CN103150904A - Bayonet vehicle image identification method based on image features - Google Patents

Abstract

The invention relates to the field of traffic image processing, in particular to an image feature-based bayonet vehicle image recognition method. It includes the steps of: establishing a template database: storing template images of different styles and different colors of vehicles, and storing vehicle attribute data, body color data and vehicle SIFT feature data obtained through data processing corresponding to each template image; Vehicle recognition: Carry out color recognition on the image to be recognized; select a template image that matches the body color from the template database according to the color recognition result; extract SIFT features from the image to be recognized and compare it with the selected template image to obtain a match with the image to be recognized template image; output the vehicle attribute data corresponding to the matched template image. The invention combines the vehicle color recognition with the recognition based on the SIFT operator principle, adds color information to the recognition process, overcomes the defect that the SIFT operator discards the color information, and can improve the recognition accuracy.

Description

A bayonet vehicle image recognition method based on image features

technical field

The present invention relates to the field of traffic image processing, and more specifically, to a method for recognizing a bayonet vehicle image based on image features.

Background technique

Vehicle Recognition System (Vehicle Recognition System, VRS) is an important part of intelligent transportation system. The generalized vehicle identification system refers to the identification of the license plate, color, manufacturer, model, special marking and other information of the vehicle. However, the currently applied vehicle identification systems basically only realize the identification of vehicles through license plate recognition, and are mainly used for speed measurement in intervals and management of entrances and exits of expressways and parking lots. Although the license plate is the only legal identification of a vehicle, the license plate recognition technology based on image processing has problems such as low accuracy and limited use conditions in application, especially for fake, fake, and blocked license plates. Therefore, it is very important to recognize other vehicle features besides the license plate. At present, there are also technologies for distinguishing large, medium, and small models, as well as technologies for vehicle brand recognition through vehicle logo recognition, but the classification is very rough, and the recognition rate of colors is not high. The identification of specific vehicle models and styles and the search for specific vehicles still rely on worker judgment and manual retrieval. This work often requires a lot of human resources and time, and the workload of searching massive image data cannot be estimated. Therefore, it is very necessary to use computers instead of manual labor to identify vehicle models, styles and more accurate color discrimination.

At present, the main vehicle image recognition methods can be divided into three categories according to the recognized objects: direct recognition, vehicle local feature recognition and image feature recognition.

Direct identification. This type of method directly recognizes human perception of vehicles, mainly including vehicle license plates and vehicle appearance information (such as color, shape, size, etc.). Although the recognition results of the former can be used as a basis for law enforcement, the accuracy rate is limited; the latter usually has special requirements for the shooting scene, and often needs to be calibrated in advance, so it cannot be accurately calculated.

Vehicle local features. Vehicle local features mainly include vehicle logo information and vehicle face information. However, the former only contains vehicle manufacturer information, while the latter is often difficult to extract and accurately define due to the complexity of vehicle face information.

Image feature recognition. Image features generally do not represent human perception information of vehicles. Therefore, the feature recognition of images often requires a template image, and the recognition is performed by calculating the matching degree between the image to be recognized and the features of the template image. Vehicle image features mainly include color distribution features (such as histograms, color moments, etc.) and image space local features. For the former, different vehicle images may correspond to the same distribution features; for the latter, represented by SIFT, SURF, and PCA-SIFT, the computational complexity is high.

Contents of the invention

The technical problem to be solved by the present invention is to provide a bayonet vehicle image recognition method based on image features with high recognition accuracy.

In order to solve the problems of the technologies described above, the technical solution of the present invention is as follows:

A bayonet vehicle image recognition method based on image features, comprising the steps of:

Create a template database:

The template images of vehicles of different styles and colors obtained by shooting are stored in the template database, and vehicle attribute data and vehicle body color data are stored corresponding to each template image in the template database;

Perform data processing on each template image in the template database, obtain vehicle SIFT feature data and store in the template database;

Recognize the image to be recognized:

Carry out color recognition on the image to be recognized;

Select a template image matching the body color from the template database according to the color recognition result;

Use the SIFT operator principle to extract features from the image to be recognized, and compare it with the selected template image to obtain a template image that matches the image to be recognized;

Output the vehicle attribute data corresponding to the matching template image.

The present invention combines vehicle color recognition with recognition based on the SIFT operator principle, selects template image data with matching color information under each vehicle type in the template database according to the color recognition result, and performs comparison based on the SIFT operator principle, The image to be recognized is matched with the template image of the corresponding color, and color information is added to the recognition process, which overcomes the disadvantage of SIFT operator discarding color information, and can improve the accuracy of recognition; moreover, it corresponds to each vehicle type in the template database Both store corresponding vehicle attribute data, and the classification of vehicles is more refined, making the recognition results more specific and the recognition accuracy higher.

One of the improvements: data processing is performed on each template image in the template database to obtain vehicle SIFT feature data and store them in the template database. The specific steps are:

Mark the image coordinates where the vehicle license plate is located in each template image;

The vehicle SIFT feature points are determined for each template image, that is, the SIFT feature point information located in the license plate marking area in the template image is eliminated, and the remaining SIFT feature information in the template image is retained as the vehicle SIFT feature data of each template image.

The present invention excludes the feature points of the license plate area in the template image, so that the interference of the feature points of the license plate area is eliminated when the image to be recognized is matched with the template image, and the possibility of being recognized as a different vehicle type can be greatly reduced, thereby improving the accuracy of recognition. Accuracy.

Improvement 2: The specific steps for color recognition of the image to be recognized are:

The car body color is divided into six categories in advance: green, yellow, red, blue, white and black, among which yellow includes yellow, orange and brown perceived by human eyes, red includes red, pink and purple perceived by human eyes, and white includes White, silver, light gray, black includes black and dark gray perceived by human eyes;

For the five colors of green, yellow, red, blue, and white, combined with the pairwise differences of r, g, and b, set the empirical threshold to draw a certain range for the r, g, b, h, s, and v values, and the statistics are to be identified The proportion of green, yellow, red, blue, and white color pixels within the body area of the image to the pixels within the body area of the image;

According to the order of green, yellow, red, blue and white, the color ratio of the image to be recognized is judged. When the ratio of the current color exceeds the experience threshold of the corresponding color, it is judged that the body of the image to be recognized is the current color. When the body color ratio of the image does not exceed the empirical thresholds of the five colors of green, yellow, red, blue, and white, it is judged that the body color of the image to be recognized is black, and the color recognition result is obtained.

The invention combines the RGB and HSV values of the image to judge the color of the vehicle body, and judges the color of the vehicle body according to a certain color sequence, so that the result of color recognition is more stable.

Improvement 3: The range of the body of the image to be recognized excludes the range of the window, the range of the exhaust grille on the front face of the car, and the range of the lights. The invention eliminates the range of the window, the range of the exhaust grid of the front face and the range of the lights when identifying the color of the vehicle body, and can reduce the disadvantages of the colors of the windows, the exhaust grill of the front face, and the lights of the vehicle to the judgment of the vehicle color. Influence, and then improve the accuracy of body color judgment.

Improvement 4: The specific steps of using the SIFT operator principle to extract features from the image to be recognized, and comparing it with the selected template image to obtain a template image that matches the image to be recognized are:

Use the SIFT operator principle to determine the SIFT feature data of the image to be recognized;

Compare the SIFT feature data of the image to be recognized with the SIFT feature data of each selected template image to obtain matching feature point pairs;

The matching degree of the two images is calculated according to the matching feature point pairs, and the template image matching the image to be recognized is determined according to the matching degree.

Improvement 5: After obtaining the matching feature point pairs, the data purification step is performed, as follows:

The position mapping relationship between each selected template image and the matching feature point pairs in the image to be recognized is used as the input value of the RANSAC algorithm, and the homography matrix of the image transformation is estimated by using the RANSAC method, and the features that do not satisfy the geometric consistency are eliminated. Point pair, obtain the reserved feature point pair as the final matching feature point pair.

In order to improve the robustness of the algorithm, the present invention uses the RANSAC algorithm to purify the data of the matching result.

Improvement 6: After obtaining the matching feature point pairs, it is also judged whether the logarithm of the matching feature point pairs corresponding to each selected template image and the image to be recognized is greater than the threshold value μ. If it is greater than the threshold value μ, perform the data purification step, otherwise directly determine The matching degree of the two images is 0. Since in the RANSAC algorithm, at least 4 pairs of matching feature points are required to estimate the homography matrix, in order to enhance the stability of the algorithm, after the matching is completed, the logarithm of the matching feature point pair is greater than the threshold μ. For data purification operations, the threshold μ is recommended to be 6.

Improvement seven: the specific steps of calculating the degree of matching of the two images according to the pair of matching feature points, and determining the template image matching the image to be recognized according to the degree of matching are:

Calculate the image matching degree IMD=N/ N ₀ according to the feature point pairs that match the image to be recognized and each selected template image, where N represents the number of matched feature point pairs, and N ₀ is the number of SIFT feature points in the image to be recognized total;

Select the maximum value IMD _max from all IMD values, compare IMD _max with the set threshold λ, and judge whether IMD _max is greater than λ, if so, judge that the template image corresponding to the IMD _max matches the image to be recognized successfully, otherwise judge that in the template database There is no template image matching the image to be recognized.

The present invention uses the relative value of the logarithm of the matched feature points as the image matching degree IMD, which can measure the matching degree between images more objectively and can improve the matching accuracy between images.

Ninth improvement: the vehicle attribute includes vehicle brand, vehicle model and vehicle year.

In the present invention, the vehicle brand, vehicle model and vehicle year are stored in the template database as vehicle attribute data to realize a finer classification of vehicles. When a matching template image is found in the template database for the image to be recognized, the The corresponding vehicle attribute data output is found in the database, which makes the recognition result more specific and greatly improves the recognition accuracy.

Compared with the prior art, the beneficial effects of the technical solution of the present invention are:

(1) The present invention combines vehicle color recognition with recognition based on SIFT operator principle, adds color information to the recognition process, overcomes the disadvantage of SIFT operator discarding color information, and can improve the accuracy of recognition.

(2) The present invention combines the RGB and HSV values of the image when recognizing the vehicle color, and judges according to a certain color order, making the recognition result more stable and reliable.

(3) The present invention stores corresponding vehicle attribute data corresponding to each vehicle type in the template database, and when the image to be recognized matches the corresponding template image in the template database, the corresponding vehicle attribute data can be output, so that the recognition result The more specific, the higher the recognition accuracy.

(4) The present invention can reduce the influence of the color of the window, the exhaust grille on the front face of the vehicle, and the vehicle light on the vehicle color determination, and can improve the accuracy of vehicle color recognition.

(5) The present invention eliminates the interference of the feature points in the license plate area during the vehicle type recognition based on the SIFT operator, and can greatly reduce the possibility of identifying different vehicle types.

(6) The present invention expresses the similarity between the image to be recognized and the template image by defining the image matching degree IMD, and uses an objective evaluation method to describe the recognition result, which can improve the accuracy of vehicle recognition.

(7) The present invention can improve the robustness of the algorithm and reduce the recognition error rate by using the RANSAC algorithm to purify the data of the matching result.

Description of drawings

FIG. 1 is a flow chart of a specific embodiment of a bayonet vehicle image recognition method based on image features in the present invention.

FIG. 2 is an example diagram of template images in the template database.

Fig. 3 is an example diagram of calculation results of SIFT feature points of a template image in the template database.

Figure 4 is an example of an image to be recognized and its SIFT features.

FIG. 5 is a schematic diagram of matching the image to be recognized in FIG. 4 with the white template image shown in FIG. 2 .

FIG. 6 is a schematic diagram of the results of data purification for the matching results in FIG. 5 .

Fig. 7 is a flow chart of a preferred embodiment of a bayonet vehicle image recognition method based on image features in the present invention.

Detailed ways

The technical solutions of the present invention will be further described below in conjunction with the accompanying drawings and embodiments.

Example 1

As shown in FIG. 1 , it is a flowchart of a specific embodiment of a bayonet vehicle image recognition method based on image features in the present invention. Referring to example 1, the specific steps of a kind of image feature-based bayonet vehicle image recognition method of this specific embodiment include:

Step S100: Establish template database:

Step S101: Store template images of vehicles of different styles and colors captured in the template database, and store vehicle attribute data and body color data corresponding to each template image in the template database, wherein the vehicle attribute data may include vehicle brand and vehicle model and vehicle year; as shown in Figure 2, it is two template images of "Toyota, Corolla, ninth generation";

Step S102: Perform data processing on each template image in the template database, obtain vehicle SIFT feature data and store it in the template database, wherein the SIFT feature data generally includes SIFT feature points and SIFT descriptors of each feature point;

Step S200: Carry out vehicle recognition on the image to be recognized:

Step S201: performing color recognition on the image to be recognized;

Step S202: Select a template image matching the body color from the template database according to the color recognition result;

Step S203: using the SIFT operator principle to extract features from the image to be recognized, and comparing it with the selected template image to obtain a template image that matches the image to be recognized;

Step S204: output the vehicle attribute data corresponding to the matched template image.

In this specific embodiment, the vehicle color recognition is combined with the recognition based on the SIFT operator principle, and according to the color recognition result, the template image data matching the color result is selected under each vehicle type in the template database for comparison based on the SIFT operator principle. Find the matching template image in the template database, and output the corresponding vehicle attribute data according to the matching template image.

In step S102, in order to eliminate the interference of the feature points of the vehicle license plate area on image matching and improve the accuracy of recognition, in the specific implementation process of this specific embodiment, the license plate area is delineated for each template image in the template database in advance, And remove the SIFT feature within the license plate area of each template image, the specific steps are as follows:

Step S1021: mark the image coordinates of the vehicle license plate in each template image;

Step S1022: Determine the vehicle SIFT feature points for each template image, that is, remove the SIFT feature point information located in the license plate marking area in the template image, and retain the rest of the SIFT feature information in the template image as the vehicle SIFT feature of each template image data. If there is a license plate in the template image, mark the image coordinates where the license plate is located, eliminate the SIFT feature point information within the coordinate range of the license plate mark, and calculate the SIFT descriptors of the remaining feature points, and finally obtain the vehicle SIFT of each template image feature data. As shown in Figure 3, it is the SIFT feature calculation result of the template image, where the starting point of the arrow in the illustration indicates the position of the feature point, the direction of the arrow indicates the main direction of the feature point, and the direction of the arrow indicates the value of the description sub-module.

In step S201, this specific embodiment customizes the color of the vehicle body based on the difference in perception of color by the human eye and in combination with the factors affected by illumination, so as to judge the color of the vehicle body according to the customized color, and combines the RGB and The HSV value judges the color of the car body, and judges the color of the car body according to a certain color sequence, making the result of color recognition more stable. Specific steps are as follows:

Step S2011: Divide the car body colors into six categories in advance: green, yellow, red, blue, white and black, wherein yellow includes yellow, orange and brown perceived by human eyes, red includes red, pink and purple perceived by human eyes, and white includes White, silver, and light gray perceived by the human eye, and black includes black and dark gray perceived by the human eye;

Step S2012: For the five colors of green, yellow, red, blue, and white, combined with the pairwise differences of r, g, and b, set empirical thresholds to draw a certain range for the values of r, g, b, h, s, and v; As shown in Table 1 below:

Table 1 Color Judgment

color h value range s value range v value range r , g , b range other conditions green 70< h ≤ 170 s> ThreS v > ThreV diff(r,g,b)> ThreRGB - yellow 20< h ≤70 s> ThreS v > ThreV diff(r,g,b)> ThreRGB - red h ≤ 20 or h > 300 s> ThreS v > ThreV diff(r,g,b)> ThreRGB - blue 170< h ≤ 260 s> ThreS v > ThreV diff(r,g,b)> ThreRGB - White - the the r , g , b are all greater than ThreWhite the black - - - - Other conditions that do not meet the above

See Table 1, where ThreS , ThreV , ThreRGB , and ThreWhite are empirical thresholds, and the recommended values are 30, 20, 15, and 160, respectively. diff( r , g , b ) represents the difference between any two values in r , g , and b value;

As shown in Table 1, when the h value of a certain pixel in the image to be recognized is between 70<h≤170, and the values of s and v are respectively greater than ThreS, ThreV, any two values of r, g, b If the difference is greater than ThreRGB, it can be judged that the pixel to be awaited is green. The judgment of other colors can be obtained in the same way.

Step S2013: Count the proportions of green, yellow, red, blue, and white color pixels within the vehicle body range on the image to be recognized; the body range excludes the window range and the exhaust grille on the front face of the vehicle range and headlight range;

Step S2014: Judging the proportions of various colors of the image to be recognized in the order of green, yellow, red, blue, and white. When the proportion of the current color exceeds the empirical threshold of the corresponding color, it is judged that the body of the image to be recognized is the current color. When the body color ratio of the image to be recognized does not exceed the empirical thresholds of the five colors of green, yellow, red, blue, and white, it is judged that the body color of the image to be recognized is black, thereby obtaining the color recognition result.

In this specific embodiment, when performing vehicle recognition on the image to be recognized, the SIFT feature data of the vehicle image is determined according to the principle of the SIFT operator, combined with the color recognition result and input into the template database for comparison, and finally the recognition result is obtained. Therefore, the specific steps of step S203 in this specific embodiment are;

S2031: Determine the SIFT feature data of the image to be recognized by using the SIFT operator principle; as shown in FIG. 4 , it is an example of the image to be recognized and its SIFT feature.

S2032: Compare the SIFT feature data of the image to be recognized with the SIFT feature data of each selected template image to obtain matching feature point pairs;

S2033: Calculate the matching degree of the two images according to the matching feature point pairs, and determine the template image matching the image to be recognized according to the matching degree.

Wherein, the matching feature point pair in step S2032 can represent its matching degree by Euclidean distance, and the specific steps are as follows:

S20321: preset the threshold ε;

S20322: For the SIFT feature point P in the image to be recognized, calculate the Euclidean distance between all the feature points in each selected template image and the feature vector descriptor of P, and find the minimum and second smallest value of the Euclidean distance value d1 and d2, and record the SIFT feature points Q1 and Q2 corresponding to d1 and d2 in the template image respectively;

； S20323: If d2 is 0, set the parameter ratio to 0; otherwise calculate

;

S20324: Compare the parameter ratio with the threshold ε, and when the ratio is smaller than the threshold ε, judge that the SIFT feature point P in the image to be recognized is successfully matched with the feature point Q1 of the template image, otherwise judge that the match between P and Q1 is unsuccessful;

S20325: Counting and recording feature point pairs matched between each selected template image and the image to be recognized. As shown in Figure 5, match the image to be recognized in Figure 4 (the color recognition result is white) with the white template image shown in Figure 2, and the result is shown in Figure 5. In the illustration, the two vehicles are connected by a straight line Feature points represent matching feature point pairs.

After the matching in step S20324 is completed, in order to improve the robustness of the algorithm, this specific embodiment can also use the RANSAC algorithm to perform data purification on the matching result, and the specific steps are as follows:

The position mapping relationship between each selected template image and the corresponding matching feature point pairs in the image to be recognized is used as the input value of the RANSAC algorithm, and the homography matrix of the image transformation is estimated by using the RANSAC method, and the ones that do not satisfy the geometric consistency are eliminated. Feature point pairs, obtain the reserved feature point pairs as the final matching feature point pairs. As shown in FIG. 6 , it is a schematic diagram of the result of data purification for the matching result in FIG. 5 .

Since in the RANSAC algorithm, at least 4 pairs of matching feature points are required to estimate the homography matrix, in order to enhance the stability of the algorithm, after the matching is completed, the logarithm of the matching feature point pairs is greater than the set threshold The data purification operation is carried out only when μ. details as follows:

After obtaining the matching feature point pairs, it is also judged whether the logarithm of the matching feature point pairs corresponding to the image to be recognized and each selected template image is greater than the threshold μ, if greater, the data purification step is performed, otherwise, the matching degree of the two images is directly judged is 0. The threshold μ is recommended to be set to 6.

In this specific embodiment, considering that an image to be recognized may need to be matched with multiple template images in actual matching, images to be recognized of different colors may also be matched with template images of different grayscales, and the number of template image feature points The number of logarithms of feature points has a certain impact on the final successful matching feature point logarithm. Therefore, in order to further improve the accuracy of recognition, this specific embodiment uses the relative value of the logarithm of matching points as the image matching degree to represent the difference between two images. In order to measure the matching degree between images more objectively. Therefore, in step S2033, the matching degree of the two images is calculated according to the matching feature points, and the specific steps of determining the template image matched with the image to be recognized according to the matching degree are as follows:

S20331: Calculate the image matching degree IMD=N/ N ₀ according to the matching feature point pairs of the image to be recognized and each selected template image, where N represents the number of matching feature point pairs, and N ₀ is the SIFT feature in the image to be recognized the total number of points;

S20332: Select the maximum value IMD _max from all IMD values, compare IMD _max with the set threshold λ, and judge whether IMD _max is greater than λ, if so, judge that the template image corresponding to IMD _max matches the image to be recognized successfully, otherwise judge the template There is no template image matching the image to be recognized in the database.

Example 2

As shown in FIG. 7 , it is a flowchart of a preferred embodiment of a bayonet vehicle image recognition method based on image features in the present invention. Referring to Figure 7, the specific steps of this preferred embodiment are as follows:

Establish a template database: pre-store template images of different models and styles of vehicles in the template database, and store the vehicle attribute data, body color, and SIFT feature data of the template images in the template database correspondingly; wherein for each template image, its first Delineate the area where the license plate is located, and delete the SIFT feature information of each template image located in the license plate area;

When it is necessary to perform vehicle recognition on the image to be recognized, first perform color recognition on the image to be recognized, and simultaneously calculate the SIFT feature of the image to be recognized;

Input the image to be recognized into the template database for comparison, first select the template image of the appropriate color under different vehicle models in the template database according to the color recognition result to perform SIFT feature matching;

Perform data purification on the matching results;

The IMD calculation is performed on the matching result after data purification, the template image corresponding to the maximum IMD value is taken, and the vehicle attribute data of the template image is taken for output.

Claims (9)

1. a bayonet vehicle image recognition method based on image features, is characterized in that, comprises the steps: Create a template database: The template images of vehicles of different styles and colors obtained by shooting are stored in the template database, and vehicle attribute data and vehicle body color data are stored corresponding to each template image in the template database; Perform data processing on each template image in the template database, obtain vehicle SIFT feature data and store in the template database; Carry out vehicle recognition on the image to be recognized: Carry out color recognition on the image to be recognized; Select a template image matching the body color from the template database according to the color recognition result; Use the SIFT operator principle to extract features from the image to be recognized, and compare it with the selected template image to obtain a template image that matches the image to be recognized; Output the vehicle attribute data corresponding to the matching template image. 2. the bayonet vehicle image recognition method based on image feature according to claim 1, is characterized in that, carries out data processing to each template image in the template database, obtains vehicle SIFT feature data and is stored in the template database specific The steps are: Mark the image coordinates where the vehicle license plate is located in each template image; The vehicle SIFT feature points are determined for each template image, that is, the SIFT feature point information located in the license plate marking area in the template image is eliminated, and the rest of the SIFT feature information in the template image is retained as the vehicle SIFT feature data of each template image. 3. the bayonet vehicle image recognition method based on image feature according to claim 1, is characterized in that, the concrete step that image to be recognized is carried out color recognition is: The car body color is divided into six categories in advance: green, yellow, red, blue, white and black, among which yellow includes yellow, orange and brown perceived by human eyes, red includes red, pink and purple perceived by human eyes, and white includes White, silver, light gray, black includes black and dark gray perceived by human eyes; For the five colors of green, yellow, red, blue, and white, combined with the pairwise differences of r, g, and b, set the empirical threshold to draw a certain range for the values of r, g, b, h, s, and v; Count the proportion of green, yellow, red, blue, and white color pixels in the body range of the image to be recognized to the pixels in the body range of the image; According to the order of green, yellow, red, blue and white, the color ratio of the image to be recognized is judged. When the ratio of the current color exceeds the experience threshold of the corresponding color, it is judged that the body of the image to be recognized is the current color. When the body color ratio of the image does not exceed the empirical thresholds of the five colors of green, yellow, red, blue, and white, it is judged that the body color of the image to be recognized is black, and the color recognition result is obtained. 4. The bayonet vehicle image recognition method based on image features according to claim 3, wherein the range of the vehicle body of the image to be recognized excludes the range of the window, the range of the front exhaust grille and the range of the lights. 5. the bayonet vehicle image recognition method based on image feature according to claim 1, is characterized in that, utilizes SIFT operator principle to be recognized image extraction feature, and compares with the selected template image, obtains and to be recognized The specific steps for matching an image to a template image are: Use the SIFT operator principle to determine the SIFT feature data of the image to be recognized; Compare the SIFT feature data of the image to be recognized with the SIFT feature data of each selected template image to obtain matching feature point pairs; The matching degree of the two images is calculated according to the matching feature point pairs, and the template image matching the image to be recognized is determined according to the matching degree. 6. the bayonet vehicle image recognition method based on image feature according to claim 5, is characterized in that, also carries out data purification step after obtaining the matching feature point pair, specifically as follows: The position mapping relationship between each selected template image and the matching feature point pairs in the image to be recognized is used as the input value of the RANSAC algorithm, and the homography matrix of the image transformation is estimated by using the RANSAC method, and the features that do not satisfy the geometric consistency are eliminated. Point pair, obtain the reserved feature point pair as the final matching feature point pair. 7. The bayonet vehicle image recognition method based on image features according to claim 6, characterized in that, after obtaining the matched feature point pairs, it is also judged that the image to be identified is matched with the matching feature points corresponding to each selected template image Whether the logarithm of the pair is greater than the threshold μ, if it is greater, perform the data purification step, otherwise directly determine the matching degree of the two images as 0. 8. The bayonet vehicle image recognition method based on image features according to any one of claims 5-7, characterized in that, the matching degree of the two images is calculated according to the matched feature points, and the matching degree is determined according to the matching degree. The specific steps of the template image matched with the image to be recognized are: Calculate the image matching degree IMD=N/ N ₀ according to the feature point pairs that match the image to be recognized and each selected template image, where N represents the number of matched feature point pairs, and N ₀ is the number of SIFT feature points in the image to be recognized total; Select the maximum value IMD _max from all IMD values, compare IMD _max with the set threshold λ, and judge whether IMD _max is greater than λ, if so, judge that the template image corresponding to the IMD _max matches the image to be recognized successfully, otherwise judge that in the template database There is no template image matching the image to be recognized. 9. The image feature-based bayonet vehicle image recognition method according to any one of claims 1-7, wherein the vehicle attributes include vehicle brand, vehicle model and vehicle year.

Images