CN109446882B - Vehicle logo feature extraction and identification method based on feature quantification of gradient direction division - Google Patents
Vehicle logo feature extraction and identification method based on feature quantification of gradient direction division Download PDFInfo
- Publication number
- CN109446882B CN109446882B CN201811032667.4A CN201811032667A CN109446882B CN 109446882 B CN109446882 B CN 109446882B CN 201811032667 A CN201811032667 A CN 201811032667A CN 109446882 B CN109446882 B CN 109446882B
- Authority
- CN
- China
- Prior art keywords
- gradient
- features
- feature
- gradient direction
- size
- 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
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
- G06V20/58—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
- G06V20/584—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads of vehicle lights or traffic lights
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- 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
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
-
- 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
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/469—Contour-based spatial representations, e.g. vector-coding
- G06V10/473—Contour-based spatial representations, e.g. vector-coding using gradient analysis
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Probability & Statistics with Applications (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a vehicle logo feature extraction and identification method based on feature quantification of gradient direction division, which comprises the steps of preprocessing a vehicle logo image shot by a bayonet system, calculating the gradient size and direction of each pixel, and storing gradient information of all pixels into corresponding gradient matrixes; dividing k gradient directions in advance, counting the gradient sizes of all pixels around each pixel in the k gradient directions, and accumulating the gradient sizes and putting the gradient sizes into k different gradient size matrixes; respectively extracting LTP features of the k gradient size matrixes, and splicing the extracted k LTP features to obtain pixel features of the original car logo image; and classifying all the characteristics in the sample through K-means to obtain an offline codebook, and finally, carrying out classification and identification on the vehicle logo image by using an SVM (support vector machine). The invention provides a specific recognition scheme aiming at the car logo recognition in the checkpoint image, the recognition result has high accuracy, and the requirements of an actual intelligent traffic system can be met.
Description
Technical Field
The invention relates to the technical field of target identification, in particular to a vehicle logo feature extraction and identification method based on feature quantification of gradient direction division.
Background
In computer vision application, vehicle logo recognition is an important part in an intelligent traffic system all the time, and is widely applied to the fields of vehicle information acquisition, traffic flow analysis, violation hit vehicle judgment, traffic order regulation and the like. In an intelligent traffic system, a high-definition bayonet system adopts an advanced photoelectric technology, an image processing technology and a pattern recognition technology to shoot an image of each passing automobile. The method for extracting and identifying the vehicle logo features aims at high-definition images of the bayonets, defaults to use the existing vehicle logo positioning technology to find the position of the vehicle logo, and then extracts and identifies the features of the vehicle logo images.
In practical situations, many bayonet images are not ideal car logo images due to factors such as illumination, size, and orientation. It is necessary to improve the feature extraction method and classification method of the car logo image to identify the car logo more accurately. Common car logo recognition methods are mainly based on research feature extraction methods, include many feature extraction methods for face recognition, and are also applied to car logo recognition, including classification based on angular point features, SIFT features, HOG features, LBP features, LQP features, POEM features, and the like, and most of the methods adopt SVMs for classification.
In order to extract more effective features and reduce the calculation cost brought by a large number of features, the invention provides a vehicle logo feature extraction and identification method based on feature quantization of gradient direction division.
Disclosure of Invention
The invention aims to make up for the defects of the prior art and provides a vehicle logo feature extraction and identification method based on feature quantization of gradient direction division.
The invention is realized by the following technical scheme:
a vehicle logo feature extraction and identification method based on feature quantization of gradient direction division comprises the following steps:
(1) extracting gradient size and gradient direction information of the car logo image:
the collected car logo samples are divided into two types, training samples and testing samples. And carrying out normalization and graying processing on the training sample, calculating the gradient size Gv and the gradient direction Go of each pixel in the sample image, and storing the gradient size and gradient direction information into a gradient matrix.
(2) Gradient direction partitioning and gradient magnitude matrix generation:
the gradient direction is divided into k ranges, and for each pixel, pixels belonging to the same range in the neighborhood thereof are stored using one matrix (called gradient magnitude matrix). There are k gradient magnitude matrices for each emblem sample and then the gradient magnitude matrix for each emblem sample is calculated.
(3) Extracting features based on the gradient size matrix:
and extracting LTP features based on the k gradient size matrixes respectively, and connecting the LTP features in the k different directions corresponding to each element in series to serve as the LTP features corresponding to the pixels of the original image.
(4) Characteristic quantization:
and (3) screening all LTP characteristics extracted from the training sample, and discarding the characteristics with too low occurrence frequency. And then clustering by using K-means to obtain N clustering centers as a characteristic codebook.
(5) And (3) identifying the car logo:
and (4) in the vehicle logo recognition process, respectively carrying out feature extraction on the vehicle logo training set and the vehicle logo testing set by using the method in the step (3) for training and classifying the SVM.
The car logo feature extraction and identification method based on feature quantification divided in the gradient direction is characterized by comprising the following steps of: in the step (1), the storage mode of the gradient magnitude and the gradient direction in the gradient matrix is described as follows:
firstly, a structural body with two elements is defined, wherein the two elements respectively refer to the gradient magnitude and the gradient direction in the step (1), and each pixel corresponds to one structural body. For each structure for which the gradient magnitude and direction are calculated, they are stored in a two-dimensional array.
The car logo feature extraction and identification method based on feature quantification divided in the gradient direction is characterized by comprising the following steps of: in the step (2), the specific steps of gradient direction division and gradient magnitude matrix generation are as follows:
the gradient direction range of the pixels obtained by calculation in the step (1) is 0-180 degrees, the gradient direction range is evenly divided into k sub-ranges, and each sub-range corresponds to a matrix and is used for storing gradient size information, namely a gradient size matrix. The matrix corresponding to the k sub-ranges is M1、M2、…MkAnd the size of the matrix is the same as the size of the resolution of the car logo image. Counting the gradient size and gradient direction of neighborhood pixels around each pixel, accumulating the gradient size of the pixels belonging to the same gradient direction range, and respectively storing the results obtained by accumulation in M1、M2、…、MkAnd finally obtaining gradient size matrixes of k different directions at the positions of the pixels in the matrixes.
The car logo feature extraction and identification method based on feature quantification divided in the gradient direction is characterized by comprising the following steps of: in the step (3), the extraction process of the LTP features is as follows:
selecting a central value icCalculating the difference between the element values of the R radius neighborhood around the adjacent R radius and the element values, and coding the neighborhood with the relative central value variation in the range of t as 1 to icCodes greater than t are 2, icCodes less than t are 0 and t is a threshold. Then serially connecting the codes obtained from all neighborhoods of the central value in sequence to obtain a code sequence Pij(i is the center value; j is 1,2,3, …, k).
The coding sequence P corresponding to each element in the k gradient size matrixesi1、Pi2、Pi3、…、PikAre connected in series to obtain the LTP characteristic P of the corresponding pixel of the original car logo imagei。
The car logo feature extraction and identification method based on feature quantification divided in the gradient direction is characterized by comprising the following steps of: in the step (4), the characteristic quantization process is as follows:
and screening all the features extracted based on the training samples, if the occurrence frequency of the features is greater than m (m is a set threshold), reserving the features, and if the occurrence frequency of the features is less than or equal to m, ignoring the features. And clustering the screened features by using K-means, and iterating for a plurality of times to obtain N clustering centers. The N clustering centers are the feature codebooks obtained by quantizing all the features.
The car logo feature extraction and identification method based on feature quantification divided in the gradient direction is characterized by comprising the following steps of: in the step (5), the identification process of the car logo comprises the following steps:
for the test sample, each emblem image is divided into blocks. In each block, extracting the characteristics of the pixels, comparing the extracted characteristics with the clustering characteristics in the characteristic codebook one by one, calculating the clustering characteristic with the minimum Euclidean distance in the codebook, and accumulating by a voting method to form an N-dimensional characteristic histogram. And sequentially splicing the characteristic histograms corresponding to each block of one car logo image to finally obtain the characteristic vector of the whole car logo image, and training and classifying the characteristic vector.
The invention has the advantages that:
1. in the identification method:
(1) the method for extracting and identifying the vehicle logo features based on the feature quantification of the gradient direction division can more comprehensively describe the vehicle logo features. Different from the traditional feature description, the method reasonably combines gradient and edge texture information by dividing different gradient direction ranges, and greatly increases feature description information.
(2) The quantization method used for the features increases the feature description information, can efficiently reduce the dimension of the features, and retains the effective information of the features of the original image.
2. In recognition effect:
the vehicle logo feature extraction and identification method based on the feature quantification of the gradient direction division can well distinguish feature information among different vehicle logos in the bayonet vehicle logo image identification, and has high identification rate.
Drawings
FIG. 1 is a flow chart of the method of the present invention.
Fig. 2 is a diagram of a feature quantification and identification process.
Detailed Description
As shown in fig. 1, a feature extraction method based on gradient direction division includes the following steps:
(1) extracting gradient size and gradient direction information of the car logo image:
the collected car logo samples are divided into two types, training samples and testing samples. And carrying out normalization and graying processing on the training sample, calculating the gradient size Gv and the gradient direction Go of each pixel in the sample image, and storing the gradient size and gradient direction information into a gradient matrix.
(2) Gradient direction partitioning and gradient magnitude matrix generation:
the gradient direction is divided into k ranges, and for each pixel, pixels belonging to the same range in the neighborhood thereof are stored using one matrix (called gradient magnitude matrix). There are k gradient magnitude matrices for each emblem sample and then the gradient magnitude matrix for each emblem sample is calculated.
(3) Extracting features based on the gradient size matrix:
and extracting LTP features based on the k gradient size matrixes respectively, and connecting the LTP features in the k different directions corresponding to each element in series to serve as the LTP features corresponding to the pixels of the original image.
In the step (1), the storage mode of the gradient magnitude and the gradient direction in the gradient matrix is described as follows:
firstly, a structural body with two elements is defined, wherein the two elements respectively refer to the gradient magnitude and the gradient direction in the step (1), and each pixel corresponds to one structural body. For each structure for which the gradient magnitude and direction are calculated, they are stored in a two-dimensional array.
In the step (2), the specific steps of gradient direction division and gradient magnitude matrix generation are as follows:
the gradient direction range of the pixels obtained by calculation in the step (1) is 0-180 degrees, the gradient direction range is evenly divided into k sub-ranges, and each sub-range corresponds to a matrix and is used for storing gradient size information, namely a gradient size matrix. The matrix corresponding to the k sub-ranges is M1、M2、…MkAnd the size of the matrix is the same as the size of the resolution of the car logo image. Counting the gradient size and gradient direction of neighborhood pixels around each pixel, accumulating the gradient size of the pixels belonging to the same gradient direction range, and respectively storing the results obtained by accumulation in M1、M2、…、MkAnd finally obtaining gradient size matrixes of k different directions at the positions of the pixels in the matrixes.
In the step (3), the extraction process of the LTP features is as follows:
selecting a central value icCalculating the difference between the element values of the R radius neighborhood around the adjacent R radius and the element values, and coding the neighborhood with the relative central value variation in the range of t as 1 to icCodes greater than t are 2, icCodes less than t are 0 and t is a threshold. Then serially connecting the codes obtained from all neighborhoods of the central value in sequence to obtain a code sequence Pij(i is the center value; j is 1,2,3, …, k).
The coding sequence P corresponding to each element in the k gradient size matrixesi1、Pi2、Pi3、…、PikAre connected in series to obtain the LTP characteristic P of the corresponding pixel of the original car logo imagei。
As shown in fig. 2, the feature quantization and identification includes the following steps:
(4) characteristic quantization:
and (3) screening all LTP characteristics extracted from the training sample, and discarding the characteristics with too low occurrence frequency. And then clustering by using K-means to obtain N clustering centers as a characteristic codebook.
(5) And (3) identifying the car logo:
and (4) in the vehicle logo recognition process, respectively carrying out feature extraction on the vehicle logo training set and the vehicle logo testing set by using the method in the step (3) for training and classifying the SVM.
In the step (4), the characteristic quantization process is as follows:
and screening all the features extracted based on the training samples, if the occurrence frequency of the features is greater than m (m is a set threshold), reserving the features, and if the occurrence frequency of the features is less than or equal to m, ignoring the features. And clustering the screened features by using K-means, and iterating for a plurality of times to obtain N clustering centers. The N clustering centers are the feature codebooks obtained by quantizing all the features.
In the step (5), the identification process of the car logo comprises the following steps:
for the test sample, each emblem image is divided into blocks. In each block, the pixels are subjected to feature extraction according to the method in claim 3, the extracted features are compared with the clustering features in the feature codebook one by one, the clustering feature with the minimum Euclidean distance in the codebook is calculated, and an N-dimensional feature histogram is formed by accumulation through a voting method. And sequentially splicing the characteristic histograms corresponding to each block of one car logo image to finally obtain the characteristic vector of the whole car logo image, and training and classifying the characteristic vector.
When the car logo is identified, samples of various car logos are firstly acquired from the bayonet system, and the number of the samples of each type of car logo is required to be 20. Then, the method is mainly carried out according to the following five steps:
step 1: extracting gradient size and gradient direction information of the car logo image: the collected car logo samples are divided into two types, training samples and testing samples. Carrying out normalization and graying processing on the training sample, calculating the gradient size Gv and the gradient direction Go of each pixel in the sample image, and storing the information of the gradient size and the gradient direction into a gradient matrix;
firstly, a structural body with two elements is defined, wherein the two elements respectively refer to the gradient magnitude and the gradient direction in the step (1), and each pixel corresponds to one structural body. For each structure with calculated gradient size and direction, they are stored in a two-dimensional array (the array size is the same as the image resolution), which facilitates the subsequent gradient direction division.
Step 2:
gradient direction partitioning and gradient magnitude matrix generation: the gradient direction is divided into k ranges, and for each pixel, pixels belonging to the same range in the neighborhood thereof are stored using one matrix (called gradient magnitude matrix). There are k gradient magnitude matrices for each emblem sample and then the gradient magnitude matrix for each emblem sample is calculated.
The gradient direction range of the pixels obtained by calculation in the step (1) is 0-180 degrees, the gradient direction range is evenly divided into k sub-ranges, and each sub-range corresponds to a matrix and is used for storing gradient size information, namely a gradient size matrix. The matrix corresponding to the k sub-ranges is M1、M2、…MkSize of matrixThe resolution of the car logo image is the same as that of the car logo image. Counting the gradient size and gradient direction of neighborhood pixels around each pixel, accumulating the gradient size of the pixels belonging to the same gradient direction range, and respectively storing the results obtained by accumulation in M1、M2、…、MkAnd finally obtaining gradient size matrixes of k different directions at the positions of the pixels in the matrixes.
And step 3:
extracting features based on the gradient size matrix: and extracting LTP features based on the k gradient size matrixes respectively, and connecting the LTP features in the k different directions corresponding to each element in series to serve as the LTP features corresponding to the pixels of the original image.
Selecting a central value icCalculating the difference between the element values of the R radius neighborhood around the adjacent R radius and the element values, and coding the neighborhood with the relative central value variation in the range of t as 1 to icCodes greater than t are 2, icCodes less than t are 0 and t is a threshold. Then serially connecting the codes obtained from all neighborhoods of the central value in sequence to obtain a code sequence Pij(i is the center value; j is 1,2,3, …, k).
The coding sequence P corresponding to each element in the k gradient size matrixesi1、Pi2、Pi3、…、PikAre connected in series to obtain the LTP characteristic P of the corresponding pixel of the original car logo imagei。
And 4, step 4:
characteristic quantization: and (3) screening all LTP characteristics extracted from the training sample, and discarding the characteristics with too low occurrence frequency. And then clustering by using K-means to obtain N clustering centers as a characteristic codebook.
And screening all the features extracted based on the training samples, if the occurrence frequency of the features is greater than m (m is a set threshold), reserving the features, and if the occurrence frequency of the features is less than or equal to m, ignoring the features. And clustering the screened features by using K-means, and iterating for a plurality of times to obtain N clustering centers. The N clustering centers are the feature codebooks obtained by quantizing all the features.
And 5: and (4) in the vehicle logo recognition process, respectively carrying out feature extraction on the vehicle logo training set and the vehicle logo testing set by using the method in the step (3) for training and classifying the SVM.
For the test sample, each emblem image is divided into blocks. In each block, the pixels are subjected to feature extraction according to the method in claim 3, the extracted features are compared with the clustering features in the feature codebook one by one, the clustering feature with the minimum Euclidean distance in the codebook is calculated, and an N-dimensional feature histogram is formed by accumulation through a voting method. And sequentially splicing the characteristic histograms corresponding to each block of one car logo image to finally obtain the characteristic vector of the whole car logo image, and training and classifying the characteristic vector.
The unique features of the invention are as follows:
1. different from the traditional feature description, the method reasonably combines gradient information and edge texture information by dividing k different gradient direction ranges, and greatly increases feature description information.
2. The quantization method used for the features increases the feature description information, can efficiently reduce the dimension of the features, and retains the effective information of the features of the original image.
3. Experiments prove that the method for extracting and identifying the vehicle logo features based on the feature quantification of the gradient direction division can well distinguish feature information among different vehicle logos in the bayonet vehicle logo image identification, and has high identification rate.
In summary, the method for extracting and identifying the car logo features based on the feature quantification of the gradient direction division can play an important role in the identification of the bayonet car logo images, and is greatly helpful for realizing the identification of the car logos.
Claims (4)
1. A car logo feature extraction and identification method based on feature quantification divided in gradient direction is characterized in that: the method comprises the following steps:
(1) extracting gradient magnitude and gradient direction information of the car logo image;
dividing the collected car logo samples into two types, namely training samples and testing samples, carrying out normalization and graying on the training samples, calculating the gradient size Gv and the gradient direction Go of each pixel in a sample image, and storing the gradient size and gradient direction information into a gradient matrix;
(2) dividing the gradient direction and generating a gradient size matrix;
the gradient direction range of the pixels obtained by calculation in the step (1) is 0-180 degrees, the gradient direction range is evenly divided into k sub-ranges, each sub-range corresponds to a matrix and is used for storing gradient size information, the matrix is called as a gradient size matrix, and the matrix corresponding to the k sub-ranges in sequence is M1、M2、…MkThe size of the matrix is the same as the resolution of the car logo image, the gradient size and the gradient direction of neighborhood pixels around each pixel are counted, the gradient sizes of the pixels belonging to the same gradient direction range are accumulated, and the accumulated results are respectively stored in M1、M2、…、MkFinally obtaining gradient size matrixes of k different directions at the positions of the pixels in the matrixes;
(3) extracting features based on the gradient size matrix;
extracting LTP features based on the k gradient size matrixes respectively, and connecting the LTP features in k different directions corresponding to each element in series to serve as the LTP features corresponding to the pixels of the original image;
the extraction process of the LTP features is as follows:
selecting a central value icCalculating the difference between the element values of the R radius neighborhood around the adjacent R radius and the element values, and coding the neighborhood with the relative central value variation in the range of t as 1 to icCodes greater than t are 2, icThe codes less than t are 0, t is threshold value, then the codes obtained from all neighborhoods of central value are connected together in series according to sequence to obtain a code sequence PijI is the central value; j ═ 1,2,3, …, k;
the coding sequence P corresponding to each element in the k gradient size matrixesi1、Pi2、Pi3、…、PikAre connected in series to obtain the LTP characteristic P of the corresponding pixel of the original car logo imagei;
(4) Quantizing the characteristics;
screening all LTP characteristics extracted from a training sample, discarding the characteristics with too low occurrence frequency, and then clustering by using K-means to obtain N clustering centers serving as a characteristic codebook;
(5) recognizing the car logo;
and (4) in the vehicle logo recognition process, respectively carrying out feature extraction on the vehicle logo training set and the vehicle logo testing set by using the method in the step (3) for training and classifying the SVM.
2. The method for extracting and identifying vehicle logo features based on feature quantification of gradient direction division according to claim 1, wherein the method comprises the following steps: in the step (1), the storage mode of the gradient magnitude and the gradient direction in the gradient matrix is described as follows:
firstly, defining a structure body with two elements, namely the gradient size and the gradient direction in the step (1), wherein each pixel corresponds to one structure body, and storing the structure bodies with the calculated gradient size and gradient direction in a two-dimensional array.
3. The method for extracting and identifying vehicle logo features based on feature quantification of gradient direction division according to claim 1, wherein the method comprises the following steps: in the step (4), the characteristic quantization process is as follows:
screening all the features extracted based on the training sample, if the occurrence frequency of the features is greater than m and m is a set threshold value, reserving the features, if the occurrence frequency of the features is less than or equal to m, neglecting, clustering the screened features by using K-means, and iterating for a plurality of times to obtain N clustering centers, wherein the N clustering centers are the feature codebook obtained by quantizing all the features.
4. The method for extracting and identifying vehicle logo features based on feature quantification of gradient direction division according to claim 1, wherein the method comprises the following steps: in the step (5), the identification process of the car logo comprises the following steps:
according to the method, each car logo image is divided into a plurality of blocks according to a test sample, pixels are subjected to feature extraction in each block, the extracted features are compared with the clustering features in the feature codebook one by one, the clustering features with the minimum Euclidean distance in the codebook are calculated, an N-dimensional feature histogram is formed through the voting method in an accumulating mode, the feature histograms corresponding to each block of one car logo image are sequentially spliced, finally, the feature vector of the whole car logo image is obtained, and the feature vector is trained and classified.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811032667.4A CN109446882B (en) | 2018-09-05 | 2018-09-05 | Vehicle logo feature extraction and identification method based on feature quantification of gradient direction division |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811032667.4A CN109446882B (en) | 2018-09-05 | 2018-09-05 | Vehicle logo feature extraction and identification method based on feature quantification of gradient direction division |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109446882A CN109446882A (en) | 2019-03-08 |
CN109446882B true CN109446882B (en) | 2022-03-04 |
Family
ID=65530143
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811032667.4A Active CN109446882B (en) | 2018-09-05 | 2018-09-05 | Vehicle logo feature extraction and identification method based on feature quantification of gradient direction division |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109446882B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110795582A (en) * | 2019-10-31 | 2020-02-14 | 京东方科技集团股份有限公司 | Image recommendation method, system, terminal device and server |
CN111787179B (en) * | 2020-05-30 | 2022-02-15 | 上海电力大学 | Image hash acquisition method, image security authentication method and device |
CN111862178B (en) * | 2020-07-31 | 2022-09-30 | 易思维(杭州)科技有限公司 | Improved LBP feature extraction method |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8705866B2 (en) * | 2010-12-07 | 2014-04-22 | Sony Corporation | Region description and modeling for image subscene recognition |
CN106156374A (en) * | 2016-09-13 | 2016-11-23 | 华侨大学 | A kind of view-based access control model dictionary optimizes and the image search method of query expansion |
CN106650731A (en) * | 2016-12-23 | 2017-05-10 | 中山大学 | Robust license plate and logo recognition method |
CN106886599A (en) * | 2017-02-28 | 2017-06-23 | 北京京东尚科信息技术有限公司 | Image search method and device |
CN108122000A (en) * | 2017-11-28 | 2018-06-05 | 合肥工业大学 | A kind of automobile logo identification method of feature based study |
-
2018
- 2018-09-05 CN CN201811032667.4A patent/CN109446882B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8705866B2 (en) * | 2010-12-07 | 2014-04-22 | Sony Corporation | Region description and modeling for image subscene recognition |
CN106156374A (en) * | 2016-09-13 | 2016-11-23 | 华侨大学 | A kind of view-based access control model dictionary optimizes and the image search method of query expansion |
CN106650731A (en) * | 2016-12-23 | 2017-05-10 | 中山大学 | Robust license plate and logo recognition method |
CN106886599A (en) * | 2017-02-28 | 2017-06-23 | 北京京东尚科信息技术有限公司 | Image search method and device |
CN108122000A (en) * | 2017-11-28 | 2018-06-05 | 合肥工业大学 | A kind of automobile logo identification method of feature based study |
Non-Patent Citations (1)
Title |
---|
基于Adaboost和SVM的车标识别方法研究;闫向阳;《中国优秀硕士学位论文全文数据库》;20180430 * |
Also Published As
Publication number | Publication date |
---|---|
CN109446882A (en) | 2019-03-08 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108510467B (en) | SAR image target identification method based on depth deformable convolution neural network | |
CN109101924B (en) | Machine learning-based road traffic sign identification method | |
CN106650731B (en) | Robust license plate and vehicle logo recognition method | |
CN109446882B (en) | Vehicle logo feature extraction and identification method based on feature quantification of gradient direction division | |
CN105320705B (en) | The search method and device of similar vehicle | |
Bianco et al. | Logo recognition using cnn features | |
Manivannan et al. | HEp-2 cell classification using multi-resolution local patterns and ensemble SVMs | |
Hsieh et al. | Aircraft type recognition in satellite images | |
CN109344880B (en) | SAR image classification method based on multiple features and composite kernels | |
Ali et al. | Urdu natural scene character recognition using convolutional neural networks | |
Islam et al. | Real-time bangla license plate recognition system using faster r-cnn and ssd: A deep learning application | |
CN110826415A (en) | Method and device for re-identifying vehicles in scene image | |
Dorbe et al. | FCN and LSTM based computer vision system for recognition of vehicle type, license plate number, and registration country | |
Manzoor et al. | Vehicle make and model recognition using random forest classification for intelligent transportation systems | |
Jana et al. | Intra-class recognition of fruits using color and texture features with neural classifiers | |
Peng et al. | Vehicle classification using sparse coding and spatial pyramid matching | |
CN107153827B (en) | Identification processing method and device for hand back vein image | |
JP2008251029A (en) | Character recognition device and license plate recognition system | |
CN107273889B (en) | License plate recognition method based on statistics | |
WO2017129015A1 (en) | Vehicle type recognition method and apparatus | |
CN108122000B (en) | Vehicle logo recognition method based on feature learning | |
CN112257730A (en) | Plant pest image identification method, device, equipment and storage medium | |
Cui et al. | Coarse to fine patches-based multitemporal analysis of very high resolution satellite images | |
Prates et al. | An adaptive vehicle license plate detection at higher matching degree | |
Han et al. | Image recognition by learned linear subspace of combined bag-of-features and low-level features |
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 |