AI Smart Network Camera System

We were approached by a new-start up who had developed advanced software for capture and interpretation of live video images. The software utilised unique AI algorithms to differentiate and interpret light and shade, movement characteristics and human, animal and vehicle profiles for interrogation and reporting purposes.

The entrepreneurs wished to explore the possibility of implementing the software in devices which would tackle unique problems in challenging environments, particularly where there were limited networking capabilities, and where systems would be required to work in harsh conditions, semi-autonomously for long periods of time.

Utilising our significant experience within the development of rail diagnostics equipment, we developed two variant multi-camera housings. The respective units were intended to fulfil 2 separate niches with the wildlife management and asset infrastructure management markets. The products were developed with IP rating, durability and anti-vandalism as primary considerations. Ease of manufacture and enclosure reliability in challenging environments was also a primary requirement. Consideration was also given to multiple deployment configurations, using custom bracketry, tethers, bolts and other mounting methods to ensure maximum mounting flexibility in a variety of intended locations.

A particular challenge was the sourcing of optical-glass products suitable for the proposed light conditions, camera lens/sensor types and cost requirements. We applied our learnings from the development of comparable rail-analysis systems and sourced suitable advanced optical products from a trusted supply partner.

One of the variant camera systems was intended to be deployed in remote locations to monitor and report on animal management and poaching problems. We recognised that the systems would require suitable ‘camouflage’ or disguise, so as to not disturb wildlife, or to be obviously evident to poachers or thieves. We explored a number of options including printed Cordura overcovers. Our final solution utilised a mix of tether points for recovered foliage, and a dip transfer printed disruptive pattern, which could be tailored to the proposed deployment environment.

Multiple prototype units were engineered and delivered to tight deadlines. The units were 5-axis machined from stock material with fully functional seals, clamping faces, lenses, mounting configurations and heat management. We also supplied operational internal camera mounts, cable management and other required paraphernalia.

Live field trailing was subsequently conducted on both the ‘Wildlife’ and ‘Asset Management’ products, with view to debugging and optimising for full commercialisation.