Vehicle identity recognition method and device and electronic equipment

Abstract

The invention provides a vehicle identity recognition method and device and electronic equipment, and relates to the technical field of image processing. The method comprises the steps that when vehicle identity recognition is carried out, a target image of a to-be-recognized vehicle is acquired, and multi-dimensional features of the to-be-recognized vehicle and confidence coefficients corresponding to the features are acquired based on the target image; and based on the confidence corresponding to each feature, determining a plurality of target features of which the confidence is greater than a preset confidence threshold from the multi-dimensional features, and matching the plurality of target features with the respective multi-dimensional features of a plurality of preset vehicles to obtain an identity recognition result of the to-be-recognized vehicle. In this way, the multiple target features with the high confidence degree serve as the basis of vehicle identity recognition and are used for vehicle identity recognition, and the accuracy of the vehicle identity recognition result can be effectively improved.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to a vehicle identification method, device and electronic equipment. Background Art

[0002] Vehicle identity recognition is of great significance for various transportation applications such as highway toll evasion inspection and accident vehicle query.

[0003] At present, when identifying a vehicle, the license plate is considered to be the physical ID card of the vehicle, so the vehicle identity is usually identified based on the license plate. However, since the license plate is easily damaged, blocked, or even duplicated, the accuracy of the vehicle identity recognition result is low. Summary of the invention

[0004] The present application provides a vehicle identity recognition method, device and electronic device to solve the problem of low accuracy of vehicle identity recognition results in the prior art, thereby effectively improving the accuracy of vehicle identity recognition results.

[0005] In a first aspect, the present application provides a vehicle identification method, the method comprising:

[0006] Get the target image of the vehicle to be identified.

[0007] Based on the target image, multi-dimensional features of the vehicle to be identified and the confidence level corresponding to each of the features are obtained.

[0008] Based on the confidences corresponding to the features, a plurality of target features having confidences greater than a preset confidence threshold are determined from the multi-dimensional features.

[0009] The multiple target features are matched with the multi-dimensional features of each of the multiple preset vehicles to obtain an identity recognition result of the vehicle to be recognized.

[0010] According to a vehicle identification method provided by the present application, the multiple target features are matched with the multi-dimensional features of multiple preset vehicles to obtain the identification result of the vehicle to be identified, including:

[0011] The multiple target features are matched with the multi-dimensional features of each of the multiple preset vehicles, and a candidate vehicle including the multiple target features is determined from the multiple preset vehicles.

[0012] In the case where there are multiple candidate vehicles, based on the preemption coefficients of the multiple target features, a first target feature corresponding to the maximum preemption coefficient is determined from the multiple target features, and the preemption coefficient of the target feature is used to characterize the rarity of the target feature.

[0013] The first target feature is matched with a preset association feature combination corresponding to multiple candidate vehicles to determine a target association feature of the first target feature; the multiple target features and the target association feature are matched with the multi-dimensional features of each of the candidate vehicles respectively to obtain multiple correlation scores.

[0014] The identity recognition result is determined according to the multiple correlation scores.

[0015] According to a vehicle identification method provided by the present application, determining the identification result according to the multiple correlation scores includes:

[0016] In the case that there is a correlation score greater than a preset threshold value among the multiple correlation scores, the candidate vehicle corresponding to the correlation score greater than the preset score threshold value is determined as the identity recognition result.

[0017] When there are at least two correlation scores greater than the preset threshold among the multiple correlation scores, the target feature corresponding to the second largest preemption coefficient is determined as the new first target feature, and the multiple target features, the first target feature, the target association feature of the first target feature, the new first target feature and the new target association feature of the first target feature are matched with the preset association feature combination corresponding to multiple candidate vehicles until there is a correlation score greater than the preset score threshold among the multiple correlation scores obtained, and the candidate vehicle corresponding to the correlation score greater than the preset score threshold is determined as the identity recognition result.

[0018] According to a vehicle identification method provided by the present application, the first target feature is matched with a preset association feature combination corresponding to a plurality of candidate vehicles to determine a target association feature of the first target feature; the plurality of target features and the target association feature are matched with the multi-dimensional features of each of the candidate vehicles respectively to obtain a plurality of correlation scores, including:

[0019] The first target feature is matched with the preset associated feature combination to determine at least one associated feature combination including the first target feature; and the features other than the first target feature in the associated feature combination corresponding to the maximum confidence are determined as the target associated features.

[0020] The multiple target features and the target-associated features are matched with the multi-dimensional features of each candidate vehicle to obtain multiple correlation scores.

[0021] When at least two correlation scores among the multiple correlation scores are greater than the preset score threshold, the features other than the first target feature in the association feature combination corresponding to the second largest confidence are determined as new target association features of the first target feature.

[0022] The multiple target features, the target association features and the new target association features are matched with the multi-dimensional features of the multiple candidate vehicles respectively, until at least one association feature combination is traversed to obtain the multiple correlation scores.

[0023] According to a vehicle identification method provided by the present application, the method further includes:

[0024] For each target feature, determine from the multiple preset vehicles a first number of preset vehicles whose corresponding multidimensional feature includes the target feature, and a second number of preset vehicles whose feature type includes the feature type to which the target feature belongs in the feature type to which the corresponding multidimensional feature belongs, and based on the first number and the second number, determine the isolation coefficient corresponding to the target feature.

[0025] The preemption coefficient of the target feature is determined according to the first number, the second number, the confidence of the target feature, and the isolation coefficient corresponding to the target feature.

[0026] According to a vehicle identification method provided by the present application, the multiple target features and the target-associated features are matched with the multi-dimensional features of each candidate vehicle to obtain multiple correlation scores, including:

[0027] For each of the candidate vehicles, determine the weight corresponding to each second target feature in the multiple target features and the target association features, the number of times each second target feature appears in the multi-dimensional features of the multiple candidate vehicles, the number of times each second target feature appears in the multi-dimensional features of the candidate vehicles, and the isolation coefficient corresponding to each second target feature.

[0028] The correlation score corresponding to the candidate vehicle is determined based on the weight corresponding to each second target feature, the number of times each second target feature appears in the multi-dimensional features of the multiple candidate vehicles, the number of times each second target feature appears in the multi-dimensional features of the candidate vehicle, and the isolation coefficient corresponding to each second target feature.

[0029] According to a vehicle identification method provided by the present application, the method further includes:

[0030] For each of the candidate vehicles, based on the minimum support and minimum confidence criteria, feature association calculation is performed on the multi-dimensional features of the candidate vehicle to obtain multiple initial association feature combinations and initial confidences corresponding to each initial association feature combination.

[0031] According to the initial confidences corresponding to the initial association feature combinations, a plurality of first association feature combinations having initial confidences greater than a preset threshold are determined from the plurality of initial association feature combinations.

[0032] For each of the first associated feature combinations, a confidence correction coefficient corresponding to the first associated feature combination is determined based on the confidence levels of at least two features included in the first associated feature combination and the number of times each feature appears in the multidimensional features corresponding to the multiple preset vehicles, and the initial confidence level corresponding to the first associated feature combination is corrected based on the confidence correction coefficient to obtain a corrected confidence level.

[0033] Among the plurality of first associated feature combinations, a first associated feature combination whose corresponding corrected confidence level is greater than the preset threshold is determined as the preset associated feature combination corresponding to the candidate vehicle.

[0034] In a second aspect, an embodiment of the present application further provides a vehicle identification device, the device comprising:

[0035] The first acquisition unit is used to acquire a target image of a vehicle to be identified.

[0036] The second acquisition unit is used to acquire the multi-dimensional features of the vehicle to be identified and the confidence level corresponding to each feature based on the target image.

[0037] The first determination unit is used to determine, based on the confidence corresponding to each of the features, a plurality of target features whose confidence is greater than a preset confidence threshold from the multi-dimensional features.

[0038] The first processing unit is used to match the multiple target features with the multi-dimensional features of each of the multiple preset vehicles to obtain an identity recognition result of the vehicle to be recognized.

[0039] According to a vehicle identification device provided by the present application, the first processing unit is specifically used for:

[0040] The multiple target features are matched with the multi-dimensional features of each of the multiple preset vehicles, and a candidate vehicle including the multiple target features is determined from the multiple preset vehicles.

[0041] In the case where there are multiple candidate vehicles, based on the preemption coefficients of the multiple target features, a first target feature corresponding to the maximum preemption coefficient is determined from the multiple target features, and the preemption coefficient of the target feature is used to characterize the rarity of the target feature.

[0042] The first target feature is matched with a preset association feature combination corresponding to multiple candidate vehicles to determine a target association feature of the first target feature; the multiple target features and the target association feature are matched with the multi-dimensional features of each of the candidate vehicles respectively to obtain multiple correlation scores.

[0043] The identity recognition result is determined according to the multiple correlation scores.

[0044] According to a vehicle identification device provided by the present application, the first processing unit is specifically used for:

[0045] In the case that there is a correlation score greater than a preset threshold value among the multiple correlation scores, the candidate vehicle corresponding to the correlation score greater than the preset score threshold value is determined as the identity recognition result.

[0046] When there are at least two correlation scores greater than the preset threshold among the multiple correlation scores, the target feature corresponding to the second largest preemption coefficient is determined as the new first target feature, and the multiple target features, the first target feature, the target association feature of the first target feature, the new first target feature and the new target association feature of the first target feature are matched with the preset association feature combination corresponding to multiple candidate vehicles until there is a correlation score greater than the preset score threshold among the multiple correlation scores obtained, and the candidate vehicle corresponding to the correlation score greater than the preset score threshold is determined as the identity recognition result.

[0047] According to a vehicle identification device provided by the present application, the first processing unit is specifically used for:

[0048] The first target feature is matched with the preset associated feature combination to determine at least one associated feature combination including the first target feature; and the features other than the first target feature in the associated feature combination corresponding to the maximum confidence are determined as the target associated features.

[0049] The multiple target features and the target-associated features are matched with the multi-dimensional features of each candidate vehicle to obtain multiple correlation scores.

[0050] When at least two correlation scores among the multiple correlation scores are greater than the preset score threshold, the features other than the first target feature in the association feature combination corresponding to the second largest confidence are determined as new target association features of the first target feature.

[0051] The multiple target features, the target association features and the new target association features are matched with the multi-dimensional features of the multiple candidate vehicles respectively, until at least one association feature combination is traversed to obtain the multiple correlation scores.

[0052] According to a vehicle identification device provided by the present application, the device also includes a second determination unit and a third determination unit.

[0053] The second determination unit is used to determine, for each target feature, from the multiple preset vehicles, a first number of preset vehicles whose corresponding multi-dimensional feature includes the target feature, and a second number of preset vehicles whose feature type includes the feature type to which the target feature belongs, and determine the isolation coefficient corresponding to the target feature based on the first number and the second number.

[0054] The third determination unit is used to determine the preemption coefficient of the target feature according to the first number, the second number, the confidence of the target feature, and the isolation coefficient corresponding to the target feature.

[0055] According to a vehicle identification device provided by the present application, the first processing unit is specifically used for:

[0056] For each of the candidate vehicles, determine the weight corresponding to each second target feature in the multiple target features and the target association features, the number of times each second target feature appears in the multi-dimensional features of the multiple candidate vehicles, the number of times each second target feature appears in the multi-dimensional features of the candidate vehicles, and the isolation coefficient corresponding to each second target feature.

[0057] The correlation score corresponding to the candidate vehicle is determined based on the weight corresponding to each second target feature, the number of times each second target feature appears in the multi-dimensional features of the multiple candidate vehicles, the number of times each second target feature appears in the multi-dimensional features of the candidate vehicle, and the isolation coefficient corresponding to each second target feature.

[0058] According to a vehicle identification device provided by the present application, the device also includes a calculation unit, a fourth determination unit, a fifth determination unit, and a sixth determination unit.

[0059] The calculation unit is used to perform feature association calculation on the multi-dimensional features of each candidate vehicle based on the minimum support and minimum confidence criteria to obtain multiple initial association feature combinations and initial confidences corresponding to each initial association feature combination.

[0060] The fourth determining unit is used to determine, from the multiple initial correlation feature combinations, a plurality of first correlation feature combinations whose initial confidences are greater than a preset threshold value according to the initial confidences corresponding to the respective initial correlation feature combinations.

[0061] The fifth determination unit is used to determine, for each of the first associated feature combinations, a confidence correction coefficient corresponding to the first associated feature combination based on the confidence levels of at least two features included in the first associated feature combination and the number of times each feature appears in the multidimensional features corresponding to the multiple preset vehicles, and to correct the initial confidence levels corresponding to the first associated feature combination based on the confidence correction coefficient to obtain a corrected confidence level.

[0062] The sixth determination unit is used to determine, among the multiple first associated feature combinations, a first associated feature combination whose corresponding corrected confidence level is greater than the preset threshold as the preset associated feature combination corresponding to the candidate vehicle.

[0063] In a third aspect, an embodiment of the present application further provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, a vehicle identification method as described in any one of the above is implemented.

[0064] In a fourth aspect, an embodiment of the present application further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a vehicle identification method as described in any one of the above.

[0065] In a fifth aspect, an embodiment of the present application further provides a computer program product, including a computer program, which, when executed by a processor, implements any of the vehicle identification methods described above.

[0066] The vehicle identification method, device and electronic device provided by the present application, when performing vehicle identification, obtains a target image of the vehicle to be identified, and based on the target image, obtains the multi-dimensional features of the vehicle to be identified and the confidence corresponding to each feature; then based on the confidence corresponding to each feature, multiple target features with confidence greater than a preset confidence threshold are determined from the multi-dimensional features, and the multiple target features are matched with the multi-dimensional features of multiple preset vehicles to obtain the identification result of the vehicle to be identified. In this way, multiple target features with a high confidence level are used as the basis for vehicle identification, and used for vehicle identification, which can effectively improve the accuracy of the vehicle identification result. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] In order to more clearly illustrate the technical solutions in the present application or the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0068] Figure 1 A schematic diagram of a vehicle identification method provided in an embodiment of the present application;

[0069] Figure 2 A flowchart of a method for matching multiple target features with respective multi-dimensional features of multiple preset vehicles provided in an embodiment of the present application;

[0070] Figure 3 A schematic diagram of the structure of a vehicle identification device provided in an embodiment of the present application;

[0071] Figure 4 A schematic diagram of the physical structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0072] In order to make the purpose, technical solutions and advantages of this application clearer, the technical solutions in this application will be clearly and completely described below in conjunction with the drawings in this application. Obviously, the described embodiments are part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0073] In the embodiments of the present application, "at least one" means one or more, and "more" means two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone, where A and B can be singular or plural. In the text description of the present application, the character " / " generally indicates that the associated objects before and after are in an "or" relationship.

[0074] The technical solution provided in the embodiments of the present application can be applied to vehicle identification scenarios. For example, identification of vehicles on the road, vehicle identification in parking fee collection systems, etc. At present, vehicle identification is usually performed based on license plates. However, since license plates are prone to damage, obstruction, or even duplication, the accuracy of vehicle identification results is low.

[0075] In order to improve the accuracy of the vehicle identity recognition results, in an embodiment of the present application, it is possible to consider using features of different dimensions of the vehicle as the basis for vehicle identity recognition for vehicle identity recognition; in addition, considering that features of different dimensions have different confidence levels, features with higher confidence levels are selected from the features of different dimensions of the vehicle as the basis for vehicle identity recognition for vehicle identity recognition, which can effectively improve the accuracy of the vehicle identity recognition results.

[0076] The vehicle identification method provided by the present application will be described in detail below through the following specific embodiments. It is understandable that the following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments.

[0077] Embodiment 1

[0078] Figure 1 The present invention provides a flow chart of a vehicle identification method according to an embodiment of the present invention. The vehicle identification method can be executed by software and / or hardware devices, such as electronic devices. For example, see Figure 1 As shown, the vehicle identification method may include:

[0079] S101, obtaining a target image of a vehicle to be identified.

[0080] For example, when acquiring the target image of the vehicle to be identified, the target image of the vehicle to be identified can be directly captured; or the target image of the vehicle to be identified sent by other electronic devices can be received; of course, the target image of the vehicle to be identified can also be searched and acquired from the current storage. The specific settings can be made according to actual needs. Here, the embodiment of the present application does not further limit the method of acquiring the target image.

[0081] S102: Based on the target image, obtain multi-dimensional features of the vehicle to be identified and the confidence level corresponding to each feature.

[0082] For example, in the embodiment of the present application, the multi-dimensional features can be shown in Table 1 below, and can include features under different dimensions. For example, the features under the vehicle dimension can include specific body color, vehicle brand, ..., and vehicle model. The features under the dimension of in-car items can include specific pendants, ornaments, ..., and water cups, and the features under the dimension of in-car personnel can include specific drivers, co-pilots, ..., and back row personnel. The specific settings can be made according to actual needs. Here, the embodiment of the present application is only illustrated by taking the three different dimensions of vehicle dimension, in-car item dimension, and in-car personnel as examples, but it does not mean that the embodiment of the present application is limited to this.

[0083] Table 1

[0084] Dimensions feature feature feature feature vehicle Body color Vehicle Brand … Vehicle Model Car Items Widgets Ornaments … Water cup People in the car Driver Co-pilot … Back row

[0085] It is understandable that in the embodiment of the present application, when the identity of the vehicle to be identified is performed, the multi-dimensional features of the vehicle to be identified are usually compared with the multi-dimensional features of multiple preset vehicles whose identities have been determined to determine the identity recognition result of the vehicle to be identified. In the embodiment of the present application, in order not to be limited by the image shooting angle, the features of multiple preset vehicles in different dimensions can be obtained in advance, and the features in all dimensions can be obtained as much as possible, so as to cover the features at different shooting angles to the greatest extent through multi-dimensional features, thereby improving the accuracy of the identity recognition result.

[0086] However, in actual operation, it is difficult to obtain multi-dimensional features based on a target image. For example, if the target image is a picture of the side of the vehicle to be identified, fewer features can be extracted based on the target image. For example, specific vehicle body color features and driver features can be extracted, but it is difficult to extract specific vehicle model features, license plate number features, etc. Therefore, when obtaining the multi-dimensional features of the vehicle to be identified based on the target image, in addition to extracting the features of the vehicle to be identified from the target image, it is also necessary to fill in the features of the vehicle to be identified based on the multi-dimensional features of multiple preset vehicles to obtain the multi-dimensional features of the vehicle to be identified.

[0087] For example, in the embodiment of the present application, it is assumed that the multi-dimensional feature of the preset vehicle can be recorded as m*n, where m represents m dimensions and n represents n subdivided features under one dimension. The multi-dimensional feature m*n can be regarded as a vehicle identity template corresponding to the preset vehicle. For example, the vehicle identity template can be represented by matrix A:

[0088]

[0089] For example, in combination with the above Table 1, the vehicle represents a dimension, and the body color, vehicle brand, ..., and vehicle model are multiple subdivision features under the vehicle dimension; the items in the car represent a dimension, and the pendant features, ornament features, ..., and water cup features are subdivision features under the dimension of items in the car; the people in the car represent a dimension, and the driver features, co-pilot features, ..., and back seat features are subdivision features under the dimension of people in the car.

[0090] By way of example, in an embodiment of the present application, for each preset vehicle, in addition to extracting the multi-dimensional features m*n shown in the above matrix A, the confidence corresponding to each feature in the multi-dimensional features m*n can also be obtained as a reference in the subsequent vehicle identification process, which can be represented by the text T, T = [a11&p11,a12&p12,…,amn&pmn], wherein a11 represents the feature of the 1st row and 1st column in the matrix A, p11 represents the confidence corresponding to the feature a11, a12 represents the feature of the 1st row and 2nd column in the matrix A, p12 represents the confidence corresponding to the feature a12,…,amn represents the feature of the mth row and nth column in the matrix A, and pmn represents the confidence corresponding to the feature amn.

[0091] It should be noted that in the embodiment of the present application, different features in the above-mentioned multidimensional features m*n may be extracted by different algorithms, and the different algorithms can be integrated in different image analysis systems. After extracting the corresponding features, the different image analysis systems can send the extracted features and the confidence levels corresponding to the features through an interface to the electronic device that executes the technical solution of the present application. The electronic device can analyze and fuse the features received from different image analysis systems and the confidence levels corresponding to the features to obtain the above-mentioned multidimensional features m*n.

[0092] It is understandable that when obtaining the multi-dimensional features of a preset vehicle, the multi-dimensional features are not fixed, but a process of continuous filling and updating. For example, for the vehicle personnel dimension, in addition to obtaining the driver features, co-pilot features, ..., and back row personnel features as shown in Table 1, if the facial features, jewelry features, ..., and clothing features of the vehicle personnel are further obtained, the features under the vehicle personnel dimension shown in Table 1 can also be updated based on the acquired facial features, jewelry features, ..., and clothing features, so that the multi-dimensional features of the preset vehicle are richer. For example, please refer to the following Table 2:

[0093] Table 2

[0094] Face beard Eye … Eyebrow accessories Glasses earrings … necklace … … … … … Clothes and hats hat scarf … Jacket

[0095] As shown in Table 2, facial features may include beard, eyes, ..., eyebrows, accessories features may include glasses, earrings, ..., necklaces, and clothing features may include hats, scarves, ..., tops. Assuming that the facial features, accessories features, ..., and clothing features are j*k features, then A corresponding to the preset vehicle may be updated to A=(m+j)*(n+k), and the updated matrix is ​​used as the vehicle identity template corresponding to the preset vehicle.

[0096] For example, when obtaining the multi-dimensional features of a preset vehicle, the license plate of the vehicle that is successfully identified is Zhejiang A12345. When the vehicle is identified for the first time, the information that can be extracted from the image is filled in sequentially, for example, the body color is filled into the features under the corresponding dimension, and the confidence corresponding to the features is saved together to obtain the original identity information A1 of the vehicle. Since the effective information obtained from each capture of the vehicle is inconsistent, some are the features of the front of the vehicle, and some are the features of the rear of the vehicle. For example, the matrix corresponding to the vehicle obtained based on the image captured for the i-th time is A i , if A i If the confidence corresponding to a feature in is greater than the confidence corresponding to the feature of the vehicle obtained based on the image captured at the i-1th time, the feature obtained at the i-1th time is replaced; if A i If a feature is a new feature, then the matrix corresponding to the vehicle obtained based on the image captured at the i-1th time is A i-1 This feature is added to realize the filling and updating of the feature. By continuously filling and updating, the features of the vehicle can be gradually expanded, thereby improving the accuracy of vehicle identification.

[0097] For example, if the front view of the vehicle is not included in the image captured at the i-th time, the interior pendant cannot be identified, and the interior pendant feature is set to 0; and the interior pendant feature is identified based on the image captured at the i-1th time, which includes the front view of the vehicle. In this case, if the confidence corresponding to the feature at the i-1th time is greater than the confidence corresponding to the feature at the i-th time, there is no need to update the feature identified at the i-1th time using the feature identified at the i-1th time; on the contrary, if the confidence corresponding to the feature at the i-1th time is less than the confidence corresponding to the feature at the i-th time, it is necessary to update the feature identified at the i-1th time using the feature identified at the i-1th time.

[0098] It should be noted that in the embodiment of the present application, for vehicles without license plates or vehicles without accurate identification, virtual numbers are generated, and the corresponding features are filled in sequence. The matrix A corresponding to the latest vehicle is obtained by expansion and replacement according to the replacement principle. i .

[0099] Normally, the multi-dimensional features of multiple preset vehicles used as a benchmark for vehicle identity recognition are pre-acquired. In the vehicle identity recognition scenario, when acquiring the multi-dimensional features of the vehicle to be identified, it is necessary to use the multi-dimensional features of each preset vehicle as a benchmark to fill in the features extracted based on the target image, so that the multi-dimensional features of the vehicle to be identified after filling are consistent with the feature types of the multi-dimensional features of the preset vehicle, that is, if the multi-dimensional features of the preset vehicle include body color features, then the corresponding multi-dimensional features of the vehicle to be identified also need to include body color features.

[0100] For example, assuming that the multi-dimensional feature of the preset vehicle is m*n, the multi-dimensional feature m*n of the vehicle to be identified can be obtained accordingly. The feature types of these two multi-dimensional features are consistent, and the multi-dimensional feature m*n of the vehicle to be identified can be regarded as a vehicle identity template corresponding to the vehicle to be identified, and the vehicle identity template can also be represented by a matrix B similar to the above-mentioned matrix A.

[0101] By way of example, in an embodiment of the present application, for a vehicle to be identified, in addition to obtaining the multidimensional features m*n of the vehicle to be identified, the confidence corresponding to each feature in the multidimensional features m*n can also be obtained, which is used as a reference in the subsequent vehicle identification process, and can be represented by the text T1, T1 = [b11&p11, b12&p12,…, bmn&pmn], wherein b11 represents the feature of the 1st row and 1st column in the matrix B, p11 represents the confidence corresponding to the feature b11, b12 represents the feature of the 1st row and 2nd column in the matrix B, p12 represents the confidence corresponding to the feature b12,…, bmn represents the feature of the mth row and nth column in the matrix B, and pmn represents the confidence corresponding to the feature bmn.

[0102] After obtaining the multi-dimensional features of the vehicle to be identified and the confidence level corresponding to each feature, the following S103 can be executed:

[0103] S103. Based on the confidence corresponding to each feature, determine multiple target features whose confidence is greater than a preset confidence threshold from the multi-dimensional features.

[0104] The value of the preset confidence threshold can be set according to actual needs. Here, the embodiment of the present application does not further limit the specific value of the preset confidence threshold.

[0105] Taking into account the different confidence levels corresponding to different features, for example, the confidence level of the license plate features identified on a sunny day is higher than the confidence level of the vehicle color features identified on a cloudy day. Therefore, in order to further improve the accuracy of the vehicle identity recognition results, in an embodiment of the present application, features with higher confidence levels can be selected as the basis for vehicle identity recognition, and matched with the multi-dimensional features of multiple preset vehicles, that is, the following S104 is executed, which can further improve the accuracy of the vehicle identity recognition results.

[0106] S104, matching the multiple target features with the multi-dimensional features of the multiple preset vehicles to obtain an identity recognition result of the vehicle to be recognized.

[0107] For example, if among the multi-dimensional features of the vehicle to be identified, multiple target features with a confidence level greater than a preset confidence threshold include feature a, feature b, and feature c, then feature a, feature b, and feature c are used as the basis for identifying the vehicle to be identified, and are matched with the multi-dimensional features of multiple preset vehicles to determine the identity recognition result of the vehicle to be identified.

[0108] It can be seen that in the embodiment of the present application, when performing vehicle identification, the target image of the vehicle to be identified is obtained, and based on the target image, the multi-dimensional features of the vehicle to be identified and the confidence corresponding to each feature are obtained; then based on the confidence corresponding to each feature, multiple target features with confidence greater than a preset confidence threshold are determined from the multi-dimensional features, and the multiple target features are matched with the multi-dimensional features of multiple preset vehicles to obtain the identification result of the vehicle to be identified. In this way, multiple target features with a larger confidence are used as the basis for vehicle identification, and used for vehicle identification, which can effectively improve the accuracy of the vehicle identification result.

[0109] Based on the above Figure 1 In order to facilitate understanding of how to match multiple target features with the multi-dimensional features of multiple preset vehicles in the above S104 to obtain the identity recognition result of the vehicle to be recognized, the following will be described. Figure 2 The second embodiment shown describes in detail the technical solution of how to match multiple target features with the multi-dimensional features of multiple preset vehicles to obtain the identity recognition results of the vehicle to be recognized.

[0110] Embodiment 2

[0111] Figure 2 A flowchart of a method for matching multiple target features with multi-dimensional features of multiple preset vehicles provided in an embodiment of the present application, which method can also be executed by software and / or hardware devices. Figure 2 As shown, the method may include:

[0112] S201. Match multiple target features with respective multi-dimensional features of multiple preset vehicles, and determine a candidate vehicle including the multiple target features from the multiple preset vehicles.

[0113] Taking multiple target features including feature a, feature b and feature c as an example, the three target features can be matched with the multi-dimensional features of each preset vehicle, and the preset vehicle including the three target features can be determined from multiple preset vehicles, and the preset vehicle including the three target features can be determined as a candidate vehicle.

[0114] For example, when the number of determined candidate vehicles is one, the candidate vehicle may be directly determined as the identity recognition result of the vehicle to be recognized.

[0115] When the number of candidate vehicles determined is multiple, the following S202 is executed:

[0116] S202. When there are multiple candidate vehicles, based on the preemption coefficients of the multiple target features, determine a first target feature corresponding to the maximum preemption coefficient from the multiple target features, and the preemption coefficient of the target feature is used to characterize the rarity of the target feature.

[0117] Among them, the rarer the target feature, the larger the corresponding preemption coefficient.

[0118] For example, in an embodiment of the present application, when determining the preemption coefficients of multiple target features, considering that the method for determining the preemption coefficient of each target feature is similar, the determination of the preemption coefficient of a certain target feature will be taken as an example to describe how to determine the preemption coefficients of multiple target features.

[0119] For example, when determining the preemption coefficient of a certain target feature, you can first determine from multiple preset vehicles a first number of preset vehicles whose corresponding multi-dimensional features include the target feature, and a second number of preset vehicles whose feature types include the feature type to which the target feature belongs, and determine the isolation coefficient corresponding to the target feature based on the first number and the second number; then determine the preemption coefficient of the target feature based on the first number, the second number, the confidence level of the target feature, and the isolation coefficient corresponding to the target feature.

[0120] Assuming that the number of preset vehicles is 20, the target feature is red body color, and the feature type to which the target feature belongs is body color, then the first number can be understood as the number of preset vehicles among the 20 preset vehicles whose corresponding multi-dimensional features include red body color. Assuming that there are 3 preset vehicles whose body colors are all red, the first number can be determined to be 3; the second number can be understood as the number of preset vehicles among the 20 preset vehicles whose corresponding multi-dimensional features include body color in the feature type. Assuming that the feature types to which the multi-dimensional features of the 20 preset vehicles belong include body color, the second number can be determined to be 20.

[0121] For example, in the embodiment of the present application, when determining the isolation coefficient corresponding to the target feature based on the first quantity and the second quantity, the isolation coefficient can be determined according to Determine the isolation coefficient corresponding to the target feature.

[0122] Wherein, S(B(i,j)) represents the isolation coefficient corresponding to the target feature B(i,j), C(B(i,j)) represents the first number of preset vehicles whose corresponding multi-dimensional features include the target feature, T B(i,j) Indicates the feature type to which the target feature belongs, C(T B(i,j) ) represents the second number of preset vehicles of the feature type to which the target feature belongs, including the feature type to which the target feature belongs.

[0123] For example, in the embodiment of the present application, when determining the preemption coefficient of the target feature based on the first quantity, the second quantity, the confidence of the target feature, and the isolation coefficient corresponding to the target feature, the preemption coefficient of the target feature can be determined according to Determine the preemption coefficient of the target feature. Wherein, R represents the preemption coefficient of the target feature, N represents the number of multiple preset vehicles, b(B(i,j)) represents the confidence corresponding to the target feature B(i,j), and S(B(i,j)) represents the isolation coefficient corresponding to the target feature B(i,j).

[0124] Taking multiple target features including feature a, feature b and feature c as an example, combined with the above description, the preemption coefficients of feature a, feature b and feature c can be calculated. Assuming that the preemption coefficient of feature b is the largest, the preemption coefficient of feature a is second, and the preemption coefficient of feature c is the smallest, feature b+feature a+feature c can be used to name the vehicle to be identified, that is, feature b+feature a+feature c is used as the T-name of the vehicle to be identified.

[0125] After calculating the preemption coefficients of feature a, feature b, and feature c, considering that feature b has the largest preemption coefficient and is the rarest, a progressive search method can be used to determine the target-associated features of feature b, so as to expand features a, feature b, and feature c, and use the target-associated features of features a, feature b, feature c, and feature b as the basis for identity recognition, and match them with the multi-dimensional features of multiple candidate vehicles, that is, execute the following S203.

[0126] S203, matching the first target feature with a combination of preset associated features corresponding to a plurality of candidate vehicles to determine a target associated feature of the first target feature; and matching the plurality of target features and the target associated features with the multi-dimensional features of each candidate vehicle respectively to obtain a plurality of correlation scores.

[0127] It can be understood that in the embodiment of the present application, the first target feature is matched with a preset association feature combination corresponding to multiple candidate vehicles, and a progressive search method is adopted to determine the target association feature of feature b. Only one association feature can be searched each time and used as the target association feature of feature b. Of course, two association features can also be searched each time and used as the target association feature of feature b. The specific setting can be made according to actual needs. Here, the embodiment of the present application only takes the example of searching only one association feature each time and using it as the target association feature of feature b for explanation, but it does not mean that the embodiment of the present application is limited to this.

[0128] For example, in an embodiment of the present application, when obtaining preset association feature combinations corresponding to multiple candidate vehicles, for each candidate vehicle, the multi-dimensional features of the candidate vehicle can be first obtained, and based on the minimum support and minimum confidence criteria, the Apriori algorithm is used to perform feature association calculations on the multi-dimensional features to obtain multiple initial association feature combinations corresponding to the candidate vehicle, and the multiple initial association feature combinations each correspond to their own confidence. After obtaining multiple initial association feature combinations, the multiple initial association feature combinations can be first screened according to a basic confidence, and from the multiple initial association feature combinations, multiple first association feature combinations with confidence greater than the basic confidence are screened, and the multiple first association feature combinations meet the minimum support and minimum confidence requirements.

[0129] It should be noted that, in the embodiment of the present application, after screening out multiple first associated feature combinations whose confidences are greater than the basic confidences, the multiple first associated feature combinations are not directly determined as the preset associated feature combinations corresponding to the candidate vehicles, but the confidences corresponding to each of the multiple first associated feature combinations are further corrected to obtain new confidences corresponding to each of the multiple first associated feature combinations; and based on the new confidences corresponding to the multiple first associated feature combinations, the first associated feature combinations whose corresponding confidences are greater than the preset threshold are screened out from the multiple first associated feature combinations, and are determined as the preset associated feature combinations corresponding to the candidate vehicles.

[0130] By way of example, in an embodiment of the present application, when correcting the confidences corresponding to each of a plurality of first associated feature combinations, for each first associated feature combination, a confidence correction coefficient corresponding to the first associated feature combination can be determined based on the confidences of each of at least two features included in the first associated feature combination and the number of times each appears in the multidimensional features corresponding to a plurality of preset vehicles, and the initial confidence corresponding to the first associated feature combination is corrected based on the confidence correction coefficient to obtain a corrected confidence; and the first associated feature combination among the plurality of first associated feature combinations whose corresponding corrected confidence is greater than a preset threshold is determined as the preset associated feature combination corresponding to the candidate vehicle.

[0131] Assume that the first associated feature combination M (m1, m2, ..., m t ) includes t features, t is a positive integer greater than or equal to 2, then the first associated feature combination M (m1, m2, ..., m t ) corresponding to the confidence correction coefficient, we can use Determine the first associated feature combination M(m1, m2, ..., m t ) corresponding to the confidence correction coefficient, where δ represents the first associated feature combination M(m1, m2,…, m t ) corresponding to the confidence correction coefficient, γ represents the effective coefficient of the first associated feature combination M, M(m1, m2,…, m t )∈T, then γ=1, Then γ=0; Represents the i-th feature m i The corresponding confidence level, C(m i ) represents the i-th feature m i The number of times it appears in the multi-dimensional features corresponding to multiple preset vehicles, N represents the number of preset vehicles, and K represents the adjustment parameter.

[0132] After determining the first associated feature combination M (m1, m2, ..., m t) After the corresponding confidence correction coefficient δ, we can calculate the confidence correction coefficient according to P′ M =δP M Determine the first tandem feature combination M (m1, m2, ..., m t ) is the corrected confidence. Among them, P′ M Represents the first associated feature combination M(m1, m2,…, m t ) Corrected confidence, P M Indicates the initial confidence corresponding to the first associated feature combination.

[0133] It is understandable that, usually, a preset combination of associated features corresponding to each of the multiple preset vehicles is predetermined, rather than just a preset combination of associated features corresponding to each of the candidate vehicles. In addition, each preset vehicle can also be named. For example, in an embodiment of the present application, when naming a preset vehicle, multiple features with a confidence level greater than a preset confidence threshold can be determined from the multi-dimensional features corresponding to the preset vehicle, and the preemption coefficient of each of the multiple features with a confidence level greater than the preset confidence threshold can be calculated. The specific calculation method can refer to the above-mentioned calculation method for calculating the preemption coefficient of each of the multiple target features, and the embodiment of the present application will not be repeated here.

[0134] After calculating the preemption coefficients of multiple features with confidence greater than the preset confidence threshold, the multiple features can be sorted in order from large to small, and the sorting results are used as the T-name of the preset vehicle. It should be noted that if the feature names overlap, the numbers 1 and 2 are added after the preset vehicle according to the confidence level. For example, the license plate feature is usually not repeated and is the most rare feature. Therefore, the preemption coefficient R value will be the highest. Adding the body color feature and vehicle model feature, the corresponding preset vehicle T-name can be recorded as: Zhejiang A12345+pink red flag+H7.

[0135] After obtaining the preset associated feature combinations corresponding to the multiple candidate vehicles, the first target feature can be matched with the preset associated feature combinations corresponding to the multiple candidate vehicles to determine the identity recognition result of the vehicle to be recognized.

[0136] For example, when matching the first target feature with the preset associated feature combination corresponding to multiple candidate vehicles, the first target feature can be matched with the preset associated feature combination first to determine at least one associated feature combination including the first target feature; and the features other than the first target feature in the associated feature combination corresponding to the maximum confidence are determined as target associated features. Afterwards, the multiple target features and the target associated features are matched with the multi-dimensional features of each candidate vehicle respectively to obtain multiple correlation scores. In the case where only one correlation score among the multiple correlation scores is greater than the preset score threshold, the preset vehicle corresponding to the correlation score is determined as the identity recognition result; in the case where at least two correlation scores among the multiple correlation scores are greater than the preset score threshold, the features other than the first target feature in the associated feature combination corresponding to the second largest confidence are determined as new target associated features of the first target feature; the multiple target features, the target associated features and the new target associated features are matched with the multi-dimensional features of each of the multiple candidate vehicles respectively until at least one associated feature combination is traversed to obtain multiple correlation scores.

[0137] When multiple target features and target-related features are matched with the multi-dimensional features of each candidate vehicle respectively to obtain multiple correlation scores, the description is given by taking as an example how to match multiple target features and target-related features with the multi-dimensional features of any candidate vehicle to determine their corresponding correlation scores.

[0138] For example, when matching multiple target features and target-associated features with the multi-dimensional features of a candidate vehicle, first determine the weight corresponding to each second target feature in the multiple target features and target-associated features, the number of times each second target feature appears in the multi-dimensional features of each of the multiple candidate vehicles, the number of times each second target feature appears in the multi-dimensional features of the candidate vehicle, and the isolation coefficient corresponding to each second target feature; and determine the correlation score corresponding to the candidate vehicle based on the weight corresponding to each second target feature, the number of times each second target feature appears in the multi-dimensional features of each of the multiple candidate vehicles, the number of times each second target feature appears in the multi-dimensional features of the candidate vehicle, and the isolation coefficient corresponding to each second target feature.

[0139] For example, when determining the weight corresponding to the i-th second target feature, it can be based on Determine the weight corresponding to the i-th second target feature. Where W i represents the weight corresponding to the i-th second target feature, n(t i ) represents the first number of preset vehicles whose corresponding multi-dimensional features include the i-th second target feature, and N represents the number of preset vehicles.

[0140] For example, when determining the correlation score corresponding to the candidate vehicle according to the weight corresponding to each second target feature, the number of times each second target feature appears in the multi-dimensional features of each of the plurality of candidate vehicles, the number of times each second target feature appears in the multi-dimensional features of the candidate vehicle, and the isolation coefficient corresponding to each second target feature, the correlation score corresponding to the candidate vehicle can be calculated according to Determine the correlation scores corresponding to the candidate vehicles, thereby determining a plurality of correlation scores.

[0141] Where d represents the text of the vehicle identity template corresponding to the candidate vehicle, T1 represents the text of the vehicle identity template corresponding to the vehicle to be identified, Score(T s ,d) represents the correlation score corresponding to the candidate vehicle, W i represents the weight corresponding to the i-th second target feature, Q represents the total number of multiple target features and target-related features, Among them, f i represents the number of times the i-th second target feature appears in the multi-dimensional features of the candidate vehicle, qf i represents the number of times the i-th second target feature appears in the text of the vehicle identity template corresponding to the candidate vehicle, k1, k2, b are adjustable parameters, K = k1 (1-b + b * L d / L ave ), L d , L ave is the length of d and the average length of the text corresponding to all candidate vehicles.

[0142] Taking feature b as the first target feature, feature b can be first matched with preset associated feature combinations corresponding to multiple candidate vehicles. Assuming that the associated feature combinations including feature b include associated feature combination (b, d), associated feature combination (b, e), associated feature combination (b, f), and associated feature combination (b, g), and the confidence corresponding to the associated feature combination (b, d) is 90%, the confidence corresponding to the associated feature combination (b, e) is 93%, the confidence corresponding to the associated feature combination (b, f) is 70%, and the confidence corresponding to the associated feature combination (b, g) is 80%. Then, the associated feature combination corresponding to the maximum confidence is feature e in (b, e) as the target associated feature of feature b, and features a, feature b, feature c and the target associated feature e of feature b are matched with the multi-dimensional features of each candidate vehicle respectively to obtain multiple correlation scores.

[0143] When only one correlation score among multiple correlation scores is greater than the preset score threshold, the preset vehicle corresponding to the correlation score is determined as the identity recognition result; when at least two correlation scores among multiple correlation scores are greater than the preset score threshold, feature d in the associated feature combination (b, d) corresponding to the second largest confidence is used as the new target associated feature; feature a, feature b, feature c, feature e and feature d are respectively matched with the multi-dimensional features of multiple candidate vehicles until the associated feature combination (b, f) is traversed to obtain multiple correlation scores.

[0144] After traversing to the last associated feature combination including the first target feature, multiple target features and their target associated features are matched with the multi-dimensional features of the multiple candidate vehicles respectively, and the identity recognition result is determined according to the obtained multiple correlation scores, that is, the following S204 is executed:

[0145] S204: Determine an identity recognition result according to the multiple correlation scores.

[0146] For example, when there is one correlation score greater than a preset threshold among multiple correlation scores, the candidate vehicle corresponding to the correlation score greater than the preset score threshold is determined as the identity recognition result; when there are at least two correlation scores greater than the preset threshold among multiple correlation scores, the target feature corresponding to the second largest preemption coefficient is determined as the new first target feature, and the multiple target features, the first target feature, the target association feature of the first target feature, the new first target feature, and the target association feature of the new first target feature are matched with the preset association feature combination corresponding to the multiple candidate vehicles until there is one correlation score greater than the preset score threshold among the obtained multiple correlation scores, and the candidate vehicle corresponding to the correlation score greater than the preset score threshold is determined as the identity recognition result of the vehicle to be identified.

[0147] Assume that feature a, feature b, feature c, target-associated feature e, target-associated feature d, target-associated feature g and target-associated feature f of feature b are matched with the multi-dimensional features of multiple candidate vehicles to obtain multiple correlation scores. If only one correlation score among the multiple correlation scores is greater than a preset threshold, the candidate vehicle corresponding to the correlation score greater than the preset score threshold is determined as the identity recognition result; if at least two correlation scores among the multiple correlation scores are greater than the preset threshold, feature a with the second largest preemption coefficient among features a, feature b and feature c is determined as the new first target feature, and the target-associated feature of feature a is determined. Features a, feature b, feature c, target-associated feature e, target-associated feature d, target-associated feature g and target-associated feature f of feature b, and the target-associated feature of feature a are matched with the multi-dimensional features of multiple candidate vehicles until one correlation score among the multiple correlation scores obtained is greater than the preset score threshold, and the candidate vehicle corresponding to the correlation score greater than the preset score threshold is determined as the identity recognition result.

[0148] It can be seen that in the embodiment of the present application, when determining the identification result of the vehicle to be identified, multiple target features with greater confidence can be matched with the multi-dimensional features of multiple preset vehicles, and a candidate vehicle including multiple target features can be determined from multiple preset vehicles; when the number of candidate vehicles is multiple, based on the preemption coefficients of multiple target features, the first target feature corresponding to the maximum preemption coefficient is determined from the multiple target features, that is, the first target feature with a higher degree of rarity, and the first target feature with a higher degree of rarity is matched with the preset association feature combination corresponding to multiple candidate vehicles to determine the target association feature of the first target feature, so as to expand the multiple target features, and the multiple target features and the target association features are matched with the multi-dimensional features of each candidate vehicle respectively to obtain multiple correlation scores; and then the identification result is determined according to the multiple correlation scores. In this way, the first target feature with a greater degree of confidence and its target association feature are used as the basis for vehicle identification, which can not only effectively improve the accuracy of the vehicle identification result, but also give priority to the target feature with the maximum preemption coefficient as the first target feature, which can further improve the efficiency of vehicle identification.

[0149] The vehicle identification device provided in the present application is described below. The vehicle identification device described below and the vehicle identification method described above can be referenced to each other.

[0150] Figure 3 This is a schematic diagram of a vehicle identification device provided in an embodiment of the present application. For example, see Figure 3 As shown, the vehicle identification device 30 may include:

[0151] The first acquisition unit 301 is used to acquire a target image of a vehicle to be identified.

[0152] The second acquisition unit 302 is used to acquire multi-dimensional features of the vehicle to be identified and the confidence level corresponding to each feature based on the target image.

[0153] The first determination unit 303 is used to determine, based on the confidence corresponding to each feature, a plurality of target features whose confidence is greater than a preset confidence threshold from the multi-dimensional features.

[0154] The first processing unit 304 is used to match the multiple target features with the multi-dimensional features of the multiple preset vehicles to obtain an identity recognition result of the vehicle to be recognized.

[0155] Optionally, the first processing unit 304 is specifically configured to:

[0156] The plurality of target features are matched with the respective multi-dimensional features of the plurality of preset vehicles, and a candidate vehicle including the plurality of target features is determined from the plurality of preset vehicles.

[0157] In the case where there are multiple candidate vehicles, based on the preemption coefficients of the multiple target features, a first target feature corresponding to the maximum preemption coefficient is determined from the multiple target features, and the preemption coefficient of the target feature is used to characterize the rarity of the target feature;

[0158] Matching the first target feature with a combination of preset associated features corresponding to the plurality of candidate vehicles to determine a target associated feature of the first target feature; matching the plurality of target features and the target associated features with the multi-dimensional features of each candidate vehicle respectively to obtain a plurality of correlation scores;

[0159] Based on the multiple correlation scores, an identification result is determined.

[0160] Optionally, the first processing unit 304 is specifically configured to:

[0161] When one of the multiple correlation scores is greater than a preset threshold, the candidate vehicle corresponding to the correlation score greater than the preset score threshold is determined as the identity recognition result;

[0162] When there are at least two correlation scores greater than a preset threshold among multiple correlation scores, the target feature corresponding to the second largest preemption coefficient is determined as a new first target feature, and the multiple target features, the first target feature, the target association feature of the first target feature, the new first target feature, and the target association feature of the new first target feature are matched with the preset association feature combination corresponding to multiple candidate vehicles until there is a correlation score greater than the preset score threshold among the multiple correlation scores obtained, and the candidate vehicle corresponding to the correlation score greater than the preset score threshold is determined as the identity recognition result.

[0163] Optionally, the first processing unit 304 is specifically configured to:

[0164] The first target feature is matched with a preset associated feature combination to determine at least one associated feature combination including the first target feature; and the features other than the first target feature in the associated feature combination corresponding to the maximum confidence are determined as target associated features.

[0165] Multiple target features and target-related features are matched with the multi-dimensional features of each candidate vehicle to obtain multiple correlation scores.

[0166] When at least two correlation scores among the multiple correlation scores are greater than a preset score threshold, the features other than the first target feature in the association feature combination corresponding to the second largest confidence are determined as new target association features of the first target feature.

[0167] The multiple target features, target-associated features, and new target-associated features are matched with the multi-dimensional features of the multiple candidate vehicles respectively, until at least one associated feature combination is traversed to obtain multiple correlation scores.

[0168] Optionally, the vehicle identification device 30 further includes a second determination unit and a third determination unit.

[0169] A second determination unit is used to determine, for each target feature, from a plurality of preset vehicles, a first number of preset vehicles whose corresponding multi-dimensional feature includes the target feature, and a second number of preset vehicles whose feature types include the feature type to which the target feature belongs, and determine an isolation coefficient corresponding to the target feature based on the first number and the second number;

[0170] The third determination unit is used to determine the preemption coefficient of the target feature according to the first number, the second number, the confidence of the target feature and the isolation coefficient corresponding to the target feature.

[0171] Optionally, the first processing unit 304 is specifically configured to:

[0172] For each candidate vehicle, determine the weight corresponding to each second target feature in the multiple target features and target-associated features, the number of times each second target feature appears in the multi-dimensional features of the multiple candidate vehicles, the number of times each second target feature appears in the multi-dimensional features of the candidate vehicles, and the isolation coefficient corresponding to each second target feature.

[0173] The correlation score corresponding to the candidate vehicle is determined based on the weight corresponding to each second target feature, the number of times each second target feature appears in the multi-dimensional features of multiple candidate vehicles, the number of times each second target feature appears in the multi-dimensional features of the candidate vehicle, and the isolation coefficient corresponding to each second target feature.

[0174] Optionally, the vehicle identification device 30 further includes a calculation unit, a fourth determination unit, a fifth determination unit, and a sixth determination unit.

[0175] The calculation unit is used to perform feature association calculation on the multi-dimensional features of each candidate vehicle based on the minimum support and minimum confidence criteria to obtain multiple initial association feature combinations and initial confidences corresponding to each initial association feature combination.

[0176] The fourth determining unit is used to determine, from multiple initial correlation feature combinations, multiple first correlation feature combinations whose initial confidences are greater than a preset threshold value according to the initial confidences corresponding to the respective initial correlation feature combinations.

[0177] The fifth determination unit is used to determine, for each first associated feature combination, a confidence correction coefficient corresponding to the first associated feature combination based on the confidence levels of at least two features included in the first associated feature combination and the number of times each feature appears in the multidimensional features corresponding to multiple preset vehicles, and to correct the initial confidence level corresponding to the first associated feature combination based on the confidence correction coefficient to obtain a corrected confidence level.

[0178] The sixth determination unit is used to determine, among the plurality of first associated feature combinations, a first associated feature combination whose corresponding corrected confidence level is greater than a preset threshold as a preset associated feature combination corresponding to the candidate vehicle.

[0179] The vehicle identity recognition device 30 provided in the embodiment of the present application can execute the technical solution of the vehicle identity recognition method in any of the above-mentioned embodiments. Its implementation principle and beneficial effects are similar to the implementation principle and beneficial effects of the vehicle identity recognition method. Please refer to the implementation principle and beneficial effects of the vehicle identity recognition method, and no further details will be given here.

[0180] Figure 4 A schematic diagram of the physical structure of an electronic device provided in an embodiment of the present application, such as Figure 4As shown, the electronic device may include: a processor 410, a communication interface 420, a memory 430 and a communication bus 440, wherein the processor 410, the communication interface 420 and the memory 430 communicate with each other through the communication bus 440. The processor 410 may call the logic instructions in the memory 430 to execute the vehicle identification method, which includes: obtaining a target image of a vehicle to be identified; based on the target image, obtaining multi-dimensional features of the vehicle to be identified and the confidence corresponding to each feature; based on the confidence corresponding to each feature, determining multiple target features with confidence greater than a preset confidence threshold from the multi-dimensional features; matching the multiple target features with the multi-dimensional features of multiple preset vehicles to obtain the identification result of the vehicle to be identified.

[0181] In addition, the logic instructions in the above-mentioned memory 430 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art, and the computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk.

[0182] On the other hand, the present application also provides a computer program product, which includes a computer program, and the computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the vehicle identity recognition method provided by the above-mentioned methods, and the method includes: obtaining a target image of the vehicle to be identified; based on the target image, obtaining multi-dimensional features of the vehicle to be identified and the confidence corresponding to each feature; based on the confidence corresponding to each feature, determining multiple target features whose confidence is greater than a preset confidence threshold from the multi-dimensional features; matching the multiple target features with the multi-dimensional features of multiple preset vehicles to obtain the identity recognition result of the vehicle to be identified.

[0183] On the other hand, the present application also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the vehicle identification method provided by the above-mentioned methods, the method comprising: obtaining a target image of a vehicle to be identified; based on the target image, obtaining multi-dimensional features of the vehicle to be identified and the confidence corresponding to each feature; based on the confidence corresponding to each feature, determining from the multi-dimensional features a plurality of target features whose confidence is greater than a preset confidence threshold; matching the plurality of target features with the multi-dimensional features of a plurality of preset vehicles to obtain an identity identification result of the vehicle to be identified.

[0184] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0185] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0186] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A vehicle identification method, characterized in that: include: Acquire a target image of a vehicle to be identified; Based on the target image, obtaining multi-dimensional features of the vehicle to be identified and the confidence level corresponding to each of the features; Based on the confidence corresponding to each of the features, determining a plurality of target features whose confidence is greater than a preset confidence threshold from the multi-dimensional features; The multiple target features are matched with the multi-dimensional features of each of the multiple preset vehicles to obtain an identity recognition result of the vehicle to be recognized.

2. The vehicle identification method according to claim 1, characterized in that: The matching of the plurality of target features with the respective multi-dimensional features of the plurality of preset vehicles to obtain the identity recognition result of the vehicle to be recognized includes: Matching the plurality of target features with the multi-dimensional features of each of the plurality of preset vehicles, and determining a candidate vehicle including the plurality of target features from the plurality of preset vehicles; In the case where there are multiple candidate vehicles, based on the preemption coefficients of the multiple target features, a first target feature corresponding to the maximum preemption coefficient is determined from the multiple target features, and the preemption coefficient of the target feature is used to characterize the rarity of the target feature; Matching the first target feature with a combination of preset associated features corresponding to a plurality of candidate vehicles to determine a target associated feature of the first target feature; matching the plurality of target features and the target associated feature with the multi-dimensional features of each of the candidate vehicles respectively to obtain a plurality of correlation scores; The identity recognition result is determined according to the multiple correlation scores.

3. The vehicle identification method according to claim 2, characterized in that: Determining the identity recognition result according to the multiple correlation scores includes: In the case where there is a correlation score greater than a preset threshold value among the multiple correlation scores, determining the candidate vehicle corresponding to the correlation score greater than the preset score threshold value as the identity recognition result; When there are at least two correlation scores greater than the preset threshold among the multiple correlation scores, the target feature corresponding to the second largest preemption coefficient is determined as the new first target feature, and the multiple target features, the first target feature, the target association feature of the first target feature, the new first target feature and the new target association feature of the first target feature are matched with the preset association feature combination corresponding to multiple candidate vehicles until there is a correlation score greater than the preset score threshold among the multiple correlation scores obtained, and the candidate vehicle corresponding to the correlation score greater than the preset score threshold is determined as the identity recognition result.

4. The vehicle identification method according to claim 2, characterized in that: matching the first target feature with a combination of preset associated features corresponding to a plurality of candidate vehicles to determine a target associated feature of the first target feature; The multiple target features and the target-associated features are matched with the multi-dimensional features of each candidate vehicle to obtain multiple correlation scores, including: Matching the first target feature with the preset associated feature combination to determine at least one associated feature combination including the first target feature; determining the feature other than the first target feature in the associated feature combination corresponding to the maximum confidence as the target associated feature; Matching the multiple target features and the target-associated features with the multi-dimensional features of each of the candidate vehicles to obtain multiple correlation scores; When at least two correlation scores among the multiple correlation scores are greater than the preset score threshold, determining the feature other than the first target feature in the association feature combination corresponding to the second largest confidence as a new target association feature of the first target feature; The multiple target features, the target association features and the new target association features are matched with the multi-dimensional features of the multiple candidate vehicles respectively, until at least one association feature combination is traversed to obtain the multiple correlation scores.

5. The vehicle identification method according to any one of claims 2 to 4, characterized in that: The method further comprises: For each of the target features, determining, from the plurality of preset vehicles, a first number of preset vehicles whose corresponding multidimensional features include the target feature, and a second number of preset vehicles whose feature types include the feature type to which the target feature belongs, and determining an isolation coefficient corresponding to the target feature based on the first number and the second number; The preemption coefficient of the target feature is determined according to the first number, the second number, the confidence of the target feature, and the isolation coefficient corresponding to the target feature.

6. The vehicle identification method according to any one of claims 2 to 4, characterized in that: The step of matching the multiple target features and the target-associated features with the multi-dimensional features of each candidate vehicle to obtain multiple correlation scores includes: For each of the candidate vehicles, determine the weight corresponding to each second target feature in the multiple target features and the target association features, the number of times each second target feature appears in the multidimensional features of each of the multiple candidate vehicles, the number of times each second target feature appears in the multidimensional features of the candidate vehicles, and the isolation coefficient corresponding to each second target feature; The correlation score corresponding to the candidate vehicle is determined based on the weight corresponding to each second target feature, the number of times each second target feature appears in the multi-dimensional features of the multiple candidate vehicles, the number of times each second target feature appears in the multi-dimensional features of the candidate vehicle, and the isolation coefficient corresponding to each second target feature.

7. The vehicle identification method according to any one of claims 2 to 4, characterized in that: The method further comprises: For each of the candidate vehicles, based on the minimum support and minimum confidence criteria, feature association calculation is performed on the multi-dimensional features of the candidate vehicle to obtain a plurality of initial association feature combinations and initial confidences corresponding to each initial association feature combination; According to the initial confidences corresponding to the initial correlation feature combinations, determining a plurality of first correlation feature combinations having initial confidences greater than a preset threshold from the plurality of initial correlation feature combinations; For each of the first associated feature combinations, a confidence correction coefficient corresponding to the first associated feature combination is determined according to the confidences of the at least two features included in the first associated feature combination and the number of times each feature appears in the multi-dimensional features corresponding to the plurality of preset vehicles, and an initial confidence corresponding to the first associated feature combination is corrected according to the confidence correction coefficient to obtain a corrected confidence; Among the plurality of first associated feature combinations, a first associated feature combination whose corresponding corrected confidence level is greater than the preset threshold is determined as the preset associated feature combination corresponding to the candidate vehicle.

8. A vehicle identification device, characterized in that: include: A first acquisition unit, used to acquire a target image of a vehicle to be identified; A second acquisition unit, configured to acquire, based on the target image, multi-dimensional features of the vehicle to be identified and confidence levels corresponding to the features; A first determining unit, configured to determine, from the multi-dimensional features, a plurality of target features whose confidences are greater than a preset confidence threshold based on the confidences corresponding to the features; The first processing unit is used to match the multiple target features with the multi-dimensional features of each of the multiple preset vehicles to obtain an identity recognition result of the vehicle to be recognized.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the vehicle identification method according to any one of claims 1 to 7 is implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the vehicle identification method according to any one of claims 1 to 7 is implemented.

Images