基于深度置信网络的地表分类算法

Surface Classification Algorithm Based on Depth Belief Network

复杂地形环境下的数据特征维度往往较大以及数据不平衡,传统地形识别研究使用的浅层算法如Softmax、支持向量机(Support Vector Machine,SVM)等面对复杂地形时表征能力下降分类精度不理想。论文在研究了传统方法和深度学习理论后,使用深度置信网络(Deep Belief Network,DBN)与Softmax进行有效结合后应用于地形识别研究,利用中心对称局部二值模式与颜色直方图结合获得特征,实验证明论文提出的算法比传统算法展现了更好的分类效果。

In complex terrain environment,the data feature dimension is usually large and the data is not balanced.The tradi⁃tional shallow algorithms such as Softmax and Support Vector Machine(SVM)used in terrain recognition research decrease the rep⁃resentation ability and the classification accuracy is not ideal when facing complex terrain.In this paper,after studying the tradition⁃al methods and deep learning theory,Deep Belief Network(DBN)and Softmax are used for effective combination of terrain recogni⁃tion research,using the centrosymmetric local binary mode and color histogram to obtain features.Experimental results show that the proposed algorithm has better classification effect than the traditional algorithm.