Artificial Target Recognition with Multi-Resolution Analysis and Wavelet Holder Constant
-
摘要: 从分形布朗运动的平均Holder数出发,提取人造目标与自然场景的相对距离,用小波 变换提高抗畸变及干扰能力,通过不同分辨率下的小波Holder数的计算,并进行直线拟合,根 据直线的斜率及线性度可区分目标和背景,提出了一种多分辨分析相对距离的目标识别算法 (MRRD).实验表明,该方法可识别用单纯的分形特征参数难以识别的目标,是一种有效的识别 人造目标的方法,具有算法效率高,便于实时处理的特点.Abstract: The average Holder constant of fractional Brownian motion is described, and the different relative distance between the target and complex background is extracted. It is good in noise immunity to use wavelet translation. The wavelet Holder constants which are linear interpolated are calculated in a serial of different multi-resolution images, the target is recognized by detection of the linearity error. A novel algorithm of target recognition with multi-resolution analysis and relative distance (MRRD) is proposed. Experiments show that this algorithm is suitable for identifying targets which are difficult to recognize in fractal feature parameter, and that it has simple computation and is convenient for real-time processing.
-
计量
- 文章访问数: 2959
- HTML全文浏览量: 132
- PDF下载量: 1007
- 被引次数: 0