Video Analytics At Zircon

Video Analytics, an element of Computer Vision, refers to the use of software to automatically analyse video footage, either live or recorded, in order to recognise or detect a specific pattern, object or behaviour.

With ongoing research and development into the application of Machine Learning and other Artificial Intelligence techniques, the potential applications of Video Analytics only continue to grow.

Here at Zircon, we have been fortunate to be involved in numerous projects examining the application of Video Analytics to reliably deliver solutions to a range of problems. As a result, we have built up niche expertise in this field.

Areas of Application

Public Safety

Read More

Object Detection

Read More

Asset Detection and Recognition

Read More

Protection of Assets

Read More

Public Safety

Preserving or increasing public safety will always be a point of interest for a great many industries, and it is for this reason that there has been such high levels of interest in the potential benefits of applying Video Analytics.

As a result of our participation in research projects focused around improving passenger safety at the Platform Train Interface, or PTI, we have found that by utilising techniques such as measuring the statistical analysis of changes to an area, utilising combinations of background extraction and virtual triplines or optical flow, it is possible to:

  • Detect incidents where passengers have become trapped in doors or between train and platform, and are subsequently dragged along the platform
  • Detect incidents where passengers have encroached into the area of the track
  • Detect incidents where passengers are behaving in a manner that may put either themselves or other members of the public in danger

Whilst this research has been focused onto the rail sector, there is no reason as to why its applications should be limited there.

Demonstration of Drag Detection in Action During an Incident in Rome

Learn More About PTI Cam

Object Detection

Alongside Facial Recognition, Object Detection is possibly one of the most widely recognised uses of Computer Vision and Video Analytics.

Whilst the main uses of Object Detection tend to fall within the boundary of the security and defence sectors, in reality, there is much more potential to this technology. For example we have been a part of research and development projects investigating potential applications of this technology for the rail and highways sectors.

As part of these investigations, we have been working with Machine Learning techniques such as Google TensorFlow and Intel OpenVINO, in order to:

  • Detect and raise awareness of unauthorised persons within the boundary of railway tracks
  • Detect and track the movement of predefined objects through several frames of video footage
  • Use cameras to detect people breaking the law whilst driving.

Asset Detection and Recognition

In instances where assets are dispersed across a large area, such as with the road and rail networks, it is essential to maintain up to date details in regards to their locations.

As bizarre as it may sound, it is not uncommon for the recorded location of assets on plans to be inaccurate. For example, environmental factors may have prevented the asset from being placed in the location detailed in the plan.

Video Analytic techniques, similar to those that can be utilised for Object Detection using libraries, such as TensorFlow or OpenCV, can make it possible to utilise video feeds to both detect and recognise predefined asset structures. Should you combine this with location technologies, such as Differential GPS and visual odometry, it becomes feasible to gather accurate location data for each identified asset.

Of course, there is more potential to this type of technology than simply gaining data on asset locations. For example, by knowing the locations where an asset is expected to be, and by continually looking for that asset, if it stops being detected from the video feed the environment may have changed so that the view of that asset is obstructed and may need clearing.

Protection of Assets

As a direct result of factors, such as weather, human interference and general wear and tear, larger assets such as infrastructure will gradually degrade over time.

Should this degradation remain unchecked for too long, the scale and repair cost of the resulting damage can be substantial. Whilst regular manual checks and inspections help to reduce this, they too carry a significant cost and introduce potential safety risks. Having the ability to utilise Video Analytics for constant assessment could help to minimise this cost and safety aspect, especially in cases where assets are widely dispersed or hard to access.

Through the application of techniques such as edge detection, blob detection and Object Detection to identify the pantograph and overhead line, we have been able to utilise Video Analytics in order to:

  • Track the ‘zigzag’ movement of overhead lines to detect for anomalous events caused by slack in the lines and grooves in the pantograph
  • Identify instances of arcing between overhead lines and the pantograph
  • Inspect the integrity of rolling stock and highlight damage to rooftop equipment and areas of wear and tear.
  • Identify the appearance of graffiti on railway assets

Looking For More?

Our experience of Computer Vision and working with Video Analytic techniques is continually expanding. If you would like to hear more about our work in this area, or feel that we may be able to provide you with advice or assistance, we would love to hear from you.