Feature Tracking and Motion Factorization from Monocular Video
نویسندگان
چکیده
Determining shape and structure of moving objects from a single-camera video of their motion has been an active research area. Good, robust solutions to this problem enable, for example, the detection of hand gestures and facial expressions of a human subject by observation from a single camera. The aim of this project is to study how feature tracking and motion factorization can be coupled to implement a complete, robust motion capture system that requires monocular input. We examine the popular KLT tracker, and some ways of making it more robust; and see how it works in conjunction with the Tomasi-Kanade factorization algorithm for rigid body motion.
منابع مشابه
Factorization Method Using Interpolated Feature Tracking via Projective Geometry
The factorization method by Tomasi and Kanade simultaneously recovers camera motion and object shape from an image sequence. This method is robust because the solution is linear by assuming the orthographic camera model. However, the only feature points that are tracked throughout the image sequence can be reconstructed, it is difficult to recover whole object shape by the factorization method....
متن کاملSimultaneous Compaction and Factorization of Sparse Image Motion Matrices
Matrices that collect the image coordinates of point features tracked through video – one column per feature – have often low rank, either exactly or approximately. This observation has led to many matrix factorization methods for 3D reconstruction, motion segmentation, or regularization of feature trajectories. However, temporary occlusions, image noise, and variations in lighting, pose, or ob...
متن کاملApplication of 3D Human Motion in the Sports based on Computer Aided Analysis
3D human motion tracking is in recent years the field of machine vision is a very important research direction, it has a wide range of applications, such as human-computer interaction, intelligent animation, video surveillance and other. Currently about three-dimensional human motion tracking research mostly based multicast video, monocular video due to the depth information of the lack of the ...
متن کامل3-D Facial Pose and Gaze Point Estimation Using a Robust Real-Time Tracking Paradigm
Facial pose and gaze point are fundamental to any visually directed human-machine interface. In this paper, we propose a system capable of tracking a face and estimating the 3-D pose and the gaze point all in a real-time video stream of the head. This is done by using a 3-D model together with multiple triplet triangulation of feature positions assuming an affine projection. Using feature-based...
متن کاملRobust Fruit Counting: Combining Deep Learning, Tracking, and Structure from Motion
We present a novel fruit counting pipeline that combines deep segmentation, frame to frame tracking, and 3D localization to accurately count visible fruits across a sequence of images. Our pipeline works on image streams from a monocular camera, both in natural light, as well as with controlled illumination at night. We first train a Fully Convolutional Network (FCN) and segment video frame ima...
متن کامل