基于SVM的资源三号测绘卫星影像多特征分类

ZY-3 Satellite Image Multi-feature Classification Based on SVM

针对传统分类方法精度不高、感兴趣目标分类不理想等缺陷,采用多特征组合的支持向量机影像分类方法,利用颜色矩、颜色集和灰度共生矩阵进行特征提取,总体精度、Kappa系数和混淆矩阵作为评价指标对单一特征、组合特征的不同分类结果进行分析。实验结果表明,该方法有效地解决了单数据源分类不完整、精度低等问题,对高维输入向量具有较高的推广力。

To avoid the drawbacks of traditional classification method that low precision and bad classification result of the interested targets,a method for image classification based on support vector machine(SVM)was proposed by making use of multi-feature.We utilized color moment,color sets and gray level co-occurrence matrices(GLCM) to extract the feature.The different classification results of single feature and multifeature were analyzed by the evaluation index of overall accuracy,kappa coefficient and confusion matrix.The experimental results show that the problem that incomplete and low precision of single data classification can be effectively solved.The method for high dimensional vector has high generalization ability and has obtained better practical effect.