CN111626230B - Vehicle logo identification method and system based on feature enhancement - Google Patents
Vehicle logo identification method and system based on feature enhancement Download PDFInfo
- Publication number
- CN111626230B CN111626230B CN202010472817.4A CN202010472817A CN111626230B CN 111626230 B CN111626230 B CN 111626230B CN 202010472817 A CN202010472817 A CN 202010472817A CN 111626230 B CN111626230 B CN 111626230B
- Authority
- CN
- China
- Prior art keywords
- feature
- image
- map
- self
- features
- 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
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/62—Text, e.g. of license plates, overlay texts or captions on TV images
- G06V20/625—License plates
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/08—Detecting or categorising vehicles
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Abstract
The invention discloses a car logo recognition method and system based on feature enhancement, belonging to the technical field of image recognition, wherein the method comprises the steps of preprocessing a car logo image to obtain a processed image, and extracting the integral feature and the self-symmetry similarity feature of the processed image; processing the processed image to obtain a saliency map, obtaining a local saliency block according to the saliency map, and extracting the local saliency block to obtain local saliency block features; then fusing the characteristics to obtain final characteristics; and identifying the final characteristics to obtain a car logo identification result. The system is characterized in that the preprocessing unit and the feature fusion unit are respectively connected with the feature extraction unit, and the feature fusion unit is connected with the identification unit. The invention aims to overcome the defect of low vehicle logo recognition accuracy in the prior art, can improve the vehicle logo recognition accuracy, further can realize efficient vehicle logo recognition under the condition of few training samples, and has better vehicle logo recognition effect.
Description
Technical Field
The invention relates to the field of image recognition, in particular to a car logo recognition method and system based on feature enhancement.
Background
In an intelligent transportation system, information recognition of vehicles is an important component, and among them, identification of vehicle logos plays an important role. The vehicle logo is a significant mark of a vehicle, has strong representativeness, is not easy to damage or tamper, can well represent vehicle information, and can play an important supporting role in a traffic monitoring system, vehicle tracking and vehicle identification. In the prior art, the identification of the car logo is mostly based on the improvement of some classical algorithms, and the actual road image often does not have a good identification effect. In a traffic system, images of passing vehicles captured by a road monitoring system are often influenced by various external factors, so that the image quality is low, and the problems of uneven illumination, shadow, noise, blur and the like occur; in addition, the car logo image obtained based on the actual road image positioning has smaller resolution, and the recognition effect is influenced; moreover, although the car logo positioning technology is gradually mature, the car logo cannot be positioned accurately, the car logo image obtained by segmentation is subjected to position deviation or small occupation ratio in the image, and redundant background information also has a certain influence on the identification effect.
In view of the above problems, some solutions are also provided in the prior art, such as the invention and creation names: the scheme discloses a vehicle logo identification method and a vehicle logo identification system (application date: 2015, 9, 17, and application number: 201510599186.1), and the method comprises the following steps: recognizing the position of a license plate from an image, preliminarily determining an image of the vehicle logo to be recognized according to the position relation between the license plate and the vehicle logo, then performing morphological image processing to obtain an accurate image of the vehicle logo to be recognized, judging the type of the image of the vehicle logo to be recognized, performing template matching and constant moment characteristic matching on the image to be recognized and a vehicle logo template image of a corresponding type in a standard library, weighting to calculate a final matching system, and outputting a vehicle logo recognition result according to a comparison result of a maximum final matching coefficient and a given threshold value. The invention preliminarily determines the position of the car logo according to the position of the license plate, precisely positions the car logo according to morphological transformation, realizes car logo recognition by combining template matching and feature matching, positions and recognizes layer by layer, and has simple algorithm, less memory consumption and high processing speed. However, the disadvantages of this solution are: in the scheme, the car logo image is limited to a certain extent when the template matching coefficient is calculated, and when the car logo is deviated in position or small in proportion in the image, the matching result can be greatly influenced.
In addition, the invention and creation name is: a car logo recognition based on point pair characteristics (application date: 2015, 8, 14 days; application number: 2015105004906) discloses a car logo recognition method based on point pair characteristics, which comprises the steps of preprocessing a car logo pattern, and forming a standard pattern of the car logo through judgment, binarization and normalization of front and back backgrounds; then, based on the standard pattern, extracting skeleton regions in the front background and the back background, and extracting characteristic point pairs by randomly taking points to form standard point pairs; judging the effectiveness of the characteristic point pairs through vehicle logo samples intercepted from the actual bayonet images, storing the effective point pairs into a database to form a characteristic point pair database, and calculating the judging threshold values of various vehicle logos; and extracting the characteristic point pair template in the database, and performing multi-scale matching with the vehicle logo candidate area so as to identify the vehicle logo. The scheme provides a specific recognition scheme for recognizing the car logo in the checkpoint image, the recognition result accuracy is high, and the needs of an actual intelligent traffic system can be met. However, the scheme has the following defects: according to the scheme, only random point pair information in the foreground background framework is considered, the overall characteristics are ignored, and when the framework area in the car logo image is affected by local illumination or shadow, the identification effect is not stable.
In summary, how to improve the identification accuracy of the car logo is an urgent problem to be solved in the prior art.
Disclosure of Invention
1. Problems to be solved
The invention aims to overcome the defect of low identification accuracy of car logos in the prior art, and provides a car logo identification method and system based on feature enhancement, which can improve the identification accuracy of the car logos, further realize high-efficiency identification of the car logos under the condition of few training samples, and have better car logo identification effect.
2. Technical scheme
In order to solve the problems, the technical scheme adopted by the invention is as follows:
the invention relates to a vehicle logo recognition method based on feature enhancement, which comprises the steps of inputting a vehicle logo image and preprocessing the vehicle logo image to obtain a processed image; then extracting the integral characteristic and the self-symmetry similarity characteristic of the processed image; then, background elimination is carried out on the processed image to obtain an amplitude map, a saliency map is obtained through calculation according to the amplitude map, the processed image is segmented according to the saliency map to obtain local salient blocks, and then feature extraction is carried out on the local salient blocks to obtain local salient block features; then, fusing the overall characteristic, the self-symmetry similarity characteristic and the local significant block characteristic to obtain a final characteristic; and finally, identifying the final characteristics to obtain a vehicle logo identification result.
The specific process of preprocessing the car logo image comprises the following steps: normalizing the car logo image to obtain an image with a standard size, and performing graying processing on the image to obtain a processed image; the standard size of the image is w0 × h0, w0 represents the width of the image, and h0 represents the height of the image.
Furthermore, the specific process of extracting the self-symmetry similarity features of the processed image is as follows: setting a sliding window with width of m and height of n, sliding the sliding window on the processed image in the horizontal direction by step length l to obtain k sliding windows, wherein the size of k is
Will be first
Turning the images of the k sliding windows, and extracting the HOL characteristics [ f ] of the images in all the sliding windows (1) ,f (2) ,…f (k) ](ii) a In turn so as to->
For the reference feature, calculating a reference feature and
the Euclidean distance between all the HOL features is calculated, and the minimum value in the calculated Euclidean distance is used as the self-symmetry similarity feature corresponding to the reference feature; then slave->
And selecting the minimum value from the corresponding self-symmetry similarity characteristics as the self-symmetry similarity characteristics of the processed image.
Further, the specific process of obtaining the amplitude map by performing background elimination on the processed image is as follows: and filtering the processed image in the horizontal direction and the vertical direction by using an edge detection operator to obtain an amplitude map.
Furthermore, the specific process of calculating the saliency map according to the magnitude map is as follows: setting an amplitude map as W, setting g = (i, j) as pixel points in the amplitude map W, setting a neighborhood block corresponding to each pixel point as p (g), and calculating the intensity value of the neighborhood block according to the following formula:
wherein, W ij The amplitude of the pixel points in the neighborhood block;
calculating intensity values corresponding to all the pixel points to obtain an intensity map; and then calculating according to the intensity map to obtain a significance map.
Furthermore, the specific process of calculating the saliency map according to the intensity map is as follows:
calculating the centroid of the intensity graph T, wherein the position of the centroid in the horizontal direction
Position of the center of mass in the vertical direction->
Where w and h are the width and height of the intensity map, and S (x, y) is the corresponding intensity value at the (x, y) position;
then using the centroid g 0 ＝(M x ,M y ) Taking the neighborhood block of the position as a reference, calculating the similarity s and the position distance d between other neighborhood blocks and the reference neighborhood block by the following formulas, and obtaining the corresponding correlation degree f:
s＝|T(g)-T(g 0 )| 2
wherein e is x ＝t*w，e y ＝t*h；And obtaining a significance characteristic value R (g) =1-exp (-R f (p (g), p (g)) according to the correlation degree f 0 ) T) and r) are coefficients; and then obtaining a significance map according to the significance characteristic value.
Further, the final features are identified by a CRC classifier to obtain a car logo identification result.
The invention relates to a vehicle logo recognition system based on feature enhancement. Furthermore, the system comprises a preprocessing unit, a feature extraction unit, a feature fusion unit and an identification unit, wherein the preprocessing unit and the feature fusion unit are respectively connected with the feature extraction unit, and the feature fusion unit is connected with the identification unit; the feature extraction unit is used for extracting overall features, self-symmetry similarity features and local salient block features.
Furthermore, the feature extraction unit comprises an overall feature extraction module, a self-symmetry similarity feature extraction module and a local significant block feature extraction module, and the overall feature extraction module, the self-symmetry similarity feature extraction module and the local significant block feature extraction module are respectively connected with the feature fusion unit.
3. Advantageous effects
Compared with the prior art, the invention has the beneficial effects that:
(1) According to the feature enhancement-based vehicle logo identification method, the effective features of the vehicle logo image can be obtained through the overall features, the symmetric characteristics of the vehicle logo can be represented through the self-symmetric similarity features, the local information with the dense features in the vehicle logo image can be extracted through the local significant block features, namely, the overall and effective description of the vehicle logo image features can be realized through the overall features, the self-symmetric similarity features and the local significant block features, so that the vehicle logo can be identified efficiently, and the identification effect of the vehicle logo in an actual scene is further improved.
(2) According to the vehicle logo recognition system based on feature enhancement, the feature extraction unit can be used for extracting the overall features, the self-symmetry similarity features and the local significant block features of the vehicle logo image, so that the effective feature information of the vehicle logo image can be obtained, and the recognition effect of the vehicle logo can be improved.
Drawings
FIG. 1 is a schematic flow chart of a vehicle logo recognition method based on feature enhancement according to the present invention;
fig. 2 is a schematic structural diagram of a car logo system based on feature enhancement according to the present invention.
The reference numerals in the schematic drawings illustrate:
100. a pre-processing unit; 200. a feature extraction unit; 210. an overall feature extraction module; 220. a self-symmetry similarity feature extraction module; 230. a local significant block feature extraction module; 300. a feature fusion unit; 400. and a recognition unit.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some embodiments of the present invention, but not all embodiments; moreover, the embodiments are not relatively independent, and can be combined with each other according to needs, so that a better effect is achieved. Thus, the following detailed description of the embodiments of the present invention, presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
For a further understanding of the invention, reference should be made to the following detailed description taken in conjunction with the accompanying drawings and examples.
Example 1
Referring to fig. 1, the method for identifying a vehicle logo based on feature enhancement of the present invention specifically includes the following steps:
1) Preprocessing car logo images
Inputting a car logo image and preprocessing the car logo image to obtain a processed image; specifically, the input car logo image is a car logo image shot by various camera devices; the specific process of preprocessing the car logo image comprises the following steps: firstly, normalizing the car logo image to obtain an image with a standard size, and then performing graying processing on the image to obtain a processed image; where the standard size of an image is w0 × h0, w0 denotes the width of the image, h0 denotes the height of the image, and w0 and h0 are pixel values. It should be noted that, since the different sizes of the input car logo images may affect the feature extraction, the car logo images are reduced or enlarged to a standard size, which is 64 × 64 in this embodiment; in addition, the image is converted into a single-channel gray image from a three-channel color image after the graying processing.
2) Feature extraction
i) The method comprises the following steps of extracting and processing the overall characteristics of an image based on the HOL characteristics, wherein the specific process comprises the following steps: firstly, dividing a processed image into J multiplied by J small areas, combining each Z multiplied by Z small areas into a block, and overlapping the blocks. In this embodiment, the processed image is divided into 8 × 8 small regions, each 2 × 2 small regions are combined into one block, and the overlapping rate between adjacent blocks is 50%. And then filtering the image by using a Gabor filtering template with 12 directions and the same size, processing each pixel point in the image to obtain 12 amplitude values, and selecting the minimum amplitude value and the corresponding direction thereof for the next operation.
Further, for each small region, accumulating the amplitude of each pixel point in corresponding 12 directions according to the direction of each pixel point to calculate the feature vector of each small region; for each block, the feature vectors of 2 × 2 small regions included in the block are connected in series to obtain the features of the block, and finally the features of all the blocks are connected in series to obtain the overall features of the processed image.
Further, the self-symmetry similarity feature of the processed image is extracted, and the specific process is as follows: setting a sliding window with width m and height n, wherein n = h0; sliding the sliding window on the processed image in the horizontal direction by step length l to obtain k sliding windows, where k is the size of k in this embodiment
Will be first
Turning the images in the k sliding window, and extracting the HOL characteristics [ f ] of the images in all the sliding windows (1) ,f (2) ,…f (k) ](ii) a It should be noted that, specifically, pixels in the sliding window are exchanged left and right with the window central axis as an axis, and the sliding window located on the right side of the image central axis is flipped in this embodiment. Then, calculating the minimum Euclidean distance between sliding windows on two sides of the central axis of the image according to the HOL characteristics, wherein the minimum Euclidean distance is the self-symmetry similarity characteristic; i.e. in sequence with +>
For the reference feature, calculating a reference feature and
the Euclidean distance between all the HOL features is calculated, and the minimum value in the calculated Euclidean distance is used as the self-symmetry similarity feature corresponding to the reference feature; then slave->
And selecting the minimum value from the corresponding self-symmetry similarity characteristics as the self-symmetry similarity characteristics of the processed image. The method comprises the following specific steps:
with f (1) Computing HOL features for reference features
To f (k) Each is independently and f (1) Selects the minimum distance value from all the distance values>
As f (1) Corresponding self-symmetric similarity features; then with f (2) As a reference feature, an HOL feature is calculated>
To f (k) Each is independently and f (2) The minimum distance value is selected from all the distance values as f (2) Corresponding self-symmetric similarity features; according to the above-mentioned steps of cyclic calculation, until it is counted as->
Corresponding self-symmetric similarity features; it should be noted that the self-symmetry similarity feature obtained by each calculation is compared with the self-symmetry similarity feature obtained by the previous calculation, and a smaller value is selected as the preselected self-symmetry similarity feature, that is, the final self-symmetry similarity feature of the processed image is the minimum value of the obtained self-symmetry similarity features.
It is worth explaining that the symmetric information of the car logo image can be obtained according to the self-symmetric similarity characteristics, namely, the symmetric structure or the asymmetric structure of the car logo image can be used as effective judgment information to assist the classification and the identification of the car logo image, so that the identification accuracy of the car logo can be improved.
ii) the specific steps of extracting the local significant block features are as follows: the method comprises the steps of firstly carrying out background elimination on a processed image to obtain an amplitude map, calculating according to the amplitude map to obtain a saliency map, then segmenting the processed image according to the saliency map to obtain local salient blocks, and then extracting the local salient blocks to obtain local salient block characteristics. It should be noted that the specific process of obtaining the amplitude map by performing background removal on the processed image is as follows: filtering the processed image in the horizontal direction and the vertical direction to obtain an amplitude map; in this embodiment, a Prewitt edge detection operator and a direction template Q are used to eliminate the background of the processed image, and specifically,
Q＝[-1 1]
wherein P and P T The templates in the vertical and horizontal directions of the Prewitt edge detection operator are respectively, and Q is a direction template. W is a group of x And W y Are the resulting amplitude maps in the horizontal and vertical directions. Recalculating W x And W y The intensity value E of (a) is,
wherein, W ij Is the amplitude at pixel point (i, j); w is to be x Corresponding intensity value and W y And comparing the corresponding intensity values, and selecting the amplitude diagram corresponding to the smaller value for further processing.
The specific process of calculating the saliency map according to the amplitude map comprises the following steps: let the amplitude map be W, let g = (i, j) be pixel points in the amplitude map W, and the neighborhood block corresponding to each pixel point is p (g), and it should be noted that the neighborhood block is an image block of a × a size with the pixel point as the center, and the size of the neighborhood block in this embodiment is 17 × 17.
The intensity of p (g) is calculated in the manner of
Wherein W ij The amplitude of the pixel points in the neighborhood block; and calculating the intensity of the neighborhood blocks of all the pixel points to obtain an intensity graph T with the same size as the amplitude graph W. In order to calculate the intensity of the pixel neighborhood block corresponding to the edge pixel point, the amplitude map W is expanded to the periphery and is filled with 0. It is worth to be noted that the intensity map can highlight the dense features with significance in the image, and meanwhile, since the significance region in the image is composed of a plurality of neighborhood blocks, the distance and similarity between the neighborhood blocks can also reflect the possibility that different neighborhood blocks belong to the significance region. When the distance between the adjacent domain blocks is smaller and the similarity is larger, the possibility that the adjacent domain blocks belong to the significant region is larger.
To represent the possibility that the neighborhood blocks corresponding to different pixel points belong to a salient region, the quality of the intensity map is first computedPosition of heart, center of mass in horizontal direction
Position in the vertical direction>
Wherein w and h are the width and height of the intensity map, and w = w0, h = h0; s (x, y) is the corresponding intensity value at the (x, y) position; with centroid g0= (M) x ,M y ) The neighborhood block of the position is taken as a reference, and the similarity between other neighborhood blocks and the reference neighborhood block is calculated through a distance formula s And the position distance d, and obtaining the corresponding correlation degree f, wherein e x And e y Related to the width and height of the image, e x ＝t*w，e y ＝t*h。
The specific calculation distance formula is as follows:
s＝|T(g)-T(g 0 )| 2
and obtaining a significance characteristic value R (g) =1-exp (-R f (p (g), p (g)) according to the correlation degree f 0 ) T) and r) are coefficients to control the eigenvalues in the saliency region to be within the desired range; t =0.1, r =20 in the present example; and obtaining a final saliency map according to the saliency characteristic values. And then mapping the saliency region of the saliency map onto a processed image to obtain local saliency blocks through segmentation, specifically, binarizing the saliency map to obtain angular points of the saliency map, obtaining a circumscribed rectangular frame of the saliency region based on the relative positions of the angular points, mapping the positions of the circumscribed rectangular frame onto the processed image to segment the local saliency blocks, and finally, performing HOL feature extraction on the local saliency blocks to obtain local saliency block features. It is worth mentioning that for the global feature, the self-symmetric similarity featureAnd the extraction sequence of the local significant block features has no specific requirement, for example, the local significant block features can be extracted first, and then the overall features and the self-symmetry similarity features can be extracted.
3) Feature fusion
Fusing the overall characteristic, the self-symmetry similarity characteristic and the local significant block characteristic to obtain a final characteristic; specifically, the overall characteristic, the self-symmetry similarity characteristic and the local salient block characteristic are connected in series in a characteristic splicing mode to obtain a final characteristic.
4) Feature identification
The CRC classifier identifies the final features to obtain a car logo identification result; it is worth to be noted that, in the invention, a training set and a test set are adopted to train and classify the CRC classifier, specifically, the training set and the test set respectively comprise a plurality of car logo images, and then based on the steps, the final characteristics corresponding to the car logo images can be obtained; and then training and classifying the CRC classifier by using the final characteristics corresponding to the car logo images in the training set and the test set. The CRC classifier may then identify the corresponding emblem image based on the final characteristics of the emblem image. It is worth to be noted that the CRC classifier of the invention can realize rapid classification under a small training sample, and meets the real-time and efficient vehicle logo recognition process to a certain extent.
According to the feature enhancement-based vehicle logo identification method, the effective features of the vehicle logo image can be obtained through the overall features, the symmetrical features of the vehicle logo can be represented through the self-symmetrical similarity features, the local information with the dense features in the vehicle logo image can be extracted through the local significant block features, namely, the overall effective description of the vehicle logo image features can be realized through the overall features, the self-symmetrical similarity features and the local significant block features, so that the vehicle logo can be identified efficiently, and the identification effect of the vehicle logo in an actual scene is further improved.
The invention discloses a vehicle logo recognition system based on feature enhancement, which comprises a preprocessing unit 100, a feature extraction unit 200, a feature fusion unit 300 and a recognition unit 400, wherein the preprocessing unit 100 and the feature fusion unit 300 are respectively connected with the feature extraction unit 200, and the feature fusion unit 300 is connected with the recognition unit 400; it should be noted that the preprocessing unit 100 is configured to perform normalization processing and graying processing on the car logo image to obtain a processed image. The feature extraction unit 200 is configured to extract an overall feature, a self-symmetric similarity feature, and a local significant block feature, and specifically, the feature extraction unit 200 includes an overall feature extraction module 210, a self-symmetric similarity feature extraction module 220, and a local significant block feature extraction module 230, where the overall feature extraction module 210, the self-symmetric similarity feature extraction module 220, and the local significant block feature extraction module 230 are respectively connected to the feature fusion unit 300.
In addition, the feature fusion unit 300 is configured to fuse the overall feature, the self-symmetric similarity feature, and the local significant block feature to obtain a final feature; the recognition unit 400 is configured to recognize the final features to obtain a car logo recognition result, and specifically, the recognition unit 400 recognizes the final features by using a CRC classifier, which is worth explaining that the quick and efficient recognition of the car logo image can be achieved by using the CRC classifier.
According to the feature enhancement-based vehicle logo recognition system, the overall features, the self-symmetry similarity features and the local significant block features of the vehicle logo image can be extracted through the feature extraction unit 200, so that effective feature information of the vehicle logo image can be obtained, and the recognition effect of the vehicle logo can be improved.
The invention has been described in detail hereinabove with reference to specific exemplary embodiments thereof. It will, however, be understood that various modifications and changes may be made without departing from the scope of the invention as defined in the appended claims. The detailed description and drawings are to be regarded in an illustrative rather than a restrictive sense, and any such modifications and variations, if any, are intended to fall within the scope of the invention as described herein. Furthermore, the background is intended to be illustrative of the present development and significance of the technology and is not intended to limit the invention or the application and field of application of the invention.
Claims (7)
1. A car logo identification method based on feature enhancement is characterized by comprising
Inputting a car logo image and preprocessing the car logo image to obtain a processed image;
extracting the integral characteristic and the self-symmetry similarity characteristic of the processed image; the specific process is as follows: setting a sliding window with width m and height n, sliding the sliding window on the processed image in the horizontal direction by step length l to obtain k sliding windows, wherein the size of k is
Will be first
Turning the images in the k sliding window, and extracting the HOL characteristics [ f ] of the images in all the sliding windows (1) ,f (2) ,…f (k) ](ii) a In sequence with f (1) ，f (2) ,…/>
For the reference feature, calculating a reference feature and
the Euclidean distance between all the HOL features is calculated, and the minimum value in the calculated Euclidean distance is used as the self-symmetry similarity feature corresponding to the reference feature; then from f (1) ，f (2) ,…/>
Selecting the minimum value from the self-symmetry similarity characteristics corresponding to the self-symmetry similarity characteristics as the self-symmetry similarity characteristics of the processed image;
background elimination is carried out on the processed image to obtain an amplitude map, a saliency map is obtained through calculation according to the amplitude map, and the specific process of obtaining the saliency map through calculation according to the amplitude map comprises the following steps: setting an amplitude map as W, setting g = (i, j) as pixel points in the amplitude map W, setting a neighborhood block corresponding to each pixel point as p (g), and calculating the intensity value of the neighborhood block according to the following formula:
wherein, W ij The amplitude of the pixel points in the neighborhood block;
obtaining an intensity map according to the intensity values corresponding to all the pixel points; then calculating according to the intensity map to obtain a saliency map; the specific process of calculating the significance map according to the intensity map comprises the following steps:
calculating the centroid of the intensity graph T, wherein the position of the centroid in the horizontal direction
Position of the center of mass in the vertical direction->
Where w and h are the width and height of the intensity map, and S (x, y) is the corresponding intensity value at the (x, y) location;
then using the centroid g 0 ＝(M x ,M y ) Taking the neighborhood block of the position as a reference, calculating the similarity s and the position distance d between other neighborhood blocks and the reference neighborhood block by the following formula, and obtaining the corresponding correlation degree f:
s＝|T(g)-T(g 0 )| 2
wherein e is x ＝t*w，e y = t × h; and obtaining a significance characteristic value R (g) =1-exp (-R f (p (g), (g) according to the correlation degree f 0 ) ) where t and r are coefficients; then obtaining a saliency map according to the saliency characteristic value;
segmenting the processed image according to the saliency map to obtain local salient blocks, and then extracting the features of the local salient blocks to obtain the features of the local salient blocks;
fusing the overall characteristic, the self-symmetry similarity characteristic and the local significant block characteristic to obtain a final characteristic;
and identifying the final characteristics to obtain a car logo identification result.
2. The method for recognizing the car logo based on the feature enhancement as claimed in claim 1, wherein the specific process of preprocessing the car logo image is as follows: normalizing the car logo image to obtain an image with a standard size, and performing graying processing on the image to obtain a processed image; the standard size of the image is w0 × h0, w0 represents the width of the image, and h0 represents the height of the image.
3. The vehicle logo recognition method based on the feature enhancement as claimed in claim 1, wherein the specific process of obtaining the amplitude map by performing the background elimination on the processed image is as follows: and filtering the processed image in the horizontal direction and the vertical direction by using an edge detection operator to obtain an amplitude map.
4. The vehicle logo recognition method based on the feature enhancement as claimed in any one of claims 1 to 3, wherein a CRC classifier is used for recognizing the final features to obtain a vehicle logo recognition result.
5. A vehicle logo recognition system based on feature enhancement is characterized in that the vehicle logo recognition method based on feature enhancement is adopted in any one of claims 1 to 4.
6. The vehicle logo recognition system based on the feature enhancement is characterized by comprising a preprocessing unit, a feature extraction unit, a feature fusion unit and a recognition unit, wherein the preprocessing unit and the feature fusion unit are respectively connected with the feature extraction unit, and the feature fusion unit is connected with the recognition unit; the feature extraction unit is used for extracting overall features, self-symmetry similarity features and local salient block features.
7. The system according to claim 6, wherein the feature extraction unit comprises a global feature extraction module, a self-symmetric similarity feature extraction module and a local significant block feature extraction module, and the global feature extraction module, the self-symmetric similarity feature extraction module and the local significant block feature extraction module are respectively connected to the feature fusion unit.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010472817.4A CN111626230B (en) | 2020-05-29 | 2020-05-29 | Vehicle logo identification method and system based on feature enhancement |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010472817.4A CN111626230B (en) | 2020-05-29 | 2020-05-29 | Vehicle logo identification method and system based on feature enhancement |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111626230A CN111626230A (en) | 2020-09-04 |
| CN111626230B true CN111626230B (en) | 2023-04-14 |
Family
ID=72258078
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010472817.4A Active CN111626230B (en) | 2020-05-29 | 2020-05-29 | Vehicle logo identification method and system based on feature enhancement |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111626230B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105160330A (en) * | 2015-09-17 | 2015-12-16 | 中国地质大学（武汉） | Vehicle logo recognition method and vehicle logo recognition system |
| CN105205444A (en) * | 2015-08-14 | 2015-12-30 | 合肥工业大学 | Vehicle logo identification method based on dot pair characteristics |
| WO2019000653A1 (en) * | 2017-06-30 | 2019-01-03 | 清华大学深圳研究生院 | Image target identification method and apparatus |
| CN110472658A (en) * | 2019-07-05 | 2019-11-19 | 哈尔滨工程大学 | A kind of the level fusion and extracting method of the detection of moving-target multi-source |
-
2020
- 2020-05-29 CN CN202010472817.4A patent/CN111626230B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105205444A (en) * | 2015-08-14 | 2015-12-30 | 合肥工业大学 | Vehicle logo identification method based on dot pair characteristics |
| CN105160330A (en) * | 2015-09-17 | 2015-12-16 | 中国地质大学（武汉） | Vehicle logo recognition method and vehicle logo recognition system |
| WO2019000653A1 (en) * | 2017-06-30 | 2019-01-03 | 清华大学深圳研究生院 | Image target identification method and apparatus |
| CN110472658A (en) * | 2019-07-05 | 2019-11-19 | 哈尔滨工程大学 | A kind of the level fusion and extracting method of the detection of moving-target multi-source |
Non-Patent Citations (1)
| Title |
|---|
| 抗模糊特征提取策略下的车标识别;贺敏雪等;《中国图象图形学报》(第03期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111626230A (en) | 2020-09-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2019196130A1 (en) | Classifier training method and device for vehicle-mounted thermal imaging pedestrian detection | |
| CN110414507B (en) | License plate recognition method and device, computer equipment and storage medium | |
| WO2019196131A1 (en) | Method and apparatus for filtering regions of interest for vehicle-mounted thermal imaging pedestrian detection | |
| CN112819094B (en) | Target detection and identification method based on structural similarity measurement | |
| Huang et al. | Vehicle detection and inter-vehicle distance estimation using single-lens video camera on urban/suburb roads | |
| CN103699905B (en) | Method and device for positioning license plate | |
| CN106650731B (en) | Robust license plate and vehicle logo recognition method | |
| CN110866430B (en) | License plate recognition method and device | |
| CN105678213B (en) | Dual-mode mask person event automatic detection method based on video feature statistics | |
| CN109460722B (en) | Intelligent license plate recognition method | |
| Prates et al. | Brazilian license plate detection using histogram of oriented gradients and sliding windows | |
| Islam et al. | Automatic vehicle number plate recognition using structured elements | |
| CN111027544B (en) | MSER license plate positioning method and system based on visual saliency detection | |
| CN111915583A (en) | Vehicle and pedestrian detection method based on vehicle-mounted thermal infrared imager in complex scene | |
| CN112699267A (en) | Vehicle type recognition method | |
| WO2013026205A1 (en) | System and method for detecting and recognizing rectangular traffic signs | |
| Asgarian Dehkordi et al. | Vehicle type recognition based on dimension estimation and bag of word classification | |
| Dhar et al. | Bangladeshi license plate recognition using adaboost classifier | |
| WO2017161544A1 (en) | Single-camera video sequence matching based vehicle speed measurement method and system | |
| CN107977608B (en) | Method for extracting road area of highway video image | |
| CN113053164A (en) | Parking space identification method using look-around image | |
| CN110516666B (en) | License plate positioning method based on combination of MSER and ISODATA | |
| Hommos et al. | Hd Qatari ANPR system | |
| CN111626230B (en) | Vehicle logo identification method and system based on feature enhancement | |
| CN111860113A (en) | Lane line detection method and system |
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 |