Computer Vision Algorithms in Surveillance Cameras
Science may never tell the humanity that it is waiting around the next corner, but the researchers came up with the technology closest to this expression: a computer program that turns an ordinary digital camera into a periscope.
At the demonstration of the “computational periscope”, an American team from Boston University showed that it could see the details of objects hidden from eyes, analyzing the shadows that they cast on the nearest wall.
Vivek Goyal, an electrical engineer from this university, said he hopes that such an innovation will lead to the creation of robots capable of better orientation, and will be able to increase the safety of cars without a driver.
He assured that he himself was not excited particularly by observation and did not want to frighten those around him, but now he is already able to see that there is a child in a parked car on the other side of the road, or to see some things around the corner of the intersection. He is confident that such a study can have a significant impact on safety. The only problem is that the ability to see rounded corners has taken at least a decade for modern researchers. Moreover, although scientists have achieved significant success in this area, the equipment used so far is highly specialized and expensive.
In his latest study, Goyal and his team used a standard digital camera and a middle class laptop. He noted that they did not use complex equipment. According to him, everyone can carry this regular camera in the pockets.
Computer vision algorithms process
Researchers from the journal “Nature” described the process of the technology. The scientist directs the digital camera on the vague shadows, which the object casts on the nearest wall. If the wall were a mirror, the task would be simple, but a matte wall scatters light in all directions, so the reflected image is blurred. Further, the task of computing is to turn a matte wall into a mirror.
They found that when the camera could not capture an object around the corner, their algorithms could use a combination of light and shadow at different points on the wall to restore what was inaccessible to view. In the tests, the program put together hidden images of the characters of a video game including details such as the eyes and mouth, along with colored stripes and the letters “BU”.
Given the relative simplicity of the program and equipment, Goyal believes that people can learn the same trick. In a draft blog written for Nature, he said that people could even learn to look around the corner with their own eyes; it does not need any superhuman abilities. In addition, the program currently takes about 48 seconds to process a hidden scene from a digital image, but Goyal believes that this can be done much faster with more computer power.
In an accompanying article, Martin Laurenzis, an imaging expert at the French-German Research Institute of Saint-Louis, explains that with a few images, this technique can track the movement of hidden targets. This technology can also be much more widely used, he said, in microscopy, medical imaging and monitoring in hazardous environments such as chemical plants and nuclear power plants.