CN-117253255-B - Palm correction method based on non-homologous binocular

CN117253255BCN 117253255 BCN117253255 BCN 117253255BCN-117253255-B

Abstract

A non-homologous binocular palm correction method comprises the following steps of S1, respectively detecting a palm of a first image I ir and a second image I rgb to respectively obtain a palm region ROI ir of the first image I ir and a palm region ROI rgb of the second image I rgb , wherein the first image and the second image are non-homologous images, S2, estimating the palm orientation of the first image I ir or the second image I rgb by adopting a deep learning method, S3, if the palm orientation does not meet preset requirements, re-executing the step S1, and S4, correcting the first image and the second image according to the palm orientation. According to the invention, the non-homologous binocular system formed by the first camera and the second camera is utilized to directly process the first image and the second image, so that the quality of image acquisition is improved, and the palm orientation is utilized to correct, so that the palm with a larger angle can be processed, and meanwhile, the correction accuracy is greatly improved.

Inventors

HUANG LONGXIANG
GUO HUWEI
HUANG JIAWEN
WANG BO
ZHU LI
LV FANGLU

Assignees

深圳市光鉴科技有限公司

Dates

Publication Date: 20260512
Application Date: 20220609

Claims (9)

1. A non-homologous binocular palm correction method, comprising the steps of: step S1, for a first image And a second image Respectively detecting palms to respectively obtain the first images Is in the palm region of (2) And the second image Is in the palm region of (2) Wherein the first image and the second image are non-homologous images; step S2, adopting a deep learning method to carry out the first image Or the second image Is estimated from the palm orientation; Step S3, if the palm orientation does not meet the preset requirement, the step S1 is executed again; step S4, aiming at the first image according to the palm orientation And the second image Correcting; The step S4 includes: step S41, dividing an image into a first stretching area, a first compression area and a first patching area on the second image according to the palm orientation, the gray level value and the distance from the edge, and copying the areas onto the first image, wherein the first image is a texture image of the palm, and the second image is a vein image of the palm; step S42, on the first image, fine tuning the first stretching region, the first compression region and the first repairing region according to texture characteristics to respectively obtain a second stretching region, a second compression region and a second repairing region, and copying the second stretching region, the second compression region and the second repairing region to the second image; and S43, stretching the second stretching region with the same amplitude on the first image and the second image, compressing the second compression region with the same amplitude, and repairing the second repairing regions respectively.
2. The non-homologous binocular palm correcting method according to claim 1, further comprising, prior to step S1: S0, respectively carrying out distortion correction on the original first image and the original second image, and carrying out polar correction to obtain a corrected first image And a corrected second image 。
3. The non-homologous binocular palm correcting method of claim 1, wherein in step S1, the first images are also respectively corrected And the second image According to the palm area And the palm area Image segmentation is performed and non-palm regions are zeroed out.
4. A non-homologous binocular palm correcting method according to claim 1, wherein in step S2, only the first image is corrected And the second image One of which performs palm orientation estimation.
5. The non-homologous binocular palm correcting method according to claim 1, wherein in step S2 the first image is corrected And the second image And respectively performing deep learning to respectively obtain palm orientation estimation, and obtaining the palm orientation through cross verification.
6. The non-homologous binocular palm correcting method according to claim 1, wherein in step S2 the first image is processed And the second image And simultaneously, deep learning is carried out to obtain the palm orientation.
7. The non-homologous binocular palm correcting method according to claim 1, wherein the step S2 comprises: Step S21 of acquiring the first images respectively And the second image Key points on the table; step S22, calculating three-dimensional space information of the key points by utilizing parallax; and S23, calculating the palm orientation according to the three-dimensional space information.
8. A non-homologous binocular-based palm correcting device, comprising: A processor; a memory module having stored therein executable instructions of the processor; Wherein the processor is configured to perform the steps of the non-homologous binocular palm correction method of any one of claims 1 to 7 via execution of the executable instructions.
9. A computer-readable storage medium storing a program, characterized in that the program when executed implements the steps of a non-homologous binocular palm correction method according to any one of claims 1 to 7.

Description

Palm correction method based on non-homologous binocular Technical Field The invention relates to the field of palm recognition cameras, in particular to a non-homologous binocular palm correction method. Background Because the lines of the palm of the person are more stable, the palm brushing identification is a biological identification technology which is more stable and safer than the face brushing identification, and the person identity can be identified through the palm brushing, so that the palm brushing identification device is used in the fields of security check, payment, identity identification and the like. The brush palm identification is a technology with wide application prospect. The existing non-contact palm recognition equipment is divided into two types, namely palm print recognition function and palm vein recognition function, wherein the other type is used for recognizing palm vein information, so that the system is greatly simplified, the palm vein information can be well ensured by using lower resolution ratio due to the fact that palm veins belong to coarse granularity characteristics, but the palm vein recognition precision is correspondingly limited, in addition, the palm vein imaging quality is greatly influenced by a light source, palm distance and gesture, and unstable images are easy to collect. Especially, in the prior art, the image binarization processing mode is adopted to carry out contrast processing with the original image, so that the quality of some data is improved, but the commercial application requirements can not be met. Disclosure of Invention Therefore, the invention utilizes the non-homologous binocular system formed by the first camera and the second camera to directly process the first image and the second image, improves the quality of image acquisition, and corrects the image by utilizing the palm orientation, so that the palm with a larger angle can be processed, and the correction accuracy is greatly improved. In a first aspect, the present invention provides a non-homologous binocular palm correction method, which is characterized by comprising the following steps: Step S1, detecting palms of a first image I ir and a second image I rgb respectively to obtain a palms region ROI ir of the first image I ir and a palms region ROI rgb of the second image I rgb respectively, wherein the first image and the second image are non-homologous images; Step S2, estimating the palm orientation of the first image I ir or the second image I rgb by adopting a deep learning method; Step S3, if the palm orientation does not meet the preset requirement, the step S1 is executed again; And S4, correcting the first image I ir and the second image I rgb according to the palm orientation. Optionally, the non-homologous binocular palm correction method is characterized by further comprising, before the step S1: And S0, respectively carrying out distortion correction on the original first image and the original second image, and carrying out polar line correction to obtain a corrected first image I ir and a corrected second image I rgb. Optionally, in the step S1, the method further performs image segmentation on the first image I ir and the second image I rgb according to the palm region ROI ir and the palm region ROI rgb, and zeroes out the non-palm region. Optionally, the non-homologous binocular palm correcting method is characterized in that in the step S2, palm orientation estimation is performed on only one of the first image I ir and the second image I rgb. Optionally, in the step S2, the first image I ir and the second image I rgb are respectively subjected to deep learning to obtain palm orientation estimation, and the palm orientation is obtained through cross verification. Optionally, in the step S2, the method is characterized in that the first image I ir and the second image I rgb are subjected to deep learning at the same time to obtain the palm orientation. Optionally, the non-homologous binocular palm correction method is characterized in that the step S2 includes: Step S21, key points on the first image I ir and the second image I rgb are respectively acquired; step S22, calculating three-dimensional space information of the key points by utilizing parallax; and S23, calculating the palm orientation according to the three-dimensional space information. Optionally, the non-homologous binocular palm correction method is characterized in that the step S4 includes: step S41, dividing an image into a first stretching area, a first compression area and a first patching area on the second image according to the palm orientation, the gray level value and the distance from the edge, and copying the areas onto the first image, wherein the first image is a texture image of the palm, and the second image is a vein image of the palm; step S42, on the first image, fine tuning the first stretching region, the first compression region and the first repairing region according to