噪声。四种改进的图像复原技术被给出,分别为基于Hopfield神经网络和数学形态学的图像复原技术、基于小波神经网络和数学形态学的图像复原技术、基于暂态混沌神经网络和数学形态学的图像复原技术及基于细胞神经网络和数学形态学的图像复原技术。图像分割部分提出了两种新的图像分割技术和两种改进的基于神经网络的图像分割技术,分别为基于中心灰度级信息矩阵法的图像分割技术、基于级间灰度级信息矩阵法的图像分割技术、基于Hopfield神经网络的图像分割技术及基于暂态混沌神经网络的图像分割技术。为了验证从图像处理角度提出的海杂波抑制方法,本文应用由IPIX航海雷达采集的无目标和有目标的两组海浪图像数据进行仿真实验,并对实验的结果做了细致的比较和分析,证实本文方法能够有效抑制海杂波。更多还原

【Abstract】 The signal of sea clutter has a very strong correlation, especially manifested in the effect of sea spike. This correlation makes the traditional method of frequency-domain filtering based on power spectrum of sea clutter have not a so good suppressive effect. In order to suppress the sea clutter effectively, a method to suppress the sea clutter from the aspect of image processing is presented, which avoids the problem of sea clutter correlation emerged in traditional method. This method first makes use of image restoration technology to restore the ocean wave image obtained by the marine radar, and filters the noise in this ocean wave image to obtain the sea clutter image. And then it takes advantage of image segmentation technology to judge whether there is target in the sea clutter image. If does, separate them.The image restoration part presents image restoration technique based on neural networks and mathematical morphology. It inherits fitting character and convergence property of neural networks. Besides, it presents a new type of morphology amoeba structuring element to overcome the disadvantages of traditional filter methods and the noise of image has been smoothed effectively on the premise of keeping image edge information. Four types of improved image restoration technique are given, including one based on Hopfield neural network and mathematical morphology, one based on wavelet neural network and mathematical morphology, one based on transiently chaotic neural network and mathematical morphology, and one based on cellular neural network and mathematical morphology.The image segmentation part presents two new types of image segmentation technique and two improved ones based on neural networks, including one based on gray level information matrix method, which is obtained by minimizing the sum of distance values that is from every within-class gray level to central gray level, one based on gray level information matrix method, which is obtained by minimizing the sum of distance values that is from one within-class gray level to another, one based on Hopfield neural network, and one based on transiently chaotic neural network.In order to verify the method of sea clutter suppression presented from the angle of image processing, two groups of data that one group has a target and the other not, which is collected by the IPIX marine radar is used to conduct simulation experiment. The experiment results have been analyzed and compared carefully, which show this method can suppress sea clutter effectively.更多还原