CN117649688A - Near infrared image-based image recognition method, system and storage medium - Google Patents
Near infrared image-based image recognition method, system and storage medium Download PDFInfo
- Publication number
- CN117649688A CN117649688A CN202311523764.4A CN202311523764A CN117649688A CN 117649688 A CN117649688 A CN 117649688A CN 202311523764 A CN202311523764 A CN 202311523764A CN 117649688 A CN117649688 A CN 117649688A
- Authority
- CN
- China
- Prior art keywords
- image
- data
- target
- preset
- contour
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 53
- 210000003462 vein Anatomy 0.000 claims abstract description 56
- 238000012545 processing Methods 0.000 claims abstract description 49
- 238000013136 deep learning model Methods 0.000 claims abstract description 13
- 238000000701 chemical imaging Methods 0.000 claims abstract description 12
- 238000013441 quality evaluation Methods 0.000 claims description 19
- 230000009467 reduction Effects 0.000 claims description 19
- 238000001303 quality assessment method Methods 0.000 claims description 18
- 238000012549 training Methods 0.000 claims description 16
- 238000013519 translation Methods 0.000 claims description 11
- 238000003708 edge detection Methods 0.000 claims description 6
- 230000008447 perception Effects 0.000 claims description 5
- 238000002329 infrared spectrum Methods 0.000 description 18
- 238000004458 analytical method Methods 0.000 description 14
- 238000005516 engineering process Methods 0.000 description 11
- 230000000694 effects Effects 0.000 description 9
- 238000003384 imaging method Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 7
- 230000008569 process Effects 0.000 description 7
- 238000004364 calculation method Methods 0.000 description 6
- 238000001514 detection method Methods 0.000 description 5
- 230000008859 change Effects 0.000 description 4
- 239000000463 material Substances 0.000 description 4
- 230000003287 optical effect Effects 0.000 description 4
- 230000003595 spectral effect Effects 0.000 description 4
- 239000008280 blood Substances 0.000 description 3
- 210000004369 blood Anatomy 0.000 description 3
- 210000004204 blood vessel Anatomy 0.000 description 3
- 238000003745 diagnosis Methods 0.000 description 3
- 238000009499 grossing Methods 0.000 description 3
- 238000005286 illumination Methods 0.000 description 3
- 239000012535 impurity Substances 0.000 description 3
- 239000011159 matrix material Substances 0.000 description 3
- 238000012544 monitoring process Methods 0.000 description 3
- 238000011282 treatment Methods 0.000 description 3
- 102000001554 Hemoglobins Human genes 0.000 description 2
- 108010054147 Hemoglobins Proteins 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 239000003814 drug Substances 0.000 description 2
- 229940079593 drug Drugs 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 230000004927 fusion Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012827 research and development Methods 0.000 description 2
- 230000011218 segmentation Effects 0.000 description 2
- 210000004872 soft tissue Anatomy 0.000 description 2
- 238000001228 spectrum Methods 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 210000001519 tissue Anatomy 0.000 description 2
- 238000001429 visible spectrum Methods 0.000 description 2
- 238000010521 absorption reaction Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000017531 blood circulation Effects 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 239000011248 coating agent Substances 0.000 description 1
- 238000000576 coating method Methods 0.000 description 1
- 238000002059 diagnostic imaging Methods 0.000 description 1
- 238000005315 distribution function Methods 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 230000001815 facial effect Effects 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 238000010191 image analysis Methods 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000013077 scoring method Methods 0.000 description 1
- 230000007723 transport mechanism Effects 0.000 description 1
- 230000006438 vascular health Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/14—Vascular patterns
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Vascular Medicine (AREA)
- Human Computer Interaction (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses an image recognition method, an image recognition system and a storage medium based on near infrared images, which are applied to the technical field of image processing, can realize vein recognition of a target object and effectively improve the accuracy and reliability of vein recognition. The method comprises the following steps: acquiring a plurality of groups of target object images with different wavelengths through a preset near infrared multispectral imaging camera; scoring and voting the target object image by a preset quality scoring algorithm to obtain a target wave band image; performing background removal processing on the target wave band image through a preset deep learning model to obtain foreground image data; calculating the outline of the target object according to the foreground image data to obtain target outline data; calculating according to foreground image data by a preset Gaussian difference method to obtain a difference image; and extracting pixel data meeting a preset pixel threshold value in the contour range according to the differential image and the target contour data, so as to obtain a target recognition result of the vein to be recognized according to the pixel data.
Description
Technical Field
The present invention relates to the field of image processing technologies, and in particular, to an image recognition method, system and storage medium based on near infrared images.
Background
A near infrared spectrum camera is an imaging device capable of sensing near infrared bands, which operates in the near infrared region between the visible and infrared spectrums. The near infrared spectrum camera has important application in the fields of agriculture, food industry, drug research and development, medical diagnosis, environmental monitoring, material analysis, industrial detection and the like, and can help to realize nondestructive analysis and imaging. In addition, the near infrared vein recognition technology is a biological recognition technology that scans and analyzes a vein network of the back of the palm of a human body using light in a near infrared spectrum range. Since light in the near infrared spectral range is able to penetrate skin and soft tissue to some extent, but is absorbed by hemoglobin in blood, this property allows near infrared light to be absorbed by the venous network, thereby forming a palm back venous image. However, in the related art, the near-infrared palmar dorsal venous blood vessel recognition technology also faces a plurality of challenges, such as larger influence of illumination conditions, large physiological change of palms, insufficient sample number, high dependence of hardware equipment and the like. Therefore, the above technical problems need to be solved.
Disclosure of Invention
In order to solve at least one of the above technical problems, the present invention provides an image recognition method, system and storage medium based on near infrared images, which can realize vein recognition of a target object and effectively improve the accuracy and reliability of vein recognition.
In one aspect, an embodiment of the present invention provides an image recognition method based on near infrared images, including the following steps:
acquiring a plurality of groups of target object images with different wavelengths through a preset near infrared multispectral imaging camera; the target object image comprises veins to be identified;
scoring and voting the target object image through a preset quality scoring algorithm to obtain a target wave band image;
performing background removal processing on the target band image through a preset deep learning model to obtain foreground image data;
calculating the outline of the target object according to the foreground image data to obtain target outline data;
calculating according to the foreground image data by a preset Gaussian difference method to obtain a difference image;
and extracting pixel data meeting a preset pixel threshold value in a contour range according to the differential image and the target contour data, so as to obtain a target recognition result of the vein to be recognized according to the pixel data.
According to some embodiments of the invention, the preset quality scoring algorithm includes a contrast language-image pre-training image quality assessment algorithm, a natural image quality assessment algorithm, and a no-reference image quality assessment algorithm;
scoring and voting the target object image through a preset quality scoring algorithm to obtain a target band image, wherein the scoring and voting comprise the following steps:
performing first quality scoring on the target object image through the non-reference image quality assessment algorithm to obtain first scoring data;
performing a second quality scoring on the target object image through the natural image quality evaluation algorithm to obtain second scoring data;
performing third quality scoring on the target object image through the contrast language-image pre-training image quality evaluation algorithm to obtain third scoring data;
determining the target wave band image according to the first scoring data, the second scoring data and the third scoring data in a preset voting mode; the target wave band images comprise images with optimal image quality in the target object images with different wavelengths.
According to some embodiments of the present invention, the background removing processing is performed on the target band image by a preset deep learning model to obtain foreground image data, including:
Inputting the target band image into a preset image perception model to identify pixel points, and obtaining an image mask; the image mask comprises a foreground pixel point and a background pixel point;
performing preset image processing on the target wave band image according to the image mask to obtain the foreground image data; the preset image processing includes reserving pixel points corresponding to the foreground pixel points in the target band image, and removing the pixel points corresponding to the background pixel points.
According to some embodiments of the invention, the calculating the outline of the target object according to the foreground image data to obtain target outline data includes:
performing edge detection according to the foreground image data to obtain first contour data;
performing contour approximation processing through a Lamerger-Laplace-Puck algorithm according to the first contour data to obtain second contour data;
and carrying out contour reduction by a preset contour reduction algorithm according to the second contour data to obtain the target contour data.
According to some embodiments of the invention, the performing contour reduction according to the second contour data by a preset contour reduction algorithm to obtain the target contour data includes:
Traversing adjacent edges in the second contour data according to a preset direction to obtain a plurality of groups of adjacent edge data; the adjacent edge data comprises a first adjacent edge and a second adjacent edge;
taking a common point of the first adjacent side and the second adjacent side as an initial vertex, and calculating a rotation angle to obtain a common point translation direction;
translating the common point according to the translation direction of the common point to obtain a target vertex;
and connecting all the target vertexes to obtain the target contour data.
According to some embodiments of the invention, the calculating the difference image according to the foreground image data by a preset gaussian difference method includes:
carrying out Gaussian blur processing on the foreground image data to obtain Gaussian blur data; the Gaussian blur data comprises a first Gaussian blur result and a second Gaussian blur result, and the blur scales of the first Gaussian blur result and the second Gaussian blur result are different;
and carrying out difference according to the first Gaussian blur result and the second Gaussian blur result to obtain the difference image.
According to some embodiments of the invention, the extracting pixel data meeting a preset pixel threshold in a contour range according to the differential image and the target contour data to obtain a target recognition result of the vein to be recognized according to the pixel data includes:
Determining the preset pixel threshold value through an Ojin algorithm;
obtaining pixel points in the differential image, wherein the pixel value of the pixel points in the contour range of the target contour data is larger than the preset pixel threshold value, so as to obtain target pixel points;
outputting the target recognition result according to the target pixel point; wherein the target recognition result includes pixel coordinates or a fused image.
On the other hand, the embodiment of the invention also provides an image recognition system based on the near infrared image, which comprises:
the first module is used for acquiring a plurality of groups of target object images with different wavelengths through a preset near infrared multispectral imaging camera; the target object image comprises veins to be identified;
the second module is used for scoring and voting the target object image through a preset quality scoring algorithm to obtain a target wave band image;
the third module is used for carrying out background removal processing on the target band image through a preset deep learning model to obtain foreground image data;
a fourth module, configured to calculate a target object contour according to the foreground image data, to obtain target contour data;
a fifth module, configured to calculate, according to the foreground image data, a difference image by using a preset gaussian difference method;
And a sixth module, configured to extract, according to the differential image and the target contour data, pixel data that satisfies a preset pixel threshold value within a contour range, so as to obtain a target recognition result of the vein to be recognized according to the pixel data.
On the other hand, the embodiment of the invention also provides an image recognition system based on the near infrared image, which comprises:
at least one processor;
at least one memory for storing at least one program;
the at least one program, when executed by the at least one processor, causes the at least one processor to implement the near infrared image-based image recognition method as described in the above embodiments.
In another aspect, an embodiment of the present invention further provides a computer storage medium, in which a program executable by a processor is stored, where the program executable by the processor is used to implement the image recognition method based on near infrared images according to the above embodiment.
According to the image identification method based on the near infrared image, provided by the embodiment of the invention, the method has at least the following beneficial effects: according to the embodiment of the invention, a plurality of groups of target object images with different wavelengths are acquired through a preset near infrared multispectral imaging camera, and the target object images are scored and voted through a preset quality scoring algorithm so as to determine target wave band images. In the embodiment of the invention, the target object image comprises veins to be identified. Then, the embodiment of the invention carries out background removal processing on the target wave band image through the preset deep learning model so as to reduce the interference of a background area on vein recognition and reduce corresponding calculation cost, thereby obtaining the foreground image data. Then, the embodiment of the invention calculates the outline of the target object according to the foreground image data to obtain the target outline data. Meanwhile, according to the foreground image data, a corresponding difference image is obtained through calculation through a preset Gaussian difference method. Finally, according to the differential image and the target contour data, the pixel data meeting the preset pixel threshold value in the contour range are extracted, so that the target recognition result of the vein to be recognized is obtained according to the corresponding pixel data, namely, the recognition result of the relevant vein to be recognized in the target object image is obtained according to the pixel data meeting the preset pixel threshold value, vein recognition of the target object is achieved, and the accuracy and reliability of vein recognition are effectively improved.
Drawings
Fig. 1 is a flowchart of an image recognition method based on near infrared images according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of an image quality evaluation flow of a reference-free image quality evaluation algorithm according to an embodiment of the present invention;
FIG. 3 is a schematic flow chart of calculating an image quality score by a contrast language-image pre-training image quality assessment algorithm according to an embodiment of the present invention;
fig. 4 is a schematic diagram of contour reduction by a preset contour reduction algorithm according to an embodiment of the present invention;
fig. 5 is a schematic diagram of target contour data after contour reduction by a preset contour reduction algorithm according to an embodiment of the present invention;
fig. 6 is a schematic diagram of a differential image obtained by calculation by a preset gaussian differential method according to an embodiment of the present invention;
fig. 7 is a schematic diagram of a target recognition result of a palmar dorsal vein according to an embodiment of the present invention;
fig. 8 is a schematic diagram of an image recognition system based on near infrared images according to an embodiment of the present invention;
fig. 9 is a schematic block diagram of an image recognition system based on near infrared images according to an embodiment of the present invention.
Detailed Description
The embodiments described in the present application should not be construed as limitations on the present application, but rather as many other embodiments as possible without inventive faculty to those skilled in the art, are intended to be within the scope of the present application.
In the following description, reference is made to "some embodiments" which describe a subset of all possible embodiments, but it is to be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments and can be combined with one another without conflict.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of the present application only and is not intended to be limiting of the present application.
Before describing embodiments of the present application, related terms referred to in the present application will be first described.
Near infrared band: refers to the electromagnetic band of wavelengths ranging between visible and far infrared. Wherein the near infrared band has a wavelength range of about 700 nm to 2500 nm. Accordingly, light in the near infrared band has higher penetrability, and can penetrate some materials and tissues, so that the light can play an important role in some applications, such as the fields of medical imaging, biological detection, food quality detection, material analysis and the like.
Near infrared spectrum camera: the camera is used for acquiring and analyzing images in the near infrared spectrum range, and can receive and record radiation in the near infrared spectrum range through a sensor and an optical filter with characteristics.
Near infrared vein recognition technology: is a biometric technique that uses light in the near infrared spectral range to scan and analyze a venous network on the back of the human palm. Since light in the near infrared spectral range can penetrate skin and soft tissues to some extent, it is absorbed by hemoglobin in blood. This property allows near infrared light to be absorbed by the vein network, thereby forming a palmar dorsal vein image.
A near infrared spectrum camera is an imaging device capable of sensing near infrared bands, which operates in the near infrared region between the visible and infrared spectrums. The near infrared spectrum camera has important application in the fields of agriculture, food industry, drug research and development, medical diagnosis, environmental monitoring, material analysis, industrial detection and the like, and can help to realize nondestructive analysis and imaging. In addition, the near infrared vein recognition technology uses light in the near infrared spectrum to scan and analyze vein networks of the back of the palm of the human body, thereby forming a palm back vein image. However, in the related art, the near-infrared palmar dorsal venous blood vessel recognition technology also faces a plurality of challenges, such as larger influence of illumination conditions, large physiological change of palms, insufficient sample number, high dependence of hardware equipment and the like. Therefore, the above technical problems need to be solved.
An embodiment of the invention provides an image recognition method, an image recognition system and a storage medium based on near infrared images, which can realize vein recognition of a target object and effectively improve the accuracy and reliability of vein recognition. Referring to fig. 1, the method of the embodiment of the present invention includes, but is not limited to, step S110, step S120, step S130, step S140, step S150, and step S160.
Specifically, the method application process of the embodiment of the invention includes, but is not limited to, the following steps:
s110: acquiring a plurality of groups of target object images with different wavelengths through a preset near infrared multispectral imaging camera; the target object image comprises veins to be identified.
S120: and scoring and voting the target object image by a preset quality scoring algorithm to obtain a target wave band image.
S130: and performing background removal processing on the target band image through a preset deep learning model to obtain foreground image data.
S140: and calculating the outline of the target object according to the foreground image data to obtain target outline data.
S150: and calculating according to foreground image data by a preset Gaussian difference method to obtain a difference image.
S160: and extracting pixel data meeting a preset pixel threshold value in the contour range according to the differential image and the target contour data, so as to obtain a target recognition result of the vein to be recognized according to the pixel data.
In the working process of the specific embodiment, the embodiment of the invention firstly acquires a plurality of groups of target object images with different wavelengths through a preset near infrared multispectral imaging camera. Specifically, the target object image in the embodiment of the invention includes veins to be identified of the corresponding target object. According to the embodiment of the invention, the target object is acquired by adopting a plurality of different wave bands, so that the target object image with corresponding wave bands is obtained, and the target object image with better imaging quality can be selected for processing. Illustratively, blood has stronger absorption characteristics to 700-1000 nm near infrared light than other biological tissues, and the embodiment of the invention captures near infrared images of the back of the palm through a near infrared spectrum camera based on a MEMS (micro electro mechanical system) chip, wherein 10 wave band near infrared images are captured at a time. The near infrared spectrum camera in the embodiment of the invention can extract a large number of information layers from a plurality of near infrared spectrums, so that the information of a target object can be better acquired. The filter of the camera is based on the Fabry-Perot optical cavity principle, is designed into a series of coating reflectors, is arranged on an MEMS component, and enables the Fabry-Perot optical cavity to be spatially changed by controlling the change of the voltage applied to the upper mirror support so as to only allow the light with the required infrared wavelength to pass through, thereby realizing image shooting with multiple groups of wavelengths. When the near infrared spectrum camera shoots the palmback image of the target object, the near infrared spectrum camera can obtain 10 groups of near infrared palmback images with wavelengths between 713 and 920nm, namely the target object image, so as to adapt to different shooting environments and provide conditions for preferentially selecting shooting results.
Then, the embodiment of the invention scores and votes the target object image through a preset quality scoring algorithm to obtain a target band image. Specifically, in the embodiment of the invention, when the image acquisition is performed on the target object, such as the palm back, a plurality of groups of target object images with different wavelengths are acquired. Therefore, the embodiment of the invention needs to select the image with the optimal wave band from the target object images with the different wavelengths so as to perform subsequent image recognition analysis. Correspondingly, the embodiment of the invention adopts a mode of scoring and voting the target object image by a preset quality scoring algorithm, so that the target object image with the optimal score is selected, and the target wave band image is obtained. Further, in the embodiment of the invention, background removal processing is carried out on the target band image through a preset deep learning model, so that foreground image data is obtained. It is easy to understand that the embodiment of the invention provides cleaner and easier-to-process image data for subsequent image processing and analysis tasks by performing background removing operation on the target band image. In the embodiment of the invention, the background of the obtained picture is generally close to pure black when a target object (such as the palm back of the palm) is shot by the spectrum camera, but is limited by factors such as shooting environment, illumination condition, shooting angle and the like, and clutter elements and noise in the background can interfere with image processing and analysis tasks to influence the accuracy of a result. Accordingly, embodiments of the present invention reduce these disturbances by removing the background, which may reduce the computational cost of subsequent processing and analysis, while focusing the processing on the object of interest.
Further, according to the embodiment of the invention, the outline of the target object is calculated according to the foreground image data to obtain target outline data, and a difference image is calculated according to the foreground image data by a preset Gaussian difference method. Specifically, according to the embodiment of the invention, the edge calculation is performed on the foreground image data through a preset algorithm for searching the contour, so that the boundary data of the target object is obtained, and the target contour data is obtained. Meanwhile, in the embodiment of the invention, the detail information in the image is improved and the blurring effect is reduced by calculating the difference data of the foreground image data through a preset Gaussian difference method, so that a corresponding difference image is obtained. Finally, according to the differential image and the target contour data, the embodiment of the invention extracts the pixel data meeting the preset pixel threshold value in the contour range, thereby obtaining the target recognition result of the vein to be recognized according to the extracted pixel data. It is easy to understand that the embodiment of the invention can effectively enable the position of the vein to be identified in the target object to be more clearly presented in a Gaussian difference mode, and meanwhile, the pixels outside the outline can be eliminated through the target outline data so as to relieve the problem that the image edge is interfered by noise and impurity information, thereby effectively improving the accuracy and reliability of vein identification of the target object and effectively relieving the problem of edge interference.
In some embodiments of the present invention, the preset quality scoring algorithms include a contrast language-image pre-training image quality assessment algorithm, a natural image quality assessment algorithm, and a no-reference image quality assessment algorithm. Accordingly, in the embodiment of the present invention, the target object image is scored and voted by a preset quality scoring algorithm to obtain a target band image, including but not limited to the following steps:
and performing first quality scoring on the target object image through a non-reference image quality assessment algorithm to obtain first scoring data.
And performing second quality scoring on the target object image through a natural image quality evaluation algorithm to obtain second scoring data.
And performing third quality scoring on the target object image through a contrast language-image pre-training image quality evaluation algorithm to obtain third scoring data.
And determining the target wave band image according to the first scoring data, the second scoring data and the third scoring data in a preset voting mode. The target wave band image comprises images with optimal image quality in a plurality of groups of target object images with different wavelengths.
In this embodiment, the present embodiment of the present invention selects the optimal band Image, i.e., the target band Image, by scoring and voting the target object Image by comparing the language-Image Pre-training Image quality evaluation algorithm (Contrastive Language-Image Pre-training Image Quality Assessment, CLIP-IQA), the natural Image quality evaluation algorithm (Natural Image Quality Evaluator, NIQE), and the no-reference Image quality evaluation algorithm (Blind/Referenceless Image Spatial QUality Evaluator, brique), respectively. Specifically, in the embodiment of the invention, first quality scoring is performed on a target object image through a non-reference image quality evaluation algorithm to obtain first scoring data. Referring to fig. 2, the non-reference image quality evaluation algorithm in the embodiment of the present invention can automatically evaluate the subjective quality of an image without reference to the image. In addition, the quality evaluation standard of the non-reference image quality evaluation algorithm depends on a statistical mode of the image in a spatial domain, and after the image is normalized, the pixel brightness distribution of the natural image is similar to a Gaussian distribution, while the pixel brightness distribution of the non-natural or distorted image does not conform to the Gaussian distribution. Therefore, the embodiment of the invention takes the difference between the distribution curve and the ideal Gaussian curve as the standard for measuring the distortion degree of the image. Accordingly, the lower the BRISQUE (no reference image quality assessment algorithm) score, the higher the image's degree of restoration to the object.
Further, the embodiment of the invention carries out second quality scoring on the target object image through a natural image quality evaluation algorithm to obtain second scoring data. Specifically, the natural image quality evaluation algorithm in the embodiment of the invention is based on constructing a series of features for measuring the image quality, and fitting a multidimensional Gaussian distribution model according to the features. Wherein these features are extracted from some simple and highly regular natural scene. It should be noted that the model is intended to measure the difference between the image to be measured and a multidimensional distribution, which is constructed from the same features extracted from a series of normal natural images. The formula for calculating the image quality by the image quality evaluation algorithm in the embodiment of the invention is shown in the following formula (1):
wherein v is 1 Representing the mean value of the input image, v stand Mean value of standard image is represented, sigma 1 Covariance matrix of input image, Σ stand Representing the covariance matrix of the standard image.
Further, the embodiment of the invention carries out third quality scoring on the target object image through a contrast language-image pre-training image quality evaluation algorithm to obtain third scoring data. Specifically, the contrast language-image pre-training image quality assessment algorithm (CLIP-IOA) in the embodiments of the present invention is an image quality assessment algorithm based on a contrast language-image pre-training model, which combines multi-modal understanding and characterization capabilities of contrast language-image pre-training to assess the quality of an image based on the contrast learning of natural language and images. In addition, the contrast language-image pre-training model (CLIP) in the embodiment of the invention is a multi-modal deep learning model, and text and images can be embedded into a common semantic space after pre-training. Referring to fig. 3, in the contrast language-image pre-training image quality assessment model, the task of the model is to compare the quality of an image with a text description, by constructing a pair of "good" and "bad" prompts, calculating the ratio of cosine similarity between a text vector and an image vector, and then calculating an image score by a Softmax function, so as to obtain third scoring data.
Finally, according to the embodiment of the invention, the target wave band image is determined according to the first scoring data, the second scoring data and the third scoring data in a preset voting mode. Specifically, the target wave band image in the embodiment of the invention comprises a plurality of groups of images with optimal image quality in target object images with different wavelengths, namely the embodiment of the invention votes scoring data calculated by three algorithms to perform preferential selection so as to obtain the images with optimal image quality in the target object images with different wavelengths. In an exemplary embodiment, after three image quality scores of the first score data, the second score data and the third score data are obtained through calculation, an image with the optimal quality score is selected through a preset voting mode. For example, if there are two or more scoring methods that consider the best quality of an image in a certain band, the image in that band is selected for analysis. In addition, if three methods select three images with different wave bands, the three images calculate Min-Max normalized scores of three quality scores and add, and the image with the lowest score is selected for analysis.
In some embodiments of the present invention, background removal processing is performed on the target band image through a preset deep learning model to obtain foreground image data, including but not limited to the following steps:
Inputting the target wave band image into a preset image perception model to identify pixel points, and obtaining an image mask. The image mask comprises a foreground pixel point and a background pixel point.
And carrying out preset image processing on the target band image according to the image mask to obtain foreground image data. The preset image processing comprises the steps of reserving pixel points corresponding to foreground pixel points in the target band image and removing pixel points corresponding to background pixel points.
In this embodiment, the embodiment of the present invention firstly inputs the target band image into a preset image perception model to perform pixel point recognition, so as to obtain an image mask including a foreground pixel point and a background pixel point. Then, the embodiment of the invention performs preset image processing on the target band image according to the image mask to obtain foreground image data, and specifically, in the preset image processing process, the embodiment of the invention firstly reserves the pixel points corresponding to the foreground pixel points in the target band image, and then removes the pixel points corresponding to the background pixel points to extract and obtain the foreground image data. In the embodiment of the invention, the image background is eliminated in an image semantic segmentation mode. The image segmentation process in the embodiment of the present invention may be regarded as classifying each pixel point in the image into a certain object class in the image. In addition, in the background elimination process in the embodiment of the invention, the pixel points on the image are divided into two categories, namely a foreground category and a background category, so that the pixel points belonging to the background category are eliminated. For example, for the palm back image shot by a spectrum camera in a specific environment, the image background is relatively simple, a better image background removing effect can be obtained by using a pre-trained ISNet (Instance Segmentation Network) model, and the model running speed is higher. Correspondingly, the embodiment of the invention inputs the images into the pre-trained ISNet model, namely the preset image perception model, by importing the images with the selected wavelengths. The result of the output of the ISNet model in the embodiment of the present invention is a two-dimensional array of image masks (masks), where the elements on the array are 0 (representing background pixels) or 1 (representing foreground pixels). Then, the embodiment of the invention can remove the background of the target band image by taking out the points representing the foreground pixels, so as to obtain foreground image data.
In some embodiments of the present invention, the target object contour is calculated from the foreground image data to obtain target contour data, including but not limited to the following steps:
and performing edge detection according to the foreground image data to obtain first contour data.
And performing contour approximation processing according to the first contour data through a Lamerger-Target-Puck algorithm to obtain second contour data.
And carrying out contour reduction by a preset contour reduction algorithm according to the second contour data to obtain target contour data.
In this embodiment, the embodiment of the present invention first performs edge detection according to foreground image data to obtain first contour data. Specifically, the embodiment of the invention firstly searches the contour by utilizing the intensity difference of the edge in the foreground image data to obtain first contour data. For example, the embodiment of the invention can perform edge detection on the foreground image data through a Sobel (Sobel) algorithm and a Canny (Canny) algorithm to obtain the boundary of a target object, such as the outline data of the palm back. It should be noted that, after obtaining the edge image, the task of contour tracing (edge detection) in the embodiment of the present invention is to find continuous boundary points from the edge image. Accordingly, in the embodiment of the present invention, the contour tracing algorithm starts from a point of the foreground image data, traces along the boundary pixels until returning to the starting point, and forms a closed contour. In addition, in addition to finding the outline, the embodiment of the invention also calculates hierarchical structure information of the outline. Then, the embodiment of the invention performs contour approximation processing on the first contour data through a Lamerger-Tiger-Puck algorithm to obtain second contour data. It will be readily appreciated that finding a contour algorithm will typically involve a large number of contour points, which may lead to increased storage and processing costs, as well as possibly affecting the efficiency of subsequent analysis. Therefore, the embodiment of the invention reduces the problem of increasing the storage and processing cost by simplifying the outline, and improves the analysis and processing efficiency. In addition, in the process of simplifying the first contour data, not only the number of vertices in the contour is reduced, and the processing and calculating costs are reduced, but also the main shape and structural characteristics of the target object, such as the back of the hand, need to be kept, so as to avoid excessive information loss, i.e. the overall shape of the object needs to be kept unchanged as much as possible. Illustratively, in the embodiment of the invention, the contour approximation processing is performed through a Ramer-Douglas-Peucker (RDP) algorithm, so that the vertex number of a given threshold value is reduced to simplify the broken line, and the second contour data is obtained. After a starting point and an ending point of a curve are given, embodiments of the present invention first find the vertex furthest from the line connecting the two reference points, referred to as the "furthest point". When the distance to the "furthest point" is less than the threshold, all vertices between the starting point and the ending point are automatically ignored and the curve is made a straight line. Accordingly, when the distance to the "furthest point" exceeds the threshold, embodiments of the present invention recursively repeat the step of taking the "furthest point" as a new reference point and continue the process described above. Finally, according to the embodiment of the invention, contour reduction is performed through a preset contour reduction algorithm according to the second contour data, so as to obtain target contour data. In particular, after the simplified profile (i.e., the second profile data) is calculated, the profile typically contains boundaries of the image, which often contain some noise, impurities, or insignificant information. Accordingly, such information may interfere with subsequent image analysis and processing, resulting in unstable or inaccurate recognition results. In addition, the vein identified in the embodiment of the present invention, i.e. the vein to be identified, will not generally appear on the boundary of the target object, such as the palm boundary. Therefore, the embodiment of the invention reduces the influence of unstable pixels on the boundary by reducing the outline, thereby better capturing the internal structure of the target object.
In some embodiments of the present invention, the profile is reduced according to the second profile data by a preset profile reduction algorithm to obtain target profile data, including but not limited to the following steps:
traversing adjacent edges in the second contour data according to a preset direction to obtain a plurality of groups of adjacent edge data. The adjacent edge data comprises a first adjacent edge and a second adjacent edge.
And taking the common point of the first adjacent side and the second adjacent side as an initial vertex, and calculating the rotation angle to obtain the translation direction of the common point.
And translating the common point according to the translation direction of the common point to obtain a target vertex.
And connecting all the target vertexes to obtain target contour data.
In this embodiment, the embodiment of the present invention first traverses adjacent edges in the second contour data according to a preset direction to obtain a plurality of sets of adjacent edge data. Specifically, in the embodiment of the invention, the adjacent edge data includes a first adjacent edge and a second adjacent edge. In the embodiment of the present invention, the preset direction refers to a traversing direction, for example, traversing adjacent edges on the contour in a counterclockwise direction or a clockwise direction. Then, the embodiment of the invention takes the common point of the first adjacent side and the second adjacent side as an initial vertex, calculates a corresponding rotation angle, and obtains the translation direction of the common point, and translates the common point according to the translation direction of the common point to obtain a target vertex. Finally, the embodiment of the invention connects all the target vertexes, and builds the reduced target object outline, thus obtaining the target outline data. For example, referring to fig. 4 and 5, in the process of performing contour reduction on the second contour data, the embodiment of the present invention first traverses adjacent edges on the second contour data in a counterclockwise direction to obtain several sets of adjacent edge data. As shown in fig. 4, for the common vertex of the adjacent edges, the edges arranged in front of the vertex in the counterclockwise direction in the embodiment of the present invention are called: edge 1 (first adjacent edge), the edge that is arranged behind the vertex is called: edge 2 (second adjacent edge). Then, the embodiment of the invention calculates the rotation angle of each group of edges, takes the common point of the adjacent edges as the vertex, rotates the edge 1 by half of the rotation angle of the two edges, and the direction of the rotated edge 1 is the translation direction of the common point. Correspondingly, the embodiment of the invention calculates the translation direction of each common point (vertex), translates each vertex by a certain distance according to the direction, and then connects the translated vertices, thereby obtaining the reduced outline, namely the target outline data, as shown in fig. 5. It is easy to understand that after the profile is reduced by adopting the method, the embodiment of the invention also detects whether the intersection exists between the line segments in the profile, and if the line segments do not intersect, the final profile reduction result is obtained. Otherwise, if the line segments cross, the embodiment of the invention eliminates the cross of the line segments by the methods of exchanging vertexes between the line segments, reducing translation distance and the like.
In some embodiments of the present invention, the difference image is calculated according to foreground image data by a preset gaussian difference method, including but not limited to the following steps:
and carrying out Gaussian blur processing on the foreground image data to obtain Gaussian blur data. The Gaussian blur data comprises a first Gaussian blur result and a second Gaussian blur result, and the blur scales of the first Gaussian blur result and the second Gaussian blur result are different.
And carrying out difference according to the first Gaussian blur result and the second Gaussian blur result to obtain a difference image.
In this embodiment, the embodiment of the present invention first performs gaussian blur processing on foreground image data to obtain gaussian blur data. Specifically, when the foreground image data is processed through Gaussian blur in the embodiment of the invention, two groups of Gaussian blur are calculated, namely, the Gaussian blur data in the embodiment of the invention comprises a first Gaussian blur result and a second Gaussian blur result, and the blur scales of the first Gaussian blur result and the second Gaussian blur result are different. According to the embodiment of the invention, through a mode of carrying out Gaussian blur processing on foreground image data, noise in an image is reduced, details of the image are smoothed, and a Gaussian function (namely a normal distribution function) is utilized to carry out weighted average on the image, so that the change of pixel values in the image is reduced. The embodiment of the invention takes Gaussian blur processing as a preprocessing step, and prepares for acquiring vein pixels by analyzing edge and texture information in an image through Gaussian difference. Illustratively, the gaussian blur processing in the embodiment of the present invention includes two steps of generating a gaussian kernel and applying convolution. In the embodiment of the invention, the gaussian blur kernel, i.e. the filter or the window, is a two-dimensional weight matrix, and the weight value is gradually reduced in the shape of gaussian distribution at the center highest point of the gaussian kernel. Accordingly, the size of the gaussian kernel determines the degree of blurring. Of these, smaller gaussian kernels result in a lighter smoothing effect, while larger gaussian kernels result in a more intense smoothing effect. Next, the embodiment of the invention carries out convolution operation on the generated Gaussian kernel and each pixel of the image and surrounding pixels. In the embodiment of the invention, the convolution operation is to multiply the kernel with the corresponding element of the image area and add all the results to obtain a new pixel value. Accordingly, this new pixel value is the average value of the image over the area, but the weights of surrounding pixels will gradually decrease according to a gaussian distribution, thereby achieving a smoothing effect.
Further, the embodiment of the invention carries out difference according to the first Gaussian blur result and the second Gaussian blur result to obtain a difference image. Specifically, the embodiment of the invention can better display the pixels of the vein to be identified in the target object by carrying out the difference mode on two groups of Gaussian blur, namely the first Gaussian blur result and the second Gaussian blur result. For example, the embodiment of the present invention may implement the differential calculation by subtracting the first gaussian blur result from the second gaussian blur result, so as to obtain a differential image, as shown in fig. 6. It is easy to understand that in the embodiment of the invention, detailed information in an image is improved and a blurring effect is reduced by means of performing gaussian differential processing. In the embodiment of the invention, the Gaussian difference is used for processing the original gray level image to reduce the blurring degree of the blurred image and make the edges and textures in the image more prominent. In addition, in the embodiment of the invention, the high-frequency information in the image, namely the details with smaller scale, can be suppressed by using the Gaussian kernel to carry out convolution. However, embodiments of the present invention preserve spatial information in a frequency range common to both images by subtracting a blurred image in one image from another image. Therefore, the effect of Gaussian difference is similar to a special filter, only the information in the reserved frequency range in the original image is allowed to pass through, and the information in other frequency ranges is filtered, so that the specific frequency range in the image can be highlighted, the edge and texture details of the image are further enhanced, the image enhancement is more targeted, and the specific features focused in the image processing can be better displayed.
In some embodiments of the present invention, pixel data satisfying a preset pixel threshold within a contour range is extracted according to a differential image and target contour data, so as to obtain a target recognition result of a vein to be recognized according to the pixel data, including but not limited to the following steps:
and determining a preset pixel threshold value through an Ojin algorithm.
And acquiring pixel points in the differential image, wherein the pixel value of the pixel points in the contour range of the target contour data is larger than a preset pixel threshold value, so as to obtain the target pixel points.
And outputting a target identification result according to the target pixel point. Wherein the target recognition result includes pixel coordinates or a fused image.
In this embodiment, the embodiment of the present invention first determines, by using the oxford algorithm, a corresponding preset pixel threshold. Specifically, the Otsu's algorithm in the embodiments of the present invention is an image threshold determination algorithm. According to the embodiment of the invention, a proper threshold value, namely a preset pixel threshold value, is selected through an Ojin algorithm, and a threshold value, namely the preset pixel threshold value, is found through analyzing the gray level histogram of the image, so that the image is divided into a foreground and a background, and the inter-class variance between the foreground and the background is increased to the greatest extent. Then, the embodiment of the invention obtains the pixel points in the differential image, wherein the pixel value of the pixel points is larger than the preset pixel threshold value in the contour range of the target contour data, so as to obtain the target pixel points. It is readily understood that many pixels located at the edges of an image may be misinterpreted as venous pixels, as the edges of the image may be disturbed by noise and impurity information. Therefore, the embodiment of the invention can relieve the problem of misjudgment through palm contour information. According to the embodiment of the invention, the pixels outside the outline are removed by using the reduced palm outline, namely the target outline data, and only the pixels in the outline are reserved as real vein blood vessel image pixels, namely the target pixel points, so that the blood vessel structures of veins to be identified in the target object image can be more accurately identified and extracted, and the edge interference is effectively eliminated. Finally, the embodiment of the invention outputs the target recognition result according to the target pixel point. Specifically, after the target pixel point is extracted, the embodiment of the present invention outputs a recognition result of the vein to be recognized according to the corresponding target pixel point, for example, outputs the pixel coordinates of the corresponding target pixel point or outputs a fusion image, and displays the recognition result in an image fusion manner, as shown in fig. 7. In addition, the embodiment of the invention can also display the identified vein contours for further analysis and processing by utilizing the mode of drawing the image contours and simplifying the contours, thereby providing visual information and providing a useful basis for further research.
It will be readily appreciated that infrared spectral imaging in embodiments of the present invention provides more spectral information that can be used to analyze and identify characteristics of different objects. Accordingly, the embodiment of the invention provides a low-cost and convenient solution for realizing near infrared spectrum imaging by using the low-cost near infrared multispectral imaging camera based on the MEMS chip. In addition, the spectral imaging camera used in the embodiment of the invention can shoot near infrared pictures with 10 wavelengths from 713-920 nm at one time, namely target object images, can be better adapted to the scene actually shot, and ensures that better imaging effect can be achieved when shooting in different environments. It should be noted that, the near infrared spectrum imaging in the embodiment of the invention realizes the target object, such as the metacarpal vein image recognition technology, and has wide application space in the fields of medical treatment, biological recognition and the like. For example, by the aid of the metacarpal vein image recognition technology in the medical field, non-invasive monitoring, diagnosis and treatment can be realized in the medical field, so that more information is provided for making accurate medical decisions, and medical experience and treatment effect are improved. In medical record management application, the metacarpal vein image can be used for authentication of patient identity and access control of medical records, and safety and reliability of medical data of patients are ensured. In addition, the dorsal metacarpal vein image can be used for regular health detection to help monitor vascular health, blood flow conditions, etc., thereby early finding possible problems. As another example, in the biometric authentication application, since veins are distributed under the skin, it is more difficult to tamper with internal structural information of the human body than external information such as fingerprint and face, thereby having high security. In addition, relative to face recognition, palm vein recognition often circumvents personal privacy concerns because palm vein images are not as easy to acquire as facial images.
Referring to fig. 8, an embodiment of the present invention further provides an image recognition system based on near infrared images, including:
the first module 210 is configured to acquire a plurality of sets of target object images with different wavelengths through a preset near infrared multispectral imaging camera. The target object image comprises veins to be identified.
And a second module 220, configured to score and vote on the target object image by using a preset quality scoring algorithm, so as to obtain a target band image.
And a third module 230, configured to perform background removal processing on the target band image through a preset deep learning model, so as to obtain foreground image data.
And a fourth module 240, configured to calculate the contour of the target object according to the foreground image data, so as to obtain target contour data.
A fifth module 250, configured to calculate a difference image according to the foreground image data by a preset gaussian difference method.
A sixth module 260, configured to extract, according to the differential image and the target contour data, pixel data that satisfies a preset pixel threshold within the contour range, so as to obtain, according to the pixel data, a target recognition result of the vein to be recognized.
Referring to fig. 9, an embodiment of the present invention further provides an image recognition system based on near infrared images, including:
At least one processor 310.
At least one memory 320 for storing at least one program.
The at least one program, when executed by the at least one processor 310, causes the at least one processor 310 to implement the near infrared image-based image recognition method as described in the above embodiments.
An embodiment of the present invention also provides a computer-readable storage medium storing computer-executable instructions for execution by one or more control processors, e.g., to perform the steps described in the above embodiments.
Those of ordinary skill in the art will appreciate that all or some of the steps, systems, and methods disclosed above may be implemented as software, firmware, hardware, and suitable combinations thereof. Some or all of the physical components may be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application specific integrated circuit. Such software may be distributed on computer readable media, which may include computer storage media (or non-transitory media) and communication media (or transitory media). The term computer storage media includes both volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data, as known to those skilled in the art. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital Versatile Disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium which can be used to store the desired information and which can be accessed by a computer. Furthermore, as is well known to those of ordinary skill in the art, communication media typically embodies computer readable instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transport mechanism and includes any information delivery media.
While the preferred embodiment of the present invention has been described in detail, the present invention is not limited to the above embodiment, and various equivalent modifications and substitutions can be made by those skilled in the art without departing from the spirit of the present invention, and these equivalent modifications and substitutions are intended to be included in the scope of the present invention as defined in the appended claims.
Claims (10)
1. The image recognition method based on the near infrared image is characterized by comprising the following steps of:
acquiring a plurality of groups of target object images with different wavelengths through a preset near infrared multispectral imaging camera; the target object image comprises veins to be identified;
scoring and voting the target object image through a preset quality scoring algorithm to obtain a target wave band image;
performing background removal processing on the target band image through a preset deep learning model to obtain foreground image data;
calculating the outline of the target object according to the foreground image data to obtain target outline data;
calculating according to the foreground image data by a preset Gaussian difference method to obtain a difference image;
and extracting pixel data meeting a preset pixel threshold value in a contour range according to the differential image and the target contour data, so as to obtain a target recognition result of the vein to be recognized according to the pixel data.
2. The near infrared image-based image recognition method of claim 1, wherein the preset quality scoring algorithm comprises a contrast language-image pre-training image quality assessment algorithm, a natural image quality assessment algorithm, and a no-reference image quality assessment algorithm;
scoring and voting the target object image through a preset quality scoring algorithm to obtain a target band image, wherein the scoring and voting comprise the following steps:
performing first quality scoring on the target object image through the non-reference image quality assessment algorithm to obtain first scoring data;
performing a second quality scoring on the target object image through the natural image quality evaluation algorithm to obtain second scoring data;
performing third quality scoring on the target object image through the contrast language-image pre-training image quality evaluation algorithm to obtain third scoring data;
determining the target wave band image according to the first scoring data, the second scoring data and the third scoring data in a preset voting mode; the target wave band images comprise images with optimal image quality in the target object images with different wavelengths.
3. The near infrared image-based image recognition method according to claim 1, wherein the performing background removal processing on the target band image by a preset deep learning model to obtain foreground image data comprises:
inputting the target band image into a preset image perception model to identify pixel points, and obtaining an image mask; the image mask comprises a foreground pixel point and a background pixel point;
performing preset image processing on the target wave band image according to the image mask to obtain the foreground image data; the preset image processing includes reserving pixel points corresponding to the foreground pixel points in the target band image, and removing the pixel points corresponding to the background pixel points.
4. The near infrared image-based image recognition method according to claim 1, wherein the calculating the target object contour according to the foreground image data to obtain target contour data comprises:
performing edge detection according to the foreground image data to obtain first contour data;
performing contour approximation processing through a Lamerger-Laplace-Puck algorithm according to the first contour data to obtain second contour data;
And carrying out contour reduction by a preset contour reduction algorithm according to the second contour data to obtain the target contour data.
5. The near infrared image-based image recognition method according to claim 4, wherein the performing contour reduction according to the second contour data by a preset contour reduction algorithm to obtain the target contour data comprises:
traversing adjacent edges in the second contour data according to a preset direction to obtain a plurality of groups of adjacent edge data; the adjacent edge data comprises a first adjacent edge and a second adjacent edge;
taking a common point of the first adjacent side and the second adjacent side as an initial vertex, and calculating a rotation angle to obtain a common point translation direction;
translating the common point according to the translation direction of the common point to obtain a target vertex;
and connecting all the target vertexes to obtain the target contour data.
6. The near infrared image-based image recognition method according to claim 1, wherein the calculating the difference image according to the foreground image data by a preset gaussian difference method comprises:
carrying out Gaussian blur processing on the foreground image data to obtain Gaussian blur data; the Gaussian blur data comprises a first Gaussian blur result and a second Gaussian blur result, and the blur scales of the first Gaussian blur result and the second Gaussian blur result are different;
And carrying out difference according to the first Gaussian blur result and the second Gaussian blur result to obtain the difference image.
7. The near infrared image-based image recognition method according to claim 1, wherein extracting pixel data satisfying a preset pixel threshold value within a contour range according to the differential image and the target contour data to obtain a target recognition result of the vein to be recognized according to the pixel data comprises:
determining the preset pixel threshold value through an Ojin algorithm;
obtaining pixel points in the differential image, wherein the pixel value of the pixel points in the contour range of the target contour data is larger than the preset pixel threshold value, so as to obtain target pixel points;
outputting the target recognition result according to the target pixel point; wherein the target recognition result includes pixel coordinates or a fused image.
8. An image recognition system based on near infrared images, comprising:
the first module is used for acquiring a plurality of groups of target object images with different wavelengths through a preset near infrared multispectral imaging camera; the target object image comprises veins to be identified;
the second module is used for scoring and voting the target object image through a preset quality scoring algorithm to obtain a target wave band image;
The third module is used for carrying out background removal processing on the target band image through a preset deep learning model to obtain foreground image data;
a fourth module, configured to calculate a target object contour according to the foreground image data, to obtain target contour data;
a fifth module, configured to calculate, according to the foreground image data, a difference image by using a preset gaussian difference method;
and a sixth module, configured to extract, according to the differential image and the target contour data, pixel data that satisfies a preset pixel threshold value within a contour range, so as to obtain a target recognition result of the vein to be recognized according to the pixel data.
9. An image recognition system based on near infrared images, comprising:
at least one processor;
at least one memory for storing at least one program;
the at least one program, when executed by the at least one processor, causes the at least one processor to implement the near infrared image-based image recognition method of any one of claims 1 to 7.
10. A computer storage medium in which a processor-executable program is stored, wherein the processor-executable program is for implementing the near infrared image-based image recognition method according to any one of claims 1 to 7 when executed by the processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311523764.4A CN117649688A (en) | 2023-11-15 | 2023-11-15 | Near infrared image-based image recognition method, system and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311523764.4A CN117649688A (en) | 2023-11-15 | 2023-11-15 | Near infrared image-based image recognition method, system and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117649688A true CN117649688A (en) | 2024-03-05 |
Family
ID=90048716
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311523764.4A Pending CN117649688A (en) | 2023-11-15 | 2023-11-15 | Near infrared image-based image recognition method, system and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117649688A (en) |
-
2023
- 2023-11-15 CN CN202311523764.4A patent/CN117649688A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110852160B (en) | Image-based biometric identification system and computer-implemented method | |
| US9317761B2 (en) | Method and an apparatus for determining vein patterns from a colour image | |
| Syarif et al. | Enhanced maximum curvature descriptors for finger vein verification | |
| JP5107045B2 (en) | Method for identifying a pixel representing an iris in an image acquired for the eye | |
| US20060222212A1 (en) | One-dimensional iris signature generation system and method | |
| CN108985210A (en) | A kind of Eye-controlling focus method and system based on human eye geometrical characteristic | |
| Frucci et al. | WIRE: Watershed based iris recognition | |
| Das et al. | A new efficient and adaptive sclera recognition system | |
| KR101315646B1 (en) | Method and apparatus for finger vein pattern extraction using guided gabor filter | |
| JP2007188504A (en) | Method for filtering pixel intensity in image | |
| Kang et al. | The biometric recognition on contactless multi-spectrum finger images | |
| Abidin et al. | Iris segmentation analysis using integro-differential and hough transform in biometric system | |
| Ashok et al. | Detection of retinal area from scanning laser ophthalmoscope images (SLO) using deep neural network | |
| Parikh et al. | Effective approach for iris localization in nonideal imaging conditions | |
| Zhou et al. | A comprehensive multimodal eye recognition | |
| Zheng | Static and dynamic analysis of near infra-red dorsal hand vein images for biometric applications | |
| Proença | Iris recognition in the visible wavelength | |
| CN112861588B (en) | Living body detection method and device | |
| Zheng et al. | Near-infrared Image Enhancement Method in IRFPA Based on Steerable Pyramid. | |
| Shashidhara et al. | A novel approach to circular edge detection for iris image segmentation | |
| CN117649688A (en) | Near infrared image-based image recognition method, system and storage medium | |
| CN114582008A (en) | Living iris detection method based on two wave bands | |
| Crihalmeanu et al. | Multispectral ocular biometrics | |
| Lin et al. | A novel framework for automatic 3D face recognition using quality assessment | |
| Ukai et al. | Facial skin blood perfusion change based liveness detection using video images |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |