Object and Event Recognition in Wireless Multimedia Sensor Networks

Object and Event Recognition in Wireless Multimedia Sensor NetworksClint MuellerCS441

Wireless Multimedia Sensor Networks (WMSN)

•Network of wirelessly interconnected sensor nodes equipped with multimedia devices (cameras and microphones) capable to retrieve video and audio streams, still images, as well as scalar sensor data.

Uses

•Military•Law-Enforcement reports•Traffic Control•Advanced health care delivery•Industrial Process Control•Automated assistance to elderly

telemedicine•Autonomous processing vehicles (cars,

robots, etc.)

Challenges

•High bandwidth•Real time delivery•Jitter and frame loss rate•Object Recognition Techniques

▫Reduce information sent to sink node•Nodes besides scalar nodes

▫Multimedia sensors▫Processing hubs▫Storage hubs

Network Architecture

•Divided into three reference models▫Single-tier flat architecture▫Single-tier clustered▫Multi-tier architecture

Single-tier flat architecture

•Homogeneous sensor nodes•All have same capabilities and

functionalities•Multimedia processing is distributed

among all the nodes, and prolongs network life time

Single-tier clustered architecture•Heterogeneous sensors•Each cluster relay data to cluster head•Cluster head performs the data

processing•Cluster head is connected to sink node or

gateway

Multi-tier architecture

•Heterogeneous sensors•First tier deployed with scalar sensors

perform motion detection•Second tier perform object detection or

object recognition•Third tier perform object tracking,

connected to sink node or gateway•Each tier may have central hub to

perform data processing and communicate with higher tier

Physical Layer

•Ultra-Wide band (UWB)▫Data rate up to 250 Mbps and range of 10

meters▫Immunity to multipath propagation and

precise positioning▫Enable low power consumption▫High data-rate or short range

communication

MAC Layer

•Multi-channel MAC protocols which can effectively utilize the available channel capacity

•QoS based protocols•COM-MAC

▫On-demand multi-channel contention-free▫Better support for high data rates

MAC Protocals

Grey rows indicate that the MAC protocol is designed for WSNs but not specifically for WMSNs.

Routing Layer•Multipath routing because network needs to

use all of its bandwidth•Ant-based Service-aware (ASAR)

▫Addresses the routing scheme between the cluster head and sink node in which a cluster head transfers the different classes of data

•Two-Phase Greedy Forwarding (TFGF)▫Geographic▫Supports multipath transmission by repeatedly

executing the algorithm to find more on-demand node-disjoint routing paths

Routing Protocols

Transport Layer

•Features of WMSN▫Traffic: continues, event driven, query

driven, hybrid▫QoS: reliability, real-time delivery, fairness▫High redundancy: reliability against packet

loss•Queue based Congestion Control Protocol

with Priority Support (QCCP-PS)▫Deal with congestion▫Based on hop-by-hop approach

Transport Protocols

Application Layer

•Multimedia processing and source coding techniques

•Effective communication with other application programs

•Traffic management and admission control

•Applications:▫Predictive Video Coding (PVC)▫Multiple Descritption Coding (MDC)

Coverage

•Multimedia sensors do not have omni-directional coverage

•Snapshot▫Automated calibration protocol▫Calabrates the location, orientation, and

range of camera sensor•Knowing overlapping areas between

cameras allows exploiting the redundancy and can be used to track moving objects in the environment

Object Recognition

•Edge Detection▫Edges: boundary between two dissimilar

regions in an image•Address Event Representation (AER)

▫Extracts and outputs only a few features of interest from the scene

▫Only pixels which realize a difference in light intensity generate the events

Object Recognition

Original, Edge Detection, and AER

Conclusion

•Do not stream the video but focus on tracking different events

•WSN protocols cannot be used in WMSN since WMSN has more constraints

•Object recognition is import in order to reduce the amount of data sent

•Coding techniques are used to reduce the redundancies in frames

References

• Image Recognition Traffic Patterns for Wireless Multimedia Sensor Networks – Ruekn Zilan, Jose M. Barcelo-Ordinas and Bulent Tavli; web http://www.academia.edu/4186077/Image_Recognition_Traffic_Patterns_for_Wireless_Multimedia_Sensor_Networks

• Wireless Multimedia Sensor Networks: Current Trends and Future Directions – Islam Almalkawi, Manel Zapata, Jamal Al-Karaki and Julian Morillo-Pozo; web http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3231118/