September 19, 2016

Making Unmanned Surveillance Systems Truly Reliable

by Puneet Gupta (Senior Manager, System Solutions)

Video surveillance systems often find applications in areas where manual monitoring of the surveillance feed is either not feasible or is prohibitively costly (ex: toll plazas, hazardous material leaks, underground rigs, parking lots or train stations). In such cases, automation techniques are used to substitute human intervention:

  1. Alerts or events based on captured video, sensors or physical connectors that can detect unexpected or undesirable occurrences (ex: increased methane level, unusually high traffic, unexpected motion, etc.)
  2. Automated actions based on the detected alerts and events (ex: evacuation alarms, calls to police helplines, locking of gates, etc.)

Reliability of event detection determines the real success of this mechanism. Machine vision algorithms like Vehicle or People Counting, Automatic License Plate Recognition (ALPR), Object Tracking, Motion Detection etc. are typically be used to generate video-based events.

  • Confidence on reported events increases with on-field training & manual overrides during the training period.
  • Accuracy of detection increases with pre-processing of video to improve video quality (de-noising, contrast enhancement, de-hazing, image stabilization, etc.)

Another key requirement for such an unmanned installation is the hardware robustness. Surveillance scanners, IP cameras and Digital Video Recorders (DVRs) often operate in harsh environmental conditions (extreme temperatures, heavy rains, snowfall, chemical fumes, etc.). It is imperative therefore that the embedded processor, peripherals, connectors & boards function stably under such environmental conditions. Despite having the best video processing & machine vision algorithms, the surveillance system cannot guarantee reliable performance if the hardware fails under stress.

Ittiam’s neonCaster hardware (Clove, Magna) and pre-configured systems can be good candidates for realizing such unmanned surveillance systems.

  • Hardware based on Sitara AM57x (AM5728/AM5718) or DM816x (DM8168/DM8169) processors, with support for single/multi-channel Full-HD (1080p60) video.
    • Industrial grade hardware optimized for size, weight and power
    • Proven to work continuously for months without needing to be restarted
  • Video Streaming & Recording Media SDKs (adroitRecord+) to do live capture, encode and streaming & recording of video
    • Encodes audio and video in high quality using standard formats like AAC and H.264
    • Enables low latency video streaming over wired/wireless networks
    • Decodable on any Media Player App on a mobile/desktop device.
  • Analytics, Vision & Graphics Media SDKs (adroitVista)
    • Pre-processing algorithms such as Contrast Enhancement, De-noising, De-interlacing, etc.
    • Vision algorithms such as Motion Detection, Object Tracking, Camera Tampering Detection, etc.
  • Field-friendly system management
    • Fail-safe field firmware upgrades
    • Built-in self tests, extensive diagnostics for remote troubleshooting

Do check out our website for more details!

For more information, reach out to us @