Our solution utilizes a decade of engineering novelties we built using our custom signal and image processing algorithms. We have developed these algorithms for binocular stereo, photometric stereo, geo-registration, high speed unknown-baseline image matching (correspondence), shape-from-shading, structure-from-motion, target detection and identification, optical system inversion and reflectivity function estimation. We can combine all of these competencies into an affordable stereo optics package installed on every cab, limo and major business vehicle driving in the congested pedestrian/cyclist streets in any major city. autonomousvision.com
Our visual technology combines Global Positioning System data from the car and individuals walking or cycling to gain a general idea of proximity detection and alert requirements. Upon receiving a visual detection of a collision from the car optical sensors with a pedestrian holding an active Safe Street mobile alert app device. The car instantly recognizes the collision up to 150 feet away and transmits an alert warning using a Bluetooth Low Energy protocol message "iBeacon" to alert the cyclist or pedestrian of a potential dangerous vehicle impact.
Our goal is to minimize false alerts with the use of our high tech visual optical integration package and photo processing algorithms. We can detect moving pedestrians and cyclist from other cars or non-moving objects at intersections which allows for a high probability of positive collision alerts to be sent to both drivers and pedestrians/cyclist.
The science behind the innovation: The core technology is an algorithm that accepts a stereo image pair and calculates semi-dense stereo disparity suitable for fast, real-time range-measurement, obstacle detection and related applications. Disparity can be directly translated into depth with a simple equation, depending on the lens calibration. I will include an equation at the bottom of this email. This disparity algorithm does not require epi-polar constraints like most stereo algorithms, adding significant capability not available from other technologies. This algorithm can operate with variable baseline, variable focal length, variable focus, independent pan/tilt/zoom lenses, unknown baseline, multiple baseline and partially or non-calibrated configurations. Most stereo algorithms, LIDAR and structured light depth sensors either only operate with fixed baseline or require precise calibration and rigid mounting of the cameras and lenses and must avoid slight mechanical movement of the lenses.