Methods and systems for video processing

Patent metadata

Field	Value
Publication number	US-12586339-B2
Application number	US-202318163247-A
Country	US
Kind code	B2
Filing date	Feb 1, 2023
Priority date	Aug 19, 2020
Publication date	Mar 24, 2026
Grant date	Mar 24, 2026

Abstract

A method and a system for video processing, including: determining a target bounding box from an image frame in a video; determining a classification result of a subject in the target bounding box by processing the target bounding box using a recognition model; and determining one or more target image frames to be saved in the video based on the classification result.

First claim

What is claimed is:

1 . A method for video processing, comprising:
determining a target bounding box from an image frame in a video; obtaining at least one candidate bounding box corresponding to at least one moving target from the image frame; obtaining a motion detection tracking sequence of each target frame in the video; and screening a candidate bounding box corresponding to a motion detection tracking sequence meeting a first preset condition, comprising:
obtaining a minimum value of at least one number of intelligent detection frames of the at least one moving target in at least one historical image frame, the number of intelligent detection frame indicating a count of times that at least one associated bounding box corresponding to the candidate bounding box from the at least one historical image frame in the video is processed by a recognition model;
determining whether the number of the minimum value is greater than one;
selecting a candidate bounding box corresponding to the minimum value as the target bounding box if the number of minimum values is not greater than one;
obtaining a maximum value of background difference values of the at least one candidate bounding box corresponding to the minimum value if the number of minimum values is greater than one, the background difference value indicating an absolute value of a pixel difference between a current frame image and a background image; and
selecting a candidate bounding box corresponding to the maximum value of background difference values as the target bounding box;
determining a classification result of a subject in the target bounding box by processing the target bounding box using the recognition model; and determining one or more target image frames to be saved in the video based on the classification result.

2 . The method of claim 1 , wherein the obtaining the at least one candidate bounding box from the image frame includes:
obtaining the background image of the image frame; determining a foreground image of the image frame based on the background image and the image frame; determining the at least one candidate bounding box based on the foreground image.

3 . The method of claim 1 , wherein the determining the one or more target image frames to be saved in the video based on the classification result, includes:
determining the one or more target image frames to be saved based on the image frame corresponding to determining that the classification result of the subject includes a specific type, and that the target bounding box satisfies a second condition.

4 . The method of claim 3 , wherein the second condition is related to a count of occurrences of the target bounding box and at least one associated bounding box corresponding to the target bounding box in the video.

5 . The method of claim 1 , further including:
determining a recognition bounding box by processing the target bounding box using the recognition model, wherein the recognition bounding box is configured to mark the subject in the target bounding box.

6 . A system for video processing, comprising:
at least one storage device including a set of instructions; and at least one processor configured to communicate with the at least one storage device, wherein when executing the set of instructions, the at least one processor is configured to direct the system to perform operations including: determining a target bounding box from an image frame in a video; obtaining at least one candidate bounding box corresponding to at least one moving target from the image frame; obtaining a motion detection tracking sequence of each target frame in the video; and screening a candidate bounding box corresponding to a motion detection tracking sequence meeting a first preset condition, comprising:
obtaining a minimum value of at least one number of intelligent detection frames of the at least one moving target in at least one historical image frame, the number of intelligent detection frame indicating a count of times that at least one associated bounding box corresponding to the candidate bounding box from the at least one historical image frame in the video is processed by a recognition model;
determining whether the number of the minimum value is greater than one;
selecting a candidate bounding box corresponding to the minimum value as the target bounding box if the number of minimum values is not greater than one;
obtaining a maximum value of background difference values of the at least one candidate bounding box corresponding to the minimum value if the number of minimum values is greater than one, the background difference value indicating an absolute value of a pixel difference between a current frame image and a background image; and
selecting a candidate bounding box corresponding to the maximum value of background difference values as the target bounding box;
determining a classification result of a subject in the target bounding box by processing the target bounding box using the recognition model; and determining one or more target image frames to be saved in the video based on the classification result.

7 . The system of claim 6 , wherein the obtaining the at least one candidate bounding box from the image frame includes:
obtaining the background image of the image frame; determining a foreground image of the image frame based on the background image and the image frame; determining the at least one candidate bounding box based on the foreground image.

8 . The system of claim 7 , wherein the obtaining the background image of the image frame includes:
determining the background image by processing the video using a Gaussian background modeling algorithm.

9 . The system of claim 6 , wherein the determining the one or more target image frames to be saved in the video based on the classification result, includes:
determining the one or more target image frames to be saved based on the image frame corresponding to determining that the classification result of the subject includes a specific type, and that the target bounding box satisfies a second condition.

10 . The system of claim 9 , wherein the second condition is related to a count of occurrences of the target bounding box and at least one associated bounding box corresponding to the target bounding box in the video.

11 . The system of claim 6 , further including:
determining a recognition bounding box by processing the target bounding box using the recognition model, wherein the recognition bounding box is configured to mark the subject in the target bounding box.

12 . A non-transitory computer readable medium, comprising executable instructions that, when executed by at least one processor, direct the at least one processor to perform a method, the method comprising:
determining a target bounding box from an image frame in a video; obtaining at least one candidate bounding box corresponding to at least one moving target from the image frame; obtaining a motion detection tracking sequence of each target frame in the video; and screening a candidate bounding box corresponding to a motion detection tracking sequence meeting a first preset condition, comprising:
obtaining a minimum value of at least one number of intelligent detection frames of the at least one moving target in at least one historical image frame, the number of intelligent detection frame indicating a count of times that at least one associated bounding box corresponding to the candidate bounding box from the at least one historical image frame in the video is processed by a recognition model;
determining whether the number of the minimum value is grea

Assignees

• Zhejiang Dahua Technology Co

Inventors

• Cheng Deqiang
• Zheng Chunhuang
• Wu Jianfeng

Classifications

• G06T2210/12

  Bounding box · CPC title

• G06V10/764

  using classification, e.g. of video objects · CPC title

• G06V10/751

  Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching · CPC title

• G06T7/194

  involving foreground-background segmentation · CPC title

• H04N7/188

  Capturing isolated or intermittent images triggered by the occurrence of a predetermined event, e.g. an object reaching a predetermined position (signal generation from motion picture films H04N5/253) · CPC title