论文部分内容阅读
针对应用传统分类器和被动学习的方法,难以满足遥感图像处理实际应用的要求这一困境,提出了一种新的基于多分类SVM的主动学习方法,与被动学习的随机选择不同,主动学习是在少量标记类别的初始训练样本集基础上,通过反复迭代主动学习的方式,得到最有利SVM分类器性能的样本为支持向量。研究表明,这种方法直接避免了大量的计算,可有效地减少样本训练时需要标记样本的数目,并取得较为理想的分类效果。
Aiming at the difficulty of applying the traditional classifier and passive learning method to meet the practical application requirements of remote sensing image processing, a new active learning method based on multi-classification SVM is proposed. Compared with the random selection of passive learning, active learning is Based on the initial training sample set with a few markup classes, the most favorable SVM classifier performance is obtained by repeated iterative active learning methods as support vector. The research shows that this method directly avoids a large number of calculations, which can effectively reduce the number of samples required for training samples and achieve better classification results.