File Download
 
 
Supplementary

Postgraduate Thesis: Motion segmentation by adaptive mode seeking and clustering consensus
  • Basic View
  • Metadata View
  • XML View
TitleMotion segmentation by adaptive mode seeking and clustering consensus
 
AuthorsPan, Guodong.
潘国栋.
 
Issue Date2012
 
PublisherThe University of Hong Kong (Pokfulam, Hong Kong)
 
AbstractThe task of multi-body motion segmentation refers to segmenting feature trajectories in a sequence of images according to their 3D motion affinity without knowing the number of motions in advance. It is critical for understanding and reconstructing a dynamic scene. This problem essentially consists of two subproblems, segmenting features and detecting the number of motions. While the state-of-the-art LBF algorithm achieves segmentation accuracy as high as 96.5%, it is still disturbed by a phenomenon called over-locality. A novel mode seeking algorithm with an adaptive distance measure is proposed to avoid this problem, and improves the accuracy to 98.1%. The LBF algorithm is incapable of detecting the number of motions itself. A randomized version of the mode seeking algorithm is presented, which could detect the number as well as preserve satisfactory segmentation accuracy. To detect the number of motions, a kernel optimization method locates it via kernel alignment. However, it suffers from over-locality and over-detects the number of motions. An intersection measure and two mutual information measures are presented to solve this problem. Using these measures, the proposed clustering consensus framework recasts the motion number detection problem to a clustering consensus problem. It extends the kernel optimization method from two-clustering consensus to multiple-clustering consensus. A large number of experiments and comparisons have been done, and convincing results are obtained.
 
AdvisorsWong, KKY
 
DegreeDoctor of Philosophy
 
SubjectComputer vision.
Computer algorithms.
Cluster analysis.
 
Dept/ProgramComputer Science
 
DC FieldValue
dc.contributor.advisorWong, KKY
 
dc.contributor.authorPan, Guodong.
 
dc.contributor.author潘国栋.
 
dc.date.hkucongregation2012
 
dc.date.issued2012
 
dc.description.abstractThe task of multi-body motion segmentation refers to segmenting feature trajectories in a sequence of images according to their 3D motion affinity without knowing the number of motions in advance. It is critical for understanding and reconstructing a dynamic scene. This problem essentially consists of two subproblems, segmenting features and detecting the number of motions. While the state-of-the-art LBF algorithm achieves segmentation accuracy as high as 96.5%, it is still disturbed by a phenomenon called over-locality. A novel mode seeking algorithm with an adaptive distance measure is proposed to avoid this problem, and improves the accuracy to 98.1%. The LBF algorithm is incapable of detecting the number of motions itself. A randomized version of the mode seeking algorithm is presented, which could detect the number as well as preserve satisfactory segmentation accuracy. To detect the number of motions, a kernel optimization method locates it via kernel alignment. However, it suffers from over-locality and over-detects the number of motions. An intersection measure and two mutual information measures are presented to solve this problem. Using these measures, the proposed clustering consensus framework recasts the motion number detection problem to a clustering consensus problem. It extends the kernel optimization method from two-clustering consensus to multiple-clustering consensus. A large number of experiments and comparisons have been done, and convincing results are obtained.
 
dc.description.naturepublished_or_final_version
 
dc.description.thesisdisciplineComputer Science
 
dc.description.thesisleveldoctoral
 
dc.description.thesisnameDoctor of Philosophy
 
dc.identifier.hkulb4819936
 
dc.languageeng
 
dc.publisherThe University of Hong Kong (Pokfulam, Hong Kong)
 
dc.relation.ispartofHKU Theses Online (HKUTO)
 
dc.rightsThe author retains all proprietary rights, (such as patent rights) and the right to use in future works.
 
dc.rightsCreative Commons: Attribution 3.0 Hong Kong License
 
dc.source.urihttp://hub.hku.hk/bib/B48199369
 
dc.subject.lcshComputer vision.
 
dc.subject.lcshComputer algorithms.
 
dc.subject.lcshCluster analysis.
 
dc.titleMotion segmentation by adaptive mode seeking and clustering consensus
 
dc.typePG_Thesis
 
<?xml encoding="utf-8" version="1.0"?>
<item><contributor.advisor>Wong, KKY</contributor.advisor>
<contributor.author>Pan, Guodong.</contributor.author>
<contributor.author>&#28504;&#22269;&#26635;.</contributor.author>
<date.issued>2012</date.issued>
<description.abstract>&#65279;The task of multi-body motion segmentation refers to segmenting feature trajectories

in a sequence of images according to their 3D motion affinity without

knowing the number of motions in advance. It is critical for understanding and

reconstructing a dynamic scene. This problem essentially consists of two subproblems,

segmenting features and detecting the number of motions. While the

state-of-the-art LBF algorithm achieves segmentation accuracy as high as 96.5%,

it is still disturbed by a phenomenon called over-locality. A novel mode seeking

algorithm with an adaptive distance measure is proposed to avoid this problem,

and improves the accuracy to 98.1%. The LBF algorithm is incapable of detecting

the number of motions itself. A randomized version of the mode seeking algorithm

is presented, which could detect the number as well as preserve satisfactory

segmentation accuracy. To detect the number of motions, a kernel optimization

method locates it via kernel alignment. However, it suffers from over-locality and

over-detects the number of motions. An intersection measure and two mutual

information measures are presented to solve this problem. Using these measures,

the proposed clustering consensus framework recasts the motion number detection

problem to a clustering consensus problem. It extends the kernel optimization

method from two-clustering consensus to multiple-clustering consensus. A large

number of experiments and comparisons have been done, and convincing results

are obtained.</description.abstract>
<language>eng</language>
<publisher>The University of Hong Kong (Pokfulam, Hong Kong)</publisher>
<relation.ispartof>HKU Theses Online (HKUTO)</relation.ispartof>
<rights>The author retains all proprietary rights, (such as patent rights) and the right to use in future works.</rights>
<rights>Creative Commons: Attribution 3.0 Hong Kong License</rights>
<source.uri>http://hub.hku.hk/bib/B48199369</source.uri>
<subject.lcsh>Computer vision.</subject.lcsh>
<subject.lcsh>Computer algorithms.</subject.lcsh>
<subject.lcsh>Cluster analysis.</subject.lcsh>
<title>Motion segmentation by adaptive mode seeking and clustering consensus</title>
<type>PG_Thesis</type>
<identifier.hkul>b4819936</identifier.hkul>
<description.thesisname>Doctor of Philosophy</description.thesisname>
<description.thesislevel>doctoral</description.thesislevel>
<description.thesisdiscipline>Computer Science</description.thesisdiscipline>
<description.nature>published_or_final_version</description.nature>
<date.hkucongregation>2012</date.hkucongregation>
<bitstream.url>http://hub.hku.hk/bitstream/10722/167212/1/FullText.pdf</bitstream.url>
</item>