Abstract
Concentration from industry and university
circles has greater than before noticeably over recent years
in the demanding area of event analysis and appreciation
from a variety of video sources counting sports,
observation, user-generated video, etc. Video event
investigation and recognition is a significant task in several
relevance such as detection of sporting places of interest,
incident detection in observation video, indexing in this
paper, we proposed technique for semantic video event
annotation that exploit global feature, local feature and
motion characteristic. with these description, video clip
can be determined as a situate of feature vectors. Then
according to dissimilar features, we train hybrid approach
based on SVM classifiers, and a bi-coded chromosome
based genetic algorithm is carry out to find optimal
classifiers and applicable most favourable weights base on
training stage. With the most favourable classifiers set and
best possible weights, the max