View-based localization in outdoor environments based on support vector learning

Visual based Localization for mobile robots with Support Vector Machines

“…These transferring functions are called kernel functions. The selection of kernels is a challenge issue, and Radial Basis Function (RBF) are the mostly used kernels nowadays [2] [13] SVM is a simple and reliable method in the area of classification and machine learning. It has many advantages over other machine learning algorithms, such as Artificial Neural Network (ANN).…”

“…As we know, ANN has been widely used in mobile robot localization. However, SVM was seldom applied in this area apart from a two-stage SVMs system in outdoor navigation [13]. Autio and Elomaa used SVMs in an indoor environment to classify the features [2].…”

“…Each block generates a Y/N value; therefore the whole frame is transferred to a position vector with 90 binary elements in this stage. Unlike an outdoor environment in [13], the most useful information of indoor images is distributed in the middle part of an image. The image processing module focuses on the part from the 120 th to the 360 th in row, and the 20 th to the 620 th in column in a 640*480 image.…”

Outdoor autonomous navigation using SURF features

“…Another navigation method is the view-based approach. [5][6][7] In the view-based approach, many images are memorized and self-location estimation is performed by template matching. In general, image data contains a very large amount of information and calculating the cost of template matching is large.…”

Place recognition-based fixed-lag smoothing for environments with unreliable GPS

“…Morita et al [3] apply a support vector machine to learn image features in key images and perform localization using learning-based frame matching. Bradley et al [4] propose another example of outdoor topological localization based on place recognition.…”