Target living body recognition method, device and computer readable storage medium thereof

Abstract

The application discloses a target living body recognition method and device and a computer readable storage medium. The method comprises the following steps: obtaining position information of a target living body by using a first neural network; obtaining living body key point information of the target living body by using a second neural network based on the position information; obtaining a score vector of the target living body based on the living body key point information, processing a historical score vector in a database and the score vector, obtaining a similarity, and obtaining a recognition result of the target living body based on the similarity, wherein the score vector is a vector generated for describing the living body key point information of the target living body, and the recognition result is used for representing whether the target living body has been successfully recognized. The application solves the technical problem that feature extraction is not comprehensive when a visual recognition living body is recognized in the related art, thereby leading to inaccurate recognition.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of image recognition, and in particular, to a target living body recognition method and device and a computer readable storage medium. BACKGROUND

[0002] At present, video monitoring technology is developing rapidly, and a large amount of data is generated in the monitoring process. Living body re-identification technology can detect and identify a specific living body from an image or a video sequence, which is of great significance to security monitoring and urban security. However, due to problems such as over-bright or over-dark illumination, blurring, serious occlusion, and diverse postures of living body pictures, the existing method directly extracts features from the detected living body, and the extracted features may not be representative, resulting in low search accuracy and search failure.

[0003] At present, no effective solution has been proposed for the above problems. SUMMARY

[0004] Embodiments of the present application provide a target living body recognition method and device and a computer readable storage medium to at least solve the technical problem that the feature extraction is not comprehensive when a living body is visually identified in the related art, resulting in inaccurate recognition.

[0005] According to an aspect of an embodiment of the present application, a target living body recognition method is provided, including: obtaining position information of a target living body by using a first neural network, wherein the first neural network is obtained by machine learning training using a plurality of sets of first training data, each set of first training data including historical position information of the target living body; obtaining living body key point information of the target living body based on the position information and by using a second neural network, wherein the second neural network is obtained by machine learning training using a plurality of sets of second training data, each set of second training data including historical living body key point information of the target living body; obtaining a score vector of the target living body based on the living body key point information, and processing historical score vectors in a database with the score vector to obtain a similarity, and based on the similarity, obtaining a recognition result of the target living body, wherein the score vector is a vector generated for describing the living body key point information of the living body, and the recognition result is used to represent whether the target living body has been successfully identified.

[0006] Optionally, the living body key point information at least includes position information of human body features and position information of human body joints.

[0007] Optionally, the score vector comprises a brightness quality score, a definition score, an occlusion degree score and a posture score, wherein the brightness quality score is a vector obtained based on brightness quality of a region where the target living body is located, the definition score is a vector obtained based on picture definition of an image of the target living body collected, the occlusion degree score is a vector obtained based on occlusion degree of the target living body by an interference object, and the posture score is a vector obtained based on a current posture of the target living body.

[0008] Optionally, before the historical score vector in the database is processed with the score vector, the method further comprises: obtaining a first feature vector by using a third neural network; performing standardization processing on the score vector to obtain a first score vector; and processing a historical feature vector to obtain the historical score vector corresponding to the historical feature vector.

[0009] Optionally, obtaining the first feature vector by using the third neural network comprises: obtaining original feature information of the target living body by using the third neural network based on the position information; and performing standardization processing on the original feature information to obtain the first feature vector.

[0010] Optionally, processing the historical score vector in the database with the score vector to obtain the similarity comprises: obtaining the similarity by using a second formula based on the first feature vector, the first score vector, the historical feature vector and the historical score vector, wherein the second formula is: S is the similarity, F1 is the first score vector, F2 is the historical score vector, W1 is the first feature vector, and W2 is the historical feature vector.

[0011] Optionally, based on the similarity, obtaining the recognition result of the target living body comprises: comparing the similarity with a preset threshold to obtain a comparison result in which the similarity is greater than the preset threshold; and selecting, as the recognition result of the target living body, a comparison result in which the value of the similarity is the highest in the comparison result in which the similarity is greater than the preset threshold.

[0012] According to another aspect of the embodiments of the present application, there is also provided a device for identifying a target living body, comprising: a first obtaining module configured to obtain position information of the target living body by using a first neural network, wherein the first neural network is obtained by machine learning using a plurality of sets of first training data, and each set of the first training data comprises historical position information of the target living body; a second obtaining module configured to obtain living body key point information of the target living body based on the position information and by using a second neural network, wherein the second neural network is obtained by machine learning using a plurality of sets of second training data, and each set of the second training data comprises historical living body key point information of the target living body; a first processing module configured to obtain a score vector of the target living body based on the living body key point information, and process a historical score vector in a database with the score vector to obtain a similarity, and obtain an identification result of the target living body based on the similarity, wherein the score vector is a vector generated for describing the living body key point information of the target living body, and the identification result is used to represent whether the target living body has been successfully identified.

[0013] Optionally, the living body key point information at least comprises position information of human body organs and position information of human body joints.

[0014] Optionally, the score vector comprises a brightness quality score, a definition score, a blocking degree score and a posture score, wherein the brightness quality score is a vector obtained based on brightness quality of a region where the target living body is located, the definition score is a vector obtained based on picture definition of an image of the target living body collected, the blocking degree score is a vector obtained based on blocking degree of the target living body blocked by an interference object, and the posture score is a vector obtained based on a current posture of the target living body.

[0015] Optionally, the device further comprises: a third obtaining module configured to obtain a first feature vector by using a third neural network before processing the historical score vector in the database with the score vector; a standardization processing module configured to perform standardization processing on the score vector to obtain a first score vector; and a second processing module configured to process a historical feature vector to obtain the historical score vector corresponding to the historical feature vector.

[0016] Optionally, the third obtaining module comprises: a first obtaining unit configured to obtain original feature information of the target living body by using the third neural network based on the position information; and a standardization processing unit configured to perform standardization processing on the original feature information to obtain the first feature vector.

[0017] Optionally, the first processing module comprises a second obtaining unit, configured to obtain the similarity based on the first feature vector, the first score vector, a historical feature vector and a historical score vector by using a second formula, wherein the second formula is: S is the similarity, F1 is the first score vector, F2 is the historical score vector, W1 is the first feature vector, and W2 is the historical feature vector.

[0018] Optionally, the first processing module comprises a third obtaining unit, configured to compare the similarity with a preset threshold to obtain a comparison result in which the similarity is greater than the preset threshold; and a selecting unit, configured to select a comparison result in which the similarity is the highest as the recognition result of the target living body.

[0019] According to another aspect of the embodiments of the present application, a computer readable storage medium is also provided, which comprises a stored computer program, wherein the computer program, when executed by a processor, controls a device where the computer readable storage medium is located to perform any of the target living body recognition methods described above.

[0020] According to another aspect of the embodiments of the present application, a processor is also provided, which is configured to execute a computer program, wherein the computer program, when executed, performs any of the target living body recognition methods described above.

[0021] In the embodiment of the present application, the position information of the target living body is acquired by using a first neural network, wherein the first neural network is obtained by machine learning training using a plurality of sets of first training data, each set of the plurality of sets of first training data comprising historical position information of the target living body; the living body key point information of the target living body is acquired based on the position information and by using a second neural network, wherein the second neural network is obtained by machine learning training using a plurality of sets of second training data, each set of the plurality of sets of second training data comprising historical living body key point information of the target living body; the score vector of the target living body is acquired based on the living body key point information, and the historical score vector in the database is processed with the score vector to acquire a similarity, and based on the similarity, an identification result of the target living body is acquired, wherein the score vector is a vector generated for describing the living body key point information of the living body, and the identification result is used to represent whether the target living body has been successfully identified. Through the target living body identification method provided in the embodiment of the present application, the purpose of judging whether the detected target living body is accurate is achieved after the living body key point information is collected, the feature vector of multiple aspects is acquired based on the information, and the vector is searched in the database, so as to realize the technical effect of improving the accuracy of visually identifying the target living body, and further solve the technical problem that the feature extraction is not comprehensive when the living body is visually identified in the related art, resulting in inaccurate identification. BRIEF DESCRIPTION OF DRAWINGS

[0022] The accompanying drawings, which are included to provide a further understanding of the present application and are incorporated in and constitute a part of this application, illustrate embodiments of the present application and serve to explain the present application. In the drawings:

[0023] Figure 1 is a flowchart of a target living body identification method according to an embodiment of the present application;

[0024] Figure 2 is a schematic diagram of a target living body identification device according to an embodiment of the present application. DETAILED DESCRIPTION

[0025] In order for those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should fall within the scope of protection of the present application.

[0026] It should be noted that the terms "first", "second", and the like in the description and claims of the application and the above drawings are used to distinguish similar objects, and do not necessarily indicate a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device that includes a series of steps or units does not necessarily limit to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0027] Embodiment 1

[0028] According to an embodiment of the application, a method embodiment of a target living body recognition method is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described herein can be executed in an order different from that shown herein.

[0029] Figure 1 is a flowchart of a target living body recognition method according to an embodiment of the application, as shown in Figure 1 The method comprises the following steps:

[0030] Step S102, obtaining position information of the target living body by using a first neural network, wherein the first neural network is obtained by machine learning using a plurality of sets of first training data, and each set of first training data in the plurality of sets of first training data comprises historical position information of the target living body.

[0031] Optionally, first, a target detection neural network is used to detect the living body position (i.e. the position information of the target living body), wherein the target detection neural network used includes but is not limited to YOLO network, R-CNN series network, SSD, etc.

[0032] It should be noted that the living body refers to a living plant and animal body, including but not limited to: human living body, animal living body.

[0033] Step S104, obtaining living body key point information of the target living body based on the position information and by using a second neural network, wherein the second neural network is obtained by machine learning using a plurality of sets of second training data, and each set of second training data in the plurality of sets of second training data comprises historical living body key point information of the target living body.

[0034] It should be noted that the living body detection model can use a neural network model such as an SSD, a RetinaNet, a YOLO series network, and the like; or can use an artificial design feature combined with a machine learning model such as a HOG feature combined with an SVM model.

[0035] In step S106, a score vector of the target living body is obtained based on the living body key point information, and a historical score vector in the database is processed with the score vector to obtain a similarity, and based on the similarity, an identification result of the target living body is obtained, wherein the score vector is a vector generated for describing the living body key point information of the target living body, and the identification result is used to represent whether the target living body has been successfully identified.

[0036] It should be noted that in the living body database, a plurality of living body features and corresponding score vectors of a living body are stored. In the database retrieval process, a single feature W2 and a corresponding score vector F2 in the database are set. The similarity between the query living body feature and the database living body feature is calculated, and the score vector and the feature vector are combined for calculation.

[0037] As can be seen from the above, in the embodiment of the present application, first, the position information of the target living body can be obtained by using the first neural network, wherein the first neural network is obtained by using a plurality of sets of first training data through machine learning training, and each set of first training data in the plurality of sets of first training data includes historical position information of the target living body; then the living body key point information of the target living body can be obtained based on the position information and by using the second neural network, wherein the second neural network is obtained by using a plurality of sets of second training data through machine learning training, and each set of second training data in the plurality of sets of second training data includes historical living body key point information of the target living body; finally, the score vector of the target living body can be obtained based on the living body key point information, and the historical score vector in the database is processed with the score vector to obtain a similarity, and based on the similarity, an identification result of the target living body is obtained, wherein the score vector is a vector generated for describing the living body key point information of the living body, and the identification result is used to represent whether the target living body has been successfully identified. Through the target living body identification method provided by the embodiment of the present application, the purpose of judging whether the detected target living body is accurate is achieved after the living body key point information is collected, the feature vector of multiple aspects is obtained based on this, and the vector is retrieved in the database, so as to realize the technical effect of improving the accuracy of visually identifying the target living body, and further solve the technical problem that the feature extraction is not comprehensive in the related art when the living body is visually identified, resulting in inaccurate identification.

[0038] As an optional embodiment, the living body key point information at least includes position information of living body five senses and position information of living body joint points.

[0039] In the optional embodiment, the human key point extraction network is used to extract the human key points, and the human key point extraction network includes but is not limited to OpenPose, CPN, HRNet, etc.

[0040] It should be noted that the extracted living body key points include but are not limited to 17 key points, which are nose, left and right eyes, left and right ears, left and right shoulders, left and right elbows, left and right wrists, left and right hips, left and right knees, and left and right ankles.

[0041] As an optional embodiment, the score vector includes a brightness quality score, a definition score, an occlusion degree score, and a posture score. The brightness quality score is a vector obtained based on the brightness quality of the target living body region, the definition score is a vector obtained based on the picture definition of the collected target living body image, the occlusion degree score is a vector obtained based on the occlusion degree of the target living body being occluded by the interference object, and the posture score is a vector obtained based on the current posture of the target living body.

[0042] In the optional embodiment, the score vector can be calculated, and the score vector includes brightness, definition, occlusion degree, and posture scores.

[0043] (1) The brightness quality score can be calculated according to a specified appropriate brightness range interval. The closer the living body is to the center of the interval within the specified interval, the higher the score. A brightness quality score calculation formula is as follows:

[0044] Wherein, f1 is the brightness score, v is the average gray value of the living body image, T1 and T2 are the set threshold values, which can be T1 = 128 and T2 = 128.

[0045] It should be noted that the brightness quality score is calculated according to the appropriate brightness interval, and other calculation methods such as Gaussian function can be used.

[0046] (2) The definition score calculation method can use gradient algorithms such as Brenner gradient and Laplacian gradient. The higher the gradient, the clearer the image, and the more angular. A definition score calculation formula is as follows:

[0047] Wherein, f2 is the definition score, w and h are the width and height of the living body image, x is in the range of [0, w], y is in the range of [0, h], and p(x, y) represents the pixel value in the range of [0, 1].

[0048] (3) The occlusion degree score (i.e. the occlusion score score) can be classified by using a neural network classifier to classify whether the identified living body is occluded. The last layer of the neural network is a softmax layer that can output an occlusion probability value. The occlusion degree score is denoted as f3.

[0049] (4) The pose score is calculated by the live key point result. Specifically, a template method can be used, that is, a standard front live template is established, the key point positions of the live template are set, then the detected key points are scaled to the standard of the template size, for each key point, the distance value of the corresponding key point of the standard template is calculated, and the maximum value is truncated. For example, the detected joint key point is (x1, y1), and the joint key point position in the template is (x2, y2). The calculation formula is as follows:

[0050] (5) The above scores are combined to form a score vector and are L2 normalized, denoted as F1 (i.e. the first score vector).

[0051] It should be noted that when the recognition object is a human, the score vector can include brightness, sharpness, occlusion degree, pose score, and can also include human clothing attribute type (including shirt, trousers, shoes, hat, mask, etc.), human skin color, face feature, etc.

[0052] As an optional embodiment, before processing the historical score vector in the database and the score vector, the method further includes: obtaining a first feature vector by using a third neural network; performing standardization processing on the score vector to obtain a first score vector; and processing the historical feature vector to obtain a historical score vector corresponding to the historical feature vector.

[0053] In the above optional embodiment, the pedestrian feature is extracted using a feature extraction network (i.e. the third neural network), and ResNet, MobileNet, etc. can be used, wherein the extracted live feature vector is denoted as W1 (i.e. the first feature vector).

[0054] As an optional embodiment, obtaining the first feature vector by using the third neural network includes: obtaining original feature information of the target live body based on the position information by using the third neural network; and performing standardization processing on the original feature information to obtain the first feature vector.

[0055] As an optional embodiment, processing the historical score vector in the database and the score vector to obtain the similarity includes: obtaining the similarity based on the first feature vector, the first score vector, the historical feature vector, and the historical score vector by using a second formula, wherein the second formula is: S is the similarity, F1 is the first score vector, F2 is the historical score vector, W1 is the first feature vector, and W2 is the historical feature vector.

[0056] It should be noted that the modulus of F1 (i.e., the first score vector), F2 (i.e., the historical score vector), W1 (i.e., the first feature vector), and W2 (i.e., the historical feature vector) is set to 1, the calculation result S is the similarity, and the range is [0, 1]. The greater S is, the higher the similarity between the to-be-queried living body and the database living body is. Finally, the living body with the highest similarity greater than a set threshold is selected as the search result, and the similarity threshold can be set to 0.75.

[0057] As an optional embodiment, based on the similarity, the identification result of the target living body is obtained, including: comparing the similarity with a preset threshold to obtain a comparison result with a similarity greater than the preset threshold; and selecting the comparison result with the highest similarity value in the comparison result with the similarity greater than the preset threshold as the identification result of the target living body.

[0058] In the above optional embodiment, the obtained similarity is compared with a preset similarity threshold to obtain a result with a similarity greater than the preset similarity threshold, and the comparison result with the highest similarity data is selected as the identification result of the target living body, which indicates whether the identification is successful.

[0059] As can be seen from the above, by using the target living body identification method of the embodiment of the present application, the human shape is detected by using a detection neural network, the living body key points are extracted by using a living body key point network, and then the human shape features are extracted by using a feature extraction network. Then, the scores of brightness, sharpness, occlusion degree, and symmetry are calculated. Finally, the calculated score vector and feature vector are compared with the database to obtain a search result, and the target living body score vector and feature vector are also combined for living body matching, thereby improving the search accuracy for the target living body.

[0060] Embodiment 2

[0061] According to another aspect of the embodiment of the present application, a target living body identification device is also provided, Figure 2 is a schematic diagram of the target living body identification device according to the embodiment of the present application, as Figure 2 shown, including a first acquisition module 21, a second acquisition module 23, and a first processing module 25. The target living body identification device will be described below.

[0062] The first acquisition module 21 is configured to acquire position information of a target living body by using a first neural network, wherein the first neural network is obtained by machine learning training using a plurality of sets of first training data, and each set of first training data in the plurality of sets of first training data includes historical position information of the target living body.

[0063] The second acquisition module 23 is configured to acquire the vital key point information of the target living body based on the position information and by using a second neural network, wherein the second neural network is obtained by machine learning training using a plurality of sets of second training data, and each set of the second training data includes historical vital key point information of the target living body.

[0064] The first processing module 25 is configured to acquire a score vector of the target living body based on the vital key point information, process the historical score vector in the database and the score vector, acquire a similarity, and acquire an identification result of the target living body based on the similarity, wherein the score vector is a vector generated for describing the vital key point information of the living body, and the identification result is used to represent whether the target living body has been successfully identified.

[0065] It should be noted that the first acquisition module 21, the second acquisition module 23, and the first processing module 25 correspond to steps S102 to S106 in Embodiment 1, and the above modules have the same instances and application scenarios as the corresponding steps, but are not limited to the content disclosed in Embodiment 1. It should be noted that the above modules as part of the device can be executed in a computer system such as a set of computer executable instructions.

[0066] As can be seen from the above, in the embodiment of the present application, first, the first acquisition module 21 can acquire the position information of the target living body by using the first neural network, wherein the first neural network is obtained by machine learning training using a plurality of sets of first training data, and each set of the first training data includes historical position information of the target living body; then the second acquisition module 23 can acquire the vital key point information of the target living body based on the position information and by using the second neural network, wherein the second neural network is obtained by machine learning training using a plurality of sets of second training data, and each set of the second training data includes historical vital key point information of the target living body; finally, the first processing module 25 can acquire the score vector of the target living body based on the vital key point information, process the historical score vector in the database and the score vector, acquire the similarity, and acquire the identification result of the target living body based on the similarity, wherein the score vector is a vector generated for describing the vital key point information of the living body, and the identification result is used to represent whether the target living body has been successfully identified. Through the target living body identification method provided in the embodiment of the present application, the purpose of judging whether the detected target living body is accurate after collecting the vital key point information and acquiring the feature vector of the multiple aspects based on the vector and searching the vector in the database is achieved, thereby realizing the technical effect of improving the accuracy of visually identifying the target living body, and further solving the technical problem that the feature extraction is not comprehensive when the living body is visually identified in the related art, resulting in inaccurate identification.

[0067] Optionally, the living body key point information at least includes position information of human body organs and position information of human body joint points.

[0068] Optionally, the score vector includes a brightness quality score, a definition score, an occlusion degree score and a posture score, wherein the brightness quality score is a vector obtained based on brightness quality of a region where the target living body is located, the definition score is a vector obtained based on image definition of an image of the target living body collected, the occlusion degree score is a vector obtained based on occlusion degree of the target living body by the interference object, and the posture score is a vector obtained based on a current posture of the target living body.

[0069] Optionally, the apparatus further includes a third obtaining module configured to obtain the first feature vector by using a third neural network before the historical score vector in the database is processed with the score vector; a standardization processing module configured to perform standardization processing on the score vector to obtain the first score vector; and a second processing module configured to process the historical feature vector to obtain the historical score vector corresponding to the historical feature vector.

[0070] Optionally, the third obtaining module includes a first obtaining unit configured to obtain original feature information of the target living body by using the third neural network based on the position information; and a standardization processing unit configured to perform standardization processing on the original feature information to obtain the first feature vector.

[0071] Optionally, the first processing module includes a second obtaining unit configured to obtain the similarity by using a second formula based on the first feature vector, the first score vector, the historical feature vector and the historical score vector, wherein the second formula is: S is the similarity, F1 is the first score vector, F2 is the historical score vector, W1 is the first feature vector, and W2 is the historical feature vector.

[0072] Optionally, the first processing module includes a third obtaining unit configured to compare the similarity with a preset threshold to obtain a comparison result in which the similarity is greater than the preset threshold; and a selecting unit configured to select, as the recognition result of the target living body, a comparison result in which the similarity has the highest value in the comparison result in which the similarity is greater than the preset threshold.

[0073] Embodiment 3

[0074] According to another aspect of the embodiments of the present application, a computer readable storage medium is provided, which includes a stored computer program, wherein the computer program, when executed by a processor, controls the device where the computer readable storage medium is located to perform the target living body recognition method in any of the above embodiments.

[0075] Embodiment 4

[0076] According to another aspect of the embodiments of the present application, a processor is provided, which is configured to execute a computer program. The computer program is configured to execute any of the above-mentioned target living body recognition methods when executed.

[0077] The above-mentioned serial numbers of the embodiments of the present application are only for description, and do not represent the advantages or disadvantages of the embodiments.

[0078] In the above-mentioned embodiments of the present application, the description of each embodiment has its own focus, and the parts not described in detail in a certain embodiment can be referred to the relevant description of other embodiments.

[0079] In several embodiments provided in the present application, it should be understood that the disclosed technical contents can be implemented by other manners. Among them, the above-mentioned apparatus embodiment is only schematic, for example, the division of the units can be a logical function division, and in actual implementation, there can be another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units or modules shown or discussed can be indirect coupling or communication connection through some interfaces, units or modules, which can be electrical or other forms.

[0080] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, that is, they can be located in one place, or they can be distributed on multiple units. According to actual needs, part or all of the units can be selected to achieve the purpose of the present embodiment scheme.

[0081] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be realized in the form of hardware or in the form of software functional unit.

[0082] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application, essentially or in other words, the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, including a number of instructions to make a computer device (which can be a personal computer, a server or a network device, etc.) execute all or part of the steps of the methods described in various embodiments of the present application. The aforementioned storage medium includes: a U disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a mobile hard disk, a magnetic disk or an optical disk, and various media that can store program codes.

[0083] The above is only the preferred embodiment of the present application, it should be pointed out that, for those skilled in the art, without departing from the principles of the present application, can make a number of improvements and refinements, these improvements and refinements should also be considered as the protection scope of the present application.

Claims

1. A method for identifying a living target, characterized in that, include: The location information of the target living body is obtained by using a first neural network, wherein the first neural network is trained by machine learning using multiple sets of first training data, and each set of first training data includes: the historical location information of the target living body; Based on the location information, the live key point information of the target living body is obtained using a second neural network, wherein the second neural network is trained by machine learning using multiple sets of second training data, and each set of second training data includes: historical live key point information of the target living body; The score vector of the target live body is obtained based on the live body key point information, and the historical score vector in the database is processed with the score vector to obtain the similarity. Based on the similarity, the recognition result of the target live body is obtained. The score vector is a vector generated to describe the live body key point information of the target live body, and the recognition result is used to characterize whether the target live body has been successfully recognized. The identification result of the target liveness based on the similarity includes: The similarity is compared with a preset threshold, and the comparison result with a similarity greater than the preset threshold is obtained; The comparison result with the highest similarity value among the comparison results with a similarity greater than a preset threshold is selected as the identification result of the target liveness; Before processing the historical score vector in the database and the score vector, the method further includes: obtaining a first feature vector using a third neural network; standardizing the score vector to obtain a first score vector; and processing the historical feature vector to obtain the historical score vector corresponding to the historical feature vector. The similarity is obtained by processing the historical score vector in the database with the score vector, including: obtaining the similarity based on the first feature vector, the first score vector, the historical feature vector, and the historical score vector using a second formula, wherein the second formula is: S is the similarity, F1 is the first score vector, F2 is the historical score vector, W1 is the first feature vector, and W2 is the historical feature vector.

2. The method according to claim 1, characterized in that, The living body key point information includes at least: the location information of the living body's facial features and the location information of the living body's joints.

3. The method according to claim 1, characterized in that, The score vector includes: brightness quality score, sharpness score, occlusion score, and pose score. The brightness quality score is a vector obtained based on the brightness quality of the area where the target live body is located. The sharpness score is a vector obtained based on the image sharpness of the captured image of the target live body. The occlusion score is a vector obtained based on the degree of occlusion of the target live body by interfering objects. The pose score is a vector obtained based on the current pose of the target live body.

4. The method according to claim 1, characterized in that, The first feature vector is obtained using a third neural network, including: Based on the location information, the original feature information of the target living organism is obtained through the third neural network; The original feature information is standardized to obtain the first feature vector.

5. A device for identifying a living target, characterized in that, include: The first acquisition module is used to acquire the location information of the target living body using a first neural network, wherein the first neural network is trained by machine learning using multiple sets of first training data, and each set of first training data includes: the historical location information of the target living body. The second acquisition module is used to acquire the live key point information of the target living body based on the location information and using a second neural network. The second neural network is trained by machine learning using multiple sets of second training data. Each set of second training data includes: historical live key point information of the target living body. The first processing module is used to obtain the score vector of the target live body based on the live body key point information, process the historical score vector in the database with the score vector to obtain the similarity, and obtain the recognition result of the target live body based on the similarity. The score vector is a vector generated to describe the live body key point information of the target live body, and the recognition result is used to characterize whether the target live body has been successfully recognized. The first processing module includes: a third acquisition unit, used to compare the similarity with a preset threshold and acquire comparison results with similarity greater than the preset threshold; and a selection unit, used to select the comparison result with the highest similarity value among the comparison results with similarity greater than the preset threshold as the identification result of the target liveness. The device further includes: a third acquisition module, used to acquire a first feature vector using a third neural network before processing the historical score vector in the database and the score vector; a standardization processing module, used to standardize the score vector to acquire the first score vector; and a second processing module, used to process the historical feature vector to acquire the historical score vector corresponding to the historical feature vector. The first processing module includes a second acquisition unit, configured to acquire the similarity based on a first feature vector, a first score vector, historical feature vectors, and historical score vectors using a second formula, wherein the second formula is: S is the similarity, F1 is the first score vector, F2 is the historical score vector, W1 is the first feature vector, and W2 is the historical feature vector.

6. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein the computer program, when executed by a processor, controls the device containing the computer-readable storage medium to perform the target living body identification method according to any one of claims 1 to 4.

7. A processor, characterized in that, The processor is used to run a computer program, wherein the computer program executes the target living body identification method according to any one of claims 1 to 4.

Images