Vehicle Camera Visibility Extension via Cumulative Virtual Image

The patent divides the obstacle detection task into two independent stages: (1) generating a basic virtual image from a single real image through pixel correlation with a planar surface model, and (2) extending visibility by superimposing multiple basic virtual images to form a cumulative virtual image. This segmentation allows each stage to be optimized independently, achieving extended visibility without requiring complex multi-camera systems.

The patent transforms the 3D obstacle detection problem into a 2D virtual image generation problem by correlating pixels with a planar surface model. This dimensionality change simplifies the measurement process while maintaining accuracy, as the planar surface assumption provides sufficient constraints for distance calculation from a single image.

The patent pre-establishes a planar surface model of the ground before image acquisition. This preliminary action provides a reference framework that allows accurate virtual image generation even when vehicle displacement between images is minimal, eliminating the need for large stereoscopic base that would otherwise be required.

The patent creates a basic virtual image that is a correlated copy of the real image, transformed according to the planar surface model. This copying process generates a top-down view that preserves distance information without requiring physical camera displacement, enabling accurate reconstruction at low speeds.

The patent merges multiple basic virtual images into a single cumulative virtual image through superimposition. This combining process extends the visibility area beyond the original camera field of view, as each new basic virtual image adds previously unseen areas while overlapping regions provide redundant distance measurements.

The patent continuously accumulates virtual images as the vehicle moves, maintaining an ever-expanding cumulative virtual image. This continuous accumulation ensures that the visibility area grows with vehicle displacement, providing extended coverage without requiring the camera to pan or tilt.

The patent replaces expensive high-precision sensors with a standard camera and computational algorithms. By using a single image acquisition apparatus with a planar surface model, the system achieves accurate obstacle detection without requiring costly specialized sensors, making the solution more economically viable.

The patent substitutes mechanical sensor precision requirements with computational image processing. Instead of relying on high-precision hardware sensors, the system uses pixel correlation algorithms and planar surface modeling to extract accurate distance measurements from standard camera images.