Computer Algorithm can analyze surveillance footage and report for suspicious activity

SBC mounted camera can identify the suspicious activity from footage

  •  
  •  
  •  
  • 1
  •  
  •  
  •  
  •  
  •  
  •  
  •  
  •  
  •  
    1
    Share

Generally, surveillance cameras are not developed in a  way that can process suspicious activities from the footage. Identifying such kind of activities are completely manual and that takes a long time to process the footage and identify.

To solve this problem, Binghamton University, State University of New York published a new research paper names as “Kerman: A Hybrid Lightweight Tracking Algorithm to Enable Smart Surveillance as an Edge Service”, which won the best paper award at the 2019 IEEE CCNC conference.

Associate Professor of Electrical and Computer Engineering Yu Chen and his team at Binghamton University developed a hybrid lightweight tracking algorithm known as Kerman (Kernelized Kalman filter) that uses Single Board Computers (SBC) mounted camera with enhanced recognition of people and tracking of the objects that seem to be suspicious.

The algorithm is able to detect the person, track their movements and recognize their behavior while processing the videos.

The algorithm proposed for human object tracking, which is coupled with a lightweight Convolutional Neural Network (L-CNN) for high performance. The proposed Kerman algorithm has been implemented on a couple of single board computers (SBC) as edge devices and validated using real-world surveillance video streams. The experimental results are promising that the Kerman algorithm is able to track the object of interest with decent accuracy at a resource consumption affordable by edge devices.

“The Kerman algorithm enables the smart cameras at the edge (the source of data generation) to raise an alert as soon as something suspicious is detected in the incoming video streams,” said Chen.

The research team introduced SBCs to be implemented in decentralized computing platforms, which distributes the workload among multiple Fog computing nodes, instead of to one centralized server. Because of decentralized computing, the video does not need to be transferred to one remote server, making the surveillance system more agile and robust. Data processing can then be processed and analyzed in a more effective and timely manner.

The algorithm does not identify, track or record the activities of anyone, thereby maintaining a high level of privacy within a secure system. Future models of this algorithm will take advantage of more advanced hardware and security mechanisms to ensure that this surveillance system is evolutionary and maintains high performance in the lifespan.