Localization of autonomous vehicles via ground image recognition

A method of localizing a host traversing a road surface, wherein the host is configured with a camera system mounted thereto, and wherein the camera system comprises a stereo camera oriented to view the road surface, an image processor, and a communications module, the method comprising:

• obtaining at least one reference image of a road surface using said stereo camera, said stereo camera configured to capture images of at least one road feature from at least two different perspectives and provide three dimensional imaging thereof;
• determining a pose of respective reference images using said image processor;
• compiling a plurality of reference images of the road surface into a reference mosaic using said image processor;
• obtaining at least a first sample image of the road surface using said stereo camera;
• determining a pose of the first sample image using said image processor;
• selecting a subset of the reference mosaic that corresponds to the pose of the first sample image;
• performing a registration, wherein micro-features in the first sample image and the selected subset of the reference mosaic are identified and compared using said image processor the micro-features comprising micro-features less than 5 cm in a first dimension, the micro-features comprising:
• contours of embedded pieces of granular material of the road surface,
• road surface pores,
• road surface cracks,
• road surface holes, and/or
• road surface fissures;
• determining a pose of the host by matching corresponding microfeatures identified in both the selected subset of the reference mosaic and the first sample image using said image processor.

A camera system for localizing a host vehicle, the camera system comprising:

• an imaging device configured to capture images of at least one road feature from at least two different perspectives and provide three dimensional imaging thereof, the imaging device mounted to said host vehicle, wherein the imaging device is faces downward to obtain sample images of a road surface;
• an image processor;
• database of reference images of the road surface, wherein a given reference image is associated with a known location;
• and
• a communications module in networked connection with said image processor and database;
• wherein the camera system is configured to identify micro-features in the road surface from the sample images, match the micro-features identified in the sample images with micro-features in the reference images of the road surface, and localize the host vehicle in space based on the known location associated with one or more reference images that match the sample images, wherein the micro-features comprise micro-features less than 5 cm in a first dimension, the micro-features comprising:
• contours of embedded pieces of granular material of the road surface,
• road surface pores,
• road surface cracks,
• road surface holes, and/or
• road surface fissures.