用于特征提取的Gabor滤波器参数设计

Parameter Design of Gabor Filters for Feature Extraction

Gabor滤波器的参数设计是Gabor特征提取过程中的一个重要环节。根据坦克的外形特点及对方向的敏感性对其进行了针对性的参数设置,实验结果证明得到了很好的滤波效果。根据Gabor核函数的窗口特性,提出了一种自适应的基于Gabor滤波器的特征提取算法。该算法应用Gabor滤波器的多尺度特性与样本图像进行卷积,对Gabor域响应进行局部极大值提取,经过该方法得到的特征点有效地减少了数据冗余,并且具有较好的图像表征能力。

For feature extraction based Gabor filters, the design of filter parameter is very important. In this paper, parameters are designed according to the appearance characters and the sensitiveness of the orientations of tank. Experimental re suits show that the method is effective. Besides, according to the characteristics of Gabor kernels function window, an adaptive feature extraction algorithm is put forward based on Gabor filter. In this algorithm, features are extracted with the multi-scale recognition of Gabor filter and the value of the extracted features are the maximum value in the defined windows. It is proved by the experiments that the novel feature extraction algorithm is effective both in data redundancy and image rep resentation.