Thermal 3D Perception for Autonomous Navigation

We propose a novel method for the 3D thermal imaging contrast improvement over an order of magnitude that will revolutionize thermal imaging and enable long-range 3D perception with uncooled Long Wave Infrared (LWIR) sensors. Our solution is completely passive, and thus, it is suitable for stealth and covert operations. The alternative technical approaches to 3D perception in the dark are active: LiDar, SWIR Structured Light, and Time-of-Flight (ToF) ranging. Conventional binocular stereo fails long-distance ranging in low-contrast environments. Our multi-sensor system enables robust long-range 3D perception making bimodal devices such as ENVG unnecessary. Our technology mitigates the limitations of the uncooled thermal sensors: low sensitivity, low resolution, and high thermal inertia. The novel sensor configuration, calibration methods, and image processing algorithms improve the microbolometer-based 3D perception exceeding that of the traditional stereo build with the most advanced sensors, even those limited only by the photon detectors' shot noise. Additional interscene accumulation of the intrascene correlations will result in the system's Noise Equivalent Temperature Difference (NETD) of 1 mK and better. The proposed technology originates from our ongoing research of long-range 3D perception with RGB images. Our LWIR 3D research funded by a U.S. Air Force Phase I SBIR resulted in 0.077pix disparity resolution, exceeding the state-of-the-art by 6.5 times. Both 3D sensing modalities implement our patented image processing method of Differential Rectification. This technology will enable new classes of applications where robust long-range, day/night, all-weather 3D perception is needed, such as navigation of high-speed autonomous and semi-autonomous vehicles, ground and aerial, as well as a head-mounted depth-sensing system for dismounted personnel.