摘要
文中首先对红外图像进行预处理，然后根据分形理论提取了红外图像的９种分形特征，用神经网络进行目标识别。这９种分形特征包括基于分形维数的特征，基于Ｈｕｒｓｔ指数的分形特征和基于缝隙的分形特征。最后进行了计算机仿真实验，实验表明基于分形特征红外图像识别方法是可行的，并取得了较好的结果。
In this paper, firstly the infrared image was preprocessed, and 9 features were extracted from infrared image based on fractal theory, then neural network was used for target recognition. The 9 fractal features include the features based on fractal dimesion, the feature based on Hurst exponent and the feature based on lacunarity. The features based on fractal dimension include the fractal dimensions of original image, high gray value image, low gray value image, horizontal smoothing image, vertical smoothing image, mutil fractal dimension and mutil scale fractal dimension. Because the difference box counting algorithm and the algorithms of computing Hurst exponent and lacunarity feature are highly efficient, the velocity of feature extracting and recognition is quick enough. The computer simulation results show that the method based on fractal feature for infrared image recognition is available and gives better result.
出处
《红外与激光工程》
EI
CSCD
1999年第1期20-24,共5页
Infrared and Laser Engineering
关键词
多重分形
神经网络
目标识别
红外图像
Multi fractal\ \ Multi scale fractal\ \ Hurst exponent\ \ Neural network Target recognition\ \ Infrared image
