Technology

Foresight’s powerful and mature detection and tracking technology is derived from field-proven homeland security technology that has been deployed worldwide for almost two decades in critical international facilities including borders, nuclear plants, and airports.

Foresight’s new vision for autonomous vehicles is achieved by fusing stereo visible-light cameras with thermal long-wave infrared sensors that provide depth perception and obtain a clear 3D view. The IP-protected technology is based on 3D video analysis, advanced algorithms for image processing and sensor fusion, achieving an unprecedented accurate obstacle detection in harsh lighting and weather conditions including complete darkness, rain, haze, fog and glare. The technology generates a 3D point cloud of data that enables analysis of different terrains and 3D sensor fusion.

Foresight’s technology scans the road ahead to foresee possible collisions with vehicles, pedestrians, animals and other obstacles. The technology can be customized to address the challenges of ADAS and autonomous vehicle markets, including object detection under diverse road, lighting and weather conditions.

technology building blocks:

Thermal Long-wave infrared

Thermal long-wave infrared (LWIR)  cameras rely on heat emission of different objects, as opposed to other existing sensors that rely on reflective light. LWIR enables detection in harsh lighting and weather conditions. In addition, due to its thermal signature capabilities, the use of LWIR allows a native detection of pedestrians and animals.

Stereoscopic Technology

Stereoscopic vision technology uses two synchronized cameras to generate a depth map. Depth map allows to detect any object and to get its accurate size, location and distance.

Monocular vision object detection technology relies on deep neural networks (DNN) object recognition. Using DNN for object recognition will always encounter corner cases where there is an unknown object to the trained network.

Foresight’s solution combines the advantages of stereoscopic technology with DNN object recognition to provide accurate obstacle detection and classification.

Automatic Calibration

The main issue in stereo vision is to achieve a precise depth map. Foresight has come up with a unique solution using a dedicated algorithm to successfully asses relative pose estimation, known as automatic calibration. Automatic calibration is fundamental for creating an accurate stereoscopic 3D perception.

Multispectral Fusion

Foresight’s technology combines stereo visible-light cameras and stereo long-wave infrared thermal cameras to provide accurate obstacle detection in harsh lighting and weather conditions. The fusion between the two stereoscopic channels addresses corner-case scenarios while reducing false alerts.

3D Point cloud

Foresight’s technology generates an accurate depth map which is converted to a high-resolution 3D point cloud that provides rich and accurate per pixel information. Point cloud provides 3D raw data that can be used for obstacle detection, terrain analysis and autonomous vehicle sensor fusion. The use of stereoscopic technology offers a non-emitting and cost-effective solution for generating a high resolution point cloud.