Active Learning Svm-based Intelligent Vehicle Obstacle Detection

Many current intelligent vehicles have perception systems that are primarily hand-tuned, which makes them difficult to adapt to new tasks and environments. Many supervised learning algorithms can automatically adjust the parameters of complex perception systems which offer a practical solution to this problem and have been widespread concerned in the field of robotic perception.This thesis uses SVM algorithm which is a machine learning methods used on classification issues to solve obstacle detection problem in field of intelligent vehicle. SVM shows many unique advantages in resolving the small sample, non-linear and high-dimensional pattern recognition problems, also its ability to promote is better than some of the traditional learning methods.An important difficulty in using machine learning techniques for large scale robotics problems comes from the fact that most supervised algorithms require labeled data for training. Large data sets occur naturally in outdoor robotics applications, and labeling is most often an expensive process. This makes the direct application of learning techniques to realistic perception problems in our domain impractical.In light of this situation, the thesis proposes the active learning methods on the basis of the introduction of the traditional SVM algorithm. Active learning methods can address the data labeling problem by analyzing the unlabeled data and automatically selecting for labeling only those examples that are important for the classification problem of interest, therefore reduces the number of samples required for training and the price of labeling.We have done a lot of experiments using the data from the real world and demonstrate that active learning SVM algorithm results in significant reductions in the amount of data labeling required and at the same time ensure the accuracy of classification, so as to enhance the self- adaptive ability of the intelligent vehicle system.