A vision based on surveillance system using a wireless sensor network
Niklas Lundell · KTH Publication Database DiVA (KTH Royal Institute of Technology) · 2010
The aim of this thesis project is to implement a vision based surveillance system in a wireless sensor network running the Contiki operating system. After a comparative review of experimental cameras the decision fell on the Cmucam3. Hardware and software to integrate the camera with a Tmote Sky mote were created. In this way, the camera can be controlled with the Contiki operation system and images can be transferred to the sensors memory for further processing and forwarding to other network nodes. Due to the lack of memory on the sensor nodes, the prospect of image compression was investigated. To save energy, the system uses an energy efficient onboard light sensor to detect ambivalent changes that would occur in the presence of, for instance, intruders or fire. When these sensors detect changes, the more energy consuming camera gets activated. The detection mechanism uses a Cusum algorithm which filters out irrelevant disturbances while maintaining high sensitivity to unexpected changes. An evaluation of the system using multiple motes running the energy efficient detection algorithm demonstrates rapid detection and transmission times of about 2 seconds.