Artificial intelligence helps CCTV turn night into day

POWERFUL artificial intelligence software can transform CCTV footage taken at night under street lighting into colour ‘daytime’ images.

The AI technology – developed by a Berlin-based start-up – is being hailed as breakthrough which will transform security in cities.

Unlike standard image-enhancing software, the Brighter AI technique uses cloud-based ‘deep learning’ to generate a realistic daylight image from an infra-red night photograph.

A combination of small infra-red cameras – which can be incorporated into street lights – and the AI algorithms is currently being trailed at locations in Germany. If successful, the developers expect the technology to be widely adopted in cities and in security applications. It can even be embedded in the rear-view mirrors of cars, so that the nighttime ‘reflection’ can viewed as if it were daytime.

‘At the moment, our focus is on successfully concluding the various pilot projects’, Brighter AI chief Marian Gläser told Lux. ‘Over time, the technology can also be applied in other fields such as the automotive sector.’ Other application areas result for aviation and aerospace systems as well as for image-based diagnosis methods in medicine.

The basic technology was developed by an interdisciplinary team of more than 15 engineers at the company, a spinoff from automotive lighting manufacturer Hella.

‘We’re particularly proud of our first spinoff Brighter AI. It speaks highly of the innovative climate here on site but it is also exemplary of the opportunities and freedoms that Hella provides us with’, said Timon Rupp, head of the Hella Incubator. ‘Thanks to newly independent companies such as Brighter AI, promising technologies and business models can quickly establish themselves on the market. We at Hella continue to be available as close strategic partner to these young entrepreneurs.’

 

 

Comments 2

Yes, that is very interesting! I love the idea but I am not sure how it exactly works. @Alan, I did some digging on them e.g. NVIDA press release and it looks like they developed an 'expert' system on day images with deep learning. It seems they give actually additional details that can not be captured with any camera, based on the content of the image.

Currently, professional digital cameras can produce wonderful "daylight" images in sub-moonlight levels of illuminance. I would be interested to know what the "deep learning" adds to the image. If the data (pixels) are there, standard image enhancing software, like in a camera, is available. If the data is not there, the AI is adding what may, or may not, be accurate.

Leave your comment