基于提升小波变换和分形维数的声纳图像识别

分形理论在图像的纹理识别中得到了广泛应用,由于分形维数不能反映图像的空间信息,容易造成误识别。针对该问题并结合声纳图像的特点,通过提升结构构造了Haar小波,并将提升小波变换同分形理论相结合,利用小波分解的多分辨率特点和分形维数的多尺度特性,提高图像的识别率。采用Levenberg-Marquardt(L-M)算法优化的BP神经网络对不同信噪比的声纳图像进行分类识别。实验结果表明,文中方法不论在识别率还是识别时间上均优于传统纹理识别方法。

Sonar image recognition based on lifting scheme and fractal dimension

Fractal dimension has been widely used in the recognition of the texture images,but it lacks the ability to describe spatial information of images. In Considering the characteristic of a sonar image,the paper uses the lifting scheme to construct the Haar wavelet,and relates the lifting scheme with fractal d-imension. Amalgamation of multi-scales characteristics of wavelet transform and fractal dimension increased the recognition rate. LMBP neural network is used to recognize the sonar images of different SNR. The results show that the new method has a higher classification rate and is more efficient than traditional methods.