融合分数维特征的水岸彩色图像水岸界线识别  

Recognition of near bank scene boundary between water surface and bank based on fractal dimension

作　　者：邵承会[1] 程光明[2] 唐可洪[2] 阚君武[1] 杨志刚[1] 

机构地区：[1]吉林大学链传动研究所,吉林长春130025 [2]吉林大学机械科学与工程学院,吉林长春130025

出　　处：《光电工程》2005年第3期78-81,共4页Opto-Electronic Engineering

基　　金：国家 863 计划资助项目(2002AA423150)

摘　　要：针对复杂的水岸景物彩色图像水岸界线识别,提出了融合多特征的分数维方法。该方法用提取边界来降低维数;再将横向细长窗口内的 Lebesgue 测度(μmin =260)和水岸界线形状位置等特征作为限定条件,只对符合条件点应用基于 ε-blanket 方法的覆盖技术进行迭代计算,找到分数维数最接近 1 的窗口;用经过部分改造的最小二乘法拟合窗口内符合条件点即为水岸界线。这样可以大幅度减少计算量,需迭代计算像素点数量仅为原算法的 0.88%。对多种自然条件下水岸图像 1026 幅进行识别,完全识别率达 85.3%,反应出该方法的鲁棒性与实用性。Fractal features were applied lonely in many image analyses, then their results were integrated with that of other techniques. So these methods were not only time-consumer but also involving conflicting and incommensurate objectives. A hybrid feature fractal method is proposed to recognize the boundary between water surface and bank, which contains complex near shore scenes. The method firstly extracts boundary to reduce the dimensionality of the original input pattern, and then takes Lebesque measure (μmin = 260) and other basic features (figure and location) in a lateral thin and long window to reduce the iterative calculation pixels. Finally the reduced blanket technique, which based on Mandelbrot's ε-blanket idea, is adopted and applied to recognize the boundary between water surface and the bank based on the edge image. Because edge detection reduces image dimensionality, the total iterative computation pixels are only 0.88 percent of that of the original algorithm. Experimental results demonstrate the effectiveness of the proposed technique. Applied to a great many near bank images in diverse weather conditions, the rate of well recognition is 85.3%.

关 键 词：图像识别 分数维数 水岸界线 边缘提职 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]