Two-dimensional laser feature extraction based on improved successive edge following

Successive edge following (SEF) has been widely used to describe the environmental characteristics based on a two-dimensional (2D) laser range finder due to its simplicity. However, the segmentation accuracy of the regular SEF for different distances is very low. And besides that, the regular SEF sometimes fails to characterize the corner features in the continuous segmentation. To solve these problems, we propose an improved SEF approach, which combines 2D polar radius–arc, adaptive threshold in a region to divide different radius data into groups more reasonably. Region growing is also discussed to localize the corner based on variance of segmentation and its fitting. Finally, all features described by line segments are used to build the map of the environment. As compared to the regular SEF, our approach is information-lossless. Our experimental results show that the segmentation accuracy is improved, while the segmentation number is more realistic and the feature detection accuracy is increased, as compared with the regular SEF.