Mobile Robot Localization and Mapping by Scan Matching using Laser Reflection Intensity of the SOKUIKI Sensor

A Review of Point Cloud Registration Algorithms for Mobile Robotics

“…Another study for 3D point clouds also concludes that moment grid is better than 3D shape context, spin image, shell image and local covariance [Bosse and Zlot, 2009b]. Usually, ICP is done using only geometric features, but some works also present results using the intensity remission from an Hokuyo [Yoshitaka et al, 2006] and from specialized system using three different wavelengths [Godin et al, 1994]. Laser range finders are also combined with cameras to add color information to measured points [Schutz et al, 1998, Druon et al, 2006.…”

“…To cope with stereo reconstruction noise, Diebel et al [2004] remove points with distance and surface angle to neighbors larger than two times the median of all distances and surface angles within the point cloud. When using laser's remission intensity, which is not invariant to distance and angle, Yoshitaka et al [2006] propose to keep points only close to the laser to reduce the impact of distance on the intensity value. For color images, points with low saturation value tend to be gray and are removed before applying clustering technique based on hue [Druon et al, 2006].…”

“…One way is to express descriptors in the feature space using a conversion factor. This was used with surface orientations Hügli, 2002, Bosse andZlot, 2009b], surface orientations and color [Schutz et al, 1998], color [Johnson and Kang, 1997] and laser intensity [Godin et al, 1994, Yoshitaka et al, 2006. The other way around is also possible (i.e., expressing feature positions in descriptor space).…”

“…More integrated solutions rely on external sensor (wheel encoders, Inertial Measurement Unit (IMU), Global Positioning System (GPS), etc.) or a fusion of them as first guess [Diebel et al, 2004, Yoshitaka et al, 2006. Multiple initial guesses from the same external motion model and based on particle filters are also used in robotics [Grisetti et al, 2005] and in human body motion tracking [Kim, 2010].…”

“…Historically, robotic applications mostly utilize 2D laser rangefinders 1 for indoor [Yoshitaka et al, 2006] and outdoor Zlot, 2009b, 2008] localization. With the need for 3D localization, systems based on rotating 2D laser rangefinders recently received an increasing attention for both indoor [Armesto et al, 2010] and outdoor [Segal et al, 2009] environments.…”