Corner Feature Extraction: Techniques for Landmark Based Navigation Systems

Abstract

Accurate and reliable maps generated autonomously guarantees improved localization especially in GPS denied surroundings like indoor (Hough, P.V.C, 1959). The use of odometry is not sufficient for position estimation due unbounded position errors. Therefore, since office like environments consists of planar surfaces, a 2D space model is adequate to describe the robot surroundings because objects are predominantly straight line segments and right angle corners. Coincidentally, line segments and corner representation are the two most popular methods for indoor modelling from a laser rangefinder. The focus in this paper however is corner extraction methods. A number of line and corner extraction techniques first transform scan data into Cartesian space then a linear regression method or corner extraction algorithm is applied. Some algorithms employ Hugh transform (Hough, P.V.C, 1959). & (Duda, R. O, 1972) a popular tool for line detection from scan data due to its robustness to noise and missing data. It works in sensor measurement space. However, the computational cost associated to its voting mechanism renders real-time implementation impossible. On the other hand, an early work by (Crowley, J, 1989) paved the way to subsequent line extraction methods from a range sensor. In their work, a process for extracting line segments from adjacent co-linear range measurements was presented. The Kalman filter update equations were developed to permit the correspondence of a line segment to the model to be applied as a correction to estimated position. The approach was recently extended by (Pfister, S.T et al. 2003), first providing an accurate means to fit a line segment to a set of uncertain points via maximum likelihood formalism. Then weights were derived from sensor noise models such that each point's influence on the fit is according to its uncertainty. Another interesting work is one by (Roumeliotis & Bekey, 2000), where two Extended Kalman filters are used to extract lines from the scan data. In the algorithm, one Kalman filter is used to track the line segments while the other estimates line parameters. The combination of the two filters makes it possible to detect edges and straight line segments within the sensor field of view. There are many features types one can extract from a laser sensor, and are dependent on the obstacles found in the room. If the room has chair and table, one would be tempted to extract point features from their legs. Size, shape and texture of objects contribute to the type of feature to extract from the sensor. The use of generalised algorithms is not uncommon, i.e. algorithms which extract lines from wall, point features from table legs and arcs to categorise circular objects (Mendes, & Nunes, 2004). The parameters that distinguish each extracted feature makes up the map or state estimate. The key to a successful robot pose estimation lies in its ability to effectively extract useful information about its location from observations (Li & Jilkov, 2003). Therefore we proposed an improved c...

Feature extraction from 2D laser range data for indoor navigation of aerial robot

“…Consider a full scan of 270° as an ordered sequence of measurements points ( ) can be defined in Cartesian or in polar coordinate as follow [3]: (1) represent the angle and for the distance of point . According to (1), a segmentation can be defined as bellow:…”

“…A robot is said to be fully autonomous if it is able to build a navigation map [1]. A 2D map representing the geometric feature of the robot surroundings is sufficient for the navigation of ground robot.…”

Natural corners-based SLAM with partial compatibility algorithm

“…Landmark features should be carefully considered because the uncertainties of the landmark position can affect the case of updating the position of a mobile robot as well as other stored landmarks. For this reason, the corners in an unknown environments have been chosen as landmarks by using vision sensors as in Rosten and Drummond, 7 Shen and Wang 8 and Hwang and Song, 9 while others have used sonar or laser sensors 10,11 but without the exact calculation of the distance uncertainty and angle uncertainty of the corner landmarks.…”

“…The basis of the corner extraction is line segments’ extraction from the raw sensor data. Line features have been used successfully in many experiments, 12,13 and there have also been many methods of line extraction, including the split-and-merge, 10,14 incremental probabilistic technique 15 and Hough transform. 16 Some experiments have been done by the evaluating different line extraction methods as in Nguyen et al 17 which showed that the split-and-merge and incremental method demonstrate the best performance.…”