For Surface Classification Support Vector Machine (svm) Active Learning Method Of Study

Mobile robot has been a hotspot in robotics. In the related technologies of mobile robot, perception of surroundings is the important premise for following tasks. Therefore, the study on terrain classification is great significant. Recently, machine learning combined with environment perception has gradually become a hot problem in the fields of mobile robot, machine learning and computer vision at home and abroad.This paper uses SVM algorithm to solve terrain image classification issues and implements terrain classification system based on SVM. Traditional support vector machine algorithms require a large number of labeled samples for training first, and then form a classifier to classify images. However, artificial labeling is a laborious and time consuming work, a large number of training samples also affect training speed of classifier.In light of this situation, the thesis proposes the active learning methods, which doesn't need to label all the samples which should be trained. In machine learning process, new samples which mostly conduce to improving the performance of classifier are forwardly chosen from the candidate samples set and added to the training samples set to study, so the number of training samples and the cost of marking samples can be reduced effectively.According to the theories of SVM and active learning, this paper proposes an algorithm combining active learning with SVM and applies it to the terrain classification. Finally, experiment shows that the SVM based on active learning is effective, which can effectively reduce the number of samples on the premise of keeping correctness of the classifier.