Fisheye collage transformation for road object detection or other object detection

1. A method comprising:

obtaining a fisheye image of a scene around a vehicle;

identifying multiple regions of interest in the fisheye image;

applying one or more transformations to transform and rotate one or more of the regions of interest in the fisheye image to produce one or more transformed regions;

generating a collage image comprising at least one portion based on the fisheye image and one or more portions containing the one or more transformed regions; and

performing object detection to identify one or more objects captured in the collage image;

wherein the multiple regions of interest comprise different portions of the fisheye image that are identified as potentially containing one or more objects of interest; and

wherein at least one region of interest identified as potentially containing at least one object of interest is located within another region of interest identified as potentially containing the at least one object of interest and includes a driving lane closest to the vehicle.

2. The method of claim 1 , wherein:

the at least one portion of the collage image based on the fisheye image comprises an undistorted version of a central portion of the fisheye image; and

the one or more portions of the collage image containing the one or more transformed regions comprise one or more transformed and rotated regions of interest having substantially a same orientation as the undistorted version of the central portion of the fisheye image.

3. The method of claim 2 , wherein the undistorted version of the central portion of the fisheye image and the one or more transformed regions have different scales.

4. The method of claim 1 , wherein the collage image contains multiple transformed regions associated with two or more of the multiple regions of interest in the fisheye image, the two or more regions of interest located on opposite sides of the fisheye image.

5. The method of claim 1 , wherein the collage image contains multiple transformed regions associated with two or more of the multiple regions of interest in the fisheye image.

6. The method of claim 1 , wherein:

multiple fisheye images are obtained and multiple regions of interest are identified in each of the fisheye images;

a collage image is generated for each of the fisheye images; and

the collage images have a common format, the common format including an area for an undistorted version of a central portion of each fisheye image and multiple areas for multiple transformed regions associated with the multiple regions of interest in each fisheye image.

7. The method of claim 1 , wherein performing object detection comprises performing object detection using a machine learning algorithm that has not been trained using rotated versions of objects.

8. The method of claim 1 , further comprising:

identifying a boundary associated with each of the one or more objects in the collage image.

9. The method of claim 8 , further comprising at least one of:

translating the boundary of each of the one or more objects from a collage image space to a fisheye image space; and

translating the boundary of each of the one or more objects from the collage image space to an undistorted image space.

10. The method of claim 1 , further comprising:

identifying at least one action based on the one or more objects; and

performing the at least one action.

11. The method of claim 10 , wherein the at least one action comprises at least one of:

an adjustment to at least one of: a steering of the vehicle, a speed of the vehicle, an acceleration of the vehicle, and a braking of the vehicle; and

an activation of an audible, visible, or haptic warning.

12. An apparatus comprising:

at least one processor configured to:

obtain a fisheye image of a scene around a vehicle;

identify multiple regions of interest in the fisheye image;

apply one or more transformations to transform and rotate one or more of the regions of interest in the fisheye image to produce one or more transformed regions;

generate a collage image comprising at least one portion based on the fisheye image and one or more portions containing the one or more transformed regions; and

perform object detection to identify one or more objects captured in the collage image;

wherein the at least one processor is configured to identify different portions of the fisheye image that potentially contain one or more objects of interest as the multiple regions of interest; and

wherein at least one region of interest identified as potentially containing at least one object of interest is located within another region of interest identified as potentially containing the at least one object of interest and includes a driving lane closest to the vehicle.

13. The apparatus of claim 12 , wherein:

the at least one portion of the collage image based on the fisheye image comprises an undistorted version of a central portion of the fisheye image; and

the one or more portions of the collage image containing the one or more transformed regions comprise one or more transformed and rotated regions of interest having substantially a same orientation as the undistorted version of the central portion of the fisheye image.

14. The apparatus of claim 13 , wherein the undistorted version of the central portion of the fisheye image and the one or more transformed regions have different scales.

15. The apparatus of claim 12 , wherein the collage image contains multiple transformed regions associated with two or more of the multiple regions of interest in the fisheye image, the two or more regions of interest located on opposite sides of the fisheye image.

16. The apparatus of claim 12 , wherein the collage image contains multiple transformed regions associated with two or more of the multiple regions of interest in the fisheye image.

17. The apparatus of claim 12 , wherein:

the at least one processor is configured to obtain multiple fisheye images and identify multiple regions of interest in each of the fisheye images;

the at least one processor is configured to generate a collage image for each of the fisheye images; and

the collage images have a common format, the common format including an area for an undistorted version of a central portion of each fisheye image and multiple areas for multiple transformed regions associated with the multiple regions of interest in each fisheye image.

18. The apparatus of claim 12 , wherein, to perform object detection, the at least one processor is configured to use a machine learning algorithm that has not been trained using rotated versions of objects.

19. The apparatus of claim 12 , wherein the at least one processor is further configured to identify a boundary associated with each of the one or more objects in the collage image.

20. The apparatus of claim 19 , wherein the at least one processor is further configured to at least one of:

translate the boundary of each of the one or more objects from a collage image space to a fisheye image space; and

translate the boundary of each of the one or more objects from the collage image space to an undistorted image space.

21. The apparatus of claim 12 , wherein the at least one processor is further configured to:

identify at least one action based on the one or more objects; and

perform the at least one action.

22. The apparatus of claim 21 , wherein the at least one action comprises at least one of:

an adjustment to at least one of: a steering of the vehicle, a speed of the vehicle, an acceleration of the vehicle, and a braking of the vehicle; and

an activation of an audible, visible, or haptic warning.

23. A non-transitory machine-readable medium containing instructions that when executed cause at least one processor to:

obtain a fisheye image of a scene around a vehicle;

identify multiple regions of interest in the fisheye image;

apply one or more transformations to transform and rotate one or more of the regions of interest in the fisheye image to produce one or more transformed regions;

generate a collage image comprising at least one portion based on the fisheye image and one or more portions containing the one or more transformed regions; and

perform object detection to identify one or more objects captured in the collage image;

wherein the instructions that when executed cause the at least one processor to identify the multiple regions of interest in the fisheye image comprise:

instructions that when executed cause the at least one processor to identify different portions of the fisheye image that potentially contain one or more objects of interest as the multiple regions of interest; and

wherein at least one region of interest identified as potentially containing at least one object of interest is located within another region of interest identified as potentially containing the at least one object of interest and includes a driving lane closest to the vehicle.

24. The non-transitory machine-readable medium of claim 23 , wherein:

the at least one portion of the collage image based on the fisheye image comprises an undistorted version of a central portion of the fisheye image; and

the one or more portions of the collage image containing the one or more transformed regions comprise one or more transformed and rotated regions of interest having substantially a same orientation as the undistorted version of the central portion of the fisheye image.

25. The non-transitory machine-readable medium of claim 24 , wherein the undistorted version of the central portion of the fisheye image and the one or more transformed regions have different scales.

26. The non-transitory machine-readable medium of claim 23 , wherein the collage image contains multiple transformed regions associated with two or more of the multiple regions of interest in the fisheye image, the two or more regions of interest located on opposite sides of the fisheye image.

27. The non-transitory machine-readable medium of claim 23 , wherein the collage image contains multiple transformed regions associated with two or more of the multiple regions of interest in the fisheye image.

28. The non-transitory machine-readable medium of claim 23 , wherein:

the instructions when executed cause the at least one processor to obtain multiple fisheye images, identify multiple regions of interest in each of the fisheye images, and generate a collage image for each of the fisheye images; and

the collage images have a common format, the common format including an area for an undistorted version of a central portion of each fisheye image and multiple areas for multiple transformed regions associated with the multiple regions of interest in each fisheye image.

29. The non-transitory machine-readable medium of claim 23 , wherein the instructions that when executed cause the at least one processor to perform object detection comprise:

instructions that when executed cause the at least one processor to use a machine learning algorithm that has not been trained using rotated versions of objects.

30. The non-transitory machine-readable medium of claim 23 , wherein the instructions when executed further cause the at least one processor to identify a boundary associated with each of the one or more objects in the collage image.

31. The non-transitory machine-readable medium of claim 30 , wherein the instructions when executed further cause the at least one processor to:

translate the boundary of each of the one or more objects from a collage image space to a fisheye image space; and

translate the boundary of each of the one or more objects from the collage image space to an undistorted image space.

32. The non-transitory machine-readable medium of claim 23 , wherein the instructions when executed further cause the at least one processor to:

identify at least one action based on the one or more objects; and

perform the at least one action.

33. The non-transitory machine-readable medium of claim 32 , wherein the at least one action comprises at least one of:

an adjustment to at least one of: a steering of the vehicle, a speed of the vehicle, an acceleration of the vehicle, and a braking of the vehicle; and

an activation of an audible, visible, or haptic warning.