基于多波束声呐图像的长江河道底质识别与分类

Sediment Identification and Classification of the Yangtze River Channel Based on Multi-beam

基于多波束对长江河道底质分类关键问题进行了研究,对多波束反射强度数据进行改正并对多波束声呐图像进行预处理,采用灰度共生矩阵对底质反向散射强度图像进行纹理特征提取,最后将提取底质图像样本作为自组织特征映射神经网络和随机森林两种分类方法的训练数据,使用训练好的预测分类模型对反向散射强度图像进行全图底质分类.实验结果表明,SOM与随机森林分类方法的总体分类精度分别达到了82.5%与85.4%,对底质声呐图像实现了较好的预测分类效果.

This paper studies the key issues of sediment classification in the Yangtze River channel based on multi-beam,corrects the multi-beam reflection intensity data and pre-processes the multi-beam sonar image,then uses the gray level co-occurrence matrix to characterize the texture of the backscatter intensity image of the substrate extraction.Finally,the extracted background image samples are used as training data for two classification methods,self-organizing feature mapping neural network and random forest.And the trained prediction classification model is used to classify the backscattered intensity image of the whole image.The experimental results show that the overall classification accuracy of the SOM and random forest classification methods reaches 82.5%and 85.4%,respectively.This study achieves a good prediction and classification effect for the bottom-quality sonar images.