SBIR-STTR Award

The AFOST is the centerpiece for a potentially revolutionary technical breakthrough required in the successful development of variable acuity and computationally efficienct sensors. Our strategy is to address the limitations that exist in current Scene Large format sensors, namely reducing the high data throughput associated with large format arrays allowing detection and transmittal of the small fraction of the scene that has useful data. In order to perform useful tasks with this technology, Cyan has developed on FPA processing circuitry concepts that detect scene activity and can allow Automatic Selection of Foveal region anywhere within the image without an external processor cue. This on FPA scene activity architecture can detect targets with lower noise levels than FPAs that perform target detection using an external processor.

Keywords:
Auto Fovea, Optical Flow, Clutter Rejection, Superpixel, Multiresolution

Current large format sensors are limited in reaching their full potential on compact platforms due to the need to have large image processors and support electronics to process the data coming off the FPA at very high data rates. Seeking to match the brain's computational efficiency, we draw inspiration from its neural circuits to improve the bandwidth efficiency and acuity of large format imaging sensors. All biological systems require sensory information to detect and guide themselves. Cyan is building the AFOST to be able to extract and process the needed sensory information to alert the sensor user to a potential threat and zoom in on the region of interest in a way that is very computationally efficient and reduces the bottlenecks common to very large format imaging sensors.

Keywords:
Auto Foveation Sensor, Biological Inspired Processing, Egomotion, Roic