Although The Combination Of Machine Learning And Imaging Technologies Has Greatly Helped Automate Many Processes Today, There Are Still Many Applications That Can Be Considered For These Technologies. 

The design of cameras and imaging tools equipped with artificial intelligence helps us in everyday life and insensitive moments. The data from surveillance and satellite cameras can train intelligent systems, or a combination of machine learning and hardware can build cameras as small as a grain of salt.

In addition, with the development of such intelligent systems, we will be able to place the burden of many of our complicated and torturous tasks on machines. In “Camera World” ‌ of this issue, we try to take a look at these cases.

Camera for future phones

Microscopes are ideal tools for use in medical or robotic applications, but microscopes. The ones that have been made so far do not capture high-quality images and have a limited field of view. Researchers at Princeton University and Washington University have developed a tiny camera that performs better than similar cameras.

According to the researchers, the new camera can produce full-color images whose quality is equivalent to that of a conventional camera with a lens that is 500,000 times larger.

This camera, which uses computational photography and a unique hardware design, will make it possible to make very delicate endoscopic instruments. An array of such cameras will turn any surface into a camera, which researchers call a “surface as a sensor.”

While conventional cameras focus light with curved glass or plastic lenses, the new optical system uses metasurface technology, producing like computer chips.

The surface is half a millimeter wide and is covered with one million and six hundred thousand excellent light antennas, each of which has a unique geometry. It is to be hoped that such technology in the future will make smartphones without the need to install multiple cameras.

The diffeAn important innovation in this camera is integrating optical surface and image processing algorithms. Hence, the geometry of each antenna causes the overall light wave to form correctly.

With the help of machine learning algorithms, antenna interactions are combined with light to produce the best possible color images and maximum field of view. An imposing improves the camera’s performance in natural light conditions compared to similar cameras that use supersonic.

Similar to previous cameras with laser light and in laboratory conditions were able to produce high-quality images.

The researchers used simulations to test different configurations of optical antennae. Such a simulation required a large amount of memory and time due to many antennas and the complexity of their interaction with light.

Figure 1:

Above – An example of a new micro camera. Compared to similar cameras, this camera offers more expansive image quality and field of view, and it is possible to mass-produce it at a reasonable price.

Bottom-left shows an example image from previous similar cameras. On the right, you see the output of a new camera that uses a technology called neural optics.