A new technique for automatically enhancing LIDAR (light detection and ranging) imagery to the highest detail yet is predicted to revolutionise and improve the capture of land data across the UK.
Developed by Lumiere Property, the technology is enabling LIDAR images captured from planes to be enhanced to 4 times their original resolution. Using the new algorithm, 1m resolution images can now produce more detailed 25cm equivalents.
According to the Environment Agency, around 40% of England is not surveyed using LIDAR or has only been surveyed at the lower resolution of 2m. For the most advanced land research and modelling, data needs to be current and at 1m resolution or higher. Their goal is to achieve full UK coverage by the middle of 2020.
“Currently, only small areas of the country have 25cm resolution coverage. With our new approach, which we’ve built using deep neural networks, we can generate a 25cm resolution LIDAR image for anywhere in the country that has 1m coverage giving our algorithms a detailed view of terrain and buildings over a much wider scale,” explained Geoff Smith, Software Engineer at Lumiere Property.
The Environment Agency’s primary use of LIDAR laser sensing techniques is for flood risk mapping but Lumiere’s latest breakthrough in super-resolution will make it relevant across a huge range of applications including identifying the attributes of a building.
“This higher resolution and 3D mapping gives us a more precise understanding of existing buildings and their surroundings and will allow us to quickly and cheaply calculate the development potential of huge numbers of properties to search out the best opportunities,” said Chris Rowland-Smith, Managing Director, Lumiere Property.
“For example, we may want to pinpoint any big buildings with a specific attribute such as what part of the building has a flat roof and what part has a pitched roof. Or smaller attributes like existing roof windows or different types of roof tile – the value this brings to our business is potentially huge,” he added.
