高级检索

    一种基于纹理域神经网络的彩色卫星图像分析方法

    A CLASSIFICATION METHOD BASED ON TEXTURE NEURAL NETWORK FOR COLOR REMOTE SENSING IMAGES

    • 摘要: 描述了一种利用纹理域神经网络 ,通过监督的学习方法来调整和训练权重 ,将预处理后合成的彩色卫星图像中的纹理信息块进行分类 ,从而达到在彩色卫星图像中将有用的色彩块分析出来的目的 .给出了纹理域的预处理过程 ;阐述了对原彩色图像进行颜色映射和量化的目的和方法 ;讨论了纹理域神经网络的构造方法及算法 ;最后给出了用以上方法对彩色卫星图像进行分类的实验结果 .实验证明该方法对彩色卫星图像有较好的分类作用

       

      Abstract: This paper gives a color image analysis method based on texture neural network. The texture neural network can be applied in the classification of color remote sensing images through network learning. For the convenience of texture classification, an algorithm to preprocess the original images is provided. Then the algorithm transforms RGB color range to HSV color range. By using this method, the information of the color image can be reduced. A method to construct a texture neural network is also provided. After constructing the texture neural network, it combines every eight neighboring field as an input data of this neural network. If users pick out the color fields they are interested in, they can use these vectors to treat the neural network. Finally, they can use this neural network to classify this color remote sensing image. The algorithm has been implemented on a PC platform to classify color remote sensing images. The result of the experiment demonstrates that the method is efficient and useful.

       

    /

    返回文章
    返回