Camera-radar fusion using correspondences
A method of performing sensor fusion of data obtained from a camera ( 51 ) and a supplemental sensor ( 52 ), the method comprising performing patch tracking ( 55 ) on image data ( 53 ) provided by the camera ( 51 ) to determine tracked patches ( 56 ), and performing a fusion ( 57 ) of the image data ( 53 ) obtained from the camera ( 51 ) with supplemental data ( 54 ) provided by the supplemental sensor ( 52 ) based on the tracked patches ( 56 ).
1. A method of performing sensor fusion of data obtained from a camera and a supplemental sensor, the method comprising:
performing patch tracking on image data provided by the camera to determine tracked patches;
determining, for each tracked patch, a patch scale ratio based on a change in size of the patch between two frames of the image data by calculating a determinant of a 2×2 upper left sub-matrix of an affine transformation matrix representing the change in the patch between the two frames;
determining, for each part of supplemental data provided by the supplemental sensor, an equivalent scale ratio based on a ratio of distances to an object at two different times, wherein the ratio is calculated using a radial velocity measurement and a current distance measurement;
identifying, for each part of the supplemental data, a corresponding tracked patch by matching the equivalent scale ratio of the part of supplemental data to the patch scale ratio of the tracked patch; and
performing a fusion of the image data obtained from the camera with supplemental data provided by the supplemental sensor based on the tracked patches.
2. The method of claim 1 , wherein performing patch tracking comprises performing an affine patch tracking.
3. The method of claim 1 , comprising applying part of the supplemental data provided by the supplemental sensor to a patch of the tracked patches.
4. The method of claim 1 , wherein the supplemental data is provided by the supplemental sensor in the form of a point cloud.
5. The method of claim 1 , wherein the part of the supplemental data is a point of a point cloud.
6. The method of claim 1 , wherein the supplemental sensor is a radar sensor.
7. The method of claim 1 , wherein the equivalent scale ratio is determined from distance information and radial velocity information of a point of the radar point cloud.
8. The method of claim 1 , comprising, for a point that is considered being located close to a patch of the tracked patches, comparing an equivalent scale ratio of the point with a scale ratio of the patch obtained from patch tracking to determine if the scale ratios match.
9. The method of claim 1 , comprising discarding those points of points that are considered being located close to a patch of the tracked patches, whose equivalent scale ratios do not match with a scale ratio of the patch obtained from affine patch tracking.
10. The method of claim 1 , comprising performing object segmentation on the image data captured by the camera.
11. The method of claim 1 , comprising averaging supplemental data related to patches associated with an object, and associating the averaged information with the object.
12. The method of claim 1 , wherein the supplemental data comprises Time to Collision information.
13. The method of claim 1 , wherein the sensor fusion is applied in an automotive context.
14. A device, comprising:
processing circuitry, the processing circuitry being configured to
perform patch tracking on image data provided by a camera to determine tracked patches,
determine, for each tracked patch, a patch scale ratio based on a change in size of the patch between two frames of the image data by calculating a determinant of a 2×2 upper left sub-matrix of an affine transformation matrix representing the change in the patch between the two frames,
determine, for each part of supplemental data provided by the supplemental sensor, an equivalent scale ratio based on a ratio of distances to an object at two different times, wherein the ratio is calculated using a radial velocity measurement and a current distance measurement,
identify, for each part of the supplemental data, a corresponding tracked patch by matching the equivalent scale ratio of the part of supplemental data to the patch scale ratio of the tracked patch, and
perform a fusion of the image data obtained from the camera with supplemental data provided by a supplemental sensor based on the tracked patches.
15. A system comprising a camera, a supplemental sensor, and the device of claim 14 , the device being configured to perform patch tracking on image data provided by the camera to determine tracked patches, and to perform a fusion of the image data obtained from the camera with supplemental data provided by the supplemental sensor based on the tracked patches.