Visual sensors take a cue from biological systems

Visual sensors based on biological systems are aiming to emulate the nonuniform distribution of photoreceptors found in the retinas of mammals. The first implementation of this approach was realized using CCD technology and produced a device of 2000 pixels arranged in concentric rings composed of 64 pixels each (see Vision Systems Design, May 1997, p. 8).

Visual sensors take a cue from biological systems

Visual sensors based on biological systems are aiming to emulate the nonuniform distribution of photoreceptors found in the retinas of mammals. The first implementation of this approach was realized using CCD technology and produced a device of 2000 pixels arranged in concentric rings composed of 64 pixels each (see Vision Systems Design, May 1997, p. 8).

Now, researchers at the University of Genoa (Genoa, Italy) and the Interuniversity Microelectronic Research Center (IMEC; Leuven, Belgium) have developed a CMOS chip of 8000 pixels. Each of the outermost 56 rings of the chip is covered by 128 pixels. In the 20 innermost rings, the number of photosites is halved progressively to a single photosite at the center of the device. Realized in 0.7 -µm CMOS technology, the smallest photosites are 14 µm wide, whereas the large photosites in the periphery are 250 µm wide. The overall size of the sensor is 8 mm square, each pixel is directly accessible, and the frame rate runs to 150 frames/s.

Following this implementation, a color version of the chip was built by Giulio Sandini and his colleagues at the University of Genoa. Fabrication was accomplished by depositing color filters on the photosites of the chip in a rotating pattern. Each pixel is covered with a certain color that has the two complementary colors as neighbors.

"At present, the University of Genoa is studying how to use the sensor for robot vision. The features of the sensor will allow robot control without the computational overhead required by traditional constant-resolution images but with comparable resolution and field of view," says Sandini.

Sandini an be reached via e-mail at sandini@dist.unige.it.

More in Robotics