@InProceedings{Kulic2009d,
  Title                    = {Whole body motion primitive segmentation from monocular video},
  Author                   = {Kuli\'{c}, D. and Dongheui Lee and Nakamura, Y.},
  Booktitle                = {Proceedings of the IEEE International Conference on Robotics and Automation},
  Year                     = {2009},
  Pages                    = {3166-3172},

  Abstract                 = {This paper proposes a novel approach for motion primitive segmentation from continuous full body human motion captured on monocular video. The proposed approach does not require a kinematic model of the person, nor any markers on the body. Instead, optical flow computed directly in the image plane is used to estimate the location of segment points. The approach is based on detecting tracking features in the image based on the Shi and Thomasi algorithm. The optical flow at each feature point is then estimated using the Lucas Kanade Pyramidal Optical Flow estimation algorithm. The feature points are clustered and tracked on-line to find regions of the image with coherent movement. The appearance and disappearance of these coherent clusters indicates the start and end points of motion primitive segments. The algorithm performance is validated on full body motion video sequences, and compared to a joint-angle, motion capture based approach. The results show that the segmentation performance is comparable to the motion capture based approach, while using much simpler hardware and at a lower computational effort.},
  Doi                      = {10.1109/ROBOT.2009.5152266},
  ISSN                     = {1050-4729},
  Keywords                 = {feature extraction;humanoid robots;image motion analysis;image segmentation;image sequences;mobile robots;pattern clustering;video signal processing;Lucas Kanade pyramidal optical flow estimation algorithm;Shi-Thomasi algorithm;coherent cluster;coherent movement;feature detection;feature tracking;image plane;location estimation;monocular video sequence;whole body motion primitive segmentation;Clustering algorithms;Computer vision;Hardware;Humans;Image motion analysis;Image segmentation;Kinematics;Optical computing;Tracking;Video sequences},
  Review                   = {Segmentation based on video. Uses optical flow.

If optical flow is evaluated on the entire image, many of the pixels will contain invalid estimates, due to lack of texture
or other unique features around a given pixel, occlusions between the two frames, depth discontinuities or reflection
highlights. To reduce computational burden and the number of invalid optical flow values, optical flow is not computed
for the entire image. Instead, the image is first searched for good features to track, using the Shi and Tomasi algorithm. 
This algorithm formally defines the feature�s quality based on how well it can be tracked.

Segment is declared when a cluster (in an image) starts/stops moves or changes directions. When a portion of the body is moving, optical flow occurs, which disappears at start/end of motion. 

Once reliable features are found in the image, the optical flow for those features is computed via the pyramidal version
of the Lucas Kanade algorithm [2], [27]. The Lucas Kanade algorithm solves the optical flow equation (Equation 2)
iteratively, using Newton-Raphson},
  Timestamp                = {2013.10.02}
}

{"_id":"DBzMeguqLdXMwm8Ma","bibbaseid":"kuli-lee-nakamura-wholebodymotionprimitivesegmentationfrommonocularvideo-2009","downloads":0,"creationDate":"2017-09-14T16:34:36.679Z","title":"Whole body motion primitive segmentation from monocular video","author_short":["Kulić, D.","Lee, D.","Nakamura, Y."],"year":2009,"bibtype":"inproceedings","biburl":"https://raw.githubusercontent.com/jfslin/jfslin.github.io/master/jf2lin.bib","bibdata":{"bibtype":"inproceedings","type":"inproceedings","title":"Whole body motion primitive segmentation from monocular video","author":[{"propositions":[],"lastnames":["Kulić"],"firstnames":["D."],"suffixes":[]},{"firstnames":["Dongheui"],"propositions":[],"lastnames":["Lee"],"suffixes":[]},{"propositions":[],"lastnames":["Nakamura"],"firstnames":["Y."],"suffixes":[]}],"booktitle":"Proceedings of the IEEE International Conference on Robotics and Automation","year":"2009","pages":"3166-3172","abstract":"This paper proposes a novel approach for motion primitive segmentation from continuous full body human motion captured on monocular video. The proposed approach does not require a kinematic model of the person, nor any markers on the body. Instead, optical flow computed directly in the image plane is used to estimate the location of segment points. The approach is based on detecting tracking features in the image based on the Shi and Thomasi algorithm. The optical flow at each feature point is then estimated using the Lucas Kanade Pyramidal Optical Flow estimation algorithm. The feature points are clustered and tracked on-line to find regions of the image with coherent movement. The appearance and disappearance of these coherent clusters indicates the start and end points of motion primitive segments. The algorithm performance is validated on full body motion video sequences, and compared to a joint-angle, motion capture based approach. The results show that the segmentation performance is comparable to the motion capture based approach, while using much simpler hardware and at a lower computational effort.","doi":"10.1109/ROBOT.2009.5152266","issn":"1050-4729","keywords":"feature extraction;humanoid robots;image motion analysis;image segmentation;image sequences;mobile robots;pattern clustering;video signal processing;Lucas Kanade pyramidal optical flow estimation algorithm;Shi-Thomasi algorithm;coherent cluster;coherent movement;feature detection;feature tracking;image plane;location estimation;monocular video sequence;whole body motion primitive segmentation;Clustering algorithms;Computer vision;Hardware;Humans;Image motion analysis;Image segmentation;Kinematics;Optical computing;Tracking;Video sequences","review":"Segmentation based on video. Uses optical flow. If optical flow is evaluated on the entire image, many of the pixels will contain invalid estimates, due to lack of texture or other unique features around a given pixel, occlusions between the two frames, depth discontinuities or reflection highlights. To reduce computational burden and the number of invalid optical flow values, optical flow is not computed for the entire image. Instead, the image is first searched for good features to track, using the Shi and Tomasi algorithm. This algorithm formally defines the feature�s quality based on how well it can be tracked. Segment is declared when a cluster (in an image) starts/stops moves or changes directions. When a portion of the body is moving, optical flow occurs, which disappears at start/end of motion. Once reliable features are found in the image, the optical flow for those features is computed via the pyramidal version of the Lucas Kanade algorithm [2], [27]. The Lucas Kanade algorithm solves the optical flow equation (Equation 2) iteratively, using Newton-Raphson","timestamp":"2013.10.02","bibtex":"@InProceedings{Kulic2009d,\n Title = {Whole body motion primitive segmentation from monocular video},\n Author = {Kuli\\'{c}, D. and Dongheui Lee and Nakamura, Y.},\n Booktitle = {Proceedings of the IEEE International Conference on Robotics and Automation},\n Year = {2009},\n Pages = {3166-3172},\n\n Abstract = {This paper proposes a novel approach for motion primitive segmentation from continuous full body human motion captured on monocular video. The proposed approach does not require a kinematic model of the person, nor any markers on the body. Instead, optical flow computed directly in the image plane is used to estimate the location of segment points. The approach is based on detecting tracking features in the image based on the Shi and Thomasi algorithm. The optical flow at each feature point is then estimated using the Lucas Kanade Pyramidal Optical Flow estimation algorithm. The feature points are clustered and tracked on-line to find regions of the image with coherent movement. The appearance and disappearance of these coherent clusters indicates the start and end points of motion primitive segments. The algorithm performance is validated on full body motion video sequences, and compared to a joint-angle, motion capture based approach. The results show that the segmentation performance is comparable to the motion capture based approach, while using much simpler hardware and at a lower computational effort.},\n Doi = {10.1109/ROBOT.2009.5152266},\n ISSN = {1050-4729},\n Keywords = {feature extraction;humanoid robots;image motion analysis;image segmentation;image sequences;mobile robots;pattern clustering;video signal processing;Lucas Kanade pyramidal optical flow estimation algorithm;Shi-Thomasi algorithm;coherent cluster;coherent movement;feature detection;feature tracking;image plane;location estimation;monocular video sequence;whole body motion primitive segmentation;Clustering algorithms;Computer vision;Hardware;Humans;Image motion analysis;Image segmentation;Kinematics;Optical computing;Tracking;Video sequences},\n Review = {Segmentation based on video. Uses optical flow.\n\nIf optical flow is evaluated on the entire image, many of the pixels will contain invalid estimates, due to lack of texture\nor other unique features around a given pixel, occlusions between the two frames, depth discontinuities or reflection\nhighlights. To reduce computational burden and the number of invalid optical flow values, optical flow is not computed\nfor the entire image. Instead, the image is first searched for good features to track, using the Shi and Tomasi algorithm. \nThis algorithm formally defines the feature�s quality based on how well it can be tracked.\n\nSegment is declared when a cluster (in an image) starts/stops moves or changes directions. When a portion of the body is moving, optical flow occurs, which disappears at start/end of motion. \n\nOnce reliable features are found in the image, the optical flow for those features is computed via the pyramidal version\nof the Lucas Kanade algorithm [2], [27]. The Lucas Kanade algorithm solves the optical flow equation (Equation 2)\niteratively, using Newton-Raphson},\n Timestamp = {2013.10.02}\n}\n\n","author_short":["Kulić, D.","Lee, D.","Nakamura, Y."],"key":"Kulic2009d","id":"Kulic2009d","bibbaseid":"kuli-lee-nakamura-wholebodymotionprimitivesegmentationfrommonocularvideo-2009","role":"author","urls":{},"keyword":["feature extraction;humanoid robots;image motion analysis;image segmentation;image sequences;mobile robots;pattern clustering;video signal processing;Lucas Kanade pyramidal optical flow estimation algorithm;Shi-Thomasi algorithm;coherent cluster;coherent movement;feature detection;feature tracking;image plane;location estimation;monocular video sequence;whole body motion primitive segmentation;Clustering algorithms;Computer vision;Hardware;Humans;Image motion analysis;Image segmentation;Kinematics;Optical computing;Tracking;Video sequences"],"downloads":0},"search_terms":["whole","body","motion","primitive","segmentation","monocular","video","kulić","lee","nakamura"],"keywords":["feature extraction;humanoid robots;image motion analysis;image segmentation;image sequences;mobile robots;pattern clustering;video signal processing;lucas kanade pyramidal optical flow estimation algorithm;shi-thomasi algorithm;coherent cluster;coherent movement;feature detection;feature tracking;image plane;location estimation;monocular video sequence;whole body motion primitive segmentation;clustering algorithms;computer vision;hardware;humans;image motion analysis;image segmentation;kinematics;optical computing;tracking;video sequences"],"authorIDs":[],"dataSources":["iCsmKnycRmHPxmhBd"]}