110 likes | 249 Views
Autonomous Mobile robots representation - H ough T ransforms & Floor Plane Extraction. 696415119 陳鼎文 696415135 王鴻嘉. Introduction. Hough Transforms(Straight edge extraction) Floor Plane Extraction in Artificial Environments Adaptive Floor Plane Extraction. Hough Transforms.
E N D
Autonomous Mobile robots representation - Hough Transforms & Floor Plane Extraction 696415119 陳鼎文 696415135 王鴻嘉
Introduction • Hough Transforms(Straight edge extraction) • Floor Plane Extraction in Artificial Environments • Adaptive Floor Plane Extraction
Hough Transforms • A pixel (xp,yp) in the image I is part of an edge. • Any straight line passing through (xp,yp) must satisfy : yp = m1xp+b1 • Where m1 is the slope, and b1 is the intercept with x or y axis.
Hough Transforms(cont’d) • Another pixel (xq,yq) in the image I is part of an edge. • Any straight line passing through (xq,yq) must satisfy : yq = m2xq+b2 • If m1=m2 and b1=b2 → the two line is the same • Where the line passes through both (xp,yp) and (xq,yq)
Hough Transforms(cont’d) • Voting mechanism: • Each pixel can vote for the (m,b) parameters. • The higher the score is, the stronger the edge is.
Hough Transforms(cont’d) • Algorithm: • Create a 2D array A with axes index m and b • Initialize A with all A[m,b]=0 • For all edge pixel (xp,yp) in I : If yp = mxp+bthen A[m,b]+=1 • Search A[m,b]>threshold , so that (m,b) correspond to a straight edge in I
Hough Transforms(cont’d) • Instead, we can use the transform: xcosθ + ysinθ = ρ • The range of θ is -90~+90 • The range of ρ is- ~
Floor Plane Extraction • To succeed Floor Plane Extraction , there are three assumptions: • Obstacles differ in appearance from the ground • The ground is flat and its angle to the camera is known • There are no overhanging obstacles
Floor Plane Extraction in Artificial Environments • Shakey’s artificial environment used homogeneously white floor tiles. • The walls and all obstacles were painted black. • The edge and the direction of edge provide the position and the orientation of wallsand obstacles.
Adaptive Floor Plane Extraction • We assume: the pixels at the bottom are part of the floor • Then we construct histograms based on the “floor sample.” • For example, the intensity, hue … histogram
Adaptive Floor Plane Extraction(cont’d) • If the target pixel has a hue that never occurred in“floor sample” → the corresponding hue histogram count=0 → the pixel is recognized as an obstacle • Results show in P.175 Figure 4.48