Object information query method, device, computer equipment and storage medium
By combining image features and feature identification vectors, combining social relationships and location information, directly querying the information of the target object, the rapid accuracy of object information query in computer vision technology is solved, and query efficiency and accuracy are improved.
Patent Information
- Application Number
- CN202210517236.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-05-13
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2042-05-13
AI Technical Summary
In the application scenarios of computer vision technology, how to quickly and accurately query the object information of the target object has become an important problem, and the existing technology is difficult to effectively solve this problem.
By obtaining the image of the target object, extracting the combination of image features and preset feature identification vectors, determining the similarity between the combined features and the verification features, using the social relationship and position distance between the candidate object and the target object to meet the preset conditions, and directly query the target The object information of the object avoids back-checking errors and reduces the time-consuming process of obtaining verification features.
It realizes fast and accurate object information query, improves the accuracy and speed of query, avoids back-checking errors, reduces the need for interaction with the cloud, and improves query efficiency.
Smart Images

Figure CN117112917B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of computer technology, and in particular to an object information query method, device, computer equipment, storage medium and computer program product. Background Art
[0002] With the development of computer technology, computer vision technology has been widely used, such as face recognition, palm payment, fingerprint recognition, etc. In some application scenarios of computer vision technology, the image of the target object is collected, and the object information of the target object is determined based on the image. How to accurately and quickly query the object information of the target object has become an important issue. Summary of the invention
[0003] Based on this, it is necessary to provide an object information query method, device, computer equipment, computer readable storage medium and computer program product to address the above technical problems, which can quickly and accurately query object information.
[0004] In a first aspect, the present application provides an object information query method. The method comprises:
[0005] When product interaction is performed at a point of interest, an image of the target object is acquired;
[0006] Combining the image features extracted from the image with a preset feature identification vector to obtain a combined feature;
[0007] Determine the similarity between the combined feature and each verification feature; the verification feature is a feature of a combination of an object feature and an object feature identification vector of a candidate object, and is obtained when a distance between a location of the candidate object and the point of interest meets a preset condition and an object having a close social relationship with the candidate object interacts with a product at the point of interest;
[0008] When there is a similarity that satisfies the similarity condition, the object information of the target object is queried according to the object feature identification vector corresponding to the similarity that satisfies the similarity condition.
[0009] In a second aspect, the present application also provides an object information query device. The device comprises:
[0010] An acquisition module, used to acquire an image of a target object when product interaction is performed at a point of interest;
[0011] A combining module, used for combining the image features extracted from the image with a preset feature identification vector to obtain a combined feature;
[0012] A determination module, configured to determine the similarity between the combined feature and each verification feature; the verification feature is a feature of a combination of an object feature and an object feature identification vector of a candidate object, and is obtained when a distance between the location of the candidate object and the point of interest meets a preset condition and an object having a close social relationship with the candidate object interacts with a product at the point of interest;
[0013] The query module is used to query the object information of the target object according to the object feature identification vector corresponding to the similarity that satisfies the similarity condition when the similarity satisfies the similarity condition.
[0014] In one embodiment, the determining module is further used to:
[0015] Extracting the combined feature element from the combined feature, and extracting the verification feature element from the verification feature;
[0016] sequentially determining the product values between each of the combined characteristic elements and each of the verification characteristic elements;
[0017] The sum of the product values is used as the similarity between the combined feature and the verification feature.
[0018] In one embodiment, the combined characteristic element includes a first characteristic element and a first identification element; the verification characteristic element includes a second characteristic element and a second identification element; and the determination module is further configured to:
[0019] sequentially determining first product values between each of the first characteristic elements and each of the second characteristic elements;
[0020] sequentially determining second product values between each of the first identification elements and each of the second identification elements;
[0021] Based on the first product value and the second product value, a product value between each of the combined feature elements and each of the verification feature elements is determined.
[0022] In one embodiment, the query module is further used to:
[0023] Sorting the similarities in order of magnitude;
[0024] In the sorted similarities, determine the target verification feature corresponding to the similarity whose sorted ranking satisfies the similarity condition;
[0025] Based on the object feature identification vector in the target verification feature, object information of the target object in the image is determined.
[0026] In one embodiment, the combined module is further used for:
[0027] Determining a marker vector insertion position in the image features extracted from the image;
[0028] According to the insertion position of the identification vector, the preset feature identification vector is inserted into the image feature extracted from the image to obtain a combined feature.
[0029] In one embodiment, the verification feature belongs to a feature used for verification in a feature library; the device further comprises:
[0030] The determination module is further used to determine the number of biological objects corresponding to each verification feature in the feature library; and determine the length range of the identification vector according to the number of objects;
[0031] A selection module, used for selecting a target length value within the length range of the identification vector;
[0032] A generating module is used to generate an object feature identification vector of each biological object according to the target length value; wherein the vector length of the preset feature identification vector is the target length value.
[0033] In one embodiment, the apparatus further comprises:
[0034] The acquisition module is further used to acquire social objects that have a friend relationship chain with the candidate object;
[0035] The determination module is further configured to, when the candidate object and at least one social object having the friendship chain are in the same social group, determine at least one social object as an object having a close social relationship with the candidate object.
[0036] In one embodiment, the social group includes a close social group; and the determining module is further configured to:
[0037] Obtaining group member objects in the intimate social group;
[0038] If at least one of the group member objects belongs to the social object of the friend relationship chain, obtaining the social frequency between the candidate object and at least one of the group member objects;
[0039] If the social frequency reaches a preset frequency threshold, at least one of the group member objects is determined as an object with a close social relationship.
[0040] In one embodiment, the query module is further used to:
[0041] Acquire product interaction data corresponding to the target object according to the object information;
[0042] Verifying the product interaction data through a resource transfer interaction model to obtain a verification result;
[0043] If the verification result indicates that the product interaction data has passed the verification, resources are transferred to the target object according to the object information.
[0044] In one embodiment, the image includes at least two hand images of different image types; and the combining module is further configured to:
[0045] Correcting each of the hand images according to the hand image template to obtain a corrected hand image;
[0046] Extracting features from each of the corrected hand images to obtain palm print features;
[0047] Fusing the palmprint features to obtain a fused palmprint feature;
[0048] The fused palmprint feature is combined with a preset feature identification vector to obtain a combined feature.
[0049] In one embodiment, the image comprises a facial image; and the apparatus further comprises:
[0050] An extraction module, used for extracting image features from the facial image;
[0051] A convolution module, used for performing at least two convolution processes on the image features through an estimation network to obtain a predicted mean and a predicted variance of the image features;
[0052] The acquisition module is further used to acquire network parameters used for standardization processing in the estimation network;
[0053] A standardization module, used for standardizing the image features based on the prediction mean, the prediction variance and the network parameters to obtain standardized features;
[0054] The combination module is further configured to combine the image feature with a preset feature identification vector when it is determined that the target object is a living body according to the standardized feature.
[0055] In a third aspect, the present application further provides a computer device. The computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the following steps when executing the computer program:
[0056] When product interaction is performed at a point of interest, an image of the target object is acquired;
[0057] Combining the image features extracted from the image with a preset feature identification vector to obtain a combined feature;
[0058] Determine the similarity between the combined feature and each verification feature; the verification feature is a feature of a combination of an object feature and an object feature identification vector of a candidate object, and is obtained when a distance between a location of the candidate object and the point of interest meets a preset condition and an object having a close social relationship with the candidate object interacts with a product at the point of interest;
[0059] When there is a similarity that satisfies the similarity condition, the object information of the target object is queried according to the object feature identification vector corresponding to the similarity that satisfies the similarity condition.
[0060] In a fourth aspect, the present application further provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the following steps are implemented:
[0061] When product interaction is performed at a point of interest, an image of the target object is acquired;
[0062] Combining the image features extracted from the image with a preset feature identification vector to obtain a combined feature;
[0063] Determine the similarity between the combined feature and each verification feature; the verification feature is a feature of a combination of an object feature and an object feature identification vector of a candidate object, and is obtained when a distance between a location of the candidate object and the point of interest meets a preset condition and an object having a close social relationship with the candidate object interacts with a product at the point of interest;
[0064] When there is a similarity that satisfies the similarity condition, the object information of the target object is queried according to the object feature identification vector corresponding to the similarity that satisfies the similarity condition.
[0065] In a fifth aspect, the present application further provides a computer program product. The computer program product includes a computer program, and when the computer program is executed by a processor, the following steps are implemented:
[0066] When product interaction is performed at a point of interest, an image of the target object is acquired;
[0067] Combining the image features extracted from the image with a preset feature identification vector to obtain a combined feature;
[0068] Determine the similarity between the combined feature and each verification feature; the verification feature is a feature of a combination of an object feature and an object feature identification vector of a candidate object, and is obtained when a distance between a location of the candidate object and the point of interest meets a preset condition and an object having a close social relationship with the candidate object interacts with a product at the point of interest;
[0069] When there is a similarity that satisfies the similarity condition, the object information of the target object is queried according to the object feature identification vector corresponding to the similarity that satisfies the similarity condition.
[0070] The object information query method, device, computer equipment, storage medium and computer program product described above, when performing product interaction at a point of interest, obtains an image obtained by capturing the target object. The image features extracted from the image are combined with a preset feature identification vector to obtain a combined feature. The similarity between the combined feature and each verification feature is determined. When there is a similarity that satisfies a similarity condition, the candidate object that satisfies the similarity condition has a high similarity with the target object, and the object information of the candidate object is the object information of the target object. According to the object feature identification vector corresponding to the similarity that satisfies the similarity condition, the object information of the target object in the image is determined. Since the verification feature includes the object feature identification vector, when the similarity that satisfies the similarity condition is determined, the object information of the target object can be directly queried according to the object feature identification vector in the verification feature corresponding to the similarity that satisfies the similarity condition, without the need to reversely query the object feature identification vector according to the object feature, thereby avoiding the occurrence of search errors and improving the accuracy of the query. In addition, since the verification features are obtained when the distance between the location of the candidate object and the point of interest meets the preset conditions and the object with a close social relationship with the candidate object interacts with the product at the point of interest, that is, the verification features of the candidate object have been sent to the local in advance, there is no need to interact with the cloud, and the locally stored verification features can be directly used to identify the target object, which effectively reduces the time spent on obtaining the verification features and improves the speed of querying object information. BRIEF DESCRIPTION OF THE DRAWINGS
[0071] Figure 1 An application environment diagram of an object information query method in an embodiment;
[0072] Figure 2 is a flowchart of an object information query method in one embodiment;
[0073] Figure 3 A schematic diagram of a method for determining object feature identification according to similarity in one embodiment;
[0074] Figure 4 A schematic diagram of a flow chart of a method for determining similarity in one embodiment;
[0075] Figure 5 A schematic diagram of a flow chart of a method for determining an object having a close social relationship with a candidate object in one embodiment;
[0076] Figure 6 A schematic diagram of a resource transfer method in one embodiment;
[0077] Figure 71 is a schematic diagram of a flow chart of a liveness detection method in an embodiment;
[0078] Figure 8 is a schematic diagram of an estimation network in one embodiment;
[0079] Fig. 9 is a schematic diagram of the principle of liveness detection in one embodiment;
[0080] Fig.10 A schematic diagram of the principle of resource transfer in one embodiment;
[0081] Fig.11 is a flowchart of an object information query method in another embodiment;
[0082] Fig.12 is a structural block diagram of an object information query device in one embodiment;
[0083] Fig.13 is a structural block diagram of an object information query device in one embodiment;
[0084] Fig.14 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION
[0085] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.
[0086] The object information query method provided in the embodiment of the present application can be applied to Figure 1In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or it can be placed on the cloud or other servers. When the product interaction is performed at the point of interest, the server 104 obtains the image obtained by the terminal 102 collecting the target object, and combines the image features extracted from the image with the preset feature identification vector to obtain the combined feature. Then determine the similarity between the combined feature and the verification feature. The verification feature is a feature of the combination of the object feature and the object feature identification vector of the candidate object. It is obtained when the distance between the location of the candidate object and the point of interest meets the preset conditions and the object with a close social relationship with the candidate object interacts with the product at the point of interest. When there is a similarity that meets the similarity condition, the object information of the target object is queried according to the object feature identification vector corresponding to the similarity that meets the similarity condition. Among them, the terminal 102 can be, but is not limited to, various desktop computers, laptops, smart phones, tablet computers, Internet of Things devices and portable wearable devices, and the Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart car devices, etc. The portable wearable device may be a smart watch, a smart bracelet, a head-mounted device, etc. The server 104 may be implemented as an independent server or a server cluster consisting of multiple servers.
[0087] In one embodiment, Figure 2 As shown, an object information query method is provided, which is applied to Figure 1 The server in the example is used to illustrate the following steps:
[0088] S202, when product interaction is performed at a point of interest, an image of the target object is acquired.
[0089] Among them, product interaction is the interaction between the user and the terminal when the user conducts a product transaction. For example, when the user conducts a product transaction, the user can trigger the product interaction operation through the product interaction control in the terminal, so that the terminal performs resource transfer related to the product transaction according to the product interaction operation. Point of interest (POI) can be a location point of product interaction that a user (such as a target object, a candidate object, or a social object of a candidate object) is interested in. For example, in a geographic information system, it can be a shop, a supermarket, or an apartment. The point of interest in the embodiment of the present application can be understood as a specific geographical location point corresponding to the positioning system, with a certain range interval. For example, the point of interest is a geographical location point within a specific range interval centered on a geographical location point (longitude Lon, latitude Lat) in the positioning system. The target object is an object that conducts product interaction at the point of interest. The image is an image obtained by collecting the target object, which can be a face image, palm print image, fingerprint image, or iris image of the target object.
[0090] In one embodiment, the image may be a color image or an infrared image; S202 specifically includes: when product interaction is performed at a point of interest, the target object is captured by a color image sensor to obtain a color image of light imaging; or the target object is captured by an infrared sensor to obtain an infrared image of pan-infrared light imaging.
[0091] S204: Combining the image features extracted from the image with the preset feature identification vector to obtain a combined feature.
[0092] Among them, image features are data used to represent the characteristics or properties of an image that distinguish it from other images, including features with physical meaning such as facial edges, textures, histograms, or abstract features learned through deep learning algorithms, which can be represented by image feature vectors. Image feature vectors can be multidimensional vectors, and each element in the vector can be floating-point data. The preset feature identification vector is a vector used to represent the feature identification corresponding to the image feature, which can be a zero vector. For example, the preset feature identification vector is ID = [i1 i2…i n ](n≥1), or the preset feature identification vector is [01 02…0 32 The combined feature is a feature obtained by combining the image feature and the preset feature identification vector. For example, the combined feature is A′=[v1v2…v m 01 02…0 n ](m≥1,n≥1), where [v1 v2…v m ] is the m-dimensional image feature, [01 02…0 n ] is an n-dimensional preset feature identification vector.
[0093] In one embodiment, S204 specifically includes: extracting features from the image using a feature extraction algorithm to obtain image features; and combining the image features with a preset feature identification vector to obtain a combined feature. The feature extraction algorithm includes a Sobel algorithm, a grayscale histogram algorithm, a texture extraction algorithm, and the like.
[0094] In one embodiment, S204 specifically includes: extracting features from the image through a deep learning network to obtain image features; and combining the image features with a preset feature identification vector to obtain a combined feature. The deep learning network is a neural network model obtained by training a deep learning model.
[0095] S206, determining the similarity between the combined feature and each verification feature; the verification feature is a feature of the combination of the object feature and the object feature identification vector of the candidate object, which is obtained when the distance between the location of the candidate object and the point of interest meets the preset conditions and an object with a close social relationship with the candidate object interacts with the product at the point of interest.
[0096] Among them, object features are features of the image obtained by collecting candidate objects, including features with physical meanings such as facial edges, textures, histograms, or abstract features learned through deep learning algorithms, which can be represented by object feature vectors. The object feature vector can be a multidimensional vector with the same dimension as the image feature vector, and each element in the vector can be floating-point data. The object feature identification vector is a vector used to represent the feature identification corresponding to the object feature, and has the same dimension as the preset feature identification vector. The elements in the object feature identification vector can be characters, numbers, special symbols or letters, etc. For example, the object feature identification vector is [1 10 1 0], and for another example, the object feature identification vector is [i1 i2…i 32 The verification feature is a feature used to verify the identity of the target object. It is obtained by combining the object feature of the candidate object and the object feature identification vector and can be stored in the feature library. For example, the verification feature is A′=[a1 a2…a 512 i1 i2…i 32 ], where [a1a2…a 512 ] is the object feature, [i1 i2…i 32 ] is the object feature identification vector.
[0097] Among them, the preset condition is a condition for judging the distance between the location of the candidate object and the point of interest. For example, the preset condition is that the distance between the location of the candidate object and the point of interest is less than the preset distance. For another example, the preset condition is that the ratio of the distance between the location of the candidate object and the point of interest to the preset distance is within a preset numerical range. A close social relationship refers to a relationship between users that meets the preset close social relationship conditions. For example, the preset close social relationship condition is that the relationship between users belongs to a friend relationship in a social application, and the social frequency between users reaches a preset frequency threshold. For another example, the preset close social relationship condition is that the relationship between users belongs to the same group in a social application.
[0098] Specifically, when the distance between the location of the candidate object and the point of interest meets the preset conditions, and an object with a close social relationship with the candidate object interacts with the product at the point of interest, the verification feature corresponding to the candidate object is sent to the local feature library. When the target object interacts with the product at the point of interest, the verification feature can be directly extracted from the feature library without interacting with the cloud, and the similarity between the combined feature and the verification feature can be determined.
[0099] S208, when there is a similarity that satisfies the similarity condition, query the object information of the target object according to the object feature identification vector corresponding to the similarity that satisfies the similarity condition.
[0100] Among them, the similarity condition is a condition for judging the similarity between the combined feature and the verification feature. For example, the similarity condition is that the similarity is greater than the similarity threshold; for another example, the similarity condition is that the ranking of the similarity ranking reaches a preset ranking. The object information is the information of the target object, which can be the identity information or behavior information of the target object. For example, the object information is the user ID or age, gender, location information, product interaction record, etc. of the target object in the social application.
[0101] When there is a similarity that satisfies the similarity condition, the object information of the target object is queried based on the object feature identification vector corresponding to the similarity that satisfies the similarity condition. Figure 3 As shown, S10 is the similarity that satisfies the similarity condition, the verification feature corresponding to S10 is the feature of the combination of the object feature 10 and the object feature identification vector 10, the object feature identification vector corresponding to the similarity that satisfies the similarity condition is the object feature identification vector 10, and the object information of the target object is queried based on the object feature identification vector 10.
[0102] Since the verification feature is a feature composed of the object feature and the object feature identification vector of the candidate object, the object feature and the object feature identification vector are combined and stored in the feature library. When the verification feature in the feature library is managed, the entire verification feature is inserted or deleted, and management anomalies caused by separate storage and management of the object feature and the object feature identification vector will not occur, which will cause confusion in the mapping relationship between the object feature and the object feature identification vector, and thus the object feature identification vector cannot be reversely checked based on the object feature.
[0103] In the above embodiment, when product interaction is performed at a point of interest, an image obtained by capturing the target object is obtained. The image features extracted from the image are combined with a preset feature identification vector to obtain a combined feature. The similarity between the combined feature and each verification feature is determined. When there is a similarity that satisfies a similarity condition, the candidate object that satisfies the similarity condition has a high similarity with the target object, and the object information of the candidate object is the object information of the target object. According to the object feature identification vector corresponding to the similarity that satisfies the similarity condition, the object information of the target object in the image is determined. Since the verification feature includes an object feature identification vector, when the similarity that satisfies the similarity condition is determined, the object information of the target object can be directly queried according to the object feature identification vector in the verification feature corresponding to the similarity that satisfies the similarity condition, without the need to reversely query the object feature identification vector according to the object feature, thereby avoiding search errors and improving the accuracy of the query. In addition, since the verification features are obtained when the distance between the location of the candidate object and the point of interest meets the preset conditions and the object with a close social relationship with the candidate object interacts with the product at the point of interest, that is, the verification features of the candidate object have been sent to the local in advance, there is no need to interact with the cloud, and the locally stored verification features can be directly used to identify the target object, which effectively reduces the time spent on obtaining the verification features and improves the speed of querying object information.
[0104] In one embodiment, Figure 4 As shown, S206 specifically includes the following steps:
[0105] S402, extracting the combined feature elements in the combined feature, and extracting the verification feature elements in the verification feature.
[0106] The combined feature element is the element in the combined feature vector corresponding to the combined feature. For example, the combined feature is V = [v1 v2…v 512 x1 x2…x 32 ], v1, v2, …v 512 , x1, x2, …, x 32 is a combined feature element. The check feature element is an element in the check feature vector corresponding to the check feature. For example, the check feature is A′=[a1 a2…a 512 i1 i2…i 32 ], a1, a2, …, a 512 ,i1,i2,…,i 32 To verify the characteristic elements.
[0107] S404, determining the product value between each combined characteristic element and each verification characteristic element in turn.
[0108] The server determines the product value between each combined feature element and each verification feature element in turn. Specifically, if the combined feature is V = [v1 v2…v512 x1 x2…x 32 ], the verification feature is A′=[a1 a2…a 512 i1 i2…i 32 ], the product values determined by the server are v1×a1, v2×a2…x1×i1, x2×i2…x 32 ×i 32 .
[0109] In one embodiment, the combined characteristic element includes a first characteristic element and a first identification element; the verification characteristic element includes a second characteristic element and a second identification element; S404 specifically includes: determining the first product value between each first characteristic element and each second characteristic element in turn; determining the second product value between each first identification element and each second identification element in turn; based on the first product value and the second product value, determining the product value between each combined characteristic element and each verification characteristic element.
[0110] The first feature element is the element in the image feature vector corresponding to the image feature. For example, the image feature is [v1 v2…v 512 ], the first characteristic element is v1, v2, ..., v 512 The first identification element is an element in a preset feature identification vector. For example, the preset feature identification vector is [x1 x2…x 32 ], the first identification element is x1, x2, ..., x 32 The second feature element is the element in the object feature vector corresponding to the object feature. For example, the object feature is [a1 a2…a 512 ], the second characteristic element is a1, a2, ..., a 512 The second identification element is an element in the object feature identification vector. For example, the object feature identification vector is [i1 i2…i 32 ], the second identification element is i1, i2, ..., i 32 The server first determines the first product values between each first characteristic element and each second characteristic element in sequence. Specifically, the first product values determined by the server are v1×a1, v2×a2…v 512 ×a 512 Then, the server determines the second product values between each first identification element and each second identification element in turn. Specifically, the second product values determined by the server are s1×i1, s2×i2…s 32 ×i 32 .
[0111] S406: Taking the sum of the product values as the similarity between the combined feature and the verification feature.
[0112] The server uses the sum of the product values between each combined feature element and each verification feature element as the similarity between the combined feature and the verification feature.
[0113] In one embodiment, the combined characteristic element includes a first characteristic element and a first identification element; the verification characteristic element includes a second characteristic element and a second identification element; the first product value between each first characteristic element and each second characteristic element is determined in sequence; the second product value between each first identification element and each second identification element is determined in sequence; the first product value and the second product value are summed to obtain the product value between each combined characteristic element and each verification characteristic element.
[0114] In one embodiment, the preset feature identification vector is a zero vector, and the first identification elements in the preset feature identification vector are all 0. Since the product of 0 and any numerical value is 0, the second product values between each first identification element and each second identification element are all 0. The sum of the first product value and the second product value obtained by the server is equal to the first product value, so the second product value does not contribute to the calculation of the similarity between the combined feature and the verification feature. The similarity between the combined feature and each verification feature is determined by the first product value between the first feature element and the second feature element. Therefore, although the combined feature is a combination of the image feature and the preset feature identification vector, and the verification feature is a combination of the object feature and the object feature identification vector, the similarity calculated between the combined feature and the verification feature is equal to the similarity between the image feature and the object feature, thereby ensuring the accuracy of the similarity.
[0115] In the above embodiment, the combined feature elements in the combined feature are extracted, and the verification feature elements in the verification feature are extracted, and the product values between each combined feature element and each verification feature element are determined in sequence, and the sum of the product values is used as the similarity between the combined feature and the verification feature. Therefore, when the similarity that meets the similarity condition is determined, the object feature identification vector can be quickly extracted from the verification feature corresponding to the similarity, and the object information of the target object can be queried based on the object feature identification vector, thereby improving the accuracy of the query.
[0116] In one embodiment, S206 specifically includes: forming a check feature matrix from the check features; determining a product value between the combined features and the check feature matrix; and determining a similarity between the combined features and the check features according to the product value.
[0117] Among them, the check feature matrix is a matrix composed of check features. For example, the check feature matrix is [a m,1 a m,2 …a m,512 i m,1 i m,2 …i m,32] is the check feature vector in the check feature matrix, and m is a positive integer greater than or equal to 1. The server determines the product value between the combined feature and the check feature matrix. Specifically, the server can determine the product value between the combined feature and the check feature according to formula (1). Where S is the product value between the combined feature and the check feature, that is, the similarity vector, A is the check feature matrix, V = [v1 v2…v 512 x1x2…x 32 ] is a combined feature.
[0118]
[0119] In the above embodiment, the check feature matrix is composed of the check features, the product value between the combined feature and the check feature matrix is determined, and the similarity between the combined feature and the check feature is determined based on the product value. Therefore, when the similarity that satisfies the similarity condition is determined, the object feature identification vector can be quickly extracted from the check feature corresponding to the similarity, and the object information of the target object can be queried based on the object feature identification vector, thereby improving the accuracy of the query.
[0120] In one embodiment, S208 specifically includes: sorting each similarity in order of size; determining, among the sorted similarities, the target verification feature corresponding to the similarity whose sorted rank meets the similarity condition; and determining the object information of the target object in the image based on the object feature identification vector in the target verification feature.
[0121] The target verification feature is the verification feature corresponding to the similarity of the sorted ranking that meets the similarity condition. The sorted ranking is the order in which the similarities are arranged according to size. The similarity condition can be that the sorted ranking is the highest ranking, or that the sorted ranking is within a preset range. The server sorts the similarities in order of size, and determines the target verification feature corresponding to the similarity of the sorted ranking that meets the similarity condition among the sorted similarities. For example, Figure 3 As shown, the similarities after sorting in order of size are S10, S1, S5, S4, S12, S15, S16, S9, S8, S2, S6, S13, S7, S3, S14, and S11. If the similarity condition is that the sorted rank is the highest rank, the target verification feature that meets the similarity condition is the verification feature 10, and the object information of the target object in the image is determined based on the object feature identification vector 10 in the verification feature 10.
[0122] In the above embodiment, the target verification feature corresponding to the similarity that satisfies the similarity condition is quickly determined by sorting the similarities according to the magnitude of the similarity. Since the target verification feature is a feature composed of the object feature and the object feature identification vector of the candidate object, after determining the target verification feature, the object information of the target object in the image can be quickly determined according to the object feature identification vector in the target verification feature. This can avoid the reverse query error caused by inconsistent storage order when the object feature identification vector and the object feature are stored separately, thereby ensuring the accuracy of the query.
[0123] In one embodiment, S204 specifically includes: determining an identification vector insertion position in the image features extracted from the image; and inserting a preset feature identification vector into the image features extracted from the image according to the identification vector insertion position to obtain a combined feature.
[0124] The identification vector insertion position is the insertion position of the preset feature identification vector in the image feature. For example, if the image feature is an n-dimensional vector, the identification vector insertion position can be before the first vector element of the image feature, or after the last vector element, or between any two adjacent vector elements. For example, if the image feature is an m×n-dimensional matrix, the identification vector insertion position can be element a ij (1≤i≤m,1≤j≤m). The server inserts the preset feature identification vector into the image feature extracted from the image according to the insertion position of the identification vector to obtain the combined feature.
[0125] In one embodiment, the object identification vector insertion position is determined in the object feature; and according to the object identification vector insertion position, the object feature identification vector is inserted into the object feature to obtain the verification feature.
[0126] The object identification vector insertion position is the insertion position of the object identification vector in the object feature, which is the same as the identification vector insertion position. That is, if the preset feature identification vector is inserted after the last element of the image feature, the object identification vector is inserted after the last element of the object feature; if the preset feature identification vector is inserted before the first element of the image feature, the object identification vector is also inserted before the first element of the object feature.
[0127] If the insertion position of the object identification vector in the object feature is the same as the insertion position of the preset feature identification vector in the image feature, then when determining the similarity between the combined feature and each verification feature, the object identification vector can be aligned with the preset feature identification vector, and each vector element in the object identification vector can be multiplied with each vector element in the preset feature identification vector. Therefore, when the preset feature identification vector is a zero vector, the product of the object identification vector and the preset feature identification vector can be made zero, and the determined similarity is equal to the similarity between the image feature and the object feature, thereby avoiding the determined similarity being affected by the product between the object identification vector and the preset feature identification vector.
[0128] In the above embodiment, the identification vector insertion position is determined in the image feature extracted from the image, and according to the identification vector insertion position, the preset feature identification vector is inserted into the image feature extracted from the image to obtain the combined feature. Thus, the object feature identification vector can be determined according to the similarity between the combined feature and the verification feature, and the object information can be queried according to the object feature identification vector, thereby improving the accuracy of the query.
[0129] In one embodiment, the verification feature belongs to a feature used for verification in a feature library; before S206, it also includes: determining the number of biological objects corresponding to each verification feature in the feature library; determining the length range of the identification vector based on the number of objects; selecting a target length value within the identification vector length range; generating an object feature identification vector for each biological object based on the target length value; wherein the vector length of the preset feature identification vector is the target length value.
[0130] Among them, the number of objects is the total number of biological objects corresponding to the verification features in the feature library. For example, 1000 verification features are stored in the feature library, each verification feature corresponds to a biological object, and the number of objects is 1000. The length range of the identification vector is the length range of the object feature identification vector, and any target length value selected within the length range of the identification vector can meet the identification capability of the biological object. Within the length range of the identification vector, any target length value can be selected, and the numerical value of the selected target length value is proportional to the number of biological objects that can be identified by the target length value. For example, the length range of the identification vector is [32bit, +∞], and the selected target length value can be 32bit, 64bit, etc. The server generates an object feature identification vector for each biological object based on the target length value, and the vector length of the generated object feature identification vector is the target length value. For example, the target length value is 32bit, and the generated object feature identification vector is [i1 i2…i 32 ]. The vector length of the preset feature identification vector is the target length value, which is equal to the vector length of the object feature identification vector.
[0131] Since the server can determine the length range of the identification vector based on the number of objects and select any target length value within the length range of the identification vector as the length value of the object feature identification vector, the length of the object feature identification vector is variable. When the number of biological objects corresponding to the verification features in the feature library is small, the length of the object feature identification vector can be reduced to save storage space. When the number of biological objects corresponding to the verification features in the feature library is large, the length of the object feature identification vector can be increased to meet the identification capability of biological objects.
[0132] In the above embodiment, the number of biological objects corresponding to each verification feature in the feature library is determined, the length range of the identification vector is determined according to the number of objects, and a target length value is selected within the length range of the identification vector; and the object feature identification vector of each biological object is generated according to the target length value. This can save storage space and meet the identification capability of biological objects.
[0133] In one embodiment, Figure 5 As shown, before S204, the following steps are also included:
[0134] S502: Acquire social objects that have a friend relationship chain with the candidate object.
[0135] The friend relationship chain is a relationship chain established based on the friend relationship between social users. For example, if candidate object A adds user B as a friend, user B is a social object that has a friend relationship chain with candidate object A. In one embodiment, the user identifier of the social object that has a friend relationship chain with the candidate object is stored in the friend relationship chain list of the candidate object, and the server obtains the social object that has a friend relationship chain with the candidate object from the friend relationship chain list of the candidate object.
[0136] S504: When the candidate object and at least one social object having a friend relationship chain are in the same social group, determine the at least one social object as an object having a close social relationship with the candidate object.
[0137] The social group is a group composed of social users, and the social users in the social group can be friends or non-friends. The server obtains multiple social objects that have a friend relationship chain with the candidate object. When the candidate object and some of the multiple social objects are in the same social group, at least one social object in the same social group as the candidate object is determined as an object that has a close social relationship with the candidate object. For example, the social objects that have a friend relationship chain with the candidate object A include social objects B, C, and D, where social objects B and C are in the same social group as the candidate object A, and social objects B and C are objects that have a close social relationship with the candidate object.
[0138] Since the object with a close social relationship with the candidate object has a close relationship with the candidate object and may reach the same point of interest, when the object with a close social relationship with the candidate object interacts with the product at the point of interest, the candidate object may also reach the point of interest and interact with the product at the point of interest. The server regards the social objects that are in the same social group as the candidate object and have a friend relationship chain with the candidate object as the objects with a close social relationship with the candidate object, and can predict the points of interest for the candidate object to interact with the product based on the points of interest where the object with a close social relationship with the candidate object interacts with the product, and when the distance between the location of the candidate object and the point of interest meets the preset conditions, the verification features of the candidate object are sent to the local server, so that when the target object interacts with the product at the point of interest, the locally stored verification features can be directly used to identify the target object, effectively reducing the time consumption of obtaining the verification features and improving the speed of querying object information.
[0139] In the above embodiment, social objects that have a friend relationship chain with the candidate object are obtained, and when the candidate object and at least one social object that has a friend relationship chain are in the same social group, the at least one social object is determined as an object that has a close social relationship with the candidate object. Therefore, when the target object interacts with the product at the point of interest, the locally stored verification feature can be directly used to identify the target object, which effectively reduces the time consumption of obtaining the verification feature and improves the speed of querying object information.
[0140] In one embodiment, the social group includes a close social group; S504 specifically includes: obtaining group member objects in the close social group; if at least one group member object belongs to a social object in a friend relationship chain, obtaining the social frequency between the candidate object and the at least one group member object; if the social frequency reaches a preset frequency threshold, determining at least one group member object as an object with a close social relationship.
[0141] Among them, the intimate social group is a group with an intimate relationship with the target object, and the intimate relationship can be a relative relationship, a classmate relationship, or a colleague relationship, etc. For example, the candidate object joins a social group labeled "family group", which is an intimate social group. The social frequency is the frequency of interaction through social applications. For example, the social frequency can be 10 times a day, or 20 times a month, etc. If the candidate object has a friend relationship with a group member object in the intimate social group, the social frequency between the candidate object and the group member object is obtained. If the social frequency reaches the preset frequency threshold, the group member object is determined as an object with an intimate social relationship with the candidate object. For example, the intimate social group labeled "family group" includes candidate object A, user B who has a friend relationship with candidate object A, and user C. The social frequency between user B and the candidate object is 40 times per month, and the social frequency between user C and the candidate object is 20 times per month. Assuming that the preset frequency threshold is 30 times per month, user B is an object with an intimate social relationship with the candidate object.
[0142] In the above embodiment, group member objects in the close social group are obtained; if at least one group member object belongs to a social object in the friend relationship chain, the social frequency between the candidate object and at least one group member object is obtained; if the social frequency reaches a preset frequency threshold, at least one group member object is determined as an object with a close social relationship. Therefore, the object with a close social relationship with the candidate object can be accurately determined based on the social frequency between the candidate object and each group member object in the close social group, so that when the distance between the location of the candidate object and the point of interest meets the preset conditions and the object with a close social relationship with the candidate object interacts with the product at the point of interest, the verification feature can be sent to the local server. When the target object interacts with the product at the point of interest, the verification feature can be directly obtained from the local server, which reduces the time consumed by interacting with the cloud server and improves the speed of querying object information.
[0143] In one embodiment, Figure 6 As shown, after S208, the following steps are also included:
[0144] S602: Acquire product interaction data corresponding to the target object according to the object information.
[0145] The product interaction data refers to data related to product interaction, including product interaction time, product type, number of interactive products, payment method, payment currency, etc. When the target object interacts with the product each time, the server stores the target object's object information and the generated product interaction data.
[0146] S604, verifying the product interaction data through the resource transfer interaction model to obtain a verification result.
[0147] The resource transfer interaction model is a machine learning model that verifies the legitimacy of product interaction behavior, and can be a model trained by product interaction data samples corresponding to each user of the product interaction application. In one embodiment, the training samples can include the user's payment amount, commonly used payment merchants and product types, active payment time period, geographic location information, network type, etc.
[0148] S606: If the verification result indicates that the product interaction data has passed the verification, resources are transferred to the target object according to the object information.
[0149] If the verification result indicates that the product interaction data has passed the verification, it means that the target object has a high credit rating, and the target object is transferred resources according to the object information. Or if the verification result indicates that the product interaction data has passed the verification, the password verification can be performed again, and the verification performed by the resource transfer interaction model is used as an auxiliary verification method for the password verification. Only when all verifications are passed can the resource transfer be performed, which ensures the security of the resource transfer.
[0150] In the above embodiment, the product interaction data corresponding to the target object is obtained according to the object information. The product interaction data is verified through the resource transfer interaction model. If the obtained verification result indicates that the product interaction data has passed the verification, the resource transfer is performed on the target object according to the object information. Therefore, the legitimacy of the product interaction behavior of the target object can be verified through the resource transfer interaction model, reducing the reliance on password verification, thereby avoiding the loss caused by the leakage of the password and improving the security of resource transfer.
[0151] In one embodiment, the image includes at least two hand images of different image types; S204 specifically includes: correcting each hand image according to the hand image template to obtain a corrected hand image; extracting features from each corrected hand image to obtain a palmprint feature; fusing each palmprint feature to obtain a fused palmprint feature; combining the fused palmprint feature with a preset feature identification vector to obtain a combined feature.
[0152] Among them, the image type is the type of image obtained by collecting the target object, which can be a color image or an infrared image. The color image is an image obtained by performing visible light imaging on the target object through a color image acquisition device. The infrared image is an image obtained by performing infrared imaging on the target object through an infrared image acquisition device. The hand image template is an image template used for correcting the hand image, including an infrared hand image template and a color hand image template. The hand image template includes a standard hand image that meets the palm recognition requirements. For example, the display position of the standard hand image is at the center of the hand image template, and there is no obstruction, and the tilt angle is zero. Since the display mode of the palm in the hand image obtained by collecting the target object is various, different display modes may affect the accuracy of feature extraction of the hand image. The display mode includes display position and display angle, etc. The display position is the position of the palm in the hand image, and the display angle is the angle between the palm and the plane where the hand image is located. Before extracting the features of the hand image, the display mode of the palm in each hand image is corrected by the hand image template, thereby improving the accuracy of feature extraction.
[0153] The palm print feature is a feature extracted from the corrected hand image. Since the corrected hand image includes at least two hand images of different image types, the fusion of the palm print features can make the obtained fused palm print feature contain the palm print features of the palms in the hand images of different types, further improving the accuracy of the extracted palm print features.
[0154] In the above embodiment, each hand image is corrected according to the hand image template, and features are extracted from each corrected hand image to obtain palm print features, and then each palm print feature is fused, and the obtained fused palm print features are combined with a preset feature identification vector to obtain a combined feature. Since the obtained fused palm print features have high accuracy, the accuracy of the combined features is improved.
[0155] In one embodiment, the image includes hand images of at least two different image types; S204 specifically includes: respectively combining the image features extracted from each hand image with a preset feature identification vector to obtain a combined feature corresponding to each hand image; S206 specifically includes: respectively determining the similarity between each combined feature and the verification feature; performing weighted summation on the similarity to obtain a weighted similarity; S208 specifically includes: when there is a weighted similarity that satisfies the similarity condition, querying the object information of the target object according to the object feature identification vector corresponding to the weighted similarity that satisfies the similarity condition.
[0156] Although different types of hand images are images obtained by capturing the same target object, the similarities between the combined features obtained by combining the image features extracted from different types of hand images with the preset feature identification vectors and the verification features are different. For example, the similarity between the combined features corresponding to the infrared hand image and the verification features may be higher than the similarity between the combined features corresponding to the color light hand image and the verification features. The similarities between each combined feature and the verification feature are weighted and summed, which integrates the similarities between the combined features and the verification features corresponding to different types of hand images, and improves the accuracy of the weighted similarity.
[0157] In the above embodiment, the image features extracted from each hand image are combined with the preset feature identification vector to obtain the combined features corresponding to each hand image, and then the similarity between each combined feature and the verification feature is determined, and the similarity is weighted and summed to obtain the weighted similarity. When there is a weighted similarity that satisfies the similarity condition, the object information of the target object is queried based on the object feature identification vector corresponding to the weighted similarity that satisfies the similarity condition. Since the accuracy of the weighted similarity is high, the accuracy of the object information of the queried target object is improved.
[0158] In one embodiment, the image includes a facial image; Figure 7 As shown, before S204, S702-S708 are also included, and S204 specifically includes S710:
[0159] S702: extracting image features from the facial image.
[0160] Among them, the facial image is an image obtained by capturing the face of the target object, which may be a color facial image or an infrared facial image, etc. It may be an image captured under various lighting conditions and backgrounds, and may have various resolutions, sizes, etc. This application does not limit this.
[0161] S704, performing at least two convolution processes on the image features through the estimation network to obtain a predicted mean and a predicted variance of the image features.
[0162] The estimation network is a neural network for predicting the mean and variance of image features, and may include a global pooling layer and a convolution layer. Convolution processing is based on convolution calculations performed by the convolution layer. It should be noted that the number of features of at least two convolution layers for convolution processing performed by the service device may be different. The predicted mean is the mean of the image features predicted by the estimation network. The predicted variance is the variance of the image features predicted by the estimation network.
[0163] In one embodiment, S704 specifically includes: the server performs convolution processing on the image feature through the first convolution layer in the estimation network to obtain the first convolution feature; then, the first convolution feature is input to the second convolution layer in the estimation network, and the first convolution feature is convolution processed by the second convolution layer to obtain the second convolution feature, and so on, the input convolution features are convolution processed by other convolution layers in the estimation network to obtain the predicted mean and the predicted variance. For example, Figure 8 As shown, the estimation network includes four convolutional layers. By performing convolution calculations on image features layer by layer through convolutional layers 1 to 4, the predicted mean and predicted variance can be obtained.
[0164] S706, obtaining network parameters used for standardization processing in the estimation network.
[0165] Normalization is a process of converting image features into numerical values within a specific range, so that image features of different units or orders of magnitude can be uniformly processed. Network parameters are parameters used in the estimation network to normalize image features, which can be learned by training the estimation network based on sample images.
[0166] S708, normalizing the image features based on the predicted mean, predicted variance and network parameters to obtain standardized features.
[0167] Among them, the standardized features are features obtained by standardizing the image features. In one embodiment, S708 specifically includes: the server uses the predicted mean, predicted variance and network parameters as parameters in the standardized processing algorithm, and uses the image features as independent variables of the standardized processing algorithm, thereby calculating the standardized features. Among them, the standardized processing algorithm includes but is not limited to a linear method, a broken line method, a curve method, etc. The linear method can be, for example, an extreme value method, a standard deviation method, etc.; the broken line method can be, for example, a three-broken line method, etc.; the curve method can be, for example, a semi-normal distribution method, etc.
[0168] S710: When it is determined that the target object is a living body according to the standardized features, the image features are combined with a preset feature identification vector.
[0169] In one embodiment, S710 may include: the server calculates the probability that the facial image is a living body image based on the standardized features, determines that the target object is a living body when the probability that the facial image is a living body image is greater than a probability threshold, and combines the image features with a preset feature identification vector.
[0170] Next, combine Fig. 9 The above embodiments are described in detail as follows:
[0171] The server inputs the acquired facial image into a feature extraction network, which includes multiple cascaded convolutional layers and data normalization layers. The data normalization layer includes an estimation network, which is used to determine the predicted mean and predicted variance of the image features. The server first inputs the facial image into the convolutional layer of the feature extraction network, performs convolution calculation on the facial image through the convolutional layer, and extracts the image features of the facial image. Then, the server inputs the image features into the estimation network in the data normalization layer to obtain the predicted mean and predicted variance of the image features. The data normalization layer performs normalization processing on the image features based on the predicted mean and predicted variance to obtain standardized features. The server inputs the standardized features into the next convolutional layer for feature extraction again, and then performs normalization processing through the data normalization layer. The above processing is performed multiple times to obtain the final standardized features, and the final standardized features are input into the classifier, and the classifier is used to determine whether the target object is alive.
[0172] In the above embodiment, the server acquires the facial image obtained by capturing the target object and extracts image features from the facial image. Then, the image features are standardized according to the predicted mean and predicted variance predicted by the estimation network, so that the obtained standardized features are more accurate. Therefore, according to the standardized features, it is possible to accurately determine whether the target object is a living body, so as to determine whether the target object is a real living user, and to prevent illegal users from causing losses to the target object by stealing the target object's account.
[0173] In one embodiment, Fig.10 As shown, when the target object interacts with the product at the point of interest, a resource transfer request is triggered by the payment client 100, and the server 102 responds to the resource transfer request and performs a resource transfer legitimacy verification. After the verification is passed, the resource transfer is performed according to the object information of the target object, which specifically includes the following steps:
[0174] S1002 , the payment client 100 sends a resource transfer request to the server 102 .
[0175] S1004 , in response to the resource transfer request, the server 102 sends a resource transfer legitimacy verification instruction to the payment client 100 to instruct the payment client 100 to collect a facial image of the target object.
[0176] In response to the resource transfer legitimacy verification instruction, the payment client 100 displays an image acquisition frame 1002 to acquire a face image of the target object in the image acquisition frame 1002. After the payment client 100 acquires the face image, S1006 is executed.
[0177] S1006 , the payment client 100 sends the facial image to the server 102 .
[0178] The server 102 performs a resource transfer legitimacy verification based on the face image, including liveness detection and interactive data verification. The server 102 first performs liveness detection on the target object based on the face image. The liveness detection process includes: extracting image features from the face image, and standardizing the image features to obtain standardized features. When the target object is determined to be a live object based on the standardized features, it is determined that the face image passes the liveness detection and the target object is a live object.
[0179] When the target object passes the liveness detection, the server 102 performs interactive data verification according to the object information of the target object, including: the server 102 combines the image features extracted from the facial image with the preset feature identification vector to obtain the combined feature; determine the similarity between the combined feature and each verification feature; the verification feature is the feature of the combination of the object feature and the object feature identification vector of the candidate object, which is obtained when the distance between the position of the candidate object and the point of interest meets the preset conditions and the object with a close social relationship with the candidate object interacts with the product at the point of interest; when there is a similarity that meets the similarity condition, query the object information of the target object according to the object feature identification vector corresponding to the similarity that meets the similarity condition. Obtain the product interaction data corresponding to the target object according to the object information; verify the product interaction data through the resource transfer interaction model to obtain the verification result.
[0180] S1008 , if the verification result indicates that the product interaction data has passed the verification, the server 102 transfers resources to the target object according to the object information, and sends resource transfer success information to the payment client 100 .
[0181] When the payment client 100 receives the resource transfer success information, it displays "Payment Successful", such as Fig.10 As shown in the reference numeral 1004.
[0182] In the above embodiment, the server performs liveness detection on the target object requesting resource transfer and performs data verification on the interaction data, and performs resource transfer after the verification is passed, thereby ensuring the security of resource transfer.
[0183] In one embodiment, Fig.11 As shown, the object information query method includes the following steps:
[0184] S1102, when product interaction is performed at a point of interest, an image of the target object is acquired.
[0185] S1104, determining an identification vector insertion position in the image features extracted from the image, and inserting a preset feature identification vector into the image features extracted from the image according to the identification vector insertion position to obtain a combined feature.
[0186] S1106, determining the number of biological objects corresponding to each verification feature in the feature library, and determining the length range of the identification vector according to the number of objects.
[0187] S1108, selecting a target length value within the identification vector length range, and generating an object feature identification vector of each biological object according to the target length value, wherein the vector length of the preset feature identification vector is the target length value.
[0188] S1110, extracting the combined feature elements in the combined feature, and extracting the verification feature elements in the verification feature.
[0189] S1112, determine the product values between each combined feature element and each verification feature element in turn, and use the sum of the product values as the similarity between the combined feature and the verification feature.
[0190] S1114, sorting the similarities in order of magnitude. Among the sorted similarities, determining the target verification features corresponding to the similarities whose sorted rankings satisfy the similarity condition.
[0191] S1116, determining object information of the target object in the image based on the object feature identification vector in the target verification feature.
[0192] S1118, obtaining product interaction data corresponding to the target object according to the object information, and verifying the product interaction data through the resource transfer interaction model to obtain a verification result.
[0193] S1120: If the verification result indicates that the product interaction data has passed the verification, resources are transferred to the target object according to the object information.
[0194] For the specific contents of S1102 to S1120 above, reference may be made to the specific implementation process described above.
[0195] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.
[0196] Based on the same inventive concept, the embodiment of the present application also provides an object information query device for implementing the object information query method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more object information query device embodiments provided below can refer to the limitations on the object information query method above, and will not be repeated here.
[0197] In one embodiment, Fig.12 As shown, an object information query device is provided, including: an acquisition module 1202, a combination module 1204, a determination module 1206 and a query module 1208, wherein:
[0198] An acquisition module 1202 is used to acquire an image of a target object when product interaction is performed at a point of interest;
[0199] A combining module 1204, configured to combine the image features extracted from the image with a preset feature identification vector to obtain a combined feature;
[0200] Determination module 1206, used to determine the similarity between the combined feature and each verification feature; the verification feature is a feature of the combination of the object feature and the object feature identification vector of the candidate object, which is obtained when the distance between the position of the candidate object and the point of interest meets a preset condition and an object having a close social relationship with the candidate object interacts with the product at the point of interest;
[0201] The query module 1208 is further configured to query the object information of the target object in the image according to the object feature identification vector corresponding to the similarity satisfying the similarity condition when the similarity satisfies the similarity condition.
[0202] In the above embodiment, when product interaction is performed at a point of interest, an image obtained by capturing the target object is obtained. The image features extracted from the image are combined with a preset feature identification vector to obtain a combined feature. The similarity between the combined feature and each verification feature is determined. When there is a similarity that satisfies a similarity condition, the candidate object that satisfies the similarity condition has a high similarity with the target object, and the object information of the candidate object is the object information of the target object. According to the object feature identification vector corresponding to the similarity that satisfies the similarity condition, the object information of the target object in the image is determined. Since the verification feature includes an object feature identification vector, when the similarity that satisfies the similarity condition is determined, the object information of the target object can be directly queried according to the object feature identification vector in the verification feature corresponding to the similarity that satisfies the similarity condition, without the need to reversely query the object feature identification vector according to the object feature, thereby avoiding search errors and improving the accuracy of the query. In addition, since the verification features are obtained when the distance between the location of the candidate object and the point of interest meets the preset conditions and the object with a close social relationship with the candidate object interacts with the product at the point of interest, that is, the verification features of the candidate object have been sent to the local in advance, there is no need to interact with the cloud, and the locally stored verification features can be directly used to identify the target object, which effectively reduces the time spent on obtaining the verification features and improves the speed of querying object information.
[0203] In one embodiment, the determining module 1206 is further configured to:
[0204] Extracting the combined feature element from the combined feature, and extracting the verification feature element from the verification feature;
[0205] sequentially determining the product values between each of the combined characteristic elements and each of the verification characteristic elements;
[0206] The sum of the product values is used as the similarity between the combined feature and the verification feature.
[0207] In one embodiment, the combined characteristic element includes a first characteristic element and a first identification element; the verification characteristic element includes a second characteristic element and a second identification element; and the determination module 1206 is further configured to:
[0208] sequentially determining first product values between each of the first characteristic elements and each of the second characteristic elements;
[0209] sequentially determining second product values between each of the first identification elements and each of the second identification elements;
[0210] Based on the first product value and the second product value, a product value between each of the combined feature elements and each of the verification feature elements is determined.
[0211] In one embodiment, the query module 1208 is further configured to:
[0212] Sorting the similarities in order of magnitude;
[0213] In the sorted similarities, determine the target verification feature corresponding to the similarity whose sorted ranking satisfies the similarity condition;
[0214] Based on the object feature identification vector in the target verification feature, object information of the target object in the image is determined.
[0215] In one embodiment, the combining module 1204 is further configured to:
[0216] Determining a marker vector insertion position in the image features extracted from the image;
[0217] According to the insertion position of the identification vector, the preset feature identification vector is inserted into the image feature extracted from the image to obtain a combined feature.
[0218] In one embodiment, the verification feature belongs to a feature used for verification in a feature library; Fig.13 As shown, the device also includes:
[0219] The determination module 1206 is further used to determine the number of biological objects corresponding to each verification feature in the feature library; and determine the length range of the identification vector according to the number of objects;
[0220] A selection module 1210, configured to select a target length value within the length range of the identification vector;
[0221] The generating module 1212 is used to generate an object feature identification vector of each biological object according to the target length value; wherein the vector length of the preset feature identification vector is the target length value.
[0222] In one embodiment, the apparatus further comprises:
[0223] The acquisition module 1202 is further used to acquire social objects that have a friend relationship chain with the candidate object;
[0224] The determination module 1206 is further configured to, when the candidate object and at least one social object having the friend relationship chain are in the same social group, determine at least one social object as an object having a close social relationship with the candidate object.
[0225] In one embodiment, the social group includes a close social group; the determining module 1206 is further configured to:
[0226] Obtaining group member objects in the intimate social group;
[0227] If at least one of the group member objects belongs to the social object of the friend relationship chain, obtaining the social frequency between the candidate object and at least one of the group member objects;
[0228] If the social frequency reaches a preset frequency threshold, at least one of the group member objects is determined as an object with a close social relationship.
[0229] In one embodiment, the query module 1208 is further configured to:
[0230] Acquire product interaction data corresponding to the target object according to the object information;
[0231] Verifying the product interaction data through a resource transfer interaction model to obtain a verification result;
[0232] If the verification result indicates that the product interaction data has passed the verification, resources are transferred to the target object according to the object information.
[0233] In one embodiment, the image includes at least two hand images of different image types; the combining module 1204 is further configured to:
[0234] Correcting each of the hand images according to the hand image template to obtain a corrected hand image;
[0235] Extracting features from each of the corrected hand images to obtain palm print features;
[0236] Fusing the palmprint features to obtain a fused palmprint feature;
[0237] The fused palmprint feature is combined with a preset feature identification vector to obtain a combined feature.
[0238] In one embodiment, the image comprises a facial image; and the apparatus further comprises:
[0239] An extraction module 1214, configured to extract image features from the facial image;
[0240] A convolution module 1216, configured to perform at least two convolution processes on the image feature through an estimation network to obtain a predicted mean and a predicted variance of the image feature;
[0241] The acquisition module 1202 is further used to acquire network parameters used for standardization processing in the estimation network;
[0242] A standardization module 1218, configured to perform standardization processing on the image features based on the prediction mean, the prediction variance and the network parameters to obtain standardized features;
[0243] The combining module 1204 is further configured to combine the image feature with a preset feature identification vector when it is determined that the target object is a living body according to the standardized feature.
[0244] Each module in the above object information query device can be implemented in whole or in part by software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each module.
[0245] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Fig.14 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store object information query data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, an object information query method is implemented.
[0246] Those skilled in the art will understand that Fig.14 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.
[0247] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.
[0248] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.
[0249] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.
[0250] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards of relevant countries and regions.
[0251] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.
[0252] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0253] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.
Claims
1. A method for querying object information, characterized in that: The method comprises: When product interaction is performed at a point of interest, an image of the target object is acquired; Combining the image features extracted from the image with a preset feature identification vector to obtain a combined feature; Determine the similarity between the combined feature and each verification feature; the verification feature is a feature of a combination of an object feature and an object feature identification vector of a candidate object, and is obtained when a distance between a location of the candidate object and the point of interest meets a preset condition and an object having a close social relationship with the candidate object interacts with a product at the point of interest; When there is a similarity that satisfies the similarity condition, the object information of the target object is queried according to the object feature identification vector corresponding to the similarity that satisfies the similarity condition.
2. The method according to claim 1, characterized in that Determining the similarity between the combined feature and each verification feature includes: Extracting the combined feature element from the combined feature, and extracting the verification feature element from the verification feature; sequentially determining the product values between each of the combined characteristic elements and each of the verification characteristic elements; The sum of the product values is used as the similarity between the combined feature and the verification feature.
3. The method according to claim 2, characterized in that The combined characteristic element includes a first characteristic element and a first identification element; the verification characteristic element includes a second characteristic element and a second identification element; and sequentially determining the product value between each of the combined characteristic elements and each of the verification characteristic elements includes: sequentially determining first product values between each of the first characteristic elements and each of the second characteristic elements; sequentially determining second product values between each of the first identification elements and each of the second identification elements; Based on the first product value and the second product value, a product value between each of the combined feature elements and each of the verification feature elements is determined.
4. The method according to claim 1, characterized in that: When the similarity satisfies the similarity condition, querying the object information of the target object according to the object feature identification vector corresponding to the similarity satisfying the similarity condition includes: Sorting the similarities in order of magnitude; In the sorted similarities, determine the target verification feature corresponding to the similarity whose sorted ranking satisfies the similarity condition; Based on the object feature identification vector in the target verification feature, the object information of the target object in the image is queried.
5. The method according to claim 1, characterized in that The combining of the image features extracted from the image with the preset feature identification vector to obtain the combined features comprises: Determining a marker vector insertion position in the image features extracted from the image; The preset feature identification vector is inserted into the image feature extracted from the image according to the identification vector insertion position to obtain a combined feature.
6. The method according to claim 1, characterized in that The verification feature belongs to a feature used for verification in a feature library; the method further comprises: Determining the number of biological objects corresponding to each verification feature in the feature library; Determining the length range of the identification vector according to the number of objects; Selecting a target length value within the length range of the identification vector; generating an object feature identification vector of each of the biological objects according to the target length value; Among them, the vector length of the preset feature identification vector is the target length value.
7. The method according to claim 1, characterized in that The method further comprises: Acquire a social object that has a friend relationship chain with the candidate object; When the candidate object and at least one social object having the friendship link are in the same social group, at least one social object is determined as an object having a close social relationship with the candidate object.
8. The method according to claim 7, characterized in that The social group includes a close social group; when the candidate object and at least one social object having the friend relationship chain are in the same social group, determining at least one social object as an object having a close social relationship with the candidate object includes: Obtaining group member objects in the intimate social group; If at least one of the group member objects belongs to the social object of the friend relationship chain, obtaining the social frequency between the candidate object and at least one of the group member objects; If the social frequency reaches a preset frequency threshold, at least one of the group member objects is determined as an object with a close social relationship.
9. The method according to claim 1, characterized in that: After querying the object information of the target object according to the object feature identification vector corresponding to the similarity that satisfies the similarity condition, the method further includes: Acquire product interaction data corresponding to the target object according to the object information; Verifying the product interaction data through a resource transfer interaction model to obtain a verification result; If the verification result indicates that the product interaction data has passed the verification, resources are transferred to the target object according to the object information.
10. The method according to claim 1, characterized in that The image includes at least two hand images of different image types; the image features extracted from the image are combined with a preset feature identification vector to obtain a combined feature including: Correcting each of the hand images according to the hand image template to obtain a corrected hand image; Extracting features from each of the corrected hand images to obtain palm print features; Fusing the palmprint features to obtain a fused palmprint feature; The fused palmprint feature is combined with a preset feature identification vector to obtain a combined feature.
11. The method according to claim 1, characterized in that: The image includes a facial image; before combining the image features extracted from the image with the preset feature identification vector, the method further includes: extracting image features from the facial image; Performing at least two convolution operations on the image features through an estimation network to obtain a predicted mean and a predicted variance of the image features; Acquiring network parameters used for standardization processing in the estimation network; Standardizing the image features based on the prediction mean, the prediction variance and the network parameters to obtain standardized features; The combining of the image features extracted from the image and the preset feature identification vector comprises: When the target object is determined to be a living body according to the standardized feature, the image feature is combined with a preset feature identification vector.
12. An object information query device, characterized in that: The device comprises: An acquisition module, used to acquire an image of a target object when product interaction is performed at a point of interest; A combining module, used for combining the image features extracted from the image with a preset feature identification vector to obtain a combined feature; A determination module, configured to determine the similarity between the combined feature and each verification feature; the verification feature is a feature of a combination of an object feature and an object feature identification vector of a candidate object, and is obtained when a distance between the location of the candidate object and the point of interest meets a preset condition and an object having a close social relationship with the candidate object interacts with a product at the point of interest; The query module is used to query the object information of the target object in the image according to the object feature identification vector corresponding to the similarity that satisfies the similarity condition when the similarity satisfies the similarity condition.
13. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 11 are implemented.
14. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 11 are implemented.
15. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 11 are implemented.
