Identity recognition method, computer apparatus, non-transitory computer-readable storage medium

Patent metadata

Field	Value
Publication number	US-12542004-B2
Application number	US-202318210628-A
Country	US
Kind code	B2
Filing date	Jun 15, 2023
Priority date	Dec 15, 2020
Publication date	Feb 3, 2026
Grant date	Feb 3, 2026

Abstract

An identity recognition method, a computer apparatus, a non-transitory computer-readable storage medium are provided. The method includes: acquiring a to-be-identified image including a test object; extracting multi-modal features of the test object from the to-be-identified image based on a pre-established feature extraction model, the multi-modal features including at least one face feature and one human body feature; comparing the multi-modal features to modal features included by multi-modal feature sets in a pre-established feature registry; determining a target multi-modal set corresponding to the highest comparison score from the multi-modal feature sets, wherein each of the multi-modal feature sets incorporates at least one of the face feature and the human body feature; and determining an identity information corresponding to a target face feature included by the target multi-modality set, and determining the identity information corresponding to the target face feature as an identity information of the test object.

First claim

What is claimed is:

1 . An identity recognition method, comprising:
acquiring a to-be-identified image comprising a test object; extracting multi-modal features of the test object from the to-be-identified image based on a pre-established feature extraction model, the multi-modal features comprising at least one face feature and one human body feature; comparing the multi-modal features to modal features comprised by each of at least one of multi-modal feature sets in a pre-established feature registry; and determining a target multi-modal set corresponding to the highest comparison score from the at least one of the multi-modal feature sets, wherein each of the at least one of the multi-modal feature sets incorporates at least one of the face feature and the human body feature; and determining an identity information corresponding to a target face feature comprised by the target multi-modality set, and determining the identity information corresponding to the target face feature as an identity information of the test object; before the comparing the multi-modal features to the modal features comprised by each of at least one of the multi-modal feature sets in the pre-established feature registry, the method further comprising: extracting face features corresponding to each of at least one of sample images by a pre-established face recognition feature extraction model, in an initial registration stage; determining a human body feature corresponding to each of the sample images based on the human body feature of the multi-modal features of the test object; and establishing a feature registry comprising at least one of the multi-modal feature sets of the at least one of the sample images, wherein each of the sample images in the feature registry corresponds to one multi-modal feature set comprising a corresponding face feature and a corresponding human body feature of the at least one of the multi-modal feature sets.

2 . The method according to claim 1 , wherein the comparing the multi-modal features to modal features comprised by each of at least one of the multi-modal feature sets in the pre-established feature registry; and determining the target multi-modal set corresponding to the highest comparison score from the at least one of the multi-modal feature sets, comprises:
comparing each of modal features of the multi-modal features to the modal feature comprised by each of the at least one of the multi-modal feature sets in the feature registry, to obtain a comparison score of each of the modal features; determining a weighted sum of the comparison score between each of the multi-modal feature sets and each of the modal features of the multi-modal features, with a set as a unit; and determining the target multi-modal set corresponding to the highest comparison score based on the weighted sum.

3 . The method according to claim 2 , wherein the weighted sum of the comparison score between each of the multi-modal feature sets and each of the modal features of the multi-modal features is determined based on the following formula:
S=a*d ( pf face , gf face ( t ))+(1−a) *d ( pf body , gf body ( t )) wherein S represents the weighted sum of the comparison score between each of the multi-modal feature sets and each of the modal features of the multi-modal features, pf face represents the face feature of the multi-modal features, pf body represents the human body feature of the multi-modal feature, gf face (t) represents the face feature of each of the multi-modal feature sets, gf body (t) represents the human body feature of each of the multi-modal feature sets, d represents a similarity metric function, and a represents a preset weight.

4 . The method according to claim 1 , wherein when the highest comparison score is greater than a preset threshold, after the determining the target multi-modal set corresponding to the highest comparison score from the at least one of the multi-modal feature sets, the method further comprises:
performing a weighted calculation on the face feature in the target multi-modal set and the face feature of the multi-modal features to obtain a weighted face feature; obtaining a first multi-modal feature set comprising the weighted face features and the human body feature of the multi-modal features; and entering the first multi-modal feature set into the feature registry to obtain an updated feature registry.

5 . The method according to claim 1 , wherein when the highest comparison score is greater than a preset threshold, after the determining the target multi-modal set corresponding to the highest comparison score from the at least one of the multi-modal feature sets, the method further comprises:
performing a weighted calculation on the face feature in the target multi-modal set and the face feature of the multi-modal features to obtain a weighted face feature; performing a weighted calculation on the human body feature in the target multi-modal set and the human body feature of the multi-modal features to obtain a weighted human body feature; obtaining a second multi-modal set comprising the weighted face feature and the weighted human body feature; and entering the second multi-modal set into the feature registry to obtain an updated feature registry.

6 . The method according to claim 4 , wherein the weighted face feature is obtained based on the following formula:
gf face ( t+ 1)=(1 −S )* gf face ( t ) +S*pf face wherein S represents the highest comparison score, gf face (t) represents the face feature in the target multi-modal set, pf face represents the face feature of the multi-modal features, gf face (t+1) represents the weighted face feature.

7 . The method according to claim 5 , wherein the weighted human body feature is obtained based on the following formula:
gf body ( t+ 1)=(1− S ) gf body ( t )+ S*pf body
wherein S represents the highest comparison score, gf body (t) represents the human body feature in the target multi-modal set, pf body represents the human body feature of the multi-modal features, gf body (t+1) represents the weighted human body feature.

8 . A computer apparatus, comprising a processor, wherein when executing a computer program stored in a memory, the processor is configured to implement:
acquiring a to-be-identified image comprising a test object; extracting multi-modal features of the test object from the to-be-identified image based on a pre-established feature extraction model, the multi-modal features comprising at least one face feature and one human body feature; comparing the multi-modal features to modal features comprised by each of at least one of multi-modal feature sets in a pre-established feature registry; and determining a target multi-modal set corresponding to the highest comparison score from the at least one of the multi-modal feature sets, wherein each of the at least one of the multi-modal feature sets incorporates at least one of the face feature and the human body feature; and determining an identity information corresponding to a target face feature comprised by the target multi-modality set, and determining the identity information corresponding to the target face feature as an identity information of the test object; before the comparing the multi-modal features to the modal features comprised by each of at least one of the multi-modal feature sets in the pre-established feature registry, the method further comprising: extracting face features corresponding to each of at least one of sample images by a pre-established face recognition feature extraction model, in an initial registration stage; determining a human body feature corresponding to each of the sample images based on the huma

Assignees

• Zhejiang Dahua Technology Co

Inventors

• Hao Jingsong
• Di Dening
• Zhu Shulei

Classifications

• G06V40/50

  Maintenance of biometric data or enrolment thereof · CPC title

• G06V40/172

  Classification, e.g. identification · CPC title

• G06V40/171

  Local features and components; Facial parts (eye characteristics G06V40/18); Occluding parts, e.g. glasses; Geometrical relationships · CPC title

• G06V40/70 Primary

  Multimodal biometrics, e.g. combining information from different biometric modalities · CPC title

• G06V10/806

  of extracted features · CPC title