利用多分辨分形特征进行目标识别  

TARGET RECOGNITION BASED ON MULTIRESOLUTION FRACTAL FEATURE

作　　者：符新伟[1] 余秋星[1] 李志[1] 

机构地区：[1]西北工业大学航海工程学院,22信箱西安710072

出　　处：《电子与信息学报》2003年第11期1575-1577,共3页Journal of Electronics & Information Technology

摘　　要：该文分析了湖底回波信号经小波多分辨分解后,发现不同分解尺度上的小波系数的能量分别与分解尺度在一定范围内满足幂指数关系,与分形维的定义相似。据此求得两类特征参量,并对五类回波进行了分类,取得了较为理想的识别结果。In this paper, echoes are decomposed based on wavelet multiresolution analysis. It is shown that the energy of wavelet coefficients is exponential to the decomposed scale in some degree, which is similar to the definition of fractal dimension. Hence, two sorts of features are extracted and applied to recognizing five classes of echoes. The recognition results are satisfactory.

关 键 词：多分辨分形特征 目标识别 小波系数 分解尺度 分形维 信号 

分 类 号：TN911.7[电子电信—通信与信息系统]