计算机视觉课件_LC25-sfm.pptVIP

Three-Dimensional Structure from Video Sequence Dr. Stephen (Shengyong) Chen Imperial College London A Moving Camera / Object structure for motion: automatic recovery of camera motion and scene structure from two or more images. It is a self calibration technique and called automatic camera tracking or matchmoving. Questions Correspondence (Feature extracting and stereo matching): Given a point in just one image, how does it constrain the position of the corresponding point in another image? Scene geometry (structure): Given 2D point matches in two or more images, where are the corresponding points in 3D? Camera geometry (motion): Given a set of corresponding points in two or more images, what are the camera matrices for these views? Problem Modeling Given: m images of n fixed 3D points  xij = Pi Xj , i = 1, … , m, j = 1, … , n Problem: estimate m projection matrices Pi and n 3D points Xj from the mn correspondences xij The General Procedure Structure from motion Step 1: Track Features Detect good features, Shi Tomasi, SIFT Find correspondences between frames Lucas Kanade-style motion estimation window-based correlation SIFT matching KLT tracking Structure from Motion Step 2: Estimate Motion and Structure Simplified projection model, e.g., [Tomasi 92] 2 or 3 views at a time [Hartley 00] Problem statement Structure from Motion Step 3: Refine estimates “Bundle adjustment” in photogrammetry Other iterative methods Structure from Motion Step 4: Recover surfaces (image-based triangulation, silhouettes, stereo…) Results Results Results Results Results Results Results Results Results Results Results Results Results Limitation Propose for Future * * Feb 13, 2009 ? Project Report Unknown camera viewpoints x1j x2j x3j Xj P1 P2 P3 feature tracking 3D estimation optimization (bundle adjust) visualization pipeline 2D feature selection /~stb/klt/ Good mesh - - - - 263 tracked points 860 tracked points 860 tracked points A van A van A van A van A van A van