Method and device for detecting living body in business handling and computer device

Abstract

The application relates to a live body detection method and device in business handling and a computer device, which can be used in the field of artificial intelligence technology, and can also be used in the field of financial technology or other related fields. The method comprises the following steps: extracting target video frames corresponding to pure color picture display time from an object video of a current business handling object participating in video call; determining an initial live body detection result according to the color of reflected light of the target video frames and the color of the displayed pure color picture; in the case that the initial live body detection result indicates that the current business handling object is a live body to be verified, extracting front video frames and rear video frames of the target video frames from the object video; determining face depth information features according to frame difference information of the target video frames, and determining time sequence information features according to frame difference information of the front video frames and the rear video frames; and determining a target live body detection result according to the face depth information features and the time sequence information features. The method can provide live body detection accuracy.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a method, apparatus, computer equipment, storage medium, and computer program product for liveness detection in business processing. Background Technology

[0002] In financial settings, business transactions are sometimes conducted via video calls, such as remote loan interviews. However, if spoofing occurs during these video calls, such as using user photos, printed paper, or composite images, there is a risk of fraud, which could affect the effective processing of the transaction.

[0003] In traditional technology, during business processing, the verification of a person's liveness is primarily achieved through manual review of video call content. However, with large amounts of video call data, manual review is prone to misjudgments or omissions, resulting in low accuracy in liveness detection. Summary of the Invention

[0004] Therefore, it is necessary to provide a liveness detection method, apparatus, computer equipment, computer-readable storage medium, and computer program product in business processing that can improve the accuracy of liveness detection, in order to address the above-mentioned technical problems.

[0005] Firstly, this application provides a method for liveness detection in business processing. The method includes:

[0006] Extract the target video frame corresponding to the solid color screen display time from the video of the current business transaction object participating in the video call; the solid color screen display time is used to represent the time when the object terminal of the current business transaction object displays a solid color screen during the video call, and the solid color screen displayed by the object terminal is used to illuminate the current business transaction object;

[0007] The initial liveness detection result of the current business processing object is determined based on the color of the reflected light of the target video frame and the color of the displayed solid color image;

[0008] If the initial liveness detection result indicates that the current business processing object is a live object to be verified, the preceding and following video frames of the target video frame are extracted from the object video.

[0009] Based on the frame difference information of the target video frame, the facial depth information features of the current business processing object are determined, and based on the frame difference information of the previous video frame and the subsequent video frame, the temporal information features of the current business processing object are determined.

[0010] Based on the facial depth information features and the temporal information features, the target liveness detection result of the current business processing object is determined.

[0011] In one embodiment, the solid color image is displayed for multiple durations;

[0012] The step of determining the initial liveness detection result of the current business processing object based on the color of the reflected light of the target video frame and the color of the displayed solid color image includes:

[0013] According to the order in which the solid color images are displayed, the colors of the reflected light of the target video frames are combined to obtain a first color sequence, and the colors of the displayed solid color images are combined to obtain a second color sequence.

[0014] If the first color sequence and the second color sequence are the same, the current business processing object is confirmed to be a live entity to be verified.

[0015] In one embodiment, the method further includes:

[0016] If the first color sequence and the second color sequence are different, the current business processing object is confirmed to be non-living, and a warning message is generated;

[0017] The warning message is sent to the target terminal.

[0018] In one embodiment, determining the target liveness detection result of the current business processing object based on the facial depth information features and the temporal information features includes:

[0019] The facial depth information features and temporal information features are fused using a pre-trained liveness detection model to obtain fused information features.

[0020] Perform liveness prediction processing on the fused information features to obtain the liveness prediction probability of the current business processing object;

[0021] If the predicted liveness probability is greater than the preset probability, the current business processing object is confirmed to be a live object.

[0022] In one embodiment, the pre-trained liveness detection model is trained in the following manner:

[0023] Obtain the sample object video of the sample object participating in the video call;

[0024] From the sample object video, extract the sample video frame corresponding to the display time of the sample solid color image, and extract the video frame before and after the sample video frame from the sample object video;

[0025] Based on the frame difference information of the sample video frames, the facial depth information features of the sample object are determined, and based on the frame difference information between the preceding and following video frames of the sample video frames, the temporal information features of the sample object are determined.

[0026] The facial depth information features and temporal information features of the sample object are input into the liveness detection model to be trained to obtain the liveness prediction probability of the sample object.

[0027] Based on the difference between the predicted liveness probability and the actual liveness probability of the sample object, the liveness detection model to be trained is trained to obtain a trained liveness detection model, which is used as the pre-trained liveness detection model.

[0028] In one embodiment, determining the facial depth information features of the current business processing object based on the frame difference information of the target video frame includes:

[0029] The face region is extracted from the target video frame to obtain the face video frame;

[0030] Based on the frame difference information of the facial video frames, the facial depth information features of the current business processing object are determined.

[0031] In one embodiment, before extracting the target video frame corresponding to the solid color screen display time from the object video of the current business transaction object participating in the video call, the method further includes:

[0032] The display time of the solid color image is randomly determined;

[0033] Generate a solid color image display instruction corresponding to the solid color image display time;

[0034] The solid color display instruction is sent to the target terminal; the target terminal is used to display the corresponding solid color image during the solid color display time according to the solid color display instruction during the video call.

[0035] Secondly, this application also provides a liveness detection device for business processing. The device includes:

[0036] The first extraction module is used to extract the target video frame corresponding to the solid color screen display time from the object video of the current business handling object participating in the video call; the solid color screen display time is used to represent the time when the object terminal of the current business handling object displays a solid color screen during the video call, and the solid color screen displayed by the object terminal is used to illuminate the current business handling object.

[0037] The first detection module is used to determine the initial liveness detection result of the current business processing object based on the color of the reflected light of the target video frame and the color of the displayed solid color image;

[0038] The second extraction module is used to extract the preceding and following video frames of the target video frame from the object video when the initial liveness detection result indicates that the current business processing object is a liveness to be verified.

[0039] The feature determination module is used to determine the facial depth information features of the current business processing object based on the frame difference information of the target video frame, and to determine the temporal information features of the current business processing object based on the frame difference information of the previous video frame and the subsequent video frame.

[0040] The second detection module is used to determine the target liveness detection result of the current business processing object based on the facial depth information features and the temporal information features.

[0041] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to perform the following steps:

[0042] Extract the target video frame corresponding to the solid color screen display time from the video of the current business transaction object participating in the video call; the solid color screen display time is used to represent the time when the object terminal of the current business transaction object displays a solid color screen during the video call, and the solid color screen displayed by the object terminal is used to illuminate the current business transaction object;

[0043] The initial liveness detection result of the current business processing object is determined based on the color of the reflected light of the target video frame and the color of the displayed solid color image;

[0044] If the initial liveness detection result indicates that the current business processing object is a live object to be verified, the preceding and following video frames of the target video frame are extracted from the object video.

[0045] Based on the frame difference information of the target video frame, the facial depth information features of the current business processing object are determined, and based on the frame difference information of the previous video frame and the subsequent video frame, the temporal information features of the current business processing object are determined.

[0046] Based on the facial depth information features and the temporal information features, the target liveness detection result of the current business processing object is determined.

[0047] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, performs the following steps:

[0048] Extract the target video frame corresponding to the solid color screen display time from the video of the current business transaction object participating in the video call; the solid color screen display time is used to represent the time when the object terminal of the current business transaction object displays a solid color screen during the video call, and the solid color screen displayed by the object terminal is used to illuminate the current business transaction object;

[0049] The initial liveness detection result of the current business processing object is determined based on the color of the reflected light of the target video frame and the color of the displayed solid color image;

[0050] If the initial liveness detection result indicates that the current business processing object is a live object to be verified, the preceding and following video frames of the target video frame are extracted from the object video.

[0051] Based on the frame difference information of the target video frame, the facial depth information features of the current business processing object are determined, and based on the frame difference information of the previous video frame and the subsequent video frame, the temporal information features of the current business processing object are determined.

[0052] Based on the facial depth information features and the temporal information features, the target liveness detection result of the current business processing object is determined.

[0053] Fifthly, this application also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, performs the following steps:

[0054] Extract the target video frame corresponding to the solid color screen display time from the video of the current business transaction object participating in the video call; the solid color screen display time is used to represent the time when the object terminal of the current business transaction object displays a solid color screen during the video call, and the solid color screen displayed by the object terminal is used to illuminate the current business transaction object;

[0055] The initial liveness detection result of the current business processing object is determined based on the color of the reflected light of the target video frame and the color of the displayed solid color image;

[0056] If the initial liveness detection result indicates that the current business processing object is a live object to be verified, the preceding and following video frames of the target video frame are extracted from the object video.

[0057] Based on the frame difference information of the target video frame, the facial depth information features of the current business processing object are determined, and based on the frame difference information of the previous video frame and the subsequent video frame, the temporal information features of the current business processing object are determined.

[0058] Based on the facial depth information features and the temporal information features, the target liveness detection result of the current business processing object is determined.

[0059] The aforementioned liveness detection method, apparatus, computer equipment, storage medium, and computer program product in the business processing process extracts a target video frame corresponding to the solid color image display time from the object video of the current business processing object participating in the video call. The solid color image display time represents the time during which the object terminal of the current business processing object displays a solid color image during the video call, and the solid color image displayed by the object terminal is used to illuminate the current business processing object. Then, based on the color of the reflected light of the target video frame and the color of the displayed solid color image, the initial liveness detection result of the current business processing object is determined. If the initial liveness detection result indicates that the current business processing object is a live object to be verified, the preceding and following video frames of the target video frame are extracted from the object video. Next, based on the frame difference information of the target video frame, the facial depth information features of the current business processing object are determined, and based on the frame difference information of the preceding and following video frames, the temporal information features of the current business processing object are determined. Finally, based on the facial depth information features and the temporal information features, the target liveness detection result of the current business processing object is determined. In this way, when performing liveness detection, comprehensively considering the color of the reflected light in the target video frame corresponding to the display time of the solid color image, the facial depth information features and temporal information features of the current business transaction subject, helps to improve the accuracy of the target liveness detection result, thereby improving the liveness detection accuracy. At the same time, performing an initial liveness detection based on the color of the reflected light in the target video frame, and then performing a secondary liveness detection based on the facial depth information features and temporal information features of the current business transaction subject, helps to further improve the liveness detection accuracy. Moreover, the entire process does not require manual review, thus avoiding the drawback of low liveness detection accuracy caused by manual review. Attached Figure Description

[0060] Figure 1 This is a flowchart illustrating a liveness detection method in a business process, as shown in one embodiment.

[0061] Figure 2 This is a flowchart illustrating the steps for determining the target liveness detection result in one embodiment;

[0062] Figure 3 This is a flowchart illustrating the training steps of a liveness detection model in one embodiment.

[0063] Figure 4 This is a flowchart illustrating a liveness detection method in a business process according to another embodiment;

[0064] Figure 5 This is a structural block diagram of a liveness detection device in a business process according to one embodiment;

[0065] Figure 6 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0066] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0067] It should be noted that the liveness detection method, apparatus, computer equipment, storage medium, and computer program products provided in this application for business processing can be used in the fintech field, such as performing multiple liveness checks on the current business transaction participants in a video call, which helps improve the accuracy of liveness detection during the business processing process; they can also be used in other related fields, such as in the field of artificial intelligence technology, using the color of the reflected light of the target video frame corresponding to the time of the solid color screen display in the object video, the facial depth information features and temporal information features of the current business transaction participant to perform intelligent liveness detection, the entire process does not require manual review, which helps improve the accuracy of liveness detection.

[0068] 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, data stored, data displayed, 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 the relevant data shall comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0069] In one embodiment, such as Figure 1 As shown, a liveness detection method is provided for business processing. This embodiment illustrates the application of this method to a server. It is understood that this method can also be applied to a terminal, or to a system including both a terminal and a server, and is implemented through interaction between the terminal and the server. The terminal can be, but is not limited to, various personal computers, laptops, smartphones, tablets, etc.; the server refers to a bank server, which can be implemented using a standalone server or a server cluster composed of multiple servers. In this embodiment, the method includes the following steps:

[0070] Step S101: Extract the target video frame corresponding to the solid color display time from the object video of the current business handling object participating in the video call; the solid color display time is used to represent the time when the object terminal of the current business handling object displays a solid color image during the video call, and the solid color image displayed by the object terminal is used to illuminate the current business handling object.

[0071] In financial scenarios, individuals who cannot easily visit a branch to conduct business (i.e., customers) may choose to handle transactions remotely via video call, such as remote loan interviews. The "current business processing participant" refers to the person currently participating in the video call to conduct business remotely, such as the customer.

[0072] Here, "solid color screen" refers to a red, green, or blue screen, etc.; the target terminal of the current business being processed in the video call refers to a mobile phone, tablet computer, computer, etc. The solid color screen display time refers to the time during which the solid color screen is displayed on the target terminal during the video call, such as at the 5th second, the 10th second, etc. The specific display time and duration can be adjusted according to the actual situation, and this application does not impose any specific restrictions.

[0073] The display time for a solid color image is multiple times, and each display time corresponds to a different color. The display time for a solid color image is randomly determined, and the color of the solid color image corresponding to each display time is also randomly determined. Furthermore, the colors of the solid color images corresponding to adjacent display times are different.

[0074] It's important to note that the solid-color image displayed on the target terminal is used to illuminate the current business transaction object. In other words, the solid-color image acts as a color light source, projecting light of the corresponding color onto the current business transaction object. Utilizing the selective absorption of different wavelengths of light by different materials, the color of the reflected light from the video frames containing the object can be analyzed to determine if the object is alive. For example, the color of the reflected light from the target video frame containing the current business transaction object, corresponding to the time the solid-color image was displayed, can be used to determine if the current business transaction object is alive, thus mitigating synthetic face attacks. Furthermore, the reflected light signals of a genuine live object and a fake live object (such as a user photo, printed paper, or a composite image) are different. Liveness detection is performed by determining whether the surface of the current business transaction object contains reflected light of the color corresponding to the solid-color image. For example, if the solid-color image is green, the reflected light from the surface of a genuine live object is green, while the reflected light from the surface of a fake live object is a different color.

[0075] The "object video" refers to the video captured by the camera of the object terminal during a video call. The "target video frame" refers to the video frame containing the object during the current business transaction, corresponding to the time frame of the solid color screen display. It should be noted that there are multiple solid color screen display times, and also multiple target video frames.

[0076] Specifically, the client's terminal responds to the video call request forwarded by the server from the remote customer service provider, establishes a video call connection with the remote customer service provider through the server, and simultaneously receives the solid color screen display time sent by the server. During the video call, the client's terminal captures video of the client using its front-facing camera, and displays the corresponding solid color screen on the terminal interface during the solid color screen display time to illuminate the client. Next, the client's terminal sends the captured video of the client to the server. The server extracts the video frame corresponding to the solid color screen display time from the received video and uses it as the target video frame.

[0077] Step S102: Determine the initial liveness detection result of the current business processing object based on the color of the reflected light of the target video frame and the color of the displayed solid color image.

[0078] By performing color analysis on the target video frame, the color of the reflected light in the target video frame can be obtained. For example, using a pre-trained color prediction model to predict the color of the target video frame's reflected light can yield the color of the reflected light. A pre-trained color prediction model is a neural network model used to predict the color of the reflected light in a video frame.

[0079] The display time of a solid color image is associated with a corresponding color, so the color of the solid color image can be determined by the display time of the solid color image.

[0080] The initial liveness detection result indicates whether the current business transaction recipient is a live person, such as whether the recipient is not live or whether the recipient is a live person awaiting verification. It should be noted that assuming the current business transaction recipient is a live person awaiting verification means that the recipient may be live, but it is necessary to combine this with subsequent facial depth information and temporal information features to determine whether the recipient is truly a live person.

[0081] Specifically, the server inputs the target video frame into a pre-trained color prediction model, and uses the color prediction model to predict the color of the reflected light of the target video frame. Based on the color associated with the display time of the solid color image corresponding to the target video frame, the color of the solid color image to be displayed is determined. If the color of the reflected light of the target video frame is the same as the color of the displayed solid color image, the current business processing object is confirmed as a live entity to be verified; otherwise, the current business processing object is confirmed as a non-live entity.

[0082] For example, if the reflected light of the target video frame is red, green, blue, and purple in sequence, and the solid color image displayed is red, green, blue, and purple in sequence, then the current business processing target is confirmed to be a live entity to be verified; if the reflected light of the target video frame is red, green, blue, and purple in sequence, and the solid color image displayed is red, green, yellow, and purple in sequence, then the current business processing target is confirmed to be a non-live entity.

[0083] Furthermore, the training process of the pre-trained color prediction model is as follows: the server acquires sample video frames; the sample video frames are input into the color prediction model to be trained to obtain the predicted color of the reflected light of the sample video frames; based on the difference between the predicted color of the reflected light of the sample video frames and the real color, the color prediction model to be trained is iteratively trained to obtain the trained color prediction model, which serves as the pre-trained color prediction model.

[0084] Step S103: If the initial liveness detection result indicates that the current business processing object is a live object to be verified, extract the previous and next video frames of the target video frame from the object video.

[0085] The preceding video frame refers to the video frame preceding the target video frame, and the following video frame refers to the video frame following the target video frame.

[0086] Specifically, if the server identifies the current business transaction object as a live entity to be verified, it extracts the preceding and following video frames of the target video frame from the object's video, thereby obtaining the preceding and following video frames of the target video frame.

[0087] Step S104: Determine the facial depth information features of the current business processing object based on the frame difference information of the target video frame, and determine the temporal information features of the current business processing object based on the frame difference information of the previous and subsequent video frames.

[0088] Among them, the frame difference information of the target video frame refers to the frame difference information between adjacent target video frames, specifically the absolute value of the difference between adjacent target video frames.

[0089] Among them, facial depth information features are used to represent facial depth information, specifically composed of frame difference information between adjacent target video frames. For example, if the target video frames are frame A, frame B, and frame C, then the facial depth information features refer to the frame difference information between frame A and frame B, and the frame difference information between frame B and frame C.

[0090] It's important to note that there's a significant difference in depth between a genuine, living face and a fake one. A genuine, living face conforms to standard facial depth information, while a fake one does not. Because a real face is a three-dimensional structure, different areas of the face will exhibit varying brightness and color under different lighting conditions. Fake faces, such as those found on electronic device screens or printed paper, lack these characteristics.

[0091] The frame difference information between the preceding and following video frames refers to the absolute value of the difference between the preceding and following video frames.

[0092] The temporal information feature is used to characterize the time-related changes in image features during multiple rounds of solid-color image display. Specifically, it consists of the frame difference information between preceding and subsequent video frames. For example, if the target video frames are A, B, and C, the server can first obtain the frames before and after frame A (A- / A+), the frames before and after frame B (B- / B+), and the frames before and after frame C (C- / C+). Then, by comparing the frame difference information of A- / A+, B- / B+, and C- / C+ respectively, the temporal information feature can be obtained. Finally, the temporal information feature can be used to determine whether the frame difference information between preceding and subsequent video frames is too large. Under normal circumstances, the frame difference information between preceding and subsequent video frames is less than a preset threshold. If the frame difference information between preceding and subsequent video frames is too large, such as greater than or equal to the preset threshold, it indicates that the current business processing object is not a living entity.

[0093] Specifically, the server performs frame difference processing on adjacent target video frames to obtain frame difference information between adjacent target video frames, and combines the frame difference information between adjacent target video frames to obtain the facial depth information features of the current business processing object; it also performs frame difference processing on the preceding and following video frames to obtain the frame difference information between the preceding and following video frames, and combines the frame difference information between the preceding and following video frames to obtain the temporal information features of the current business processing object.

[0094] Step S105: Determine the target liveness detection result of the current business processing object based on the facial depth information features and temporal information features.

[0095] By combining information from three dimensions—the color of the reflected light in the target video frame, facial depth information features, and temporal information features—it can be determined whether the current business transaction is being processed by a living person.

[0096] Among them, the target liveness detection result refers to the final judgment result, which is used to indicate whether the current business processing object is a real living person.

[0097] Specifically, if the facial depth information features meet the preset facial depth information and the temporal information features (i.e., the frame difference information between consecutive video frames) are less than the preset threshold, the server confirms that the current business processing object is a live person; otherwise, it confirms that the current business processing object is not a live person.

[0098] Furthermore, the server can input facial depth information features and temporal information features into a pre-trained liveness detection model to obtain the liveness prediction probability of the current business processing object; if the liveness prediction probability is greater than the preset probability, the current business processing object is confirmed to be alive; otherwise, the current business processing object is confirmed to be not alive.

[0099] It should be noted that if the current business transaction is being processed by a non-living entity, the server will notify the remote customer service representative via message to stop the current business transaction or proactively terminate the current business transaction.

[0100] The aforementioned liveness detection method in business processing involves extracting a target video frame corresponding to the solid color screen display time from the video of the current business processing participant in the video call. The solid color screen display time represents the time during which the target terminal displays a solid color screen during the video call, and the solid color screen displayed by the target terminal illuminates the current business processing participant. Then, based on the color of the reflected light from the target video frame and the color of the displayed solid color screen, the initial liveness detection result of the current business processing participant is determined. If the initial liveness detection result indicates that the current business processing participant is a live subject to be verified, the preceding and following video frames of the target video frame are extracted from the target video. Next, based on the frame difference information of the target video frame, the facial depth information features of the current business processing participant are determined, and based on the frame difference information of the preceding and following video frames, the temporal information features of the current business processing participant are determined. Finally, based on the facial depth information features and the temporal information features, the target liveness detection result of the current business processing participant is determined. In this way, when performing liveness detection, comprehensively considering the color of the reflected light in the target video frame corresponding to the display time of the solid color image, the facial depth information features and temporal information features of the current business transaction subject, helps to improve the accuracy of the target liveness detection result, thereby improving the liveness detection accuracy. At the same time, performing an initial liveness detection based on the color of the reflected light in the target video frame, and then performing a secondary liveness detection based on the facial depth information features and temporal information features of the current business transaction subject, helps to further improve the liveness detection accuracy. Moreover, the entire process does not require manual review, thus avoiding the drawback of low liveness detection accuracy caused by manual review.

[0101] In one embodiment, the solid color image is displayed multiple times. Therefore, step S102, which determines the initial liveness detection result of the current business processing object based on the color of the reflected light from the target video frame and the color of the displayed solid color image, specifically includes the following: combining the colors of the reflected light from the target video frame according to the order of the solid color image display times to obtain a first color sequence, and combining the colors of the displayed solid color images to obtain a second color sequence; if the first color sequence and the second color sequence are the same, the current business processing object is confirmed as a liveness to be verified.

[0102] The first color sequence includes the colors of reflected light from multiple target video frames, and the second color sequence includes the colors of multiple solid color images. It should be noted that since the display time of the solid color images corresponds to both the target video frames and the solid color images simultaneously, the number of colors in the first color sequence is the same as the number of colors in the second color sequence.

[0103] Specifically, the server combines the colors of the reflected light from the target video frames in the order in which the solid color images are displayed to obtain a corresponding color sequence, which is used as the first color sequence. At the same time, the server combines the colors of the displayed solid color images in the order in which the solid color images are displayed to obtain a corresponding color sequence, which is used as the second color sequence. The server then determines whether the first color sequence and the second color sequence are the same. If the first color sequence and the second color sequence are the same, the server confirms that the current business processing object is a liveness detection target.

[0104] For example, if the first color sequence is (red, green, blue, purple) and the second color sequence is (red, green, blue, purple), it means that the current business processing object may be a live entity, that is, the current business processing object is a live entity to be verified. It is necessary to combine steps S103 to S105 to determine whether the current business processing object is a real live entity.

[0105] In this embodiment, an initial liveness detection is performed on the current business processing object based on a first color sequence composed of the colors of the reflected light of the target video frame and a second color sequence composed of the colors of the displayed solid color image. The entire process comprehensively considers the colors of the reflected light of the target video frame and the colors of the displayed solid color image, and determines whether the current business processing object is a live object from the dimension of reflected light color, which helps to improve the accuracy of liveness detection.

[0106] In one embodiment, after combining the colors of the reflected light of the target video frames according to the order of the display time of the solid color images to obtain a first color sequence, and combining the colors of the displayed solid color images to obtain a second color sequence, the method further includes the following: if the first color sequence and the second color sequence are different, confirm that the current business processing object is not a live entity and generate a warning message; send the warning message to the object terminal.

[0107] The warning message is used to remind the current customer to handle the business in accordance with the regulations.

[0108] Specifically, the server compares a first color sequence composed of the colors of the reflected light in the target video frame with a second color sequence composed of the colors of the displayed solid color image. If the first color sequence and the second color sequence are different, the server confirms that the current business transaction is not a live object and generates a corresponding warning message according to the warning message template. Then, the warning message is sent to the target terminal and displayed on the target terminal, such as "Please move the correct face to the front of the screen" to remind the current business transaction to process the business according to the specifications.

[0109] For example, if the first color sequence is (red, green, blue, purple) and the second color sequence is (red, green, yellow, purple), it indicates that the current business processing target is not a living entity. In this case, the server will notify the current business processing target to process the business according to the specifications. If the current business processing target still does not process the business according to the specifications after multiple reminders, the processing of the current business will be terminated.

[0110] In this embodiment, if the first color sequence and the second color sequence are different, it is confirmed that the current business processing object is not a living entity, and the generated warning information is sent to the object terminal, which helps to remind the current business processing object and ensure the normal processing of the business.

[0111] In one embodiment, such as Figure 2 As shown, step S105 above, which determines the target liveness detection result of the current business processing object based on the facial depth information features and temporal information features, specifically includes the following steps:

[0112] Step S201: The face depth information features and temporal information features are fused using a pre-trained liveness detection model to obtain fused information features.

[0113] Step S202: Perform liveness prediction processing on the fused information features to obtain the liveness prediction probability of the current business processing object.

[0114] Step S203: If the predicted probability of a live subject is greater than the preset probability, confirm that the current business processing object is a live subject.

[0115] Among them, the pre-trained liveness detection model is a neural network model used to output the liveness prediction probability.

[0116] Among them, the fused information feature refers to the splicing feature of facial depth information features and temporal information features.

[0117] Specifically, the server inputs facial depth information features and temporal information features into a pre-trained liveness detection model. Based on an attention mechanism, the pre-trained liveness detection model determines the first weight of the facial depth information features and the second weight of the temporal information features. According to the first and second weights, the facial depth information features and temporal information features are fused to obtain fused information features. The fused information features are then subjected to convolution and fully connected processing to predict the liveness of the current business transaction object, thereby obtaining the liveness prediction probability of the current business transaction object. If the liveness prediction probability is greater than the preset probability, the current business transaction object is confirmed to be alive.

[0118] Furthermore, if the server determines that the current business application is not being processed if the probability of liveness prediction is less than or equal to the preset probability, then it will stop processing the current business application.

[0119] In this embodiment, when the current business processing object is a liveness detection subject to be verified, a pre-trained liveness detection model is used to perform a second liveness detection on the current business processing object based on facial depth information features and temporal information features. The entire process comprehensively considers facial depth information features and temporal information features, and uses a pre-trained liveness detection model to determine whether the current business processing object is a real liveness subject, which helps to further improve the accuracy of liveness detection.

[0120] In one embodiment, such as Figure 3 As shown, the liveness detection method in the business processing provided in this application also includes a training step of a pre-trained liveness detection model, specifically including the following steps:

[0121] Step S301: Obtain the sample object video of the sample object participating in the video call.

[0122] Step S302: Extract the sample video frame corresponding to the display time of the sample solid color image from the sample object video, and extract the preceding and following video frames of the sample video frame from the sample object video.

[0123] Step S303: Determine the facial depth information features of the sample object based on the frame difference information of the sample video frames, and determine the temporal information features of the sample object based on the frame difference information of the preceding and following video frames of the sample video frames.

[0124] Step S304: Input the facial depth information features and temporal information features of the sample object into the liveness detection model to be trained to obtain the liveness prediction probability of the sample object.

[0125] Step S305: Based on the difference between the predicted liveness probability and the actual liveness probability of the sample object, train the liveness detection model to be trained to obtain the trained liveness detection model, which serves as the pre-trained liveness detection model.

[0126] Among them, the solid color image display time is used to represent the time that the object terminal of the sample object displays a solid color image during the video call, and the solid color image displayed by the object terminal of the sample object is used to illuminate the sample object.

[0127] Specifically, the server retrieves sample video files of the participants in the video call from the database; extracts sample video frames corresponding to the display time of the sample solid color screen from the sample video files, and extracts the preceding and following video frames from the sample video files; performs frame difference processing on adjacent sample video frames to obtain the frame difference information of adjacent sample video frames, and combines the frame difference information of adjacent sample video frames to obtain the facial depth information features of the sample object; performs frame difference processing on the preceding and following video frames of the sample video frame to obtain the frame difference information of the preceding and following video frames of the sample video frame, and combines the frame difference information of the preceding and following video frames of the sample video frame to obtain the temporal information features of the sample object; inputs the facial depth information features and temporal information features of the sample object into the liveness detection model to be trained, and performs fusion processing on the facial depth information features and temporal information features of the sample object to obtain the fused information features of the sample object; performs liveness prediction processing on the fused information features of the sample object to obtain the liveness prediction probability of the sample object. Next, the server retrieves the actual liveness probability of the sample objects from the database. Based on the difference between the predicted liveness probability and the actual liveness probability of the sample objects, it iteratively trains the liveness detection model to be trained until the training termination condition is met. The trained liveness detection model that has reached the training termination condition is taken as the completed liveness detection model, thus obtaining the pre-trained liveness detection model.

[0128] For example, the server obtains a target loss value based on the difference between the predicted liveness probability and the actual liveness probability of the sample object. If the target loss value is greater than or equal to the preset loss value, the model parameters of the liveness detection model to be trained are adjusted according to the target loss value, and steps S304 to S305 are repeated until the loss value obtained from the trained liveness detection model is less than the preset loss value. Then, the trained liveness detection model is used as the pre-trained liveness detection model.

[0129] In this embodiment, the liveness detection model to be trained is repeatedly trained using the sample videos of the sample objects participating in the video call and the actual liveness probability of the sample objects, so as to obtain a trained liveness detection model. This helps to improve the accuracy of the liveness prediction probability output by the liveness detection model, thereby improving the liveness detection accuracy.

[0130] In one embodiment, step S104 above, which determines the facial depth information features of the current business transaction object based on the frame difference information of the target video frame, specifically includes the following: extracting the face region from the target video frame to obtain a face video frame; and determining the facial depth information features of the current business transaction object based on the frame difference information of the face video frame.

[0131] The target video frame includes a face region and a background region, and the face video frame is a video frame composed of the face region.

[0132] Specifically, the server uses a pre-trained target detection model to identify face regions from target video frames and extracts the identified face regions from the target video frames as face video frames; it performs frame difference processing on adjacent face video frames to obtain the frame difference information of adjacent face video frames, and combines the frame difference information of adjacent face video frames to obtain the face depth information features of the current business processing object.

[0133] In this embodiment, the face region is extracted from the target video frame to obtain the face video frame, and the face depth information features of the current business processing object are determined based on the frame difference information of the face video frame. In this way, by considering the face region in the target video frame to determine the face video frame, and then determining the face depth information features through the face video frame, the interference of the background region in the target video frame can be avoided, which is conducive to improving the accuracy of the determination of face depth information features.

[0134] In one embodiment, before extracting the target video frame corresponding to the solid color display time from the object video of the current business handling object participating in the video call, step S101 above further includes the following: randomly determining the solid color display time; generating a solid color display instruction corresponding to the solid color display time; sending the solid color display instruction to the object terminal; the object terminal is used to display the corresponding solid color image at the solid color display time according to the solid color display instruction during the video call.

[0135] Before each video call, the duration of the solid color screen display and the associated color are randomly determined.

[0136] Specifically, when the target terminal establishes a video call connection with the remote customer service through the server, the server randomly determines the solid color display time and the color associated with that time. Based on the solid color display time, it generates a corresponding solid color display instruction and sends it to the target terminal. During the video call, the target terminal uses its front-facing camera to capture video of the person being served. Simultaneously, based on the solid color display instruction, the terminal interface displays the corresponding solid color image at the specified solid color display time to illuminate the person being served.

[0137] In this embodiment, a corresponding solid color image display instruction is generated based on a randomly determined solid color image display time, and the solid color image display instruction is sent to the target terminal. The target terminal displays the corresponding solid color image at the solid color image display time, which facilitates the subsequent use of the color of the reflected light of the target video frame corresponding to the solid color image display time to determine whether the current business processing object is a live object. This further improves the liveness detection accuracy and avoids the defect of low liveness detection accuracy caused by relying on a fixed solid color image display time.

[0138] In one embodiment, such as Figure 4 As shown, another method for liveness detection in business processing is provided. Taking the application of this method to a server as an example, the method includes the following steps:

[0139] Step S401: Randomly determine the display time of the solid color image; generate a solid color image display instruction corresponding to the solid color image display time.

[0140] Step S402: Send the solid color screen display instruction to the target terminal of the current business transaction object participating in the video call; the target terminal is used to display the corresponding solid color screen during the solid color screen display time according to the solid color screen display instruction during the video call.

[0141] Among them, the solid color screen display time is used to indicate the time during which the target terminal of the current business transaction displays a solid color screen during the video call. The solid color screen displayed by the target terminal is used to illuminate the target of the current business transaction.

[0142] Step S403: Extract the target video frame corresponding to the solid color screen display time from the object video of the current business handling object participating in the video call.

[0143] Step S404: According to the order of the display time of the solid color images, combine the colors of the reflected light of the target video frames to obtain the first color sequence, and combine the colors of the displayed solid color images to obtain the second color sequence.

[0144] After step S404, there are two branch steps, namely step S405 and step S406.

[0145] Step S405: If the first color sequence and the second color sequence are the same, confirm that the current business processing object is a live entity to be verified.

[0146] Step S406: If the first color sequence and the second color sequence are different, confirm that the current business processing object is not alive and generate a warning message; send the warning message to the object terminal.

[0147] Step S407: Extract the face region from the target video frame to obtain the face video frame; determine the face depth information features of the current business processing object based on the frame difference information of the face video frame.

[0148] Step S408: Determine the temporal information characteristics of the current business processing object based on the frame difference information between the previous video frame and the subsequent video frame.

[0149] Step S409: Using a pre-trained liveness detection model, the facial depth information features and temporal information features are fused to obtain fused information features; the fused information features are then subjected to liveness prediction processing to obtain the liveness prediction probability of the current business processing object.

[0150] Step S410: If the predicted probability of a live subject is greater than the preset probability, confirm that the current business processing object is a live subject.

[0151] The aforementioned liveness detection method in business processing comprehensively considers the color of the reflected light in the target video frame corresponding to the display time of the solid color image, the facial depth information features, and the temporal information features of the current business processing subject during liveness detection. This helps improve the accuracy of determining the target liveness detection result, thereby improving the overall liveness detection accuracy. Furthermore, performing an initial liveness detection based on the color of the reflected light in the target video frame, followed by a secondary liveness detection based on the facial depth information features and temporal information features of the current business processing subject, further enhances the liveness detection accuracy. Moreover, the entire process does not require manual review, thus avoiding the drawback of lower liveness detection accuracy due to manual review.

[0152] In one embodiment, to more clearly illustrate the liveness detection method in business processing provided by the embodiments of this application, the following specific embodiment is used to describe the liveness detection method in business processing. In one embodiment, this application also provides a video-based liveness detection method, which involves displaying 1-5 frames of a solid color image on the terminal interface of the current business processing object according to fixed rules during the business processing process to illuminate the current business processing object; then, according to the display time of the solid color image, extracting the frame content of the corresponding video frame from the acquired object video; combining the image frame difference processing before and after the video frame transformation, temporal information features and facial depth information features can be obtained; finally, judging the reflected light color of the frame content of the acquired video frame to determine whether the current reflected content is consistent with the lighting color of the solid color image, and combining the obtained temporal information features and facial depth information features, it is possible to identify whether the current business processing object is a live person. Specifically, the following steps are included:

[0153] Step 1: During the business processing, display 1-5 frames of solid color images on the terminal interface of the current business processing object according to fixed rules to illuminate the current business processing object.

[0154] Specifically, during video service processing, the server adds additional solid color frames to the video frame count per second. For example, if the video frame count is 15 frames, adding solid color frames results in 16-18 frames per second, including 1-3 solid color frames. The solid color frame is displayed every 5 video frames (i.e., at preset time intervals, a solid color frame is displayed on the target terminal interface to illuminate the currently processing user). The color value of the solid color frame is randomly arranged according to the current time and confirmed when the call starts, and is used to illuminate the currently processing user. Utilizing the selective absorption of different wavelengths of light by different materials, the server analyzes the reflected light collected after different colors of light illuminate the face. Simultaneously, by combining image frame difference processing before and after video frame transitions, temporal information features (i.e., time-related changes in image features during multiple rounds of solid color frame displays) and facial depth information features can be obtained, thereby enabling liveness detection.

[0155] Step 2: Based on the display time of the solid color image, extract the frame content of the corresponding video frame from the captured object video.

[0156] Specifically, the server collects video of objects during the current business processing; obtains the display time of the solid color screen in step 1, and extracts the corresponding solid color video frames from the collected object video according to the display time of the solid color screen.

[0157] Step 3: By combining the image frame difference processing before and after the video frame transformation, temporal information features and face depth information features can be obtained.

[0158] Specifically, the server performs frame difference processing on the solid color video frames obtained in step 2 to obtain facial depth information features; at the same time, based on the sequence of solid color images in step 1, it obtains the preceding and following video frames of the solid color video frames, and generates temporal information features based on the frame difference information of the preceding and following video frames of the solid color video frames.

[0159] Step 4: Determine the color of the reflected light in the captured frame content to see if the current reflected content matches the lighting color of the solid color image. Combine this with the temporal information features and facial depth information features obtained in Step 3 to determine if the current business transaction subject is a live person.

[0160] Specifically, the server first judges the color of the reflected light in the collected frame content to determine whether the current reflected content is consistent with the lighting color of the solid color image, and obtains a first judgment result; if the first judgment fails, the current business processing object is confirmed to be non-living; if the first judgment passes, the server combines the temporal information features and facial depth information features obtained in step 3 to obtain a second judgment result; if the second judgment result passes, the current business processing object is confirmed to be living; if the second judgment result fails, the current business processing object is confirmed to be non-living.

[0161] The above-mentioned video-based liveness detection method can achieve the following technical effects: (1) The whole process does not require human intervention, which is conducive to improving the efficiency of liveness detection and thus improving the efficiency of secure recognition of video call content. (2) In the liveness detection process, the color of reflected light from the collected frame content is used, and combined with the temporal information features and facial depth information features obtained by analyzing the video content, which is conducive to improving the accuracy of liveness detection.

[0162] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0163] Based on the same inventive concept, this application also provides a liveness detection device for implementing the liveness detection method in business processing as described above. The solution provided by this device is similar to the implementation described in the above method; therefore, the specific limitations in one or more embodiments of the liveness detection device for business processing provided below can be found in the limitations of the liveness detection method for business processing described above, and will not be repeated here.

[0164] In one embodiment, such as Figure 5 As shown, a liveness detection device for business processing is provided, comprising: a first extraction module 510, a first detection module 520, a second extraction module 530, a feature determination module 540, and a second detection module 550, wherein:

[0165] The first extraction module 510 is used to extract the target video frame corresponding to the solid color screen display time from the object video of the current business handling object participating in the video call; the solid color screen display time is used to represent the time when the object terminal of the current business handling object displays the solid color screen during the video call, and the solid color screen displayed by the object terminal is used to illuminate the current business handling object.

[0166] The first detection module 520 is used to determine the initial liveness detection result of the current business processing object based on the color of the reflected light of the target video frame and the color of the displayed solid color image.

[0167] The second extraction module 530 is used to extract the preceding and following video frames of the target video frame from the object video when the initial liveness detection result indicates that the current business processing object is a live object to be verified.

[0168] The feature determination module 540 is used to determine the facial depth information features of the current business processing object based on the frame difference information of the target video frame, and to determine the temporal information features of the current business processing object based on the frame difference information of the previous and subsequent video frames.

[0169] The second detection module 550 is used to determine the target liveness detection result of the current business processing object based on the facial depth information features and temporal information features.

[0170] In one embodiment, the solid color image is displayed for multiple times; the first detection module 520 is further configured to combine the colors of the reflected light of the target video frame according to the order of the solid color image display times to obtain a first color sequence, and combine the colors of the displayed solid color images to obtain a second color sequence; if the first color sequence and the second color sequence are the same, the current business processing object is confirmed to be a live entity to be verified.

[0171] In one embodiment, the first detection module 520 is further configured to confirm that the current business processing object is not a living entity if the first color sequence and the second color sequence are different, and generate a warning message; and send the warning message to the object terminal.

[0172] In one embodiment, the second detection module 550 is further configured to perform fusion processing on the face depth information features and temporal information features through a pre-trained liveness detection model to obtain fused information features; perform liveness prediction processing on the fused information features to obtain the liveness prediction probability of the current business processing object; and confirm that the current business processing object is a live object if the liveness prediction probability is greater than the preset probability.

[0173] In one embodiment, the liveness detection device for business processing provided in this application further includes a model training module, used to acquire sample object videos of sample objects participating in video calls; extract sample video frames corresponding to the display time of sample solid color images from the sample object videos, and extract the preceding and following video frames from the sample object videos; determine the facial depth information features of the sample objects based on the frame difference information of the sample video frames, and determine the temporal information features of the sample objects based on the frame difference information of the preceding and following video frames of the sample video frames; input the facial depth information features and temporal information features of the sample objects into the liveness detection model to be trained to obtain the liveness prediction probability of the sample objects; train the liveness detection model to be trained based on the difference between the liveness prediction probability and the actual liveness probability of the sample objects to obtain the trained liveness detection model, which serves as the pre-trained liveness detection model.

[0174] In one embodiment, the feature determination module 540 is further configured to extract the face region from the target video frame to obtain a face video frame; and determine the face depth information features of the current business processing object based on the frame difference information of the face video frame.

[0175] In one embodiment, the liveness detection device for business processing provided in this application further includes an instruction sending module, which is used to randomly determine the solid color screen display time; generate a solid color screen display instruction corresponding to the solid color screen display time; send the solid color screen display instruction to the target terminal; and the target terminal is used to display the corresponding solid color screen according to the solid color screen display instruction during the video call.

[0176] The various modules in the liveness detection device used in the aforementioned business processing can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0177] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 6 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs stored in the non-volatile storage media. The database stores data such as the display time of solid color images and target video frames. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When executed by the processor, the computer program implements a liveness detection method in business processing.

[0178] Those skilled in the art will understand that Figure 6 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0179] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0180] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0181] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0182] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this 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), magnetic 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. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0183] The technical features of the above embodiments can be combined in any way. For the sake of brevity, 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.

[0184] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for detecting a live body in a transaction, characterized by, The method comprises: extracting a target video frame corresponding to a pure color picture display time from an object video of a current service processing object participating in a video call; the pure color picture display time is used to represent a time at which an object terminal of the current service processing object displays a pure color picture during a video call process, and the pure color picture displayed by the object terminal is used to irradiate the current service processing object; determining an initial living body detection result of the current service processing object according to a color of reflected light of the target video frame and a color of the displayed pure color picture; in a case where the initial living body detection result indicates that the current service processing object is a living body to be verified, extracting a front video frame and a rear video frame of the target video frame from the object video; determining face depth information features of the current service processing object according to frame difference information of the target video frame, and determining time sequence information features of the current service processing object according to frame difference information of the front video frame and the rear video frame; wherein the frame difference information is a differential absolute value between adjacent target video frames, and the time sequence information features are used to represent picture feature changes related to time in multiple rounds of pure color picture display; determining a target living body detection result of the current service processing object according to the face depth information features and the time sequence information features; the method further comprises: randomly determining a pure color picture display time and a color associated with the pure color picture display time; wherein the display time refers to a time point at which the pure color picture is displayed.

2. The method of claim 1, wherein, The pure color picture display time is multiple; and the determining of the initial living body detection result of the current service processing object according to the color of the reflected light of the target video frame and the color of the displayed pure color picture comprises: combining the color of the reflected light of the target video frame in the order of the pure color picture display time to obtain a first color sequence, and combining the color of the displayed pure color picture to obtain a second color sequence; in a case where the first color sequence and the second color sequence are the same, confirming that the current service processing object is a living body to be verified.

3. The method of claim 2, wherein, The method further comprises: in a case where the first color sequence and the second color sequence are not the same, confirming that the current service processing object is a non-living body, and generating warning information; sending the warning information to the object terminal.

4. The method of claim 1, wherein, The determining of the target living body detection result of the current service processing object according to the face depth information features and the time sequence information features comprises: performing fusion processing on the face depth information features and the time sequence information features through a pre-trained living body detection model to obtain fusion information features; performing living body prediction processing on the fusion information features to obtain a living body prediction probability of the current service processing object; in a case where the living body prediction probability is greater than a preset probability, confirming that the current service processing object is a living body.

5. The method of claim 4, wherein, The pre-trained living body detection model is obtained through the following manner: obtaining a sample object video of a sample object participating in a video call; extract a sample video frame corresponding to a sample pure color picture display time from the sample object video, and extract a front video frame and a rear video frame of the sample video frame from the sample object video; determine face depth information features of the sample object according to frame difference information of the sample video frame, and determine time sequence information features of the sample object according to frame difference information of the front video frame and the rear video frame of the sample video frame; input the face depth information features and the time sequence information features of the sample object into a to-be-trained live body detection model to obtain a live body prediction probability of the sample object; train the to-be-trained live body detection model according to a difference between the live body prediction probability and an actual probability of the sample object, to obtain a trained live body detection model as the pre-trained live body detection model.

6. The method of claim 1, wherein, The determining the face depth information features of the current service processing object according to the frame difference information of the target video frame includes: extracting a face region from the target video frame to obtain a face video frame; determining the face depth information features of the current service processing object according to frame difference information of the face video frame.

7. The method according to any one of claims 1 to 6, characterized in that, Before extracting the target video frame corresponding to the pure color picture display time from the object video of the current service processing object participating in the video call, the method further includes: generating a pure color picture display instruction corresponding to the pure color picture display time; sending the pure color picture display instruction to the object terminal; the object terminal is configured to display a corresponding pure color picture at the pure color picture display time according to the pure color picture display instruction in the process of the video call.

8. A liveness detection device for business processing, characterized in that, The apparatus includes: a first extraction module configured to extract a target video frame corresponding to a pure color picture display time from an object video of a current service processing object participating in a video call; the pure color picture display time is used to represent a time at which an object terminal of the current service processing object displays a pure color picture in the process of the video call, and the pure color picture displayed by the object terminal is used to irradiate the current service processing object; a first detection module configured to determine an initial live body detection result of the current service processing object according to a color of reflected light of the target video frame and a color of the displayed pure color picture; a second extraction module configured to extract a front video frame and a rear video frame of the target video frame from the object video in a case where the initial live body detection result indicates that the current service processing object is a to-be-verified live body; a feature determination module configured to determine face depth information features of the current service processing object according to frame difference information of the target video frame, and determine time sequence information features of the current service processing object according to frame difference information of the front video frame and the rear video frame; wherein the frame difference information is a difference absolute value between adjacent target video frames, and the time sequence information features are used to represent picture feature changes related to time in multiple rounds of pure color picture display; a second detection module configured to determine a target live body detection result of the current service processing object according to the face depth information features and the time sequence information features. The device further comprises a color picture random display module configured to randomly determine a pure color picture display time and a color associated with the pure color picture display time; wherein the display time refers to a time point at which the pure color picture is displayed. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-8 when the computer program is executed by the processor. The computer program, when executed by the processor, implements the steps of the method of any one of claims 1 to 7.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program, when executed by the processor, implements the steps of the method of any one of claims 1 to 7.

11. A computer program product comprising a computer program, characterized in that, The computer program, when executed by the processor, implements the steps of the method of any one of claims 1 to 7. The computer program, when executed by the processor, implements the steps of the method of any one of claims 1 to 7.

Images