System and method for providing mobile range sensing
The present invention provides an improved method for estimating range of objects in images from various distances comprising receiving a set of images of the scene having multiple objects from at least one camera in motion. Due to the motion of the camera, each of the images are obtained at different camera locations. Then an object visible in multiple images is selected. Data related to approximate camera positions and orientations and the images of the visible object are used to estimate the location of the object relative to a reference coordinate system. Based on the computed data, a projected location of the visible object is computed and the orientation angle of the camera for each image is refined. Additionally, pairs of cameras with various locations can obtain dense stereo for regions of the image at various ranges.
1. A computer implemented method for detecting range of one or more objects in a scene comprising:
detecting one or more objects from a set of images of the scene captured from one or more cameras;
selecting images from the set of images containing the detected one or more objects;
computing dense stereo for foreground objects and for background objects of a current image by:
comparing the current image with a first image from the selected images having a long baseline for distant background objects to generate a first composite image having range estimates obtained for the background objects;
comparing the current image with a second image from the selected images having a short baseline for foreground objects to generate a second composite image having range estimates obtained for the foreground objects; and
associating the first composite image with the second composite image to form a composite range image with range estimates for the foreground and background objects,
wherein baseline is measurement of a separation distance between the one or more cameras.
2. The method of claim 1 wherein the computing is performed after:
determining a projected location of the detected one or more objects based on computed metadata comprising position and orientation for the one or more camera and for the selected images;
adjusting the pose of the one or more cameras for each of the selected images based on the projected location of the detected one or more objects, and
the computing of dense stereo is based on the adjusted pose of the one or more cameras.
3. The method of claim 2 wherein said step of adjusting matches the location of the one or more objects in the selected images with the projected location of the one or more objects.
4. The method of claim 2 further comprising buffering the computed metadata of the selected images for the one or more cameras in motion.
5. The method of claim 2 further comprising rectifying the selected images based on the adjusted estimated pose of the one or more cameras.
6. The method of claim 5 wherein the rectified images are displayed through the adjusted estimation orientation of the one or more cameras.
7. The method of claim 5 further comprising computing a dense stereo image of the rectified images.
8. A non-transitory computer readable medium comprising computer readable storage medium having a computer program stored thereon for performing the method of claim 1 .
9. The method of claim 2 wherein the computing comprises calculating multiple rays from the one or more cameras to the one or more objects defining the projected location of the one or more objects.
10. The method of claim 1 wherein the one or more cameras is one camera in motion.
11. The method of claim 1 wherein the one or more cameras is a plurality of cameras, which are stationary or moving.
12. A non-transitory computer readable medium comprising computer readable storage medium having a computer program stored thereon for:
detecting one or more objects from a set of images of the scene captured from one or more cameras;
selecting images from the set of images containing the detected one or more objects;
computing dense stereo for foreground objects and for background objects of a current image by:
comparing the current image with a first image from the selected images having a long baseline for distant background objects to generate a first composite image having range estimates obtained for the background objects;
comparing the current image with a second image from the selected images having a short baseline for foreground objects to generate a second composite image having range estimates obtained for the foreground objects; and
associating the first composite image with the second composite image to form a composite range image with range estimates for the foreground and background objects,
wherein baseline is measurement of a separation distance between the one or more cameras.
13. The method of claim 12 wherein the computing comprises calculating multiple rays from the camera to the one or more objects defining the projected location of the one or more objects.