A Survey on use of Ontology for Complex Human Activity Recognition in Video Sequences

Abstract

Video Surveillance systems are ubiquitous in today’s world. Traditional Video Surveillance
Systems are being replaced by intelligent systems and this process has been happening rapidly
since decades. This work lists the Activity Recognition approaches to recognize complex human
activities. Knowledge based approaches with video sensors are brought out to be the vital
approach in comparison with the data-based approaches, because data driven models have to
train on huge data. We investigate the significance of both the approaches in handling complex
activities with a long sequence of primitive activities. Knowledge driven approaches with
Ontology models the domain expertise to support the long sequenced activity recognition.
Hence, the approaches supported by the Ontology for handling complex activities are found to
be the most suited model for the recognition of complex human activities