The camera’s ability to capture video in the same ultra-high resolution makes it unique
Chinese scientists recently announced that they created a 500-megapixel cloud camera system. According to its makers, the camera can capture an extremely detailed image of a face in a stadium full of people. Then it can use AI and cloud computing to track down that face in an instant.
The camera shoots video, too, so it can capture very high-resolution video, according to reports.
The camera system was developed by Shanghai-based Fudan University and the Changchun Chinese Academy of Sciences. It was unveiled at the 21st China International Industry Fair, Chinese state media reported on Sunday.
"It's a contradiction for traditional cameras to see clearly and widely," said Zeng Xiaoyang, head of the research team behind the camera and a professor at Fudan University. "When 20 million pixels are allocated to 10,000 people, each person has only 20 pixels, so it is impossible to see the human body clearly."
Consumer cameras typically don't go past 50 megapixels and often only offer about 25 megapixels. One of the more advanced cameras available to consumers is a Hasselblad camera that shoots 400-megapixel photos. But the fight over having the highest megapixel count has largely ended in consumer cameras since buyers realized that more megapixels doesn't necessarily mean better photos.
But this new cloud camera system is different. Cameras that take photos this large do it by stitching together multiple images, an arduous process that takes time and processing power. Some cameras astronomers use to capture the depths of space can capture images as large as a few gigapixels -- that's a few billion pixels in one image.
The researchers behind this particular project say the 500-megapixel camera is special because it also captures video. This requires a lot of computational power.
"When pixels reach the level of 500 million, especially when video images are needed, the camera system will become very large and a real-time video mosaic is basically impossible to achieve," Zang said.
The team has managed to shoot video with the same ultra-high resolution, but this requires some large data transfers. As a result, wireless transmission produces a bottleneck, according to Zang.
"Our cloud camera has a resolution of 500 million pixels, a hundred times that of 4K," he added.
But the camera isn't just capturing the image, which requires a lot of power on its own. The researchers also need it to recognize faces. In order to reduce the amount of data needed to be transmitted and improve the transmission efficiency, the camera's ability to recognize and track down targets in real time needs to be improved, Zang said.
The team of researchers also said that the camera presents privacy risks and that this kind of technology will need to be standardized through a legal framework.
Copyright (c) 2019. South China Morning Post Publishers Ltd. All rights reserved.
