Navigation in a degraded visual environment due to darkness poses a threat to the safety of ground vehicles and their operators. Systems have been developed to navigate in darkness that depend on differences between conditions such as temperature or reflectivity at various wavelengths, however adding sensors to these systems increases complexity by introducing multiple components with the potential to create problems with alignment and calibration. An approach is needed that is passive and simple to achieve widespread acceptance. Our approach uses a type of augmented display to show fused images from visible and thermal sensors that are continuously updated. Because a raw fused image creates an unnatural color appearance, we use a color transfer process based on a look-up table. We replace the false colors with a colormap derived from daytime reference images obtained from a public database using the GPS coordinates of a vehicle. Our system performs registration, colorizing, and reference updating in near real-time, which allows operators of ground vehicles to see a colored view of a nighttime scene.;The success of a color night navigation system often depends on the accuracy of the colors in the resulting image. Small regions can incorrectly adopt the color of large regions simply due to size of the regions. Therefore, we also present a method to improve the color accuracy of a night navigation system by initially splitting a fused image into two distinct sections before the colorization. Process. The two regions, generally road and sky, are processed separately to obtain improved color accuracy as compared to not splitting the regions.;Keywords: color night vision, image registration, mutual information, image fusion, night navigation, thermal video, color transfer.
展开▼
