Method for detecting collisions in video and electronic device

Patent metadata

Field	Value
Publication number	US-2022237916-A1
Application number	US-202117537023-A
Country	US
Kind code	A1
Filing date	Nov 29, 2021
Priority date	Jan 22, 2021
Publication date	Jul 28, 2022
Grant date	—

Abstract

A method for detecting collisions in a video is provided. In the method, first bounding boxes of dynamic virtual elements are acquired, wherein the dynamic virtual elements are added into a video picture; target contour points corresponding to an original target object in the video picture are identified, wherein the target contour points are positioned on a contour line of the original target object; one second bounding box is created based on each two adjacent target contour points of the original target object; and the collisions between the first bounding boxes and the second bounding boxes are detected. A device and a computer-readable storage medium are further provided.

First claim

What is claimed is:

1 . A method for detecting collisions in a video, comprising:
acquiring first bounding boxes of dynamic virtual elements, wherein the dynamic virtual elements are added into a video picture; identifying target contour points corresponding to an original target object in the video picture, wherein the target contour points are positioned on a contour line of the original target object; creating a second bounding box based on each two adjacent target contour points of the original target object; and detecting the collisions between the first bounding boxes and the second bounding boxes to determine whether the dynamic virtual elements collide with the original target.

2 . The method for detecting the collisions in the video according to claim 1 , further comprising:
determining that the dynamic virtual elements collide with the original target object in response to the first bounding boxes colliding with any one of the second bounding boxes.

3 . The method for detecting the collisions in the video according to claim 1 , wherein said identifying the target contour points corresponding to the original target object in the video picture comprises:
determining pixel points satisfying a contour condition in the video picture as original contour points; and extracting a second reference number of the target contour points every a first reference number of the original contour points.

4 . The method for detecting the collisions in the video according to claim 3 , wherein said determining the pixel points, satisfying the contour condition, in the video picture as the original contour points comprises:
traversing pixel points in the video picture; determining a currently traversed pixel point as a first original contour point in response to the currently traversed pixel point being a pixel point corresponding to the original target object, and a previously traversed pixel point adjacent to the currently traversed pixel point being not the pixel point corresponding to the original target object; and continuing to search for other original contour points based on the first original contour point.

5 . The method for detecting the collisions in the video according to claim 4 , wherein said continuing to search for the other original contour points based on the first original contour point comprises:
traversing pixel points along a first reference direction from any one of the pixel points, among pixel points in eight neighborhoods of the first original contour point, and determining a currently traversed pixel point satisfying the contour condition as an end point; traversing the pixel points along a second reference direction from a first pixel point, among the pixel points in the eight neighborhoods of the first original contour point, and determining a currently traversed pixel point satisfying the contour condition as a second original contour point, wherein the first pixel point is a pixel point reached by moving along the second reference direction from the end point, among the pixel points in the eight neighborhoods of the first original contour point; and performing the following processes cyclically: traversing the pixel points along the second reference direction from a second pixel point, among the pixel points in the eight neighborhoods of the currently determined original contour point, determining a currently traversed pixel point satisfying the contour condition as a next original contour point, and stopping the cycles in response to the determined next original contour point being the end point, wherein the second pixel point is a pixel point reached by moving along the second reference direction from a previous original contour point, among the pixel points in the eight neighborhoods of the currently determined original contour point, wherein the first reference direction and the second reference direction are a clockwise or counterclockwise direction, and the second reference direction is different from the first reference direction.

6 . The method for detecting the collisions in the video according to claim 4 , wherein said continuing to search for the other original contour points based on the first original contour point comprises:
traversing pixel points along a first reference direction from any one of the pixel points, among pixel points in eight neighborhoods of the first original contour point, and determining a currently traversed pixel point, satisfying the contour condition, as a second original contour point, wherein the first reference direction is a clockwise or counterclockwise direction; and performing the following processes cyclically: traversing the pixel points along the first reference direction from a second pixel point, among the pixel points in the eight neighborhoods of the currently determined original contour point, determining a currently traversed pixel point, satisfying the contour condition, as a next original contour point, and stopping the cycles in response to the currently traversed pixel point being the first original contour point, wherein the second pixel point is a pixel point reached by moving along the first reference direction from a previous original contour point, among the pixel points in the eight neighborhoods of the currently determined original contour point.

7 . The method for detecting the collisions in the video according to claim 1 , wherein said identifying the target contour points corresponding to the original target object in the video picture comprises:
binarizing the video picture to acquire a mask, wherein pixel values of pixel points corresponding to the original target object in the mask are defined as first pixel values and pixel values of other pixel points are defined as second pixel values; and identifying the target contour points among the pixel points of the mask.

8 . The method for detecting the collisions in the video according to claim 1 , wherein said creating one second bounding box based on each two adjacent target contour points of the original target object comprises:
determining a distance between two adjacent target contour points as a first side length of a rectangle, and determining a reference distance as a second side length of the rectangle; and creating one second bounding box a rectangular shape based on the first side length and the second side length, wherein the two adjacent target contour points are respectively disposed at center positions of opposite sides of the second bounding box.

9 . The method for detecting the collisions in the video according to claim 1 , wherein said acquiring the first bounding box of the dynamic virtual elements comprises:
identifying reference contour points corresponding to the dynamic virtual elements, wherein the reference contour points are positioned on a contour line of the dynamic virtual elements; creating one first bounding box based on each two adjacent reference contour points; said detecting the collisions between the first bounding boxes and the second bounding boxes comprises:
detecting the collisions between any one of the first bounding boxes and any one of the second bounding boxes.

10 . The method for detecting the collisions in the video according to claim 1 , wherein said detecting the collisions between the first bounding boxes and the second bounding boxes comprises:
determining a direction perpendicular to a direction of each side of each of the first bounding boxes, and a direction perpendicular to a direction of each side of each of the second bounding boxes; projecting the first bounding boxes and the second bounding boxes into each of the determined directions; and determining that the first bound

Assignees

• Beijing Dajia Internet Information Tech Co Ltd

Inventors

• Xiao Yi

Classifications

• G06V10/469

  Contour-based spatial representations, e.g. vector-coding · CPC title

• G06V20/44 Primary

  Event detection · CPC title

• G06V10/25

  Determination of region of interest [ROI] or a volume of interest [VOI] · CPC title

• G06V2201/07

  Target detection · CPC title

• G06T7/181

  involving edge growing; involving edge linking · CPC title