Method and system for image-based positioning and mapping for a road network utilizing object detection

1. A method, comprising:

obtaining, from one or more cameras associated with a vehicle travelling through a road network, a sequence of images reflecting an environment of the road network on which the vehicle is travelling, wherein each of the images has an associated camera location at which the image was recorded, wherein one or more of the camera locations are different;

generating, using at least some of the obtained images and the associated camera locations, a local map representation representing an area of the road network on which the vehicle is travelling, the generating including aggregating two or more of the obtained images together to form the local map representation;

comparing the generated local map representation with a section of a reference map, the reference map section covering the area of the road network on which the vehicle is travelling;

processing at least some of the images to perform a segmentation, a result of the segmentation being that each pixel of the processed images is allocated at least a probability value that the pixel represents a lane marking in the environment;

processing at least some of the images of the sequence of images to generate a road image representing an area of the road network within which the vehicle is travelling, wherein the pixel value for each pixel in the road image is based on the allocated probability values of the corresponding pixels in the images used to generate the road image;

processing the road image to detect and classify one or more lane marking objects within the image, wherein the one or more lane marking objects represent lane markings on one or more roads depicted in the road image; and

processing the road image using the detected and classified lane marking objects to determine the location and geometry of the lane markings represented by the one or more lane marking objects.

2. The method of claim 1 , further comprising:

identifying, based on the comparison, one or more errors in the reference map section; and

providing, when one or more errors are identified, the local map representation, or data indicative of the local map representation, to a remote server for updating the reference map section and/or generating a new reference map section.

3. The method of claim 1 , comprising determining a relative camera location for the at least some of the images used to generate the local map representation relative to a reference image, optionally wherein the reference image is the first image in the sequence of images.

4. The method of claim 1 , wherein a further result of the segmentation is that, for each of the images, each pixel is allocated an object class or object class vector indicating a probability of each object class for that pixel.

5. The method of claim 1 , comprising processing at least some of the images to detect one or more landmark objects, and generating a landmark observation for inclusion into the local map representation using the detected one or more landmark objects.

6. The method of claim 1 , comprising:

processing at least some of the images to detect one or more landmark objects within the images, wherein the landmark objects represent landmarks in the environment of the road network;

determining, for each landmark object detected in each image of the sequence of images, and using the associated camera locations of the images, a set of transformations for mapping the detected landmark object from the image in which it was detected into one or more adjacent images of the sequence of images;

generating, for each landmark object detected in each image of the sequence of images, a representation of the landmark object in a three-dimensional coordinate space;

determining a set of three-dimensional representations generated from different images that correspond to the same landmark in the environment; and

generating, from the determined set of three dimensional representations, data indicative of the landmark in the environment represented by the set of three dimensional representations.

7. The method of claim 6 , wherein the data indicative of the landmark includes information indicative of a position and orientation of the landmark in the environment.

8. The method of any of claim 6 , comprising fusing the three-dimensional representations in each set to determine a two-dimensional contour of the landmark and/or to generate a reconstruction of the landmark in the coordinate space.

9. The method of claim 1 , comprising generating a lane marking observation for inclusion into the local map representation using the one or more lane marking objects.

10. The method of claim 1 , comprising processing the plurality of images to generate an image of the area of the road network within which the vehicle is travelling for inclusion into the local map representation, wherein the image of the area of the road network is determined by projecting a plurality of different images into a three-dimensional coordinate frame using visual odometry.

11. The method of claim 1 , further comprising:

determining, based on the comparison, the geographical location and orientation of the vehicle within the road network.

12. The method of claim 1 , wherein the segmentation comprises semantic segmentation.

13. A system comprising one or more processors and a memory, one or more processors being arranged to:

obtain, from one or more cameras associated with a vehicle travelling through a road network, a sequence of images reflecting the environment of the road network on which the vehicle is travelling, wherein each of the images has an associated camera location at which the image was recorded;

generate, using at least some of the obtained images and the associated camera locations, a local map representation representing an area of the road network on which the vehicle is travelling, the generating including aggregating two or more of the obtained images together to form the local map representation;

compare the generated local map representation with a section of a reference map extracted from a map repository, the reference map section covering the area of the road network on which the vehicle is travelling;

process at least some of the images to perform a segmentation, a result of the segmentation being that each pixel of the processed images is allocated at least a probability value that the pixel represents a lane marking in the environment;

process at least some of the images of the sequence of images to generate a road image representing an area of the road network within which the vehicle is travelling, wherein the pixel value for each pixel in the road image is based on the allocated probability values of the corresponding pixels in the images used to generate the road image;

process the road image to detect and classify one or more lane marking objects within the image, wherein the one or more lane marking objects represent lane markings on one or more roads depicted in the road image; and

process the road image using the detected and classified lane marking objects to determine the location and geometry of the lane markings represented by the one or more lane marking objects.

14. A non-transitory computer readable medium comprising instructions which, when executed by a computing device comprising one or more processors, cause the computing device to perform a method comprising:

obtaining, from one or more cameras associated with the vehicle travelling through the road network, a sequence of images reflecting an environment of the road network on which the vehicle is travelling, wherein each of the images has an associated camera location at which the image was recorded, wherein one or more of the camera locations are different;

generating, using at least some of the obtained images and the associated camera locations, a local map representation representing an area of the road network on which the vehicle is travelling, the generating including aggregating two or more of the obtained images together to form the local map representation;

comparing the generated local map representation with a section of a reference map, the reference map section covering the area of the road network on which the vehicle is travelling;

processing at least some of the images to perform a segmentation, a result of the segmentation being that each pixel of the processed images is allocated at least a probability value that the pixel represents a lane marking in the environment;

processing at least some of the images of the sequence of images to generate a road image representing an area of the road network within which the vehicle is travelling, wherein the pixel value for each pixel in the road image is based on the allocated probability values of the corresponding pixels in the images used to generate the road image;

processing the road image to detect and classify one or more lane marking objects within the image, wherein the one or more lane marking objects represent lane markings on one or more roads depicted in the road image; and

processing the road image using the detected and classified lane marking objects to determine the location and geometry of the lane markings represented by the one or more lane marking objects.

15. The non-transitory computer readable medium of claim 14 , wherein the method further comprises:

determining, based on the comparison, the geographical location and orientation of the vehicle within the road network.

16. The non-transitory computer readable medium of claim 14 , wherein the method further comprises:

identifying, based on the comparison, one or more errors in the reference map section; and

providing, when one or more errors are identified, the local map representation, or data indicative of the local map representation, to a remote server for updating the reference map section and/or generating a new reference map section.

17. The non-transitory computer readable medium of claim 14 , wherein the method further comprises:

generating a lane marking observation for inclusion into the local map representation using the one or more lane marking objects.

18. A method, comprising:

obtaining, from one or more cameras associated with a vehicle travelling through a road network, a sequence of images reflecting an environment of the road network on which the vehicle is travelling, wherein each of the images has an associated camera location at which the image was recorded, wherein one or more of the camera locations are different;

generating, using at least some of the obtained images and the associated camera locations, a local map representation representing an area of the road network on which the vehicle is travelling, the generating including aggregating two or more of the obtained images together to form the local map representation;

comparing the generated local map representation with a section of a reference map, the reference map section covering the area of the road network on which the vehicle is travelling;

performing a pixel wise segmentation on the images, the pixel wise segmentation resulting in each pixel being allocated an object class or object class vector indicating a probability of each object class for that pixel;

processing at least some of the images to detect one or more lane marking objects based at least in part on the object classes or object class vectors; and

generating, using the one or more lane marking objects, a lane marking observation for inclusion into the local map representation.

19. The method of claim 18 , wherein the method further comprises:

determining, based on the comparison, the geographical location and orientation of the vehicle within the road network.

20. The method of claim 18 , wherein the method further comprises:

identifying, based on the comparison, one or more errors in the reference map section; and

providing, when one or more errors are identified, the local map representation, or data indicative of the local map representation, to a remote server for updating the reference map section and/or generating a new reference map section.