Real Time Motion and Range Estimation of Dynamic Scene Using Monocular Video

Accurate passive range estimation from single electro-optical/infrared sensors is a valuable enabling technology for many applications, particularly for defense purposes. Two applications of great interest are Sense and Avoid (SAA) for Unmanned Aerial Systems (UAS) and missile warning systems (MWS) for manned aircraft. Current passive image ranging systems rely either on stereo vision (or “stereopsis”) which provides a depth map (2.5D) of the scene in front of the device, or on maneuvering the sensor-carrying platform in a controlled manner. For objects at large distances, bearing angle has been successfully used for range estimation. However, it is not useful in many scenarios, especially when the intruder is heading towards the sensor in a fronto-parallel direction. In such circumstances, the sensor vehicle is required to make a deliberate maneuver which can be disruptive to the mission and cause inconvenience to both operators and regulators. Based on a comprehensive study of the problem, UtopiaCompression concluded that the object’s size in the image can provide additional information regarding its location, even if the object size is barely resolved. This work describes a reliable, robust and real-time range estimation technique in an Interacting-Multiple-Model – Particle-Filter framework that can combine the bearing angle and intruder size information.