Systems and methods for object recognition

Patent metadata

Field	Value
Publication number	US-2023260263-A1
Application number	US-202318305349-A
Country	US
Kind code	A1
Filing date	Apr 23, 2023
Priority date	Dec 3, 2020
Publication date	Aug 17, 2023
Grant date	—

Abstract

The present disclosure relates to systems and methods for object recognition. The systems may obtain an image associated with an object. The systems may determine, using a recognition model, an initial recognition result of the object based on the image. The initial recognition result may include a plurality of initial recognition values corresponding to a plurality of predetermined categories respectively. Each of the plurality of initial recognition values may indicate an initial probability that the object corresponds to a corresponding predetermined category. In response to determining that the initial recognition result satisfies a preset condition, for each of the plurality of predetermined categories, the systems may obtain a reference probability corresponding to the predetermined category. The systems may determine a target recognition result of the object by adjusting the plurality of initial recognition values based on a plurality of reference probabilities corresponding to the plurality of predetermined categories respectively. The target recognition result may include a plurality of target recognition values corresponding to the plurality of predetermined categories respectively. Each of the plurality of target recognition values may indicate a target probability that the object corresponds to a corresponding predetermined category.

First claim

1 . A system, comprising:
at least one storage device including a set of instructions; and at least one processor in communication with the at least one storage device, wherein when executing the set of instructions, the at least one processor is directed to perform operations including:
obtaining an image associated with an object;
determining, using a recognition model, an initial recognition result of the object based on the image, the initial recognition result including a plurality of initial recognition values corresponding to a plurality of predetermined categories respectively, each of the plurality of initial recognition values indicating an initial probability that the object corresponds to a corresponding predetermined category;
in response to determining that the initial recognition result satisfies a preset condition, for each of the plurality of predetermined categories, obtaining a reference probability corresponding to the predetermined category; and
determining a target recognition result of the object by adjusting the plurality of initial recognition values based on a plurality of reference probabilities corresponding to the plurality of predetermined categories respectively, the target recognition result including a plurality of target recognition values corresponding to the plurality of predetermined categories respectively, each of the plurality of target recognition values indicating a target probability that the object corresponds to a corresponding predetermined category.

2 . The system of claim 1 , wherein the preset condition includes that the plurality of initial recognition values in the initial recognition result are less than a predetermined threshold.

3 . The system of claim 1 , wherein the obtaining the reference probability corresponding to the predetermined category includes:
obtaining, from a reference image set, a plurality of reference images; and determining the reference probability corresponding to the predetermined category based on a plurality of reference recognition results corresponding to the plurality of reference images respectively, for each of the plurality of reference images, the reference recognition result including a plurality of reference recognition values corresponding to the plurality of predetermined categories respectively.

4 . The system of claim 3 , wherein each reference image in the reference image set corresponds to one of the plurality of predetermined categories, counts of reference images corresponding to different predetermined categories being substantially the same.

5 . The system of claim 3 , wherein for each of the plurality of reference images, a difference or a distance between a corresponding reference recognition result and the initial recognition result is less than a difference threshold or a distance threshold.

6 . The system of claim 3 , wherein for each of the plurality of reference images, a similarity between the reference image and the image is larger than a similarity threshold.

7 . The system of claim 3 , wherein the determining the reference probability corresponding to the predetermined category based on the reference recognition results corresponding to the plurality of reference images includes:
for each of the plurality of reference recognition results corresponding to the plurality of reference images respectively, determining a maximum reference recognition value in the reference recognition result; determining a first sum of maximum reference recognition values each of which corresponds to the predetermined category; determining a second sum of maximum reference recognition values corresponding to the plurality of reference recognition results respectively; and determining the reference probability corresponding to the predetermined category based on the first sum and the second sum.

8 . The system of claim 1 , wherein the determining the target recognition result of the object by adjusting the plurality of initial recognition values based on the plurality of reference probabilities corresponding to the plurality of predetermined categories respectively includes:
for each of the plurality of predetermined categories, determining a target recognition value corresponding to the predetermined category based on a product of an initial recognition value and a reference probability corresponding to the predetermined category.

9 . The system of claim 1 , wherein the determining the target recognition result of the object by adjusting the plurality of initial recognition values based on the plurality of reference probabilities corresponding to the plurality of predetermined categories respectively includes:
for each of the plurality of predetermined categories, determining a product of an initial recognition value and a reference probability corresponding to the predetermined category; determining a sum of products corresponding to the plurality of predetermined categories respectively; and for each of the plurality of predetermined categories, determining a target recognition value corresponding to the predetermined category based on the product corresponding to the predetermined category and the sum of the products corresponding to the plurality of predetermined categories respectively.

10 . The system of claim 1 , wherein the operations further include:
designating a predetermined category corresponding to a largest target recognition value in the target recognition result as a target category of the object.

11 . A method implemented on a computing device including at least one processor, at least one storage medium, and a communication platform connected to a network, the method comprising:
obtaining an image associated with an object; determining, using a recognition model, an initial recognition result of the object based on the image, the initial recognition result including a plurality of initial recognition values corresponding to a plurality of predetermined categories respectively, each of the plurality of initial recognition values indicating an initial probability that the object corresponds to a corresponding predetermined category; in response to determining that the initial recognition result satisfies a preset condition, for each of the plurality of predetermined categories, obtaining a reference probability corresponding to the predetermined category; and determining a target recognition result of the object by adjusting the plurality of initial recognition values based on a plurality of reference probabilities corresponding to the plurality of predetermined categories respectively, the target recognition result including a plurality of target recognition values corresponding to the plurality of predetermined categories respectively, each of the plurality of target recognition values indicating a target probability that the object corresponds to a corresponding predetermined category.

12 . The method of claim 11 , wherein the preset condition includes that the plurality of initial recognition values in the initial recognition result are less than a predetermined threshold.

13 . The method of claim 11 , wherein the obtaining the reference probability corresponding to the predetermined category includes:
obtaining, from a reference image set, a plurality of reference images; and determining the reference probability corresponding to the predetermined category based on a plurality of reference recognition results corresponding to the plurality of reference images respectively, for each of the plurality of reference images, the reference recognition result including a plurality of reference recognition values corresponding to

Assignees

• Zhejiang Dahua Technology Co

Inventors

• Yang Yongtao
• Zhang Xingming
• Pan Huadong
• Tang Bangjie
• Su Hui
• Yin Jun

Classifications

• G06V10/776 Primary

  Validation; Performance evaluation · CPC title

• G06V20/52

  Surveillance or monitoring of activities, e.g. for recognising suspicious objects (recognising microscopic objects G06V20/69) · CPC title

• G06V10/761

  Proximity, similarity or dissimilarity measures · CPC title

• G06V40/103 Primary

  Static body considered as a whole, e.g. static pedestrian or occupant recognition · CPC title

• G06F18/24

  Classification techniques · CPC title