Resource Transfer Video Quality Detection Method, Device, Equipment, Medium and Product

Abstract

The present application relates to the field of information security technology, and in particular to a resource transfer video quality detection method, apparatus, equipment, medium, and product, the method comprising: obtaining dual-recording videos of a resource to be detected during a resource transfer process; the dual-recording videos are generated by a robotic process automation system (RPA) based on resource transfer data obtained from multiple business systems; detecting various transfer behavior elements during the resource transfer process and obtaining compliance detection results for each transfer behavior element; the transfer behavior elements are determined based on the identity information of the transfer parties of the resource to be detected and the voice information in the dual-recording videos; obtaining a quality quantification value of the resource transfer to be detected based on the compliance detection results of each transfer behavior element; if the quality quantification value is greater than a preset threshold, determining that the quality of the dual-recording video of the resource transfer to be detected is qualified. The use of this method can improve the efficiency and quality of video information inspection.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a resource transfer video quality detection method, device, equipment, medium and product. Background Art

[0002] As people's living standards continue to improve, more and more people are purchasing financial products and agency-sold products. During the purchase of financial products and agency-sold products, the entire transaction process needs to be recorded and videotaped. Supervisors will regularly review the recordings and videos to check whether the transaction process meets regulations.

[0003] In related technologies, after recording video information, branches need to spend a considerable amount of time to review the entire video information to determine whether the transaction process complies with regulations. Based on regulations on inspection coverage and timeliness, branches are required to achieve 100% inspection coverage and complete the inspection within 5 business days after the video is recorded.

[0004] However, during the above-mentioned inspection process, branch inspectors need to log into the system to repeat the comparison, which results in repeated steps, heavy workload and easy omissions, resulting in low inspection efficiency; and human judgment is affected by the professional quality of personnel, and the video information quality inspection standards cannot be standardized and unified, and the inspection results cannot be effectively guaranteed. Summary of the Invention

[0005] Based on this, it is necessary to provide a resource transfer video quality detection method, device, equipment, medium and product that can improve the efficiency and quality of video information inspection in response to the above technical problems.

[0006] In a first aspect, the present application provides a resource transfer video quality detection method, which is applied to a dual-recording intelligent recognition platform. The method includes:

[0007] Obtain dual-recorded video of the resource transfer process of the resource to be inspected. The dual-recorded video is generated by the Robotic Process Automation (RPA) system based on resource transfer data obtained from multiple business systems.

[0008] Detect each transfer behavior element during the resource transfer process and obtain compliance detection results for each transfer behavior element; the transfer behavior element is determined based on the identity information of the transfer parties of the resource to be detected and the voice information in the dual-recorded video;

[0009] According to the compliance test results of each transfer behavior element, the quality quantitative value of the resource transfer to be tested is obtained;

[0010] If the quality quantization value is greater than the preset threshold, it is determined that the quality of the dual-recorded video of the resource transfer to be detected is qualified.

[0011] In one embodiment, if the transfer behavior element includes facial information of both parties, each transfer behavior element in the resource transfer process is detected, and compliance detection results of each transfer behavior element are obtained, including:

[0012] Extract the facial information of both parties in the dual-recorded video;

[0013] Analyze the key information in the facial information of the two parties transferring, and obtain the facial information analysis results of each party transferring;

[0014] If the facial information analysis results of both parties to the transfer are compliant, the compliance detection results of the facial information of both parties to the transfer are determined to be qualified.

[0015] In one embodiment, key information in the facial information of the two parties transferring the data is analyzed to obtain facial information analysis results of the two parties transferring the data, including:

[0016] Obtaining facial descriptors of the two parties based on key information in the facial information of the two parties being transferred;

[0017] Obtain the Euclidean distance between the facial descriptors of the two transfer parties and the corresponding standard facial descriptors in the local face database;

[0018] If the Euclidean distances are both smaller than the preset distances, it is determined that the facial information analysis results of both parties to the transfer are compliant.

[0019] In one embodiment, if the transfer behavior element includes voice information during the transfer process, detecting each transfer behavior element during the resource transfer process and obtaining compliance detection results for each transfer behavior element include:

[0020] Get the voice information in the dual-recorded video;

[0021] Extracting text information from the voice information to obtain text information corresponding to the voice information;

[0022] Match the text information with the preset resource transfer speech template;

[0023] If the match is successful, the compliance detection result of the voice information during the transfer process is that the voice detection is qualified.

[0024] In one embodiment, extracting text information from voice information to obtain text information corresponding to the voice information includes:

[0025] Input the voice information into the preset local acoustic model to obtain the pinyin information corresponding to the voice information;

[0026] Input the pinyin information into the preset local dictionary library to obtain the text corresponding to the pinyin information;

[0027] The text corresponding to the pinyin information is input into the preset local language model to obtain the text information corresponding to the voice information.

[0028] In one embodiment, the transfer behavior elements include three-sided platform information during the transfer process, and the three-sided platform information represents the identity information of the resource transfer recipient and the historical resource transfer information;

[0029] Detect each transfer behavior element during the resource transfer process and obtain compliance detection results for each transfer behavior element, including:

[0030] Get multiple frames of images from dual-recorded videos;

[0031] Through the preset convolutional neural network model, the three-dimensional table information in multiple frames of images is detected to obtain the detection results;

[0032] If the detection result shows that three-dimensional table information exists in multiple frames of images, the compliance detection result of the three-dimensional table information during the transfer process is determined to be qualified.

[0033] In one embodiment, obtaining a quality quantification value of a resource transfer to be tested based on compliance detection results of each transfer behavior element includes:

[0034] Obtain the weight of each transfer behavior factor according to its influence on the resource transfer process;

[0035] Obtaining the weighted sum of each transfer behavior factor based on the weight of each transfer behavior factor and the compliance detection result;

[0036] The weighted sum of each transfer behavior element is determined as the quality quantification value of the resource transfer to be detected.

[0037] In one embodiment, the method further comprises:

[0038] The quality quantification value of the resource transfer to be detected is sent to the RPA to instruct the RPA to generate report information of the resource transfer process based on the quality quantification value.

[0039] In one embodiment, the method further comprises:

[0040] Send detection details of each transfer behavior element in the resource transfer process to the RPA, instructing the RPA to send the detection details to the resource transfer recipient.

[0041] In a second aspect, the present application also provides a resource transfer video quality detection method, which is applied to RPA. The method includes:

[0042] Obtain resource transfer data of the resources to be inspected during the resource transfer process from multiple business systems and generate dual-recording videos;

[0043] Send a dual-recording video to the dual-recording intelligent recognition platform to instruct the dual-recording intelligent recognition platform to detect various transfer behavior elements in the resource transfer process, obtain the compliance detection results of each transfer behavior element, and obtain the quality quantification value of the resource transfer to be detected based on the compliance detection results of each transfer behavior element. When the quality quantification value is greater than a preset threshold, it is determined that the quality of the dual-recording video of the resource transfer to be detected is qualified; wherein, the transfer behavior element is determined based on the identity information of the transfer parties of the resource to be detected and the voice information in the dual-recording video.

[0044] In one embodiment, the plurality of business systems include an accounting image archive system, a personnel information query system, and a personal customer marketing management system;

[0045] Obtain resource transfer data of the resources to be inspected during the resource transfer process from multiple business systems and generate dual-recording videos, including:

[0046] Acquire voice and image data of resource transfer process from accounting image archive system;

[0047] Obtain the identity information of both parties to the transfer from the personnel information query system;

[0048] Obtaining first risk information of the resource transfer recipient and second risk information of the resource transfer from the personal customer marketing management system;

[0049] A dual-recording video is generated based on the voice data and picture data, the identity information of the transferring parties, the first risk information and the second risk information.

[0050] In a third aspect, the present application further provides a resource transfer video quality detection device, which includes:

[0051] The first acquisition module is used to obtain dual-recorded videos of the resource to be inspected during the resource transfer process; the dual-recorded videos are generated by the Robotic Process Automation (RPA) system based on resource transfer data obtained from multiple business systems;

[0052] A processing module is used to detect various transfer behavior elements in the resource transfer process and obtain compliance detection results of each transfer behavior element; the transfer behavior elements are determined based on the identity information of the transfer parties of the resource to be detected and the voice information in the dual-recorded video;

[0053] The second acquisition module is used to obtain the quality quantitative value of the resource transfer to be detected based on the compliance detection results of each transfer behavior element;

[0054] The determination module is used to determine whether the quality of the dual-recorded video of the resource transfer to be detected is qualified when the quality quantization value is greater than a preset threshold.

[0055] In a fourth aspect, the present application further provides a resource transfer video quality detection device, the device comprising:

[0056] A generation module is used to obtain resource transfer data of the resource to be detected during the resource transfer process from multiple business systems and generate dual-recording videos;

[0057] The sending module is used to send dual-recording videos to the dual-recording intelligent recognition platform, instructing the dual-recording intelligent recognition platform to detect various transfer behavior elements in the resource transfer process, obtain the compliance detection results of each transfer behavior element, and obtain the quality quantification value of the resource transfer to be detected based on the compliance detection results of each transfer behavior element. When the quality quantification value is greater than a preset threshold, it is determined that the quality of the dual-recording video of the resource transfer to be detected is qualified; wherein, the transfer behavior element is determined based on the identity information of the transfer parties of the resource to be detected and the voice information in the dual-recording video.

[0058] In a fifth aspect, the present application further provides a computer device, which includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of any one of the methods in the embodiments of the first and second aspects above are implemented.

[0059] In a sixth aspect, the present application further provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the steps of any one of the methods in the embodiments of the first and second aspects above are implemented.

[0060] In a seventh aspect, the present application further provides a computer program product, comprising a computer program, which, when executed by a processor, implements the steps of any one of the methods in the embodiments of the first and second aspects.

[0061] The above-mentioned resource transfer video quality detection method, device, equipment, medium and product obtain dual-recording videos of the resource to be detected during the resource transfer process. The dual-recording videos are generated by the robotic process automation system RPA based on resource transfer data obtained from multiple business systems, so that various transfer behavior elements in the resource transfer process can be detected and the compliance detection results of each transfer behavior element can be obtained. The transfer behavior element is determined based on the identity information of the transfer parties of the resource to be detected and the voice information in the dual-recording videos. Then, according to the compliance detection results of each transfer behavior element, the quality quantitative value of the resource transfer to be detected can be obtained. If the quality quantitative value is greater than A preset threshold is set to determine whether the quality of the dual-recording video of the resource transfer to be detected is qualified. The robotic process automation system can automatically obtain resource transfer data to generate a dual-recording video of the resource transfer process. The entire process does not require human participation, and the dual-recording video can be obtained intelligently, reducing labor costs. The dual-recording intelligent recognition platform can quickly detect the dual-recording video, which can improve the detection efficiency of the dual-recording video. At the same time, the dual-recording intelligent recognition platform can detect each transfer behavior element in the dual-recording video separately. The detection result of each transfer behavior element determines the detection result of the dual-recording video. The dual-recording video is comprehensively detected, which improves the detection quality of the dual-recording video. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 A diagram illustrating an application environment of a resource transfer video quality detection method according to an embodiment;

[0063] Figure 2 1 is a flow chart of a method for detecting resource transfer video quality in one embodiment;

[0064] Figure 3 1 is a flow chart of a method for detecting resource transfer video quality in one embodiment;

[0065] Figure 4 1 is a flow chart of a method for detecting resource transfer video quality in one embodiment;

[0066] Figure 5 1 is a flow chart of a method for detecting resource transfer video quality in one embodiment;

[0067] Figure 6 1 is a flow chart of a method for detecting resource transfer video quality in one embodiment;

[0068] Figure 7 1 is a flow chart of a method for detecting resource transfer video quality in one embodiment;

[0069] Figure 8 1 is a flow chart of a method for detecting resource transfer video quality in one embodiment;

[0070] Figure 9 1 is a flow chart of a method for detecting resource transfer video quality in one embodiment;

[0071] Figure 10 1 is a flow chart of a method for detecting resource transfer video quality in one embodiment;

[0072] Figure 11 1 is a flow chart of a method for detecting resource transfer video quality in one embodiment;

[0073] Figure 12 1 is a flow chart of a method for detecting resource transfer video quality in one embodiment;

[0074] Figure 13 2. It is a structural diagram of a resource transfer video quality detection device in one embodiment;

[0075] Figure 14 FIG. 1 is a structural diagram of a resource transfer video quality detection device in an embodiment. DETAILED DESCRIPTION

[0076] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0077] It should be noted that the resource transfer video quality detection method, device, equipment, medium and product disclosed in the present invention can be applied in the field of artificial intelligence technology, and can also be used in other technical fields besides the field of artificial intelligence technology. The present disclosure does not limit the application field of the resource transfer video quality detection method, device, equipment, medium and product.

[0078] The resource transfer video quality detection method provided in the embodiment of the present application can be applied to Figure 1 The internal structure diagram of the computer device can be as shown in FIG. Figure 1As shown. The computer device can be a server, and the computer device includes a processor, a memory and a network interface connected through a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a computer-readable storage medium and an internal memory. The computer-readable storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the computer-readable storage medium. The database of the computer device is used to store data of resource transfer videos. The network interface of the computer device is used to communicate with an external terminal through a network connection, and when the computer program is executed by the processor, a resource transfer video quality detection method is implemented. The computer device can be implemented as an independent computer device or a computer device cluster consisting of multiple computer devices.

[0079] It should be noted that the resource transfer video quality detection method provided in this application is described with different execution entities to illustrate each embodiment. Figure 2-Figure 10 The execution subject is the dual recording intelligent recognition platform on the computer equipment. Figure 11 and Figure 12 The execution subject is a robotic process automation system on a computer device, wherein the execution subject of each embodiment can also be a computer device, wherein the device can be implemented as part or all of a client or server through software, hardware, or a combination of software and hardware.

[0080] The following first describes an embodiment executed on the dual recording intelligent recognition platform side on a computer device.

[0081] In one embodiment, Figure 2 As shown, a resource transfer video quality detection method is provided, which is applied to Figure 1 The computer device in the example is used to illustrate the process, including the following steps:

[0082] S201, obtaining a dual-recording video of the resource to be detected during the resource transfer process; the dual-recording video is generated by the robotic process automation system RPA based on resource transfer data obtained from multiple business systems.

[0083] Specifically, in order to ensure the rationality of the resource transfer process, it is necessary to record the entire resource transfer process through a camera. The robotic process automation system (RPA) issues a read instruction to obtain the audio and video recordings of the resource transfer process from the camera, forming a dual-recording video of the resource transfer process.

[0084] Furthermore, it is understandable that the computer device can search in the dual-recording video library according to the identifier of the resource to be detected, and when the dual-recording video of the resource transfer process corresponding to the resource to be detected is found, the dual-recording video of the resource transfer process of the resource to be detected is downloaded through the "Download" button. The identifier of the resource to be detected can be a time identifier, a number identifier, or a keyword identifier, etc. For example, when the identifier of the resource to be detected is a time identifier, the video recording time of the resource to be detected is from 9:00 to 10:00 on October 2, 2021, and the dual-recording video within the time period is searched in the dual-recording video library according to the video recording time, or, when the identifier of the resource to be detected is a number identifier, the video recording number of the resource to be detected is 105, and the dual-recording video corresponding to the number is searched in the dual-recording video library according to the video recording number.

[0085] S202 , detecting various transfer behavior elements in the resource transfer process and obtaining compliance detection results of the various transfer behavior elements; the transfer behavior elements are determined based on the identity information of the transfer parties of the resource to be detected and the voice information in the dual-recorded video.

[0086] Specifically, the computer device can score the identity information of both parties transferring resources and the voice information in the dual-recording video during the resource transfer process through the dual-recording intelligent recognition platform. When the identity information of both parties transferring resources is compliant, the compliance test result of the identity information of both parties transferring resources is determined to be qualified; when the identity information of any party transferring resources is non-compliant, or the identity information of both parties transferring resources is non-compliant, the compliance test result of the identity information of both parties transferring resources is determined to be unqualified. When the score of the voice information in the dual-recording video is greater than the preset voice information score, the compliance test result of the dual-recording video is determined to be qualified; when the score of the voice information is less than or equal to the preset voice information score, the compliance test result of the dual-recording video is determined to be unqualified.

[0087] S203: Obtain a quality quantification value of the resource transfer to be detected based on the compliance detection results of each transfer behavior element.

[0088] Optionally, the computer device can average the compliance detection results of each transfer behavior element, and use the obtained average value as the quality quantification value of the resource transfer to be detected. Optionally, the computer device can obtain the weight of each transfer behavior element based on historical experience, calculate the weighted average of the compliance detection results of each transfer behavior element and the weight of each transfer behavior element, and use the weighted average value as the quality quantification value of the resource transfer to be detected. Optionally, the computer device can also query historical detection results similar to the compliance detection results of each transfer behavior element, and use the quality quantification value corresponding to the historical detection result as the quality quantification value of the resource transfer to be detected. This embodiment does not limit the method of obtaining the quality quantification value of the resource transfer to be detected based on the compliance detection result.

[0089] S204: If the quality quantization value is greater than a preset threshold, it is determined that the quality of the dual-recorded video of the resource transfer to be detected is qualified.

[0090] Specifically, the computer device compares the quality quantization value with a preset threshold. When the quality quantization value is greater than the preset threshold, the quality of the dual-recorded video is considered acceptable; when the quality quantization value is less than or equal to the preset threshold, the quality of the dual-recorded video is considered unacceptable. Optionally, the computer device may use a historical threshold as the preset threshold, or may determine the preset threshold based on the quality acceptance rate of the dual-recorded videos. For example, if the quality acceptance rate of the dual-recorded videos is 80%, there are 20 dual-recorded videos, 18 of which have quality quantization values greater than 90, and the remaining two dual-recorded videos have quality quantization values of 85 and 87, respectively. Therefore, 90 is used as the preset threshold.

[0091] In the above-mentioned resource transfer video quality detection method, the computer equipment obtains dual-recording videos of the resource to be detected during the resource transfer process. The dual-recording videos are generated by the robotic process automation system RPA based on the resource transfer data obtained from multiple business systems, so that various transfer behavior elements in the resource transfer process can be detected and the compliance detection results of each transfer behavior element can be obtained. The transfer behavior element is determined based on the identity information of the transfer parties of the resource to be detected and the voice information in the dual-recording videos. Then, according to the compliance detection results of each transfer behavior element, the quality quantitative value of the resource transfer to be detected can be obtained. If the quality quantitative value is greater than the preset threshold The value is used to determine whether the quality of the dual-recording video of the resource transfer to be detected is qualified. The robotic process automation system can automatically obtain resource transfer data to generate a dual-recording video of the resource transfer process. The entire process does not require human participation, and the dual-recording video can be obtained intelligently, reducing labor costs. The dual-recording intelligent recognition platform can quickly detect the dual-recording video, which can improve the detection efficiency of the dual-recording video. At the same time, the dual-recording intelligent recognition platform can detect each transfer behavior element in the dual-recording video separately. The detection result of each transfer behavior element determines the detection result of the dual-recording video. The dual-recording video is comprehensively detected, which improves the detection quality of the dual-recording video.

[0092] Figure 3 A flow chart of a resource transfer video quality detection method provided in an embodiment of the present application. The embodiment of the present application involves detecting each transfer behavior element in the resource transfer process and obtaining compliance detection results of each transfer behavior element if the transfer behavior element includes the face information of the transfer parties. Figure 2 Based on the embodiment shown, Figure 3 As shown, the above S202 may include the following steps:

[0093] S301, extracting facial information of both parties in the dual-recorded video.

[0094] Specifically, after the computer device acquires the dual-recorded video, it loads the dual-recorded video using OpenCV to obtain multiple frames of the dual-recorded video. Optionally, the multiple frames of the dual-recorded video are grayscale processed to extract facial information of the two parties transferring the data from the multiple frames of the dual-recorded video. Alternatively, partial images of the multiple frames of the dual-recorded video can be acquired at a preset frame interval, grayscale processed on the partial images, and facial information of the two parties transferring the data from the grayscale processed images can be extracted. The preset frame interval can be 3 frames, 5 frames, or 8 frames, for example.

[0095] S302: Analyze key information in the facial information of the two transferring parties to obtain facial information analysis results of the two transferring parties.

[0096] Among them, key information in facial information may be information such as the distance between the eyes, the width of the eyes, or the width and height of the mouth.

[0097] Specifically, the computer device can determine information such as the distance between the eyes, eye width, or mouth width and height of each person transferring the data based on key points of the facial information of the two parties transferring the data in the dual-recorded video, and obtain facial analysis results for each of the two parties transferring the data. For example, if the two parties transferring the data are A and B, the key points of A's facial information determine that A's eye distance is 5cm, eye width is 18cm, and mouth width is 21cm, and this information is recorded in A's facial information analysis results; the key points of B's facial information determine that B's eye distance is 4.8cm, eye width is 16.8cm, and mouth width is 23cm, and this information is recorded in B's facial information analysis results.

[0098] Optional, Figure 4 This is a flow chart of a method for detecting the quality of a resource transfer video provided by an embodiment of the present application. This embodiment of the present application involves analyzing key information in the facial information of both parties to the transfer and obtaining an optional implementation method for analyzing the facial information of both parties to the transfer. Figure 3 Based on the embodiment shown, Figure 4 As shown, the above S302 may include the following steps:

[0099] S401, obtaining facial descriptors of the two transferring parties based on key information in the facial information of the two transferring parties.

[0100] Specifically, a computer device can call the Dlib library through Python to convert key information in the facial information of the two parties into their respective facial descriptors. Key information in facial information includes 68 key features of the face, and the facial descriptor corresponding to the facial information includes a 128-dimensional facial descriptor. The Dlib library is a currently authoritative open-source facial recognition library. It uses a pre-trained deep residual network to convert the 68 key features of the face into a 128-dimensional facial descriptor for face recognition.

[0101] S402: Obtain the Euclidean distance between the facial descriptors of the transfer parties and the corresponding standard facial descriptors in the local face database.

[0102] Specifically, each facial information in the local face database is labeled. For example, when the two parties to the transfer are a customer and a customer manager, the facial descriptors of the customer and the customer manager are obtained respectively, and the standard facial descriptors consistent with the customer and customer manager labels are searched in the local face database. The Euclidean distance between the customer and the standard facial descriptor consistent with the customer label is calculated using the Euclidean distance calculation formula to obtain the first Euclidean distance; the Euclidean distance between the customer manager and the standard facial descriptor consistent with the customer manager label is calculated using the Euclidean distance calculation formula to obtain the second Euclidean distance.

[0103] S403: If the Euclidean distances are both less than the preset distances, it is determined that the facial information analysis results of both the transfer parties are compliant.

[0104] Specifically, after obtaining the first Euclidean distance and the second Euclidean distance through step S402, the first Euclidean distance and the second Euclidean distance are compared with the preset distance. If the first Euclidean distance and the second Euclidean distance are both smaller than the preset distance, the facial information analysis results of both parties to the transfer are compliant; if either the first Euclidean distance or the second Euclidean distance is greater than or equal to the preset distance, and the other is less than the Euclidean distance, the facial information analysis results of both parties to the transfer are partially compliant. If both the first Euclidean distance and the second Euclidean distance are greater than or equal to the preset distance, the facial information analysis results of both parties to the transfer are non-compliant.

[0105] In the above-mentioned resource transfer video quality detection method, the computer device can obtain the facial descriptors of each of the transfer parties based on the key information in the facial information of the transfer parties, thereby obtaining the Euclidean distance between the facial descriptors of each of the transfer parties and the corresponding standard facial descriptors in the local face library. When the Euclidean distance is less than the preset distance, it can be determined that the facial information analysis results of each of the transfer parties are compliant. In the process of calculating the Euclidean distance, the facial descriptors in the local face library all have corresponding identification information. In the calculation process, only two sets of Euclidean distances need to be calculated, and the facial analysis results can be quickly determined.

[0106] S303: If the facial information analysis results of both the transferring parties are compliant, the compliance detection results of the facial information of both the transferring parties are determined to be qualified.

[0107] Specifically, when the facial information of both parties to the transfer is compliant, the compliance test result is qualified; when the facial information of either party to the transfer is not compliant, or the facial information of both parties to the transfer is not compliant, the compliance test result is unqualified.

[0108] In the above-mentioned resource transfer video quality detection method, the computer equipment extracts the facial information of the transferring parties in the dual-recorded video, analyzes the key information in the facial information of the transferring parties, and obtains the facial information analysis results of each of the transferring parties. If the facial information analysis results of each of the transferring parties are compliant, the compliance detection results of the facial information of the transferring parties are determined to be qualified. By transferring the key information in the facial information of the transferring parties, it is possible to accurately determine whether the detection results are compliant, thereby improving the accuracy of the compliance detection results of the facial information of the transferring parties.

[0109] Figure 5 A flow chart of a resource transfer video quality detection method provided in an embodiment of the present application. The embodiment of the present application relates to an optional implementation method of detecting each transfer behavior element in the resource transfer process and obtaining compliance detection results of each transfer behavior element if the transfer behavior element includes voice information during the transfer process. Figure 2 Based on the embodiment shown, Figure 5 As shown, the above S202 may include the following steps:

[0110] S501: Acquire voice information in dual-recorded videos.

[0111] The voice information in the dual-recorded video contains the conversation between the two parties during the resource transfer process.

[0112] Specifically, the computer device can convert the dual-recording video into an audio file through a tool library, wherein the tool library can be an ffmpeg tool library, and the format of the audio file can be a pulse code modulation (PCM) format or a Moving Picture Experts Group Audio Layer III (MP3) format.

[0113] S502: Extract text information from the voice information to obtain text information corresponding to the voice information.

[0114] Optionally, the computer device may extract the text information from the voice information using a corresponding text extraction algorithm to obtain text information corresponding to the voice information. The text extraction algorithm may be a dynamic time warping (DTW) algorithm. Optionally, the computer device may input the voice information into a preset neural network model and output the text information corresponding to the voice information through calculations of the neural network model. This embodiment does not limit the method for obtaining text information from voice information.

[0115] Furthermore, it is understandable that before extracting the text information from the voice information, the silent voice information at the beginning and end of the voice information can be cut off, so as to compress the size of the voice information file.

[0116] Optional, Figure 6 The present invention provides a flowchart of a method for detecting the quality of resource transfer videos. The present invention relates to an optional implementation method for extracting text information from voice information to obtain text information corresponding to the voice information. Figure 5 Based on the embodiment shown, Figure 6 As shown, the above S502 may include the following steps:

[0117] S601: Input the voice information into a preset local acoustic model to obtain the pinyin information corresponding to the voice information.

[0118] Among them, the local acoustic model is obtained by training a large amount of speech information and pinyin information.

[0119] Specifically, the computer device can divide the voice information into multiple small segments of voice information, and regard each small segment as a frame. There are overlapping parts between the frames. Then, the linear predictive cepstral coefficient (LPCC) and Mel frequency cepstrum coefficient (MFCC) algorithms are used to convert each frame waveform into a multi-dimensional feature vector containing sound information. The audio data of the multi-bit feature vector is input into the local acoustic model. After calculation by the preset local acoustic model, the pinyin information corresponding to the voice information is output.

[0120] S602: Input the pinyin information into a preset local dictionary to obtain text corresponding to the pinyin information.

[0121] Among them, the local dictionary library includes all the correspondences between pinyin and characters.

[0122] Specifically, the computer device inputs the pinyin information obtained through step S601 into a preset local dictionary library, and finds the corresponding text information in the local dictionary library according to the order of the pinyin information. For example, if the pinyin information is "yue", the corresponding characters for this pinyin are "月", "乐", "阅", etc.

[0123] S603, Input the text corresponding to the pinyin information into a preset local language model to obtain the text information corresponding to the pinyin information.

[0124] Specifically, the computer device can input the text information corresponding to the pinyin information into the local language model. The local language model can calculate the probability of the connection relationship between two or several adjacent characters or words, and take the one with the highest probability as the text information corresponding to the pinyin information. For example, the first character is "月", "乐", "阅", etc., and the characters adjacent to the first character are "读" and "独". After calculation by the local language model, the probability of the connection relationship between the two characters "阅" and "独" is the highest, and the text information corresponding to the pinyin information obtained is "阅读".

[0125] In the above resource transfer video quality detection method, the computer device can input the voice information into a preset local acoustic model to accurately obtain the pinyin information corresponding to the voice information, so that the pinyin information can be input into a preset local dictionary library to accurately obtain the text corresponding to the pinyin information. Furthermore, the text corresponding to the pinyin information can be input into a preset local language model to accurately obtain the text information corresponding to the voice information. Through the preset local acoustic model, local dictionary library and local language model, the voice information can be accurately converted into text information, and the accuracy of the obtained text information is higher.

[0126] S503, Match the text information with a preset resource transfer speech template.

[0127] Among them, the preset resource transfer speech template is a speech template commonly used in the resource transfer process. For example, the yield of a certain financial product is a certain percentage, and a certain financial product is about the XX sector.

[0128] Specifically, by matching the text information with the keywords in the preset resource transfer speech template, the computer device can check whether the text information contains common speech. For example, "收益率 (yield)", "理财产品 (financial product)", etc. can be used as keywords in the resource transfer speech template, and the text information is matched with "收益率 (yield)", "理财产品 (financial product)", etc. and the preset resource transfer speech template to determine whether the content is included in the text information.

[0129] S504, If the match is successful, the compliance detection result of the voice information in the transfer process is that the voice detection is qualified.

[0130] Specifically, when the text information successfully matches the preset resource transfer script template, it means that during the resource transfer process, the account manager has informed the customer of the more important information in the resource transfer process. The resource transfer process is completed when the customer is aware of the risks of the financial product and the customer's own risk-bearing ability. It is a voluntary behavior and avoids the possibility of subsequent changes.

[0131] In the above-mentioned resource transfer video quality detection method, the computer equipment obtains the voice information in the dual-recorded video, extracts the text information of the voice information, obtains the text information corresponding to the voice information, and matches the text information with the preset resource transfer speech template. If the match is successful, the compliance detection result of the voice information during the transfer process is that the voice detection is qualified. By matching with the preset resource transfer speech module, it can accurately determine whether the relevant content in the voice information is compliant, thereby improving the accuracy of voice information judgment.

[0132] Figure 7 A flow chart of a resource transfer video quality detection method provided in an embodiment of the present application. The present application embodiment involves a transfer behavior element including three-sided station information during the transfer process, which represents the identity information of the resource transfer recipient and historical resource transfer information; an optional implementation method of detecting each transfer behavior element during the resource transfer process and obtaining compliance detection results of each transfer behavior element. Figure 2 Based on the embodiment shown, Figure 7 As shown, the above S202 may include the following steps:

[0133] S701: Acquire multiple frames of images in the dual-recorded video.

[0134] Specifically, after the computer device acquires the dual-recorded video, it loads the dual-recorded video through OpenCV to obtain all images of the dual-recorded video. Optionally, all images of the dual-recorded video can be used as multiple frames, or partial images of the full dual-recorded video can be acquired according to a preset frame interval and used as multiple frames of the dual-recorded video. The preset frame interval can be 3 frames, 5 frames, or 8 frames, etc.

[0135] S702, using a preset convolutional neural network model, detect the three-dimensional table information in multiple frames of images to obtain a detection result.

[0136] Specifically, partial images from the dual-recorded video are fed into a pre-set convolutional neural network model. The convolutional layer of the convolutional neural network model is responsible for calculating and extracting local features from the image. The convolution kernel scans the entire image to derive convolutional features. The fully connected layer then classifies, searches, and compares the image fragments of the local features, gradually evolving from small local feature images to large local feature images until an image of the entire three-panel TV information is found. For example, the top, bottom, front, and side views of the three-panel TV information can be classified. When a suitable feature image is found, the neuron is activated, confirming the presence of three-panel TV information in the dual-recorded video.

[0137] S703: If the detection result shows that three-dimensional table information exists in the multiple frames of image, determine that the compliance detection result of the three-dimensional table information during the transfer process is qualified.

[0138] Specifically, after the computer device detects the presence of three-sided table information in the multiple frames of images through step S702, it indicates that the detection result of the three-sided table information is qualified.

[0139] In the above-mentioned resource transfer video quality detection method, the computer equipment obtains multiple frames of images in the dual-recorded video, and through the preset convolutional neural network model, it can accurately detect the three-sided channel information in the multiple frames of images to obtain the detection results. If the detection result is that there is three-sided channel information in the multiple frames of images, the compliance detection result of the three-sided channel information in the transfer process is determined to be qualified. The convolutional neural network model can accurately determine the detection result of the three-sided channel information, thereby improving the accuracy of judging the three-sided channel information.

[0140] Figure 8 The present invention provides a flow chart of a method for detecting the quality of resource transfer videos according to an embodiment of the present invention. The present invention relates to an optional implementation method for obtaining the quality quantification value of the resource transfer to be detected according to the compliance detection results of each transfer behavior element. Figure 2 Based on the embodiment shown, Figure 8 As shown, the above S203 may include the following steps:

[0141] S801: Obtain the weight of each transfer behavior element according to the degree of influence of each transfer behavior element on the resource transfer process.

[0142] Among them, each transfer behavior element includes the facial information of the transferor and the voice information during the transfer process, as well as the three-sided platform information.

[0143] Specifically, the computer device can determine the degree of influence of each transfer behavior element on the resource transfer process based on historical experience, and obtain the weight of each transfer behavior element based on the relationship between the degree of influence and the weight. For example, the relationship between the degree of influence and the weight may include: the weight corresponding to the maximum degree of influence is 0.5, the weight corresponding to the relatively large degree of influence is 0.4, the weight corresponding to the average degree of influence is 0.3, the weight corresponding to the relatively small degree of influence is 0.2, and the weight corresponding to the minimum degree of influence is 0.1. If, during the resource transfer process, the facial information of the transfer parties and the voice information during the transfer process have a greater degree of influence on the resource transfer process, and the three-sided station information has a smaller degree of influence on the resource transfer process, then the weight corresponding to the facial information of the transfer parties is 0.4, the weight corresponding to the voice information during the transfer process is 0.4, and the weight corresponding to the three-sided station information during the transfer process is 0.2.

[0144] S802: Determine the weighted sum of each transfer behavior element as the quality quantization value of the resource transfer to be detected.

[0145] Specifically, the computer device calculates the product of the score of each transfer behavior element and the corresponding weight of each transfer behavior element, then sums the resulting products and uses the sum as the quality quantification value of the resource transfer to be detected. For example, if the facial information of the two transfer parties is scored 100 points, the corresponding weight is 0.4; the voice information during the transfer process is scored 80 points, the corresponding weight is 0.4; the three-face information during the transfer process is scored 100 points, the corresponding weight is 0.2, the weighted sum of the transfer behavior elements is 92, and the quality quantification value of the resource transfer to be detected is also 92 points.

[0146] In the above-mentioned resource transfer video quality detection method, the computer equipment can obtain the weight of each transfer behavior element according to the degree of influence of each transfer behavior element on the resource transfer process, so that the weighted sum of each transfer behavior element can be determined as the quality quantitative value of the resource transfer to be detected. A more reasonable video quality score can be obtained through the weighted sum method, which is convenient for customer managers to check omissions in the transaction process.

[0147] In another embodiment, the present application embodiment relates to an optional implementation method for generating report information. Figure 2 Based on the illustrated embodiment, the above method further includes the following step: sending a quality quantization value of the resource transfer to be detected to the RPA to instruct the RPA to generate report information of the resource transfer process based on the quality quantization value.

[0148] Specifically, after obtaining the quantified quality value of the resource transfer process, the computer device sends the quantified quality value to the RPA. The RPA can then generate a report on the resource transfer process based on the relevant information of the quantified quality value through report compilation software. For example, the report information generated by the report compilation software may include a trend chart, a bar chart, and a pivot table.

[0149] In the above-mentioned resource transfer video quality detection method, the computer device sends the quality quantization value of the resource transfer to be detected to the RPA, instructing the RPA to generate report information of the resource transfer process based on the quality quantization value, and records the relevant information of the dual-recording video in the form of a report, which can be convenient for subsequent reference.

[0150] In another embodiment, the present application embodiment relates to an optional implementation method of sending detection details. Figure 2 Based on the illustrated embodiment, the above method further includes the following steps: sending detection details of each transfer behavior element in the resource transfer process to the RPA to instruct the RPA to send the detection details to the resource transfer recipient.

[0151] Specifically, the computer device needs to send the RPA a detailed inspection of each transfer action element during the resource transfer process. The RPA can promptly send the details of the resource transfer process to the resource transfer recipient via email or text message. For example, the resource transfer process details may include the generation of dual-recording videos, dual-recording video face detection results, dual-recording video voice detection results, dual-recording video three-way information detection results, and dual-recording video quality quantification values.

[0152] In the above-mentioned resource transfer video quality detection method, the computer device sends the detection details of each transfer behavior element in the resource transfer process to the RPA, instructing the RPA to send the detection details to the resource transfer recipient. This can promptly inform the resource transfer recipient of relevant information during the resource transfer process, achieving process transparency.

[0153] In one embodiment, Figure 9 As shown, in order to facilitate understanding by those skilled in the art, the resource transfer video quality detection method is described in detail below. The method may include:

[0154] S901, obtaining dual-recorded videos of the resource to be detected during the resource transfer process;

[0155] S902, extracting facial information of both parties in the dual-recorded video;

[0156] S903, obtaining facial descriptors of the two transferring parties based on key information in the facial information of the two transferring parties;

[0157] S904, obtaining the Euclidean distance between the facial descriptors of the transfer parties and the corresponding standard facial descriptors in the local face database;

[0158] S905: If the Euclidean distances are both less than the preset distance, it is determined that the facial information analysis results of both the transfer parties are compliant;

[0159] S906, if the facial information analysis results of both the transferring parties are compliant, determining that the compliance detection results of the facial information of both parties are qualified;

[0160] S907, obtaining voice information in the dual-recorded video;

[0161] S908, extracting text information from the voice information to obtain text information corresponding to the voice information;

[0162] S909: Input the voice information into a preset local acoustic model to obtain pinyin information corresponding to the voice information;

[0163] S910, inputting the pinyin information into a preset local dictionary to obtain text corresponding to the pinyin information;

[0164] S911, inputting the text corresponding to the pinyin information into a preset local language model to obtain text information corresponding to the voice information;

[0165] S912, matching the text information with a preset resource transfer speech template;

[0166] S913, if the match is successful, the compliance detection result of the voice information in the transfer process is that the voice detection is qualified;

[0167] S914, acquiring multiple frames of images in the dual-recorded video;

[0168] S915, using a preset convolutional neural network model, detecting three-dimensional table information in the multiple frames of images to obtain a detection result;

[0169] S916: If the detection result shows that three-dimensional table information exists in the multiple frames of image, determine that the compliance detection result of the three-dimensional table information during the transfer process is qualified.

[0170] S917, obtaining the weight of each transfer behavior element according to the degree of influence of each transfer behavior element on the resource transfer process;

[0171] S918: Determine the weighted sum of the transfer behavior elements as the quality quantization value of the resource transfer to be detected.

[0172] It should be noted that for the description in the above S901-S918, reference can be made to the relevant description in the above embodiment, and the effects are similar, so this embodiment will not be repeated here.

[0173] Further, Figure 10 A flowchart illustrating the resource transfer video quality inspection method is shown below. The RPA system logs into the accounting archive system and queries the dual-recording data details for financial transactions. This data includes the transaction area code, branch number, account manager information, customer information, financial product code, product channel source, and dual-recording video. The RPA system then writes the financial code into the database. The RPA system logs into the personnel information query system and, based on the retrieved account manager and customer information, queries the account manager and customer ID information and writes this information into the database. After logging into the new accounting archive system, the RPA system downloads the retrieved dual-recording video of the financial transactions and writes it into the database. The RPA system logs into the next-generation personal customer marketing management system and, based on the retrieved financial product code, obtains financial product risk information and updates the database with this information. The RPA system logs into the next-generation personal customer marketing management system and, based on the retrieved customer information, obtains customer risk level information and updates the database with this information. The RPA system extracts the current financial code, account manager and customer ID information, dual-recording video, financial product risk information, and customer risk level information from the database and packages them to generate dual-recording data for financial transactions to be screened. The RPA system transfers the dual-recorded financial transaction data to be screened via FTP to the dual-recording intelligent recognition platform, recording the transmission status. The dual-recording intelligent recognition platform uses an AI deep learning model to verify that the three counters are in the same frame. The AI deep learning model also performs facial recognition on the customer and account manager during the transaction process. Automatic speech recognition (ASR) technology is used to recognize speech during the transaction. The AI deep learning model calculates scores based on the results of facial and voice recognition. The RPA system aggregates and compiles statistics on the success of multiple dual-recorded financial transaction data and generates reports. The RPA system also incorporates email notifications at various pre-set process nodes. Upon successful completion or exceptions, notifications are sent to designated email addresses, alerting users to the progress of the process.

[0174] In the above-mentioned resource transfer video quality detection method, the computer device obtains dual-recording videos of the resource to be detected during the resource transfer process, extracts the facial information of the two transfer parties in the dual-recording videos, obtains the facial descriptors of the two transfer parties based on the key information in the facial information of the two transfer parties, obtains the Euclidean distance between the facial descriptors of the two transfer parties and the corresponding standard facial descriptors in the local face library, and if the Euclidean distance is less than the preset distance, it is determined that the facial information analysis results of the two transfer parties are both compliant. If the facial information analysis results of the two transfer parties are both compliant, it is determined that the compliance detection results of the facial information of the two transfer parties are qualified; obtain the voice information in the dual-recording videos, extract text information from the voice information, obtain the text information corresponding to the voice information, input the voice information into the preset local acoustic model, obtain the pinyin information corresponding to the voice information, input the pinyin information into the preset local dictionary library, obtain the text corresponding to the pinyin information, and Input into the preset local language model to obtain the text information corresponding to the voice information, and match the text information with the preset resource transfer speech template. If the match is successful, the compliance detection result of the voice information in the transfer process is that the voice detection is qualified; obtain multiple frames of images in the dual-recording video, and use the preset convolutional neural network model to detect the three-sided platform information in the multiple frames to obtain the detection result. If the detection result is that there is three-sided platform information in the multiple frames, it is determined that the compliance detection result of the three-sided platform information in the transfer process is qualified. Therefore, the weight of each transfer behavior element can be obtained according to the degree of influence of each transfer behavior element on the resource transfer process, and then the weighted sum of each transfer behavior element can be determined as the quality quantitative value of the resource transfer to be detected. The dual-recording video is inspected through three aspects: face recognition, voice recognition and three-sided platform recognition. The inspection process is relatively comprehensive and is operated through RPA. The entire process does not require human participation, which makes the quality and efficiency of video information inspection higher.

[0175] The following describes an embodiment in which a robotic process automation system on a computer device is the execution entity. It is understood that the method steps in this embodiment, which uses the robotic process automation system as the execution entity, correspond to those in the aforementioned embodiment in which the dual-recording intelligent recognition platform is the execution entity. The details involved in each step are identical, so to avoid redundancy, some details will not be repeated.

[0176] In one embodiment, Figure 11 As shown, a resource transfer video quality detection method is provided, which is applied to Figure 1 The computer device in the example is used to illustrate the process, including the following steps:

[0177] S1101, obtaining resource transfer data of a resource to be detected during a resource transfer process from multiple business systems, and generating a dual-recording video.

[0178] Specifically, computer equipment can use RPA to simulate manual work in downloading all dual-recording videos in one of multiple business systems, and query information related to the resources to be tested in other business systems. For example, this information can be the identity information of the transfer parties that should appear in the dual-recording video, the product code information involved in the resource transfer process, and the yield rate of the products involved in the resource transfer process, etc. The dual-recording video of the resource to be tested can be queried in all dual-recording videos through the relevant information of the resource to be tested.

[0179] S1102, sending a dual-recording video to the dual-recording intelligent recognition platform to instruct the dual-recording intelligent recognition platform to detect various transfer behavior elements in the resource transfer process, obtain compliance detection results of each transfer behavior element, and obtain a quality quantification value of the resource transfer to be detected based on the compliance detection results of each transfer behavior element. When the quality quantification value is greater than a preset threshold, it is determined that the quality of the dual-recording video of the resource transfer to be detected is qualified; wherein, the transfer behavior element is determined based on the identity information of the transfer parties of the resource to be detected and the voice information in the dual-recording video.

[0180] Specifically, the computer device writes the dual-recording video into a JavaScript Object Notation (JSON) file for each transaction, packages each dual-recording transaction data into a ZIP file, and sends the generated ZIP file to the dual-recording intelligent recognition platform through the RPA system. The dual-recording intelligent recognition platform then verifies various behavioral elements in the dual-recording video. For example, it verifies the correct identity information of the two parties transferring resources, verifies the correctness of the voice information in the dual-recording video, and detects the presence of three-way table information in the dual-recording video. The quality quantitative value of the dual-recording video is calculated based on the results of the detection of various behavioral elements. The quality quantitative value of the dual-recording video is then compared with a preset threshold to determine whether the dual-recording video quality meets the requirements. The JSON file includes information such as the area code, branch code, transaction date, serial number, account manager name, account manager number, account manager ID image path, JSON text file name, channel source, customer name, customer number, customer ID image path, customer risk level assessment time, product code, name, yield rate, risk level, and type.

[0181] In the above-mentioned resource transfer video quality detection method, a computer device obtains resource transfer data of the resource to be detected in the resource transfer process from multiple business systems, generates a dual-recording video, and sends the dual-recording video to the dual-recording intelligent recognition platform to instruct the dual-recording intelligent recognition platform to detect various transfer behavior elements in the resource transfer process, obtain the compliance detection results of each transfer behavior element, and obtain the quality quantification value of the resource transfer to be detected based on the compliance detection results of each transfer behavior element. When the quality quantification value is greater than a preset threshold, it is determined that the quality of the dual-recording video of the resource transfer to be detected is qualified; wherein, the transfer behavior element is determined based on the identity information of the transfer parties of the resource to be detected and the voice information in the dual-recording video. By obtaining the resource to be detected from multiple business systems to generate a dual-recording video, and sending the dual-recording video to the dual-recording intelligent recognition platform, no human intervention is required, and the quality of the dual-recording video can be intelligently identified, thereby improving the efficiency of dual-recording video recognition.

[0182] Figure 12 This is a flow chart of a resource transfer video quality detection method provided in an embodiment of the present application. This embodiment of the present application involves obtaining resource transfer data from multiple business systems and generating an optional implementation method of dual recording video. Figure 11 Based on the embodiment shown, Figure 12 As shown, the above method further includes the following steps: multiple business systems include an accounting image archive system, a personnel information query system, and a personal customer marketing management system; resource transfer data of the resource to be detected during the resource transfer process is obtained from the multiple business systems to generate a dual-recording video, including:

[0183] S1201, obtaining voice data and image data of the resource transfer process from the accounting image archive system.

[0184] For example, a computer device automatically opens a browser through the RPA system, enters the URL of the accounting image archive system, and simulates entering the preset account and password on the keyboard to log in to the accounting image archive system. The computer then locates the webpage tab in the new accounting image archive system, sets the query institution, query date, and query product type, and simulates clicking the "Search" button to enter the detailed list of financial transactions. This detailed list of financial transactions displays all dual-recorded transactions for the institution on the query date. Financial products include wealth management products, mutual funds, and insurance. The traversal list unit, through the RPA system, captures the "Current Page" and "Total Pages" values on the detailed list of financial transactions. It then determines whether to turn the page. If the next page is determined, it simulates clicking "Next Page" to advance to the next page. The RPA system, in conjunction with the traversal list unit, navigates to each row in the detailed list and captures transaction data, including the region code, branch number, account manager information handling the transaction, customer information, financial product code, channel source, and dual-recorded video. The RPA system can download the voice and image data of the resource transfer process by clicking the "Download" button. After clicking the view audio link, the RPA system calls up the local Employee Assistance Program (EAP) environment and then calls a Python script to automatically detect the download process of the dual-recording video until the dual-recording video download is complete.

[0185] S1202, obtaining identity information of both parties of the transfer from a personnel information query system.

[0186] Specifically, the computer automatically opens a browser through the RPA system, enters the URL of the personnel information query system, and simulates entering the preset account and password on the keyboard to log in to the personnel information query system. When querying the personnel information query system for the account manager number and customer number information of the transaction process based on the number of the resource to be detected or the account manager number, the RPA system calls the Application Programming Interface (API) to obtain the ID card image message of each party to the transaction. By parsing the message, it extracts the base64 message bytes and converts them into ID card images. To protect customer privacy, ID card images are named using a universally unique identifier (UUID) and do not use sensitive information such as the customer's ID number as the image name.

[0187] S1203, obtaining first risk information of the resource transfer recipient and second risk information of the resource transfer from the personal customer marketing management system;

[0188] Specifically, the computer device automatically opens a browser URL through the RPA system, simulates entering user account and password information on the keyboard, and simulates mouse clicks to log in to the personal customer marketing management system. The RPA system locates the webpage tag, enters the unified product view interface, sets the wealth management product code in the product search field, simulates clicking the "Query" button, enters the wealth management product details page, and captures the expected annualized rate of return, risk level, and other information on the page through the RPA system. This information is used as the secondary risk information for resource transfer. The customer number is set in the customer search field, and a simulated click of the "Query" button is clicked to enter the customer risk details page. The RPA system captures information such as the customer's risk tolerance on the page, and uses this as the primary risk information for the resource transfer recipient.

[0189] Furthermore, it is understandable that if the content on the wealth management product details page is empty, the RPA system clicks the link to the product manual for that wealth management product, downloads the manual PDF, and parses the PDF file to extract the required performance benchmark, risk level, and other information. For example, a computer device simulates logging into the personal customer marketing management system through the RPA system. Under the Personal Risk Assessment menu, the computer searches for the customer's risk assessment by name and number, and extracts information such as the customer's risk assessment level, assessment time, and assessment channel from the page.

[0190] S1204: Generate a dual-recording video based on the voice data and the image data, the identity information of the transferring parties, the first risk information, and the second risk information.

[0191] Specifically, the computer device searches all downloaded voice and image data based on the identity information of the transferor, the first risk information, and the second risk information to obtain the corresponding dual-recorded video. For example, the RPA system enters the identity information of the transferor and the recipient in the identity information search area of the downloaded video. The identity information search area can be used to enter the ID number, and the first risk information of the resource transfer recipient and the second risk information of the resource transfer in the risk information search area. Click the "Search" button to search and obtain the dual-recorded video corresponding to the resource to be tested.

[0192] In the above-mentioned resource transfer video quality detection method, the computer equipment obtains the voice data and picture data of the resource transfer process from the accounting image archive system, obtains the identity information of the transfer parties from the personnel information query system, and obtains the first risk information of the resource transfer recipient and the second risk information of the resource transfer from the personal customer marketing management system. Therefore, a dual-recording video can be generated based on the voice data and picture data, the identity information of the transfer parties, the first risk information and the second risk information. By obtaining relevant data from different systems to generate a dual-recording video, the information generated in the dual-recording video is richer.

[0193] It should be understood that, although the steps in the flowcharts of the above embodiments are shown in sequence as indicated by the arrows, these steps are not necessarily performed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be performed in other orders. Moreover, at least a portion of the steps in the flowcharts of the above embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily performed at the same time, but can be performed at different times. The execution order of these steps or stages is not necessarily to be performed in sequence, but can be performed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0194] Based on the same inventive concept, embodiments of the present application also provide a resource transfer video quality detection device for implementing the resource transfer video quality detection method described above. The solution provided by this device is similar to the solution described in the method described above. Therefore, the specific limitations in one or more of the following embodiments of the resource transfer video quality detection device can be found in the above-mentioned limitations of the resource transfer video quality detection method and will not be further elaborated here.

[0195] In one embodiment, Figure 13 As shown, a resource transfer video quality detection device is provided, comprising: a first acquisition module 11, a processing module 12, a second acquisition module 13 and a determination module 14, wherein:

[0196] The first acquisition module 11 is used to obtain a dual-recorded video of the resource to be detected during the resource transfer process; the dual-recorded video is generated by the Robotic Process Automation system (RPA) based on resource transfer data obtained from multiple business systems;

[0197] Processing module 12 is used to detect various transfer behavior elements in the resource transfer process and obtain compliance detection results of each transfer behavior element; the transfer behavior element is determined based on the identity information of the transfer parties of the resource to be detected and the voice information in the dual-recorded video;

[0198] The second acquisition module 13 is used to obtain the quality quantitative value of the resource transfer to be detected based on the compliance detection results of each transfer behavior element;

[0199] The determination module 14 is configured to determine that the quality of the dual-recorded video of the resource transfer to be detected is qualified when the quality quantization value is greater than a preset threshold.

[0200] The resource transfer video quality detection device provided in this embodiment can execute the above method embodiment. Its implementation principle and technical effects are similar and will not be described in detail here.

[0201] In one embodiment, the processing module includes: an extraction unit, a first processing unit, and a determination unit, wherein:

[0202] An extraction unit, used to extract the facial information of the two parties in the dual-recorded video;

[0203] The first processing unit is configured to analyze key information in the facial information of the two transferring parties to obtain facial information analysis results of the two transferring parties;

[0204] The determination unit is used to determine that the compliance detection results of the facial information of the transferring parties are qualified if the facial information analysis results of both parties are compliant.

[0205] The resource transfer video quality detection device provided in this embodiment can execute the above method embodiment. Its implementation principle and technical effects are similar and will not be described in detail here.

[0206] Optionally, the above-mentioned first processing unit is specifically used to obtain the facial descriptors of each of the transferring parties based on key information in the facial information of the transferring parties; obtain the Euclidean distance between the facial descriptors of each of the transferring parties and the corresponding standard facial descriptors in the local face library; when the Euclidean distances are both less than the preset distance, determine that the facial information analysis results of each of the transferring parties are compliant.

[0207] The resource transfer video quality detection device provided in this embodiment can execute the above method embodiment. Its implementation principle and technical effects are similar and will not be described in detail here.

[0208] In one embodiment, the processing module includes: a first acquisition unit, a second processing unit, a matching unit, and a first determination unit, wherein:

[0209] A first acquiring unit, configured to acquire voice information from the dual-recorded video;

[0210] The second processing unit is used to extract text information from the voice information to obtain text information corresponding to the voice information;

[0211] A matching unit, used to match text information with a preset resource transfer speech template;

[0212] The first determining unit is configured to determine that, if the match is successful, the compliance detection result of the voice information in the transfer process is that the voice detection is qualified.

[0213] The resource transfer video quality detection device provided in this embodiment can execute the above method embodiment. Its implementation principle and technical effects are similar and will not be described in detail here.

[0214] Optionally, the above-mentioned second processing unit is specifically used to input the voice information into a preset local acoustic model to obtain the pinyin information corresponding to the voice information; input the pinyin information into a preset local dictionary library to obtain the text corresponding to the pinyin information; input the text corresponding to the pinyin information into the preset local language model to obtain the text information corresponding to the voice information.

[0215] The resource transfer video quality detection device provided in this embodiment can execute the above method embodiment. Its implementation principle and technical effects are similar and will not be described in detail here.

[0216] In one embodiment, the processing module includes: a second acquisition unit, a third processing unit, and a second determination unit, wherein:

[0217] A second acquisition unit is used to acquire multiple frames of images in the dual-recorded video;

[0218] The third processing unit is used to detect the three-sided table information in the multiple frames of images through a preset convolutional neural network model to obtain a detection result;

[0219] The second determining unit is configured to determine that the compliance detection result of the three-dimensional table information during the transfer process is qualified when the detection result shows that the three-dimensional table information exists in the multiple frames of image.

[0220] The resource transfer video quality detection device provided in this embodiment can execute the above method embodiment. Its implementation principle and technical effects are similar and will not be described in detail here.

[0221] In one embodiment, the second acquisition module includes: a third acquisition unit, a fourth acquisition unit, and a third determination unit, wherein:

[0222] A third acquisition unit is used to acquire the weight of each transfer behavior element according to the influence degree of each transfer behavior element on the resource transfer process;

[0223] a fourth obtaining unit, configured to obtain a weighted sum of each transfer behavior element according to the weight of each transfer behavior element and the compliance detection result;

[0224] The third determining unit is configured to determine the weighted sum of the transfer behavior elements as the quality quantization value of the resource transfer to be detected.

[0225] The resource transfer video quality detection device provided in this embodiment can execute the above method embodiment. Its implementation principle and technical effects are similar and will not be described in detail here.

[0226] In one embodiment, the above module further includes: a first sending module, wherein:

[0227] The sending module is used to send the quality quantization value of the resource transfer to be detected to the RPA, so as to instruct the RPA to generate report information of the resource transfer process according to the quality quantization value.

[0228] The resource transfer video quality detection device provided in this embodiment can execute the above method embodiment. Its implementation principle and technical effects are similar and will not be described in detail here.

[0229] In one embodiment, the above module further includes: a second sending module, wherein:

[0230] The second sending module is used to send the detection details of each transfer behavior element in the resource transfer process to the RPA, so as to instruct the RPA to send the detection details to the resource transfer recipient.

[0231] The resource transfer video quality detection device provided in this embodiment can execute the above method embodiment. Its implementation principle and technical effects are similar and will not be described in detail here.

[0232] In one embodiment, Figure 14 As shown, a resource transfer video quality detection device is provided, comprising: a generating module 21 and a third sending module 22, wherein:

[0233] A generating module 21 is used to obtain resource transfer data of the resource to be detected during the resource transfer process from multiple business systems and generate a dual-recording video;

[0234] The third sending module 22 is used to send the dual-recording video to the dual-recording intelligent recognition platform, instructing the dual-recording intelligent recognition platform to detect various transfer behavior elements in the resource transfer process, obtain the compliance detection results of each transfer behavior element, and obtain the quality quantification value of the resource transfer to be detected based on the compliance detection results of each transfer behavior element. When the quality quantification value is greater than a preset threshold, it is determined that the quality of the dual-recording video of the resource transfer to be detected is qualified; wherein, the transfer behavior element is determined based on the identity information of the transfer parties of the resource to be detected and the voice information in the dual-recording video.

[0235] The resource transfer video quality detection device provided in this embodiment can execute the above method embodiment. Its implementation principle and technical effects are similar and will not be described in detail here.

[0236] Each module in the aforementioned resource transfer video quality detection device can be implemented in whole or in part through software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in the computer device in the form of software, so that the processor can call and execute the corresponding operations of each module.

[0237] In one embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor implements all the contents of the above method embodiments when executing the computer program.

[0238] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, all the contents of the above method embodiments are implemented.

[0239] In one embodiment, a computer program product is provided, including a computer program. When the computer program is executed by a processor, all the contents of the above method embodiments are implemented.

[0240] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0241] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, database or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may 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 may include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processor involved in the various embodiments provided herein may be, but are not limited to, a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic unit, a data processing logic unit based on quantum computing, and the like.

[0242] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0243] The above embodiments merely illustrate several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art may make various modifications and improvements without departing from the spirit of the present invention, all of which fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. A resource transfer video quality detection method, characterized in that: Applied to a dual-recording intelligent recognition platform, the method includes: Obtain dual-recorded video of the resource to be inspected during the resource transfer process; the dual-recorded video is generated by the Robotic Process Automation (RPA) system based on resource transfer data obtained from multiple business systems; Detecting various transfer behavior elements during the resource transfer process and obtaining compliance detection results for each transfer behavior element; the transfer behavior elements are determined based on the identity information of the transfer parties of the resource to be detected and the voice information in the dual-recorded video; the transfer behavior elements include: facial information of the transfer parties, voice information during the transfer process, and three-sided channel information; the three-sided channel information represents the identity information of the resource transfer recipient and historical resource transfer information; Obtaining a quality quantification value of the resource transfer to be detected based on compliance detection results of each transfer behavior element; If the quality quantization value is greater than a preset threshold, it is determined that the quality of the dual-recorded video of the resource transfer to be detected is qualified; The detecting of each transfer behavior element in the resource transfer process and obtaining compliance detection results of each transfer behavior element include: Extracting facial information of the two parties of the transfer from the dual-recorded video; Obtaining facial descriptors of the two transferring parties based on key information in the facial information of the two transferring parties; Obtaining the Euclidean distance between the facial descriptors of the two transfer parties and the corresponding standard facial descriptors in the local face database; If the Euclidean distances are both less than the preset distance, it is determined that the facial information analysis results of the transfer parties are both compliant; If the facial information analysis results of both the transferring parties are in compliance, the compliance detection results of the facial information of both the transferring parties are determined to be qualified; Acquiring voice information from the dual-recorded video; Inputting the voice information into a preset local acoustic model to obtain pinyin information corresponding to the voice information; Input the pinyin information into a preset local dictionary library to obtain the text corresponding to the pinyin information; Inputting the text corresponding to the pinyin information into a preset local language model to obtain text information corresponding to the voice information; Matching the text information with a preset resource transfer speech template; If the match is successful, the compliance detection result of the voice information in the transfer process is that the voice detection is qualified; Acquire multiple frames of images from the dual-recorded video; Detecting the three-dimensional table information in the multiple frames of images through a preset convolutional neural network model to obtain a detection result; If the detection result is that the multiple frames of images contain three-dimensional table information, it is determined that the compliance detection result of the three-dimensional table information in the transfer process is qualified.

2. The method according to claim 1, characterized in that The obtaining, based on the compliance detection results of the transfer behavior elements, a quality quantification value of the resource transfer to be detected includes: Obtaining a weight of each transfer behavior element according to the degree of influence of each transfer behavior element on the resource transfer process; The weighted sum of the transfer behavior elements is determined as the quality quantization value of the resource transfer to be detected.

3. The method according to claim 1, characterized in that The method further comprises: The quality quantization value of the resource transfer to be detected is sent to the RPA to instruct the RPA to generate report information of the resource transfer process according to the quality quantization value.

4. The method according to claim 1, wherein The method further comprises: Send detection details of various transfer behavior elements in the resource transfer process to the RPA to instruct the RPA to send the detection details to the resource transfer recipient.

5. A resource transfer video quality detection method, characterized in that: Applied to RPA, the method includes: Obtain resource transfer data of the resources to be inspected during the resource transfer process from multiple business systems and generate dual-recording videos; The dual-recording video is sent to the dual-recording intelligent recognition platform to instruct the dual-recording intelligent recognition platform to detect the various transfer behavior elements in the resource transfer process according to the resource transfer video quality detection method according to claim 1, obtain the compliance detection results of each transfer behavior element, and obtain the quality quantization value of the resource transfer to be detected based on the compliance detection results of each transfer behavior element, and determine that the quality of the dual-recording video of the resource transfer to be detected is qualified when the quality quantization value is greater than a preset threshold; wherein, the transfer behavior element is determined based on the identity information of the transfer parties of the resource to be detected and the voice information in the dual-recording video.

6. The method according to claim 5, characterized in that The multiple business systems include an accounting image archive system, a personnel information query system, and a personal customer marketing management system; The method of obtaining resource transfer data of the resource to be detected during the resource transfer process from multiple business systems and generating dual-recording videos includes: Acquiring voice data and image data of the resource transfer process from the accounting image archive system; Obtaining the identity information of the transfer parties from the personnel information query system; Acquire first risk information of the resource transfer recipient and second risk information of the resource transfer from the personal customer marketing management system; The dual-recording video is generated according to the voice data and the picture data, the identity information of the transfer parties, the first risk information and the second risk information.

7. A resource transfer video quality detection device, characterized in that: The device comprises: The first acquisition module is used to obtain a dual-recorded video of the resource to be detected during the resource transfer process; the dual-recorded video is generated by the Robotic Process Automation system (RPA) based on resource transfer data obtained from multiple business systems; a processing module, configured to detect various transfer behavior elements in the resource transfer process and obtain compliance detection results of the transfer behavior elements; the transfer behavior elements are determined based on the identity information of the transfer parties of the resource to be detected and the voice information in the dual-recorded video; A second acquisition module is configured to acquire a quality quantification value of the resource transfer to be detected based on the compliance detection result of each transfer behavior element; A determination module, configured to determine that the quality of the dual-recorded video of the resource transfer to be detected is qualified when the quality quantization value is greater than a preset threshold; The transfer behavior elements include: facial information of the transfer parties, voice information during the transfer process, and three-sided channel information; the three-sided channel information represents the identity information of the resource transfer recipient and historical resource transfer information; The processing module is specifically used to extract the facial information of the two transferring parties from the dual-recording video; obtain the facial descriptors of the two transferring parties according to the key information in the facial information of the two transferring parties; obtain the Euclidean distance between the facial descriptors of the two transferring parties and the corresponding standard facial descriptors in the local face database; if the Euclidean distances are both less than the preset distances, determine that the facial information analysis results of the two transferring parties are both compliant; if the facial information analysis results of the two transferring parties are both compliant, determine that the compliance detection results of the facial information of the two transferring parties are qualified; obtain the voice information in the dual-recording video; input the voice information into a preset local acoustic model, and obtain the voice information. The method comprises the following steps: obtaining pinyin information corresponding to the information; inputting the pinyin information into a preset local dictionary library to obtain the text corresponding to the pinyin information; inputting the text corresponding to the pinyin information into a preset local language model to obtain text information corresponding to the voice information; matching the text information with a preset resource transfer speech template; if the match is successful, the compliance detection result of the voice information in the transfer process is that the voice detection is qualified; obtaining multiple frames of images in the dual-recorded video; detecting the three-sided channel information in the multiple frames of images through a preset convolutional neural network model to obtain a detection result; if the detection result is that the three-sided channel information exists in the multiple frames of images, determining that the compliance detection result of the three-sided channel information in the transfer process is qualified.

8. A resource transfer video quality detection device, characterized in that: The device comprises: A generation module is used to obtain resource transfer data of the resource to be detected during the resource transfer process from multiple business systems and generate dual-recording videos; A sending module is used to send the dual-recording video to the dual-recording intelligent recognition platform, instructing the dual-recording intelligent recognition platform to detect the various transfer behavior elements in the resource transfer process according to the resource transfer video quality detection method according to claim 1, obtain the compliance detection results of each transfer behavior element, and obtain the quality quantization value of the resource transfer to be detected based on the compliance detection results of each transfer behavior element. When the quality quantization value is greater than a preset threshold, it is determined that the dual-recording video quality of the resource transfer to be detected is qualified; wherein, the transfer behavior element is determined based on the identity information of the transfer parties of the resource to be detected and the voice information in the dual-recording video.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

11. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images