US 12,277,645 B2
Method and apparatus for constructing three-dimensional channel, and non-transitory computer-readable recording medium
Hong Yi, Beijing (CN); Haijing Jia, Beijing (CN); Hengzhi Zhang, Beijing (CN); Liyan Liu, Beijing (CN); and Wei Wang, Beijing (CN)
Assigned to Ricoh Company, Ltd., Tokyo (JP)
Filed by Hong Yi, Beijing (CN); Haijing Jia, Beijing (CN); Hengzhi Zhang, Beijing (CN); Liyan Liu, Beijing (CN); and Wei Wang, Beijing (CN)
Filed on Jan. 17, 2023, as Appl. No. 18/155,159.
Claims priority of application No. 202210048805.8 (CN), filed on Jan. 17, 2022.
Prior Publication US 2023/0230316 A1, Jul. 20, 2023
Int. Cl. G06T 17/00 (2006.01); G06T 7/60 (2017.01); G06V 10/25 (2022.01); H04N 23/698 (2023.01)
CPC G06T 17/00 (2013.01) [G06T 7/60 (2013.01); G06V 10/25 (2022.01); H04N 23/698 (2023.01); G06T 2207/10028 (2013.01); G06T 2207/20084 (2013.01); G06V 2201/07 (2022.01)] 9 Claims
OG exemplary drawing
 
1. A method for constructing a three-dimensional channel, the method comprising:
generating a depth image corresponding to a first panoramic image;
performing channel object detection on the first panoramic image to obtain channel object boxes in the first panoramic image, each of the channel object boxes being four annularly interconnected arc-shaped sides, and each of the arc-shaped sides being a circular arc of a great circle;
generating, based on the first panoramic image, the channel object boxes and the depth image, a three-dimensional point cloud of a channel object in the first panoramic image; and
correcting, using the first panoramic image and the channel object boxes, depth information of the depth image, before generating the three-dimensional point cloud of the channel object in the first panoramic image,
wherein correcting the depth information of the depth image using the first panoramic image and the channel object boxes includes
calculating, based on position coordinates of four corner points of the channel object box, and one or more included angles between two adjacent corner points on a unit sphere, a two-dimensional point set of the channel object box in the first panoramic image,
obtaining, based on three-dimensional space coordinates of two-dimensional points in the two-dimensional point set of the channel object box in the depth image, a three-dimensional point set corresponding to the two-dimensional point set,
performing fitting based on the three-dimensional point set to obtain a target plane, and
determining first pixels in the target plane, and second pixels in the depth image to which the first pixels correspond, and correcting depth of the second pixels using the three-dimensional space coordinates of the first pixels.