The CMU Robotics Institute has released a new technical report titled, "Scale Selection for Classification of Point-Sampled 3D Surfaces" (PDF format). Mobile robots frequently use sensors like the SICK laser scanners or sonar that return a cloud of individual range points when measuring a 3D scene. This paper presents an automated solution to the common problem of figuring out a scale for extracting meaningful features from the point cloud such as determining whether you're looking at a rock or a bush or a man-made structure. The research was done at the Vision and Mobile Robotics Lab and was sponsored by the US Army and the National Science Foundation.