CN110489587B - Tire trace image feature extraction method in local gradient direction three-value mode - Google Patents
Tire trace image feature extraction method in local gradient direction three-value mode Download PDFInfo
- Publication number
- CN110489587B CN110489587B CN201910699380.5A CN201910699380A CN110489587B CN 110489587 B CN110489587 B CN 110489587B CN 201910699380 A CN201910699380 A CN 201910699380A CN 110489587 B CN110489587 B CN 110489587B
- Authority
- CN
- China
- Prior art keywords
- image
- value
- gradient direction
- lgdtp
- tire trace
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/583—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
- G06F16/5862—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content using texture
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/40—Image enhancement or restoration by the use of histogram techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
Abstract
A tire trace image feature extraction method in a local gradient direction three-value mode comprises the steps of image preprocessing, feature extraction, feature vector determination and image retrieval. The invention provides a local gradient direction three-value mode characteristic suitable for a tire trace image, which adopts a more stable gradient direction value to replace a gray value to carry out local texture coding, carries out threshold quantization on a gradient direction angle of a central pixel, generates high-quality texture edge information and improves retrieval accuracy; and carrying out similarity calculation on the features described by the feature vectors by using Manhattan distance to obtain a search result, wherein the search accuracy is obviously superior to other texture features. The method has the advantages of clear texture edge information of the tire trace image, high retrieval accuracy, high average precision, suitability for large sample data and the like, and can be used for extracting the tire trace image characteristics.
Description
Technical Field
The invention belongs to the technical field of image processing, and particularly relates to an image feature extraction method.
Background
In general, in a traffic accident, the process of the accident is explained by utilizing the relation between the tire indentation marks and the material evidence left on site, and the responsibility of both sides is judged, and the ground tire marks are often one of the most useful mark material evidence, so that the tire mark retrieval is often used for obtaining clues in public security case breaking or traffic accident processing.
The research related to the tire trace image retrieval in China starts later, the research on the tire trace is more fresh and has more achievements, and a standard tire trace image test database does not exist in the field of the tire trace image retrieval and classification research. At present, classification retrieval is mainly carried out on tire trace images by a traditional feature combination method, and system analysis is not carried out on the characteristics of the tire trace images. For example: by analyzing SIFT transformation and Gabor wavelet principle, a tire trace image pattern recognition method based on SIFT-Gabor transformation is provided. The tire trace image retrieval method based on sparse representation and probability latent semantic analysis is provided by combining unsupervised learning and hierarchical extraction features. A combined feature extraction method based on a Non-downsampled Contourlet transform (Non-Subsampled Contourlet Transform, NSCT) and a Gray Level Co-occurrence matrix, GLCM, and training a classifier using a multi-stage support vector machine (Support Vector Machine, SVM). The method obtains the tire trace image formed by the tire and the ground through analysis, wherein the tire trace image comprises basic characteristics of a main line groove and an edge texture line. Meanwhile, the edge texture lines are complex, the edge texture lines of different types of tires have different widths and directions, the texture lines are interconnected and wound, and trace colors are relatively single.
Among the various texture image descriptors, the local binary pattern (Local Binary Patterns, LBP) is a popular and powerful image descriptor. The method has low computational complexity, does not need training and learning, is easy for engineering realization, and therefore, receives wide attention in the fields of computer vision and image processing. Based on the original LBP approach, many improved LBP approaches have been proposed by the scholars. For example: local direction pattern (Local Directional Patterns, LDP). Enhanced local direction patterns (Enhanced Local Directional Patterns, eldp. Local direction numbers (Local Directional Number, LDN), etc.
In the technical field of image processing, a technical problem to be solved urgently at present is to provide a tire trace image feature extraction method.
Disclosure of Invention
The technical problem to be solved by the invention is to overcome the defects of the prior art, and provide the tire trace image feature extraction method which has clear texture edge information, high retrieval accuracy and high average precision and is suitable for a local gradient direction three-value mode of large sample data.
The technical scheme adopted for solving the technical problems is that the method comprises the following steps:
(1) Image preprocessing
And selecting 50-80 images of 30 tire trace sample images from the tire trace image database for size normalization and graying treatment.
(2) Feature extraction
1) Determining image difference G of image along x direction by Sobel edge detection method x And image difference G in y direction y The gradient direction angle α (x, y) of each pixel in the image is determined using equation (1)
α(x,y)＝arctan(G y /G x ) (1)
2) Determining each gradient direction angle value in a 3×3 neighborhood sliding window by using a local gradient direction three-value pattern LGDTP method, wherein the local gradient direction three-value pattern LGDTP method is added with a self-defined threshold t, g i Greater than the interval [ g ] c -t,g c +t]When the value is 1, the value belongs to the interval of 0, and the value is less than the interval of-1, so as to obtain the three-value coding value.
Wherein P is the number of neighborhood pixels, R is the radius of the neighborhood, 0 < t < 2 pi, g c Is the gradient direction angle, g, of the center pixel i Is the gradient direction angle of its neighborhood pixels.
3) Decomposing the LGDTP characteristic value into positive and negative coding characteristic values, and modifying the coding value from 1 to 0 to obtain a positive coding characteristic value LGDTP P The coding value is modified to be 1 when the coding value is-1, and the other coding values are modified to be 0, so that a negative coding characteristic value L is obtainedGDTP M 。
4) Positive coded feature value LGDTP P Constructing a positive coded image instead of pixel values in the image, negative coded feature values LGDTP M Instead of pixel values in the image, a negative coded image is constructed.
(3) Determining feature vectors
1) The positive coding feature value LGDTP P Positive coded image and negative coded feature value LGDTP M The negative coded image is divided into 3×3 sub-blocks uniformly, and the sub-blocks are numbered 1 to 9 from top to bottom in order, and the pixels of each sub-block are m×n, m, n are 128 or 256.
2) Carrying out histogram statistics on gradient direction angles of pixels in each sub-block of the positive coded image, and cascading all sub-block histograms; and carrying out histogram statistics on gradient direction angles of pixels in each sub-block in the negative coding image, cascading all sub-block histograms, and cascading the sub-block histograms of the positive coding image and the negative coding image.
3) Taking the cascaded histogram as the characteristic of the whole tire trace image, and positively encoding the image LGDTP M And negative coded image LGDTP P The feature length is 256, and the feature length of the final encoded image LGDTP is 256×2×5, denoted as F LGDTP 。
4) Will F LGDTP The features are normalized according to the formula (4), and normalized feature vector values are obtained:
wherein F is c (t) is the t-th component of feature c,
is the normalized feature vector value.
(4) Image retrieval
1) Determining feature vector values for each tire trace image
And (5) respectively taking each image as a query image and each picture in the tire trace image database, and carrying out similarity measurement by a Manhattan distance measurement method:
where d is the distance length between two tire trace image features, X i ，X j A feature vector representing each tire trace;
2) The average precision P is used as a search performance evaluation index, and the following formula (6) is used for determining:
s is the number of correct images contained in the query result, K is the total number of images of the query result, and K is smaller than the number of each type of images in the tire trace image database.
In the step 1) of determining the feature vector in the step (3), m and n are positive integers capable of dividing 3 and are equal.
In step 2) of the feature extraction step (2) of the present invention, t in the formula (3) is optimally pi/6.
In the step 2) of determining the feature vector (3) according to the invention, the sub-block histograms are the histograms of the sub-areas with the reference numbers of 2,4,5,6 and 8 respectively according to the spatial distribution specificity of the texture information of the tire trace image.
The invention extracts image characteristics aiming at a tire trace image database image and a query image, and uses the query image to query the tire trace image of related types in the tire trace image database to determine the tire types. The tire trace image formed by the tire and the ground is obtained through analysis, wherein the tire trace image comprises the basic characteristics of the main line grooves and the edge texture lines. The invention provides a local gradient direction three-value mode characteristic suitable for a tire trace image, which adopts a more stable gradient direction value to replace a gray value to carry out local texture coding, and generates high-quality texture edge information by carrying out threshold quantization on a central pixel gradient direction angle, thereby improving the retrieval accuracy; and carrying out similarity calculation on the features described by the feature vectors by using Manhattan distance to obtain a search result, wherein the search accuracy is obviously superior to other texture features. The invention has the advantages of clear texture edge information of the tire trace image, high retrieval accuracy, high average precision, suitability for large sample data and the like, and can be used for extracting the tire trace image characteristics.
Drawings
Fig. 1 is a flow chart of embodiment 1 of the present invention.
Fig. 2 is a graph showing the results of comparative experiments between the method of example 1 and 6 conventional texture feature extraction methods.
Fig. 3 is a graph of a comparison experiment of the method of example 1 with 6 prior art texture feature extraction methods.
Detailed Description
The invention will be described in further detail with reference to the drawings and examples, but the invention is not limited to the following examples.
Example 1
The images of the embodiment are from a tire trace image database built by the applicant, and comprise 2400 images of 30 types of 80 images in total, and the tire trace image feature extraction method of the local gradient direction three-value mode comprises the following steps (see fig. 1):
(1) Image preprocessing
And selecting 2400 tire trace sample images of 30 types of 80 frames from a tire trace image database, and performing size normalization to 384×384 and gray-scale treatment.
(2) Feature extraction
1) Determining image difference G of image along x direction by Sobel edge detection method x And image difference G in y direction y The gradient direction angle α (x, y) of each pixel in the image is determined using equation (1)
α(x,y)＝arctan(G y /G x ) (7)
2) Determining each gradient direction angle value and local gradient in 3×3 neighborhood sliding window by using local gradient direction three-value pattern LGDTP methodThe method of the LGDTP with the three-value mode increases the self-defined threshold t, g i Greater than the interval [ g ] c -t,g c +t]When the value is 1, the value belongs to the interval of 0, and the value is less than the interval of-1, so as to obtain a three-value coded value:
where P is the number of neighborhood pixels, R is the radius of the neighborhood, t is 2π, g in this embodiment c Is the gradient direction angle, g, of the center pixel i Is the gradient direction angle of its neighborhood pixels.
3) Decomposing the LGDTP characteristic value into positive and negative coding characteristic values, and modifying the coding value from 1 to 0 to obtain a positive coding characteristic value LGDTP P The coding value is modified to be 1 when the-1 is used for coding, and the other coding values are modified to be 0 for obtaining the negative coding characteristic value LGDTP M 。
4) Positive coded feature value LGDTP P Constructing a positive coded image instead of pixel values in the image, negative coded feature values LGDTP M Instead of pixel values in the image, a negative coded image is constructed.
(3) Determining feature vectors
1) The positive coding feature value LGDTP P Positive coded image and negative coded feature value LGDTP M The negative coded image is divided into 3×3 sub-blocks uniformly, and the sub-blocks are numbered 1 to 9 from top to bottom in order, and the pixels of each sub-block are m×n, and m and n in this embodiment are 128.
2) Carrying out histogram statistics on gradient direction angles of pixels in each sub-block of the positive coded image, and cascading all sub-block histograms; and carrying out histogram statistics on gradient direction angles of pixels in each sub-block in the negative coding image, cascading all sub-block histograms, and cascading the sub-block histograms of the positive coding image and the negative coding image. The sub-block histograms are the histograms of the sub-areas with the reference numbers of 2,4,5,6 and 8 respectively according to the spatial distribution specificity of the texture information of the tire trace image.
3) Taking the cascaded histogram as the characteristic of the whole tire trace image, and positively encoding the image LGDTP M And negative coded image LGDTP P The feature length is 256, and the feature length of the final encoded image LGDTP is 256×2×5, denoted as F LGDTP 。
4) Will F LGDTP The features are normalized according to the formula (10), and normalized feature vector values are obtained:
wherein F is c (t) is the t-th component of feature c,
is the normalized feature vector value.
(4) Image retrieval
1) Determining feature vector values for each tire trace image
And (3) respectively taking each image as a query image and each picture in the tire trace image database, and carrying out similarity measurement according to (11) by using a Manhattan distance measurement method:
where d is the distance length between two tire trace image features, X i ，X j A feature vector representing each tire trace.
2) Using the average precision P as a search performance evaluation index, the following formula (12) is used for determination:
wherein S is the number of correct images contained in the query result, and K is the total number of images of the query result.
In this embodiment, the method of this embodiment and the local binary pattern LBP method, the local direction pattern LDP method, the enhanced local direction pattern ELDP method, the local direction pattern LDN method, the optimized local direction pattern OLDP method, and the local three-value pattern LTP method are adopted to perform a comparison search experiment using a manhattan distance in a tire trace image database, the local binary pattern LBP method, the local direction pattern LDP method, the enhanced local direction pattern ELDP method, the local direction pattern LDN method, the optimized local direction pattern OLDP method, and the feature dimension of the optimized local direction pattern OLDP method is 256×5, and the local three-value pattern LTP method is to decompose LTP feature values into positive and negative coding feature values, and positive coding feature values LTP P Constructing a positive coded image instead of pixel values in the image, negative coded characteristic values LTP M Constructing a negative coded image instead of pixel values in the image, positive coded characteristic values LTP P And negative coding eigenvalue LTP M The feature dimensions after cascading were 256×5×2, the average precision of each method was calculated according to 2) in step (4), and the experimental results are shown in table 1 and fig. 2. The coded image of several methods is shown in fig. 3, and the original picture, 6 methods, and the coded image of this embodiment are shown in fig. 3, wherein this embodiment is divided into positive coded feature images LGDTP P And negative coding feature image LGDTP M 。
Table 1 comparative experimental results of this example and 6 methods (k=10)
| Sequence number | Features (e.g. a character) | Dimension(s) | Average precision (k=10) |
| 1 | LBP | 1280 | 43.1％, |
| 2 | LDP | 1280 | 45％ |
| 3 | ELDP | 1280 | 49.5％, |
| 4 | LDN | 1280 | 47.3％ |
| 5 | OLDP | 1280 | 45.6％, |
| 6 | LTP | 2560 | 52.1％ |
| 7 | This embodiment | 2560 | 63.3％ |
As can be seen from table 1 and fig. 2, when K is 10, the average precision of the present embodiment is 63.3%, and the average precision of the local binary pattern LBP method, the local direction pattern LDP method, the enhanced local direction pattern ELDP method, the local direction number pattern LDN method, the optimized local direction pattern OLDP method, and the local three-value pattern LTP method is 43.1%, 45%, 49.5%, 47.3%, 45.6%, and 52.1%, respectively, so that the highest average precision is obtained by the present method. The coded image of the method of this embodiment is more clear.
Example 2
The images of the embodiment come from a tire trace image database built by the applicant, and comprise 2400 images of 30 types of 80 images in total, and the tire trace image characteristic extraction method of the local gradient direction three-value mode comprises the following steps:
(1) Image preprocessing
This step is the same as in example 1.
(2) Feature extraction
1) Determining image difference G of image along x direction by Sobel edge detection method x And image difference G in y direction y The gradient direction angle α (x, y) of each pixel in the image is determined using equation (1)
α(x,y)＝arctan(G y /G x ) (13)
2) Determining each gradient direction angle value in a 3×3 neighborhood sliding window by using a local gradient direction three-value pattern LGDTP method, wherein the local gradient direction three-value pattern LGDTP method is added with a self-defined threshold t, g i Greater than the interval [ g ] c -t,g c +t]When the value is 1, the value belongs to the interval of 0, and the value is less than the interval of-1, so as to obtain a three-value coded value:
where P is the number of neighborhood pixels, R is the radius of the neighborhood, t is pi/1 in this embodiment8，g c Is the gradient direction angle, g, of the center pixel i Is the gradient direction angle of its neighborhood pixels. Other steps in this step are the same as those of example 1.
The other steps were the same as in example 1.
Example 3
The images of the embodiment come from a tire trace image database built by the applicant, and comprise 2400 images of 30 types of 80 images in total, and the tire trace image characteristic extraction method of the local gradient direction three-value mode comprises the following steps:
(1) Image preprocessing
This step is the same as in example 1.
(2) Feature extraction
1) Determining image difference G of image along x direction by Sobel edge detection method x And image difference G in y direction y The gradient direction angle α (x, y) of each pixel in the image is determined using equation (1)
α(x,y)＝arctan(G y /G x ) (16)
2) Determining each gradient direction angle value in a 3×3 neighborhood sliding window by using a local gradient direction three-value pattern LGDTP method, wherein the local gradient direction three-value pattern LGDTP method is added with a self-defined threshold t, g i Greater than the interval [ g ] c -t,g c +t]When the value is 1, the value belongs to the interval of 0, and the value is less than the interval of-1, so as to obtain a three-value coded value:
where P is the number of neighborhood pixels, R is the radius of the neighborhood, t is 3 pi/2, g in this embodiment c Is the gradient direction angle, g, of the center pixel i Is the gradient direction angle of its neighborhood pixels. Other steps in this step are the same as those of example 1.
The other steps were the same as in example 1.
Example 4
In the image preprocessing step (1) of the above embodiments 1 to 3, 2400 pieces of tire mark sample images 30 are selected from the tire mark image database, and size normalization is performed to 768×768 pieces of tire mark sample images, and gradation is performed.
The step 1) of the step (3) of determining a feature vector in this embodiment is as follows: the positive coding feature value LGDTP P Positive coded image and negative coded feature value LGDTP M The negative coded image is divided into 3×3 sub-blocks uniformly, and the sub-blocks are numbered 1 to 9 from top to bottom in order, and the pixels of each sub-block are m×n, and m and n in this embodiment are 256. Other steps in this step are the same as those of example 1.
The other steps were the same as in example 1.
Claims (4)
1. The tire trace image feature extraction method of the local gradient direction three-value mode is characterized by comprising the following steps of:
(1) Image preprocessing
Selecting 30 types of tire trace sample images from a tire trace image database, and carrying out size normalization and graying treatment on 50-80 types of tire trace sample images;
(2) Feature extraction
1) Determining image difference G of image along x direction by Sobel edge detection method x And image difference G in y direction y The gradient direction angle α (x, y) of each pixel in the image is determined using equation (1)
α(x,y)＝arc tan(G y /G x ) (1)
2) Determining each gradient direction angle value in a 3×3 neighborhood sliding window by using a local gradient direction three-value pattern LGDTP method, wherein the local gradient direction three-value pattern LGDTP method is added with a self-defined threshold t, g i Greater than the interval [ g ] c -t,g c +t]When the value is 1, the value belongs to the interval of 0, and the value is minus 1, so as to obtain a three-value coding value:
wherein P is the number of neighborhood pixels, R is the radius of the neighborhood, 0 < t < 2 pi, g c Is the gradient direction angle, g, of the center pixel i Is the gradient direction angle of the neighborhood pixels;
3) Decomposing the LGDTP characteristic value into positive and negative coding characteristic values, and modifying the coding value from 1 to 0 to obtain a positive coding characteristic value LGDTP P The coding value is modified to be 1 when the-1 is used for coding, and the other coding values are modified to be 0 for obtaining the negative coding characteristic value LGDTP M ；
4) Positive coded feature value LGDTP P Constructing a positive coded image instead of pixel values in the image, negative coded feature values LGDTP M Constructing a negative coded image in place of pixel values in the image;
(3) Determining feature vectors
1) The positive coding feature value LGDTP P Positive coded image and negative coded feature value LGDTP M The negative coding image is divided into 3×3 sub-blocks uniformly, and the sub-blocks are numbered 1-9 from top to bottom in sequence, the pixels of each sub-region are m×n, and m and n are 128 or 256;
2) Carrying out histogram statistics on gradient direction angles of pixels in each sub-block of the positive coded image, and cascading all sub-block histograms; carrying out histogram statistics on gradient direction angles of pixels in each sub-block in the negative coding image, cascading all sub-block histograms, and cascading the sub-block histograms of the positive coding image and the negative coding image;
3) Taking the cascaded histogram as the characteristic of the whole tire trace image, and positively encoding the image LGDTP M And negative coded image LGDTP P The feature length is 256, and the feature length of the final encoded image LGDTP is 256×2×5, denoted as F LGDTP ；
4) Will F LGDTP Characterized by returning according to (4)And (3) carrying out a conversion process to obtain normalized characteristic vector values:
wherein F is c (t) is the t-th component of feature c,
is the normalized feature vector value;
(4) Image retrieval
1) Determining feature vector values for each tire trace image
And (5) respectively taking each image as a query image and each picture in the tire trace image database, and carrying out similarity measurement by a Manhattan distance measurement method:
where d is the distance length between two tire trace image features, X i ，X j A feature vector representing each tire trace;
2) The average precision P is used as a search performance evaluation index, and the following formula (6) is used for determining:
s is the number of correct images contained in the query result, K is the total number of images of the query result, and K is smaller than the number of each type of images in the tire trace image database.
2. The method for extracting tire trace image features in a local gradient direction three-value mode according to claim 1, wherein: in step 1) of determining the feature vector step (3), m and n are positive integers capable of dividing 3 and are equal.
3. The method for extracting tire trace image features in a local gradient direction three-value mode according to claim 1, wherein: in step 2) of the feature extraction step (2), t in the formula (3) is pi/6.
4. The method for extracting tire trace image features in a local gradient direction three-value mode according to claim 1, wherein: in step 2) of determining the feature vector (3), the sub-block histograms are histograms of sub-regions with the reference numerals of 2,4,5,6 and 8 respectively according to the spatial distribution specificity of the texture information of the tire trace image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910699380.5A CN110489587B (en) | 2019-07-31 | 2019-07-31 | Tire trace image feature extraction method in local gradient direction three-value mode |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910699380.5A CN110489587B (en) | 2019-07-31 | 2019-07-31 | Tire trace image feature extraction method in local gradient direction three-value mode |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110489587A CN110489587A (en) | 2019-11-22 |
| CN110489587B true CN110489587B (en) | 2023-04-28 |
Family
ID=68548845
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910699380.5A Active CN110489587B (en) | 2019-07-31 | 2019-07-31 | Tire trace image feature extraction method in local gradient direction three-value mode |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110489587B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112918956A (en) * | 2021-02-20 | 2021-06-08 | 陆伟凤 | Garbage classification system based on image recognition technology |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105389573A (en) * | 2015-12-23 | 2016-03-09 | 山东大学 | Face recognition method based on stratified reconstruction in local ternary pattern |
| EP3076367A1 (en) * | 2013-07-22 | 2016-10-05 | Zhejiang University | Method for road detection from one image |
| CN106228163A (en) * | 2016-07-25 | 2016-12-14 | 长安大学 | The local poor ternary sequential image feature that a kind of feature based selects describes method |
| CN106529504A (en) * | 2016-12-02 | 2017-03-22 | 合肥工业大学 | Dual-mode video emotion recognition method with composite spatial-temporal characteristic |
| CN107122396A (en) * | 2017-03-13 | 2017-09-01 | 西北大学 | Three-dimensional model searching algorithm based on depth convolutional neural networks |
| CN109035317A (en) * | 2018-07-04 | 2018-12-18 | 重庆邮电大学 | Illumination reversion and invariable rotary texture expression based on three value mode of gradient local |
| WO2019100436A1 (en) * | 2017-11-22 | 2019-05-31 | Zhejiang Dahua Technology Co., Ltd. | Methods and systems for face recognition |
| WO2019136894A1 (en) * | 2018-01-10 | 2019-07-18 | Zhejiang Dahua Technology Co., Ltd. | Methods and systems for face alignment |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| MXPA03003690A (en) * | 2000-10-27 | 2004-05-05 | Chiron Spa | Nucleic acids and proteins from streptococcus groups a b. |
-
2019
- 2019-07-31 CN CN201910699380.5A patent/CN110489587B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3076367A1 (en) * | 2013-07-22 | 2016-10-05 | Zhejiang University | Method for road detection from one image |
| CN105389573A (en) * | 2015-12-23 | 2016-03-09 | 山东大学 | Face recognition method based on stratified reconstruction in local ternary pattern |
| CN106228163A (en) * | 2016-07-25 | 2016-12-14 | 长安大学 | The local poor ternary sequential image feature that a kind of feature based selects describes method |
| CN106529504A (en) * | 2016-12-02 | 2017-03-22 | 合肥工业大学 | Dual-mode video emotion recognition method with composite spatial-temporal characteristic |
| CN107122396A (en) * | 2017-03-13 | 2017-09-01 | 西北大学 | Three-dimensional model searching algorithm based on depth convolutional neural networks |
| WO2019100436A1 (en) * | 2017-11-22 | 2019-05-31 | Zhejiang Dahua Technology Co., Ltd. | Methods and systems for face recognition |
| WO2019136894A1 (en) * | 2018-01-10 | 2019-07-18 | Zhejiang Dahua Technology Co., Ltd. | Methods and systems for face alignment |
| CN109035317A (en) * | 2018-07-04 | 2018-12-18 | 重庆邮电大学 | Illumination reversion and invariable rotary texture expression based on three value mode of gradient local |
Non-Patent Citations (5)
| Title |
|---|
| Automated Facial Expression Recognition Using Gradient-Based Ternary Texture Patterns;Faisal Ahmed;《Chinese Journal of Engineering》;20131124;1-8 * |
| Dynamic Scene Recognition Using Spatiotemporal Based DLTP on Spark;Md Azher Uddin;《IEEE Access》;20181031;第6卷;66123-66133 * |
| Expression recognition using directional gradient local pattern and gradient-based ternary texture patterns;Zeynab Shokoohi;《2015 2nd International Conference on Pattern Recognition and Image Analysis (IPRIA)》;20150720;1-7 * |
| 基于视频的人脸追踪与识别;赵飞飞;《中国优秀硕士学位论文全文数据库 (信息科技辑)》;20161115(第11期);I138-399 * |
| 改进特征与GPU加速的行人检测;齐美彬;《中国图象图形学报》;20180830;第23卷(第8期);1171-1180 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110489587A (en) | 2019-11-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Daniel Costea et al. | Word channel based multiscale pedestrian detection without image resizing and using only one classifier | |
| Zahran et al. | A MODIFIED LBP METHOD TO EXTRACT FEATURES FROM COLOR IMAGES. | |
| He et al. | Rotation invariant texture descriptor using local shearlet-based energy histograms | |
| CN103258202B (en) | A kind of texture characteristic extracting method of robust | |
| Ahmed et al. | Compound local binary pattern (clbp) for rotation invariant texture classification | |
| Ahmed et al. | Directional ternary pattern (DTP) for facial expression recognition | |
| Ma et al. | Integrating orientation cue with EOH-OLBP-based multilevel features for human detection | |
| Yamauchi et al. | Relational HOG feature with wild-card for object detection | |
| Sachdeva et al. | Better object recognition using bag of visual word model with compact vocabulary | |
| CN110489587B (en) | Tire trace image feature extraction method in local gradient direction three-value mode | |
| CN100345152C (en) | Face recognition method based on template matching | |
| Cheung et al. | Ultra local binary pattern for image texture analysis | |
| CN110674334B (en) | Near-repetitive image retrieval method based on consistency region deep learning features | |
| CN112784722A (en) | Behavior identification method based on YOLOv3 and bag-of-words model | |
| CN108805183B (en) | Image classification method fusing local aggregation descriptor and local linear coding | |
| CN110705569A (en) | Image local feature descriptor extraction method based on texture features | |
| Chekol et al. | Segmented character recognition using curvature-based global image feature | |
| Ni et al. | Contextualizing histogram | |
| CN112818779B (en) | Human behavior recognition method based on feature optimization and multiple feature fusion | |
| Park et al. | Image retrieval technique using rearranged freeman chain code | |
| Suri et al. | Robust face detection using circular multi block local binary pattern and integral haar features | |
| Bao et al. | Linear spatial pyramid matching using non-convex and non-negative sparse coding for image classification | |
| Tsai et al. | Image Copy-Move Forgery Detection using Color Features and Hierarchical Feature Point Matching. | |
| Liu et al. | Wavelet-energy-weighted local binary pattern analysis for tire tread pattern classification | |
| Lakshmi et al. | Plant leaf image detection method using a midpoint circle algorithm for shape-based feature extraction |
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 |