CN110245670B - Image pyramid gradient histogram feature-based skyhead identity identification method and device - Google Patents
Image pyramid gradient histogram feature-based skyhead identity identification method and device Download PDFInfo
- Publication number
- CN110245670B CN110245670B CN201910504289.3A CN201910504289A CN110245670B CN 110245670 B CN110245670 B CN 110245670B CN 201910504289 A CN201910504289 A CN 201910504289A CN 110245670 B CN110245670 B CN 110245670B
- Authority
- CN
- China
- Prior art keywords
- image
- skyhead
- tianzhu
- gradient histogram
- celestial
- 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.)
- Active
Links
Images
Classifications
-
- 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
- G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/751—Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Evolutionary Computation (AREA)
- Databases & Information Systems (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Image Analysis (AREA)
- Studio Devices (AREA)
Abstract
A skyhold identity identification method and device based on image pyramid gradient histogram features. The method comprises the steps of collecting a skyhead image by adopting portable image collecting equipment with a network function, extracting pyramid gradient histogram features of a rectangular region of the skyhead image, calculating a Chamfer distance between the pyramid gradient histogram features of the rectangular region of the extracted skyhead image and real skyhead features collected and stored in a template library based on a skyhead template library established by the pyramid gradient histogram features of the skyhead image, and performing skyhead identity authentication. The identification challenge of the celestial beads shot under different illumination environments and at different resolutions and at any angle is overcome, the robust result can be still obtained under the condition that the local features of the celestial bead image are lost, the universality and the robustness are realized, the celestial beads can be automatically, efficiently, objectively and accurately identified, and a simple, accurate and intelligent celestial bead identity identification platform is provided for celestial bead enthusiasts and collectors.
Description
Technical Field
The invention relates to the field of image matching, in particular to a skyhead identity identification method and device based on image pyramid gradient histogram features.
Background
The Tianzhu has unique totem pattern and source flow, and is a cultural carrier. The material of the celestial pearl is generally jade, and is common in jadeite, and the shape is generally a cylinder or a similar cylinder with two ends slightly thinner than the middle, and a sphere or an ellipsoid. Tianzhu generally has a particular texture. Typical beads are shown in fig. 1, 2 and 3, wherein fig. 1 is a short cylindrical bead, fig. 2 is an ellipsoidal bead, and fig. 3 is an oblong cylindrical bead. The material of the Tianzhu can be natural material or artificial synthesis. The identification of the tianzhu identity is of great significance to historical research and standard collection markets, and the tianzhu pattern is of great importance to the identification of the tianzhu identity.
Traditional day pearl identity is discerned based on expert scholars and is judged with the naked eye under the magnifying glass is supplementary, relies on self experience and subjective sensation, and its result receives internal learning and external environment influence great, and this makes traditional day pearl identity identification result inevitably have certain individual subjectivity, and manual operation efficiency is lower moreover, also should not promote in non-professional day pearl collection and hobby crowd. As the judgment basis and experience reading of experts and scholars are different, the identification results are also very different, the identification standardization of the skybutting identity cannot be carried out, and the development of the skutting collection market is hindered. In addition, the number of Tianzhu experts and scholars is limited, and the demand of Tianzhu market is rapidly increased in recent years, so that a Tianzhu identification method capable of replacing experts and scholars is urgently needed, Tianzhu can be automatically, efficiently, objectively and accurately identified, and a simple, accurate and intelligent Tianzhu identification method is provided for Tianzhu enthusiasts and collectors.
The development of computers in the field of image matching provides a research basis for automatic identification of Tianzhu. However, the celestial pearl pattern has macroscopic culture totem representation and natural trace of microscopic collision scratches, so how to overcome the interference of the microscopic collision scratches in the celestial pearl identification process is a difficult problem to consider in celestial pearl identification; the celestial beads are a non-lambertian body (the lambertian body is a complete diffuse reflector, that is, an object whose incident energy is centered on an incident point and can isotropically reflect energy around the whole hemispherical space to realize diffuse reflection), and a specular reflection high-light area appears under strong illumination, and the high-light area can partially or completely cover the curved surface stereo characteristic of an original object image, so that the difficulty of identification of celestial beads is further increased. Based on the popularization of portable photographing equipment such as mobile phones and the like, the possibility is provided for developing a celestial pearl identity identification technology based on an image matching technology. However, the resolutions of different types of mobile phones, i.e., different types of image acquisition devices, are greatly different, and the imaging details are also greatly different, so that great difficulty is caused in completing image matching; and most of the celestial beads are irregular cylinders or spheres but irregular bodies with special curved surfaces, so that different images obtained from different shooting visual angles have extremely large difference, and the celestial bead images shot at any visual angle need to be matched in actual matching, thereby further increasing the difficulty of image matching.
Disclosure of Invention
The invention provides a skyhold identity identification method and device based on an image pyramid gradient histogram, aiming at the technical problems. Based on portable image acquisition equipment with network function, such as cell-phone, portable computer, panel computer etc. provide corresponding auxiliary information, have use friendliness, convenience. The pyramid gradient histogram features of the celestial pearl image are extracted for database comparison, the celestial pearl identification challenge of different illumination environments, different resolutions and shooting at any angle is overcome, and universality and robustness are achieved.
In order to solve the technical problem, according to an aspect of the present invention, there is provided a skyhold identity identification method based on image pyramid gradient histogram features, the method including the following steps:
1) skyhead image template generation comprising:
1.1) acquiring skyhead images from a plurality of different angles;
1.2) extracting the pyramid gradient histogram characteristics of the Tianzhu image to obtain a characteristic template file;
1.3) storing the characteristic template file into a database;
2) carrying out identification of Tianzhu based on Tianzhu image template, which comprises
2.1) collecting an actual Tianzhu image;
2.2) sending the actual skatebead image to a server;
2.3) calculating the distance between the pyramid gradient histogram feature of the extracted Tianzhu image and the Tianzhu feature in the feature template file stored in the database, and identifying the identity of the Tianzhu according to a set threshold value.
Preferably, said acquiring skyline images from a plurality of different angles comprises: the video acquisition equipment acquires the skyhead video, and after the acquisition is finished, the video is subjected to frame extraction operation, so that 120 frames are selected at equal intervals when the skyhead rotates for one circle, and all view angle images of the skyhead are covered.
Preferably, the pyramid gradient histogram feature is a pyramid gradient histogram feature of a rectangular region of the celestial sphere image.
Preferably, the extracting the pyramid gradient histogram feature of the skyhead image comprises;
1.2.1) carrying out normalization preprocessing on the rectangular area: the gray level normalization is obtained by gray level histogram equalization, and is calculated as shown in formula (1) and formula (2),
wherein k represents the gray value of the rectangular area image before normalization preprocessing, nkRepresenting a grey level of rkM and N represent the height and width of the image, pr(rk) Representing the probability of occurrence, L representing the number of equalized gray levels, skRepresenting the corresponding value after histogram equalization;
the size normalization is obtained by bilinear interpolation, and the calculation is shown as a formula (3),
wherein x is1、y1、x2、y2Is a coordinate value, Q11＝(x1，y1) Represents Q11Point coordinates, like Q12＝(x1，y2)，Q21＝(x2，y1) And Q22＝(x2，y2) Respectively represent Q12、Q21And Q22Point coordinates. f (Q)11) Represents Q11Dot pixel gray scale value, f (Q)12) Represents Q12Dot pixel gray scale value, f (Q)21) Represents Q21Dot pixel gray scale value, f (Q)22) Represents Q22The dot pixel gray scale value is calculated to obtain a gray scale value f (x, y) at which the point P is (x, y).
Preferably, the extracting the pyramid gradient histogram feature of the skyhead image comprises;
1.2.2) iteratively solving a threshold value for maximizing the variance between the two classes based on the maximum inter-class variance method, namely an OTSU algorithm, and calculating as shown in a formula (4),
where t is the iterative solution threshold, ω0(t) and ω1(t) respectively represent the class probabilities,
and
representing intra-class variance;
finally, the threshold T plus the prior threshold T is solved by iteration0Obtaining a binary threshold value T as shown in a formula (5),
T＝t+T0 (5)。
preferably, the extracting the pyramid gradient histogram feature of the skyhead image comprises;
1.2.3) obtaining the gradient value and gradient direction of each pixel point based on the skyhead texture image, quantizing the gradient direction into 9 directions to obtain the features of a histogram of Gradients (HOG), calculating as shown in formula (6) -formula (9),
wherein, I represents a gray-scale image,
representing a convolution, Gx、GyAnd G denotes a gradient value, and α denotes a gradient direction.
Preferably, the extracting the pyramid gradient histogram feature of the skyhead image comprises;
1.2.4) dividing the skyhead texture image area according to the height-to-width ratio of 1:1 to obtain 4 sub-areas, and extracting gradient histogram features from the 4 sub-areas respectively.
Preferably, the extracting the pyramid gradient histogram feature of the skyhead image comprises;
1.2.5) dividing each subregion again according to the height-to-width ratio of 1:1 to obtain 16 subregions, and extracting gradient histogram features from the 16 subregions respectively.
Preferably, the extracting the pyramid gradient histogram feature of the skyhead image comprises;
1.2.6) connecting the Histograms in sequence to obtain the characteristics of Pyramid gradient Histograms (PHOG), wherein the characteristic dimension is (1+4+16) × 9 ═ 189-dimensional characteristic vector, normalizing and storing the characteristic vector as a characteristic template file.
Preferably, the performing of the identification of the celestial pearl based on the celestial pearl image template includes:
2.1) said acquiring of actual skyline images comprises: the method comprises the steps of shooting a skyline image by adopting portable image acquisition equipment with a network function, and acquiring the image in an auxiliary mode by adopting an image frame in proportion to real skylines according to information of selecting and identifying the skylines.
Preferably, the performing of the identification of the celestial pearl based on the celestial pearl image template includes:
and 2.2) sending the Tianzhu image in the image frame to a server background, and extracting the pyramid gradient histogram features of the rectangular region of the Tianzhu image.
Preferably, the performing of the identification of the celestial pearl based on the celestial pearl image template includes:
2.3) calculating the distance between the pyramid gradient histogram feature of the extracted Tianzhu image and the Tianzhu feature in the feature template file stored in the database to be a Chamfer distance, wherein the calculation is shown as a formula (10),
wherein x istAnd xmRespectively representing pyramid gradient histograms of database skylines and user uploaded skylines to be detected, namely PHOG characteristics, m represents characteristic dimension, NmRepresents a normalization constant;
the identity is identified according to a set threshold value, as shown in formula (11):
wherein, T1The identification threshold is preset, the identification threshold is represented as ID (1) and represents that the current user uploads the bead to be tested to be consistent with the database entry bead, and similarly, the identification threshold is represented as ID (0) and represents that the current user uploads the bead to be tested to be inconsistent with the database entry bead.
Preferably, step 2.2) further comprises the following steps:
2.2.1) overall recognition of the sky bead image is firstly carried out, the gradient distribution of the overall pattern is analyzed by utilizing the collected whole sky bead image, the acquired whole sky bead image is matched with the gradient distribution of the sky bead pattern in the sky bead template library to obtain a similarity value, and the step 2.2 is carried out on the whole sky bead image with the similarity value larger than a threshold value;
2.2.2) carrying out network division on the whole Tianzhu image, wherein the Tianzhu texture image area is divided according to the height-to-width ratio of 1:1 to obtain 4 sub-areas in total, and extracting gradient histogram features from the 4 sub-areas respectively;
2.2.3) dividing each subregion again according to the height-to-width ratio of 1:1 to obtain 16 subregions, and extracting gradient histogram features from the 16 subregions respectively;
2.2.4) sequentially connecting the Histograms to obtain Pyramid gradient histogram (PHOG) features, wherein the feature dimension is (1+4+16) × 9 ═ 189-dimensional feature vector, and normalizing to obtain the Pyramid gradient histogram features of rectangular areas of the skyhead images.
According to another aspect of the present invention, there is provided an image pyramid gradient histogram feature-based skyhook identification device, including:
1) a sky-bead image template generation unit comprising:
1.1) an image acquisition unit for acquiring celestial beads images from a plurality of different angles;
1.2) a feature extraction unit, which is used for extracting the pyramid gradient histogram features of the Tianzhu image to obtain a feature template file;
1.3) a storage unit, which stores the characteristic template file into a database;
2) a skyhead identity authentication unit based on a skyhead image template, comprising
2.1) an actual image acquisition unit for acquiring an actual celestial sphere image;
2.2) a data transmission unit for transmitting the actual skyline image to a server;
and 2.3) the identification unit is used for calculating the distance between the pyramid gradient histogram feature of the extracted Tianzhu image and the Tianzhu feature in the feature template file stored in the database, and identifying the identity of the Tianzhu according to a set threshold value.
The invention has the beneficial effects that:
1. for non-Lambertian bodies such as celestial beads, the method can overcome characteristic matching under different illumination environments, and particularly solves the special matching problem that the celestial beads have specular reflection high-light areas under strong light irradiation;
2. the problem that the details of the celestial pearl image are inconsistent due to the multi-resolution difference of different celestial pearl image acquisition devices is solved, and the accuracy of identity identification is improved;
3. for a cylinder-like body or a spheroid like a skylinder, different images are collected at different shooting visual angles, and the skylinder or spheroid type image matching method can match the skylinder images shot at any visual angle when the identity of the skylinder or spheroid like the skylinder is identified in practice, so that the matching problem of multi-visual-angle images is solved;
4. the identification algorithm is simple, strong in real-time performance and high in identification precision, the automatic identification of the bead identity can be realized through portable image acquisition equipment such as a mobile phone, the operation process is simple and efficient, and the identification result is objective and accurate;
5. the manual identification process of experts is replaced, and the personal information and privacy protection of collectors are facilitated;
6. the Tianzhu identity identification method is easy to popularize, is convenient for Tianzhu lovers and beginners to learn and use, and plays a positive role in promoting the historical value inheritance and the standard collection market of Tianzhu.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate exemplary embodiments of the invention and together with the description serve to explain the principles of the invention. The above and other objects, features and advantages of the present invention will become more apparent from the detailed description of the embodiments of the present invention when taken in conjunction with the accompanying drawings.
FIG. 1 is a short cylindrical sky bead;
FIG. 2 is an ellipsoidal celestial bead;
FIG. 3 is a long cylindrical sky bead;
FIG. 4 is a skyline image template video capture device;
FIG. 5 is a schematic illustration of gray level histogram equalization;
FIG. 6 is a schematic diagram of bilinear interpolation;
FIG. 7 is a schematic diagram of a histogram of gradient (HOG) feature extraction process;
FIG. 8 is a flow chart of extracting PHOG features;
fig. 9 is a schematic diagram of a user identification process based on a mobile-end portable device celestial object identification.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are for purposes of illustration only and are not to be construed as limitations of the invention. It should be noted that, for convenience of description, only the portions related to the present invention are shown in the drawings.
In addition, the embodiments of the present invention and the features of the embodiments may be combined with each other without conflict. The present invention will be described in detail below with reference to the accompanying drawings in conjunction with embodiments.
The identification of the skybutton identity based on the image is implemented in two stages: firstly, a Tianzhu image template generation stage; and secondly, performing a Tianzhu identity authentication stage based on the Tianzhu image template.
(1) Tianzhu image template generation
In order to avoid the occurrence of a high light area due to external direct illumination when a sky-pearl image is collected as far as possible, the collection environment is supplemented with multipoint soft white light.
The captured image template uses a Canon EOS 5DSR camera, providing a maximum resolution 8688 x 5792, higher than that of a typical cell phone shot.
In order to reduce the difficulty of collecting the sky pearl image at multiple visual angles, the sky pearl is placed on a turntable, and a camera keeps a motionless video collection mode to rotationally collect the sky pearl image.
The above acquisition scheme is shown in fig. 4, and fig. 4 is a view of a skyline image template video acquisition device.
After the skyhead video is collected, performing frame extraction on the video, and selecting 120 frames at equal intervals when the skyhead rotates for one circle to cover all view angle images of the skyhead.
Since the background area is selected to be gray, the sky-pearl color is composed of black, white or yellow, and then the sky-pearl and background segmentation is performed through color difference, a rectangular area of the sky-pearl image is obtained.
And finally, extracting the pyramid gradient histogram features of the rectangular region of the celestial sphere image.
1) Carrying out normalization pretreatment on the rectangular area: the gray normalization is obtained by gray histogram equalization, as shown in fig. 2, and is calculated as shown in formula (1) and formula (2); the size normalization is obtained by bilinear interpolation, as shown in fig. 3, and the calculation is as shown in formula (3).
Wherein k represents the gray value of the rectangular area image before normalization preprocessing, nkRepresenting a grey level of rkM and N represent the height and width of the image, pr(rk) Representing the probability of occurrence, L representing the number of equalized gray levels, skRepresenting the corresponding value after histogram equalization.
Wherein x is1、y1、x2、y2Is a coordinate value, Q11＝(x1，y1) Represents Q11Point coordinates, like Q12＝(x1，y2)，Q21＝(x2，y1) And Q22＝(x2，y2) Respectively represent Q12、Q21And Q22Point coordinates. f (Q)11) Represents Q11Dot pixel gray scale value, f (Q)12) Represents Q12Dot pixel gray scale value, f (Q)21) Represents Q21Dot pixel gray scale value, f (Q)22) Represents Q22The dot pixel gray scale value is calculated to obtain a gray scale value f (x, y) at which the point P is (x, y).
2) And obtaining a binarization threshold value based on an OTSU (maximum inter-class variance method) algorithm, adding a prior threshold value to obtain a final threshold value, and performing binarization to obtain a Tianzhu texture image. And (3) iteratively solving a threshold value for maximizing the variance between the two classes by using the OTSU algorithm, wherein the calculation is shown as a formula (4).
Where t is the iterative solution threshold, ω0(t) and ω1(t) respectively represent the class probabilities,
and
representing the intra-class variance.
Finally, the threshold T plus the prior threshold T is solved by iteration0And obtaining a binarization threshold value T as shown in a formula (5).
T=t+T0 (5)
3) Gradient values and gradient directions of each pixel point are obtained based on the skyhead texture image, and the gradient directions are quantized into 9 directions to obtain a histogram of Gradients (HOG) feature, as shown in fig. 7, and the calculation is as shown in formula (6) -formula (9).
Wherein, I represents a gray-scale image,
representing a convolution, Gx、GyAnd G denotes a gradient value, and α denotes a gradient direction.
4) Dividing the skyline texture image area according to the height-to-width ratio of 1:1 to obtain 4 sub-areas in total, and extracting gradient histogram features from the 4 sub-areas respectively.
5) And dividing each sub-region according to the height-to-width ratio of 1:1 to obtain 16 sub-regions, and extracting gradient histogram features from the 16 sub-regions respectively.
6) And sequentially connecting the Histograms to obtain Pyramid gradient histogram (PHOG) features, wherein the feature dimension of the Pyramid gradient histogram (PHOG) features is (1+4+16) × 9 ═ 189-dimensional feature vector, normalizing the feature vector, and storing the feature vector as a feature template file for subsequent identification of the skulls.
The flow of the algorithm for extracting the gradient histogram feature of the skyhead image is shown in fig. 5.
(2) Tianzhu identity authentication stage based on Tianzhu image template
The bead identity authentication based on the bead image template is performed based on the bead image PHOG characteristic template. The method specifically comprises the following steps:
1) a user shoots an ancient pearl image by adopting portable image acquisition equipment (such as a mobile phone) with a network function, and the user is assisted to acquire the image by selecting an image frame which identifies the proportion of the sky pearl information to the real sky pearl.
2) And (3) sending the Tianzhu image in the image frame to a server background, and extracting the pyramid gradient histogram features of the Tianzhu image rectangular area, as shown in fig. 2.
3) Calculating the distance between the pyramid gradient histogram feature of the rectangular region of the extracted Tianzhu image and the Chamfer feature of the real Tianzhu collected and stored in the database, and calculating as shown in the formula (10); and finally, identifying the identity according to a set threshold value, as shown in a formula (11).
Wherein x istAnd xmRespectively representing the PHOG characteristics of the database skyline and the skyline to be detected uploaded by a user, wherein m represents the characteristic dimension, NmRepresenting a normalization constant.
Wherein, T1The identification threshold is preset, the identification threshold is represented as ID (1) and represents that the current user uploads the bead to be tested to be consistent with the database entry bead, and similarly, the identification threshold is represented as ID (0) and represents that the current user uploads the bead to be tested to be inconsistent with the database entry bead. The calculation method can still obtain a robust result under the condition that the local features of the image are lost, and the problem that the details of the image are annihilated in a highlight area is solved.
Wherein, the step 2) also comprises the following steps:
2.1) overall recognition of the skyhead image is firstly carried out, the gradient distribution of the overall pattern is analyzed by utilizing the collected whole skyhead image, the gradient distribution of the skyhead pattern in the skyhead template library is matched to obtain a similarity value, and the step 2.2 is carried out on the whole skyhead image of which the similarity value is greater than a threshold value;
2.2) carrying out network division on the whole Tianzhu image, wherein the Tianzhu texture image area is divided according to the height-to-width ratio of 1:1 to obtain 4 sub-areas in total, and extracting gradient histogram features from the 4 sub-areas respectively;
2.3) dividing each subregion again according to the height-to-width ratio of 1:1 to obtain 16 subregions, and extracting gradient histogram features from the 16 subregions respectively;
and 2.4) sequentially connecting the Histograms to obtain Pyramid gradient histogram (photo) features, wherein the feature dimension of the Pyramid gradient histogram is (1+4+16) × 9 ═ 189-dimensional feature vector, and normalizing to obtain the Pyramid gradient histogram features of the rectangular region of the celestial sphere image.
The above-described flow is shown in fig. 6.
Establishing a template library by constructing the characteristics of the pyramid gradient histogram by adopting the image pyramid gradient histogram-based skyhold identity identification method; in the actual matching process, the Chamfer distance between the extracted pyramid gradient histogram feature of the rectangular region of the celestial pearl image and the real celestial pearl feature collected and stored by the database is calculated, so that the problem that the details of the celestial pearl image in the region are annihilated due to the fact that the celestial pearl appears in the highlight region of mirror reflection under the irradiation of strong light can be solved, and the celestial pearl identification precision is improved; the special matching problem of a specular reflection high-light area of the celestial beads under strong light irradiation is solved. When the template library is established, the mode of rotationally collecting the skyhead images is adopted, the collected images have multiple visual angles, the template library is established based on the multiple visual angles, and the skyhead images shot at any visual angles can be matched when the actual skyhead identity is identified, so that the problem of multi-visual-angle image matching is solved. By adopting multi-resolution feature extraction, the problem that the details of the celestial pearl image are inconsistent due to the multi-resolution difference of different celestial pearl image acquisition devices is solved, and the accuracy of identity identification is improved. The identification method of the skyhead identity is simple in algorithm, small in operation amount, strong in real-time performance and high in identification precision, can realize automatic identification of the skyhead identity through portable image acquisition equipment such as a mobile phone, and is simple and efficient in operation process, and objective and accurate in identification result. The identification method of the skybutting identity can replace the manual identification process of experts and scholars, and is beneficial to protecting personal information and privacy of collectors. The method for identifying the identity of the Tianzhu is easy to popularize, is beneficial to the study and use of Tianzhu lovers and beginners, and plays a positive promoting role in the inheritance of the historical value and the standardized collection market of the Tianzhu.
So far, the technical solutions of the present invention have been described with reference to the preferred embodiments shown in the drawings, but it should be understood by those skilled in the art that the above embodiments are only for clearly illustrating the present invention, and not for limiting the scope of the present invention, and it is apparent that the scope of the present invention is not limited to these specific embodiments. Equivalent changes or substitutions of related technical features can be made by those skilled in the art without departing from the principle of the invention, and the technical scheme after the changes or substitutions can fall into the protection scope of the invention.
Claims (3)
1. A skyhold identity identification method based on image pyramid gradient histogram features is characterized by comprising the following steps:
1) skyhead image template generation comprising:
1.1) acquiring skyhead images from a plurality of different angles;
1.2) extracting the pyramid gradient histogram characteristics of the Tianzhu image to obtain a characteristic template file;
1.3) storing the characteristic template file into a database;
2) carry out Tianzhu identity authentication based on Tianzhu image template, include:
2.1) collecting an actual Tianzhu image;
2.2) sending the actual skatebead image to a server;
2.3) calculating the distance between the extracted pyramid gradient histogram feature of the Tianzhu image and the Tianzhu feature in the feature template file stored in the database, and identifying the identity of the Tianzhu according to a set threshold;
wherein the content of the first and second substances,
the method for extracting the pyramid gradient histogram features of the Tianzhu image comprises the following steps:
1.2.1) carrying out normalization preprocessing on the rectangular area: the gray level normalization is obtained by gray level histogram equalization, and is calculated as shown in formula (1) and formula (2),
wherein k represents the gray value of the rectangular area image before normalization preprocessing, nkRepresenting a grey level of rkM and N represent the height and width of the image, pr(rk) Representing the probability of occurrence, L representing the number of equalized gray levels, skTo representCorresponding values after histogram equalization;
the size normalization is obtained by bilinear interpolation, and the calculation is shown as a formula (3),
wherein x is1、y1、x2、y2Is a coordinate value, Q11＝(x1,y1) Represents Q11Point coordinates, like Q12＝(x1,y2)，Q21＝(x2,y1) And Q22＝(x2,y2) Respectively represent Q12、Q21And Q22Point coordinates, f (Q)11) Represents Q11Dot pixel gray scale value, f (Q)12) Represents Q12Dot pixel gray scale value, f (Q)21) Represents Q21Dot pixel gray scale value, f (Q)22) Represents Q22A point pixel gray value, which is calculated to obtain a gray value f (x, y) of a point P ═ x, y;
1.2.2) iteratively solving a threshold value for maximizing the variance between the two classes based on the maximum inter-class variance method, namely an OTSU algorithm, and calculating as shown in a formula (4),
where t is the iterative solution threshold, ω0(t) and ω1(t) respectively represent the class probabilities,
and
representing intra-class variance;
finally, the threshold T plus the prior threshold T is solved by iteration0Obtaining a binary threshold value T as shown in a formula (5),
T＝t+T0 (5)
1.2.3) obtaining the gradient value and gradient direction of each pixel point based on the skyhead texture image, quantizing the gradient direction into 9 directions to obtain the features of a histogram of Gradients (HOG), calculating as shown in formula (6) -formula (9),
wherein I represents a gray-scale image,
representing a convolution, Gx、GyAnd G represents a gradient value, α represents a gradient direction;
the acquiring of the actual celestial bead image comprises: shooting a skyline image by adopting portable image acquisition equipment with a network function, and acquiring the image by adopting an image frame in proportion to the real skyline according to the information of selecting and identifying the skyline;
sending the Tianzhu image in the image frame to a server background, and extracting the pyramid gradient histogram characteristics of the Tianzhu image rectangular area, wherein the pyramid gradient histogram characteristics comprise:
2.2.1) overall recognition of the celestial bead image is firstly carried out, the gradient distribution of the overall pattern is analyzed by utilizing the collected whole celestial bead image, the acquired whole celestial bead image is matched with the gradient distribution of the celestial bead pattern in a celestial bead template library to obtain a similarity value, and the step 2.2 is carried out on the whole celestial bead image with the similarity value larger than a threshold value;
2.2.2) carrying out network division on the whole Tianzhu image, wherein the Tianzhu texture image area is divided according to the height-to-width ratio of 1:1 to obtain 4 sub-areas in total, and extracting gradient histogram features from the 4 sub-areas respectively;
2.2.3) dividing each subregion again according to the height-to-width ratio of 1:1 to obtain 16 subregions, and extracting gradient histogram features from the 16 subregions respectively;
2.2.4) sequentially connecting the Histograms to obtain Pyramid gradient histogram (PHOG) features, wherein the feature dimension is (1+4+16) × 9 ═ 189-dimensional feature vector, and normalizing to obtain Pyramid gradient histogram features of rectangular regions of the skyhead image;
calculating the distance between the pyramid gradient histogram feature of the extracted Tianzhu image and the Tianzhu feature in the feature template file stored in the database to be the Chamfer distance, wherein the calculation formula is shown in (10),
wherein x istAnd xmRespectively representing the pyramid gradient histogram features of the database skyline and the skyline to be detected uploaded by a user, m represents the feature dimension, NmRepresents a normalization constant;
and identifying according to a set threshold value, as shown in formula (11):
wherein, T1The identification threshold is preset, the identification threshold is represented as ID (1) and represents that the current user uploads the bead to be tested to be consistent with the database entry bead, and similarly, the identification threshold is represented as ID (0) and represents that the current user uploads the bead to be tested to be inconsistent with the database entry bead.
2. The skyhold identity recognition method based on image pyramid gradient histogram features as claimed in claim 1,
the acquiring skyline images from a plurality of different angles comprises: the video acquisition equipment acquires the skyhead video, and after the acquisition is finished, the video is subjected to frame extraction operation, so that 120 frames are selected at equal intervals when the skyhead rotates for one circle, and all view angle images of the skyhead are covered.
3. The skyhold identity recognition method based on image pyramid gradient histogram features as claimed in claim 1,
the pyramid gradient histogram feature is a pyramid gradient histogram feature of a rectangular region of the celestial sphere image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910504289.3A CN110245670B (en) | 2019-06-11 | 2019-06-11 | Image pyramid gradient histogram feature-based skyhead identity identification method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910504289.3A CN110245670B (en) | 2019-06-11 | 2019-06-11 | Image pyramid gradient histogram feature-based skyhead identity identification method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110245670A CN110245670A (en) | 2019-09-17 |
| CN110245670B true CN110245670B (en) | 2021-08-20 |
Family
ID=67886623
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910504289.3A Active CN110245670B (en) | 2019-06-11 | 2019-06-11 | Image pyramid gradient histogram feature-based skyhead identity identification method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110245670B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116226821B (en) * | 2023-05-04 | 2023-07-18 | 成都致学教育科技有限公司 | Teaching data center management system |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102646199A (en) * | 2012-02-29 | 2012-08-22 | 湖北莲花山计算机视觉和信息科学研究院 | Motorcycle type identifying method in complex scene |
| CN102779157A (en) * | 2012-06-06 | 2012-11-14 | 北京京东世纪贸易有限公司 | Method and device for searching images |
| CN103678504A (en) * | 2013-11-19 | 2014-03-26 | 西安华海盈泰医疗信息技术有限公司 | Similarity-based breast image matching image searching method and system |
| CN105404657A (en) * | 2015-11-04 | 2016-03-16 | 北京工业大学 | CEDD feature and PHOG feature based image retrieval method |
| CN106326916A (en) * | 2016-08-11 | 2017-01-11 | 电子科技大学 | Multi-scale feature estimation and high-order BING feature-based target detection method |
| CN109086764A (en) * | 2018-07-25 | 2018-12-25 | 北京达佳互联信息技术有限公司 | Station caption detection method, device and storage medium |
| CN109461141A (en) * | 2018-10-10 | 2019-03-12 | 重庆大学 | A kind of workpiece starved detection method |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2012064893A2 (en) * | 2010-11-10 | 2012-05-18 | Google Inc. | Automated product attribute selection |
| US9076227B2 (en) * | 2012-10-01 | 2015-07-07 | Mitsubishi Electric Research Laboratories, Inc. | 3D object tracking in multiple 2D sequences |
| CN109410230B (en) * | 2018-09-07 | 2022-06-17 | 南京航空航天大学 | Improved Canny image edge detection method capable of resisting noise |
-
2019
- 2019-06-11 CN CN201910504289.3A patent/CN110245670B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102646199A (en) * | 2012-02-29 | 2012-08-22 | 湖北莲花山计算机视觉和信息科学研究院 | Motorcycle type identifying method in complex scene |
| CN102779157A (en) * | 2012-06-06 | 2012-11-14 | 北京京东世纪贸易有限公司 | Method and device for searching images |
| CN103678504A (en) * | 2013-11-19 | 2014-03-26 | 西安华海盈泰医疗信息技术有限公司 | Similarity-based breast image matching image searching method and system |
| CN105404657A (en) * | 2015-11-04 | 2016-03-16 | 北京工业大学 | CEDD feature and PHOG feature based image retrieval method |
| CN106326916A (en) * | 2016-08-11 | 2017-01-11 | 电子科技大学 | Multi-scale feature estimation and high-order BING feature-based target detection method |
| CN109086764A (en) * | 2018-07-25 | 2018-12-25 | 北京达佳互联信息技术有限公司 | Station caption detection method, device and storage medium |
| CN109461141A (en) * | 2018-10-10 | 2019-03-12 | 重庆大学 | A kind of workpiece starved detection method |
Non-Patent Citations (1)
| Title |
|---|
| 《Hand posture recognition based on heterogeneous features fusion of multiple kernels learning 》;Jiangtao Cao;《Multimed Tools Appl》;20161231;第11909–11928页； * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110245670A (en) | 2019-09-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110443143B (en) | Multi-branch convolutional neural network fused remote sensing image scene classification method | |
| CN111401384B (en) | Transformer equipment defect image matching method | |
| Zhang et al. | All-weather deep outdoor lighting estimation | |
| Goodfellow et al. | Multi-digit number recognition from street view imagery using deep convolutional neural networks | |
| Raghavendra et al. | Exploring the usefulness of light field cameras for biometrics: An empirical study on face and iris recognition | |
| CN104851099A (en) | Method for image fusion based on representation learning | |
| CN109714526B (en) | Intelligent camera and control system | |
| CN109685713B (en) | Cosmetic simulation control method, device, computer equipment and storage medium | |
| CN109635726B (en) | Landslide identification method based on combination of symmetric deep network and multi-scale pooling | |
| CN110263768A (en) | A kind of face identification method based on depth residual error network | |
| CN107766864B (en) | Method and device for extracting features and method and device for object recognition | |
| CN109886153A (en) | A kind of real-time face detection method based on depth convolutional neural networks | |
| CN109902662A (en) | A kind of pedestrian recognition methods, system, device and storage medium again | |
| CN111274944A (en) | Three-dimensional face reconstruction method based on single image | |
| CN115564993A (en) | Lip print image classification algorithm based on multi-scale feature fusion and attention mechanism | |
| CN114676777A (en) | Self-supervision learning fine-grained image classification method based on twin network | |
| CN110245670B (en) | Image pyramid gradient histogram feature-based skyhead identity identification method and device | |
| CN104915641B (en) | The method that facial image light source orientation is obtained based on Android platform | |
| CN114187506A (en) | Remote sensing image scene classification method of viewpoint-aware dynamic routing capsule network | |
| CN115410240A (en) | Intelligent face pockmark and color spot analysis method and device and storage medium | |
| CN111178370B (en) | Vehicle searching method and related device | |
| CN113436735A (en) | Body weight index prediction method, device and storage medium based on face structure measurement | |
| CN108960285B (en) | Classification model generation method, tongue image classification method and tongue image classification device | |
| Proença et al. | SHREC’15 Track: Retrieval of Oobjects captured with kinect one camera | |
| CN113011506B (en) | Texture image classification method based on deep fractal spectrum network |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |