System of depth estimation and method thereof

1. A computerized method of depth map generation, the method performed by a processor and memory circuitry (PMC), the method comprising:

obtaining a pair of images comprising a first image and a second image acquired by at least one imaging device at two time points during a relative movement between a vehicle and the at least one imaging device, the first image and the second image capturing respective portions of a surface part of the vehicle and the captured portions being partially overlapped, the pair of images constituting a part of an entire image capturing the surface part;

obtaining a segmentation map corresponding to at least the part of the entire image, the segmentation map including one or more segments corresponding to one or more vehicle components comprised in the at least part of the entire image;

dividing the pair of images into one or more sections according to the segmentation map; and

for each given section of the one or more sections, generating a depth map comprising:

i) calculating a disparity map for the given section, the disparity map comprising disparity values each being indicative of difference of location between a respective pixel in the given section in the first image and a matching pixel thereof in the second image, wherein the matching pixel is searched in a range defined based on a displacement of the given section between the first image and the second image and within the same segment in the second image to which the respective pixel in the first image belongs; and

ii) computing a depth map for the given section based on the disparity map, wherein the depth map for each given section is usable to be combined to generate a depth map for the pair of images.

2. The computerized method according to claim 1 , wherein the generating a depth map further comprises estimating the displacement of the given section between the first image and the second image, thereby giving rise to an overlapping area in the given section between the first image and the second image,

wherein said calculating comprises calculating a disparity map for the overlapping area in the given section based on the displacement thereof; and

wherein said computing comprises computing a depth map for the overlapping area in the given section based on the disparity map.

3. The computerized method according to claim 1 , wherein the pair of images are divided, along a direction perpendicular to the direction of the relative movement, into one or more sections in accordance with the segmentation map.

4. The computerized method according to claim 1 , wherein the pair of images are divided into one or more sections corresponding to the one or more segments in the segmentation map.

5. The computerized method according to claim 2 , wherein the displacement of the given section between the first image and the second image is estimated with respect to one or more displacements of the given section in one or more precedent pairs of images.

6. The computerized method according to claim 2 , wherein the calculating a disparity map comprises, for each given pixel in the overlapping area in the first image:

defining a range in the second image based on the displacement and within the same segment to which the given pixel belongs;

searching, within the defined range, a matching pixel for the given pixel in the second image using a matching method; and

calculating the difference of location between the given pixel and the matching pixel, giving rise to a disparity value corresponding to the given pixel.

7. The computerized method according to claim 6 , wherein the matching method is an optimized semi-global matching method regularized by a segmentation factor.

8. The computerized method according to claim 7 , wherein the optimized semi-global matching method is regularized by adjusting a cost function thereof according to the segmentation factor, the segmentation factor being indicative of whether a candidate pixel and one or more neighboring pixels thereof belong to the same segment, and the searching comprises using the adjusted cost function to select the matching pixel from one or more candidate pixels in the defined range.

9. The computerized method according to claim 1 , wherein the surface part is an undercarriage of the vehicle.

10. The computerized method according to claim 1 , wherein the first image and the second image are at least 50% overlapped.

11. The computerized method according to claim 1 , wherein the entire image is generated by stitching a plurality of images captured by the at least one imaging device at a plurality of time points, and the segmentation map corresponds to the entire image and includes one or more segments corresponding to one or more vehicle components comprised in the surface part.

12. A computerized method of depth map generation, the method performed by a processor and memory circuitry (PMC), the method comprising:

obtaining a pair of images comprising a first image and a second image acquired by at least one imaging device at two time points during a relative movement between a vehicle and the at least one imaging device, the first image and the second image capturing respective portions of a surface part of the vehicle and the captured portions being partially overlapped, the pair of images constituting a part of an entire image capturing the surface part;

obtaining a segmentation map corresponding to at least the part of the entire image, the segmentation map including one or more segments corresponding to one or more vehicle components comprised in the at least part of the entire image;

dividing the pair of images into one or more sections according to the segmentation map; and

for each given section of the one or more sections, generating a depth map comprising:

i) calculating a disparity map for the given section, the disparity map comprising disparity values each being indicative of difference of location between a respective pixel in the given section in the first image and a matching pixel thereof in the second image, wherein the matching pixel is searched in a range defined within the same segment to which the respective pixel belongs; and

ii) computing a depth map for the given section based on the disparity map, wherein the depth map for each given section is usable to be combined to generate a depth map for the pair of images;

wherein the entire image is generated by stitching a plurality of images captured by the at least one imaging device at a plurality of time points, and the segmentation map corresponds to the entire image and includes one or more segments corresponding to one or more vehicle components comprised in the surface part; and

wherein the segmentation map is generated using a segmentation model, the segmentation model trained using a training dataset comprising a set of training images of the surface part of the vehicle pre-segmented according to the one or more vehicle components comprised therein.

13. The computerized method according to claim 1 , further comprising extracting features from the pair of images, matching the features between the first image and the second image, calculating a vertical displacement of the matched features along a direction perpendicular to the direction of the relative movement, and adjusting the pair of images according to the vertical displacement.

14. The computerized method according to claim 13 , wherein the matching pixel is searched in a range defined within the same segment, taking into consideration the matched features.

15. A computerized system of depth map generation, the system comprising a processor and memory circuitry (PMC) configured to:

obtain a pair of images comprising a first image and a second image acquired by at least one imaging device at two time points during a relative movement between a vehicle and the at least one imaging device, the first image and the second image capturing respective portions of a surface part of the vehicle and the captured portions being partially overlapped, the pair of images constituting a part of an entire image capturing the surface part;

obtain a segmentation map corresponding to at least the part of the entire image, the segmentation map including one or more segments corresponding to one or more vehicle components comprised in the at least part of the entire image;

divide the first image into one or more sections according to the segmentation map; and

for each given section in the first image, generate a depth map, comprising:

i) calculate a disparity map for the given section, the disparity map comprising disparity values, each being indicative of difference of location between a respective pixel in the given section in the first image and a matching pixel thereof in the second image, wherein the matching pixel is searched in a range defined based on a displacement of the given section between the first image and the second image and within the same segment in the second image to which the respective pixel in the first image belongs; and

ii) compute a depth map for the given section based on the disparity map, wherein the depth map for each given section is usable to be combined to generate a depth map for the pair of images.

16. The computerized system according to claim 15 , wherein, when generating a depth map, the PMC is further configured to estimate the displacement of the given section between the first image and the second image, thereby giving rise to an overlapping area in the given section between the first image and the second image, and

wherein the PMC is configured to calculate a disparity map for the overlapping area in the given section based on the displacement thereof; and compute a depth map for the overlapping area in the given section based on the disparity map.

17. The computerized system according to claim 15 , wherein the first image is divided, along a direction perpendicular to the direction of the relative movement, into one or more sections in accordance with the segmentation map.

18. The computerized system according to claim 15 , wherein the first image is divided into one or more sections corresponding to the one or more segments in the segmentation map.

19. The computerized system according to claim 16 , wherein the displacement of the given section between the first image and the second image is estimated with respect to one or more displacements of the given section in one or more precedent pairs of images.

20. The computerized system according to claim 16 , wherein the PMC is configured to calculate a disparity map by:

for each given pixel in the overlapping area in the first image:

defining a range in the second image based on the displacement and within the same segment to which the given pixel belongs;

searching, within the defined range, a matching pixel for the given pixel in the second image using a matching system; and

calculating the difference of location between the given pixel and the matching pixel, giving rise to a disparity value corresponding to the given pixel.

21. The computerized system according to claim 20 , wherein the matching method is an optimized semi-global matching method regularized by a segmentation factor.

22. The computerized system according to claim 21 , wherein the optimized semi-global matching method is regularized by adjusting a cost function thereof according to the segmentation factor, the segmentation factor being indicative of whether a candidate pixel and one or more neighboring pixels thereof belong to the same segment, and the searching comprises using the adjusted cost function to select the matching pixel from one or more candidate pixels in the defined range.

23. The computerized system according to claim 15 , wherein the PMC is further configured to extract features from the pair of images, match the features between the first image and the second image, calculate a vertical displacement of the matched features along a direction perpendicular to the direction of the relative movement, and adjust the pair of images according to the vertical displacement.

24. The computerized system according to claim 23 , wherein the matching pixel is searched in a range defined within the same segment, taking into consideration the matched features.

25. A non-transitory computer readable storage medium tangibly embodying a program of instructions that, when executed by a computer, cause the computer to perform a method of depth map generation, the method comprising:

obtaining a pair of images comprising a first image and a second image acquired by at least one imaging device at two time points during a relative movement between a vehicle and the at least one imaging device, the first image and the second image capturing respective portions of a surface part of the vehicle and the captured portions being partially overlapped, the pair of images constituting a part of an entire image capturing the surface part;

obtaining a segmentation map corresponding to at least the part of the entire image, the segmentation map including one or more segments corresponding to one or more vehicle components comprised in the at least part of the entire image;

dividing the first image into one or more sections according to the segmentation map; and

for each given section in the first image, generating a depth map comprising:

i) calculating a disparity map for the given section, the disparity map comprising disparity values, each being indicative of difference of location between a respective pixel in the given section in the first image and a matching pixel thereof in the second image, wherein the matching pixel is searched in a range defined based on a displacement of the given section between the first image and the second image and within the same segment in the second image to which the respective pixel in the first image belongs; and

ii) computing a depth map for the given section based on the disparity map, wherein the depth map for each given section is usable to be combined to generate a depth map for the pair of images.