Object Recognition in Live Video

Abstract

Video surveillance is nowadays necessary to observe places, people and things to pull out more valuable and useful information from video data and it is one of the most active research topics in computer vision. Video surveillance is having wide range of applications like access control in special areas, theft identification from distance, traffic analysis and traffic managing, abnormal behaviour detection and to provide security at high risk areas, etc. The traditional system is based on human efforts. Human operator is needed to observe all the CCTV camera screens. This method is time consuming, less efficient and error prone.

The proposed system works for accident detection from abnormal behaviour of vehicles on highways and notifying for the same. It works with the objective of producing the complete automatic intelligent system to overcome the delay proposed by the human efforts. Moving object is detected using background subtraction. These objects are classified to separate out the desired objects which are tracked to analyse the behaviour and generating the required events. The system detects accident using the vehicles stopped motion, which can be due to accident or vehicle stopped at roadside. Now from this two situations accident situation can be detected by using the classifier.