Improved video analytics and telematic data on Google Cloud

the problem

A parcel delivery client faced concerns regarding the size of the fleet and amount of video being captured and required a solution to automate the analysis of this video and highlight areas for concern as well as provide a mechanism to have summary videos available for driver training.​

Some of the other technical issues the client wanted to resolve were; how to capture and store a significant amount of video data, how to analyse hundreds of hours of videos per day and understand what was important, how to overlay vehicle telemetry data to provide insights into summary videos, and how to automate this process to make it repeatable and efficient?​

the approach

intelia developed a sample processed video in order to demonstrate the final solution and enable the client to be able to share the outcomes across the business. The consultants at intelia developed this video using the telemetry data provided by the client team. During the development of this video, the intelia team was also able to address some data quality issues that were found around the missing telemetry data provided by the business. To rectify this, some basic machine learning algorithms were applied to fill the data gap with realistic, generated data to enable the telemetry to be continuous.

the solution

The pilot video processing engine was demonstrated by intelia using Google Cloud processing technology and machine learning services. The processed video showcased several requested features including:​

  • Facial identification and blurring of recognisable faces.​
  • Telemetry overlays of speed/tilt/GPS location etc.​
  • Real-time map overlay of where the vehicle is positioned (shown on a Google maps inlay).​

the outcome

intelia worked with the client in order to deliver a scalable, secure and functional pilot Video Event Processing Engine on the Google Cloud Platform; a critical enabler of the client’s push to drive efficiency and promote safety across their fleet. The solution involved the following components:​

  • The cloud based video processing engine will automatically process video data captured by client vehicles, to create video clips of the appropriate length which correspond to a specific set of safety related or other telemetric events.
  • These clips will be created using Google Cloud processing technology and machine learning services to automatically obfuscate faces and vehicle number plates to ensure public privacy is respected.