海底声学底质分类的ELM-AdaBoost方法

Based on the Adaptive Boosting algorithm (AdaBoost) combined with the Extreme Learning Machine (ELM), by iterating, adjusting and optimizing the weights between ELM classifiers, a strong ELM - AdaBoost classifier with strong robustness and high precision is constructed, which enhances the stability of the existing ELM classifier. Taking the side - scan sonar images of the Pearl River Estuary sea area as experimental data, the three typical substrates of reef, sand and mud are classified and identified. The average classification accuracy of this method exceeds 90%, which is better than the average classification accuracy of 85.95% of a single ELM classifier, and also better than traditional classifiers such as LVQ and BP, and the time consumed for classification is also much less than that of traditional classifiers. The experimental results show that the ELM - AdaBoost method constructed in this paper can be effectively applied to the classification of seabed acoustic substrates and can meet the requirements of real - time substrate classification.

基于自适应增强算法(AdaBoost)结合极限学习机(ELM),通过迭代、调整、优化ELM分类器之间的权值,从而构建了具有强鲁棒性、高精度的ELM-AdaBoost强分类器,增强了现有的ELM分类器的稳定性。以珠江口海区侧扫声呐图像为实验数据,对礁石、砂、泥3类典型底质进行分类识别,该方法的平均分类精度超过90%,优于单一ELM分类器的平均分类精度85.95%,也优于LVQ、BP等传统分类器,且在分类所耗时间上也远少于传统分类器。实验结果表明,本文构建的ELM-Ada- Boost方法可有效应用于海底声学底质分类,可满足实时底质分类的需求。