The Embedded Intelligent Visual Analysis Service Software In The Security Surveillance Platform

In recent years, with the development of surveillance, smart visual analysis hasbecome an important part of the security surveillance platform. The smart visual analysis isaimed to analyze the videos in the platform, and find out what happened. Thus it can assistthe platform to guarantee the environments are safe. Due to complex applicationrequirements, different application environments, vague real-time video, the intelligentvisual analysis is a difficult problem to be solved in security surveillance platform.In this paper, we use three steps to improve the previous frameworks: a) integrateenvironment status detection algorithms which don’t analysis moving targets; b) addmotion detection selection module and c) combine moving targets analysis module. Withthis improved framework, the visual analysis module can automatically adapt to differentmonitoring environments, and the environment statues detection algorithm can beintegrated in the framework. This framework includes five modules: motion detectionselection module, motion detection module, moving target analysis module, featureextraction module and alarm event detection module. The motion detection selectionmodule selects the motion detection algorithm based on monitoring environment. Then themotion detection module detects the motion area in the real-time video. After that themoving target analyses module finds out moving targets, track and identify them. At thesame time the feature extraction module extracts image features what alarm event detectionmodule needed. At last the alarm event detection module combines features and movingtargets analysis result to justify whether alarms happened and sends out the alarm message.We extend the proposed framework to an intelligent ATM surveillance module toverify the proposed framework. In order to implement this module, we propose the movingtargets analysis method, fire detection algorithm, the motion detection algorithm for ATMoperating panel and some simple event alarm algorithms. By combining the proposedalgorithms and existing motion detection algorithms, the module can detect four kinds ofevents. They are perimeter intrusion detection, abandon detection, operating panel abandondetection and fire detection. With the help of this module, it proves that the proposedframework has less computational complexity and good code reusability. And theenvironment state detection algorithms can be well integrated into the framework.In addition, the intelligent ATM surveillance module has been used in some RuralCommercial Bank in Guangdong Province. In three months’ experiment, the results showthat this module achieves real-time computing speed and high alarm accuracy.