LBP-自适应增强模型的木材纹理分类

Wood Texture Classification Based on LBP-ADABOOST

针对传统木材纹理分类的准确率低且难度大的问题,依据LBP(局部二值)算子和ADABOOST(自适应增强)算法理论,提出了LBP-ADABOOST模型对木材纹理进行识别分类.通过均匀旋转不变特性与原始LBP算子相融合,提取纹理的特征值,结合自适应增强算法,从而训练得到每类纹理所对应的分类器模型参数,构造分类器,实现对木材纹理准确高效分类.实验结果表明相比于BP神经网络,SVM支持向量机等分类算法,该模型的实验结果误差率为4%左右,准确率高,实用性强.

Aiming at the issue of low categorization accuracy and tough calamity, based on LBP （local binary） operator and ADABOOST （adaptive enhancement） algorithm theory, a LBP-ADABOOST model on wood texture classification is proposed. By uniform rotation invariant features and the integration of the original LBP operator to extract texture feature values, combined with adaptive enhancement algorithms to obtain training corresponding to each type of texture classification model parameters, weconstruct a classifier, which achieves accurate and efficient wood texture classification. The experimental results show that the model error rate is about 4%, has higher accura- cy and practicality than BP Neutral Networks, Support Vector Machine.