A method, device, equipment and medium for anti-fraud identification based on video analysis

Identifying abnormal behaviors at the bidding site through video analysis technology has solved the problem that the existing technology cannot effectively identify the potential risk of bidding collusion in the bidder's personal behavior, and improved the fairness of the bidding process.

CN118865504BActive Publication Date: 2024-12-03NEW COMM INVESTMENT (CHENGDU) BIG DATA CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411345630.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-26
Publication Date
2024-12-03
Estimated Expiration
2044-09-26

AI Technical Summary

Technical Problem

The existing technology relies solely on computer and network data analysis during the bidding process, and cannot effectively identify the potential risk of collusion and abnormal behavior of bidders' personal behavior.

Method used

Using a video analysis method, the video stream at the bidding site is obtained for preprocessing and action sampling, the acceptable probability of representative frames is calculated, the abnormal representative frames are selected, and the probability of serial bid behavior is calculated through feature extraction and secondary verification. Judge abnormal behavior.

Benefits of technology

It improves the accuracy of judgment of abnormal behaviors during the bidding process, greatly improves the fairness of the market, and can effectively identify abnormal behaviors such as bid evaluation experts and bidders, bidders, and bidders, etc.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118865504B_ABST
    Figure CN118865504B_ABST
Patent Text Reader

Abstract

The present invention belongs to the technical field of bid-rigging identification, and discloses a bid-rigging identification method, device, equipment and medium based on video analysis, wherein the method comprises: obtaining a video stream of a bidding site, pre-processing the video stream of the bidding site to obtain a video image frame set; performing action sampling on the video image frame set to obtain a representative frame set; calculating the acceptable probability of each representative frame based on the video image frame set and a number of representative frames; screening out representative frames whose acceptable probability exceeds a preset probability from the representative frame set, and marking them as abnormal representative frames; performing feature extraction on the abnormal representative frames to obtain a binary image; calculating the secondary verification probability of the abnormal representative frame based on the binary image; determining the bid-rigging behavior probability of the abnormal representative frame based on the secondary verification probability of the abnormal representative frame and the acceptable probability corresponding to the abnormal representative frame. The present invention can improve the accuracy of abnormal judgment of bid-rigging behavior and improve the fairness of bidding.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of tag-matching identification, and specifically relates to an anti-tag-matching identification method, device, equipment and medium based on video analysis. Background Art

[0002] Bidding collusion refers to the act of bidding by two or more bidders or between a bidder and a tenderer through improper means during the bidding process in order to exclude other competitors, manipulate the bidding results, and damage the interests of the tenderer. This behavior violates the principle of fair competition and undermines the normal order of the bidding market.

[0003] In order to solve the above problems, collusion in bidding is avoided through electronic bidding, and judgment is made by analyzing relevant data and files on computers and networks. For example, in electronic bidding, collusion in bidding can be identified by identifying duplicate bidder IP addresses or Mac addresses, duplicate basic information of bidding suppliers, identical bidding documents, and abnormally consistent bidding quotation compositions.

[0004] Judgment is made only by analyzing relevant data and files on computers and networks, but no effective analysis is conducted on the potential risks of collusion and abnormal behavior caused by the bidders' personal behavior during the bidding process. Summary of the invention

[0005] The purpose of the present invention is to provide an anti-coordinated bidding identification method, device, equipment and medium based on video analysis, so as to solve the problem that the prior art only relies on the analysis and judgment of relevant data and files on the computer and network terminal, but does not effectively analyze the potential bid-rigging risks and abnormal behaviors caused by the bidders' personal behaviors during the bidding process.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] In a first aspect, the present invention provides an anti-fraud identification method based on video analysis, the method comprising:

[0008] Acquire a video stream of the bidding site, pre-process the video stream of the bidding site, and obtain a video image frame set;

[0009] Performing motion sampling on a video image frame set to obtain a representative frame set, wherein the representative frame set includes a plurality of representative frames, and the representative frames are used to represent the motion changes of the actors at the bidding site;

[0010] Based on the video image frame set and several representative frames, calculating the acceptability probability of each representative frame;

[0011] Select representative frames whose acceptable probability exceeds a preset probability from the representative frame set and mark them as abnormal representative frames;

[0012] Extract features from the abnormal representative frames to obtain a binary image;

[0013] Calculate the secondary verification probability of the abnormal representative frame based on the binary image;

[0014] Based on the secondary verification probability of the abnormal representative frame and the acceptable probability corresponding to the abnormal representative frame, the probability of the collusion behavior of the abnormal representative frame is determined.

[0015] Preferably, the calculation expression of the acceptable probability of each representative frame is:

[0016] = ;

[0017] In the formula, z represents the currently selected video sample, z* represents the sampled sample in the video sample z, represents the acceptable probability of each representative frame, represents the kth sample extracted from the video sample z, represents the representative frame of the sample extracted from the video sample z for the kth time, represents the representative frame of the sample z*, represents the sample extracted from the video for the kth time from the sampled sample z*, Indicated in the sample The representative frame appears below The probability of represents the probability of the representative frame of the sample z*; Indicated in the sample The representative frame appears below The probability of Indicated in the sample The representative frame appears below The probability of represents the probability of the representative frame of the sample extracted from the video sample z for the kth time; Indicates that the representative frame appears under the sample z* probability.

[0018] Preferably, selecting representative frames whose acceptable probability exceeds a preset probability from the representative frame set and marking them as abnormal representative frames includes:

[0019] Determine whether the acceptable probability of each representative frame in the representative frame set reaches a preset probability;

[0020] If yes, the representative frame is marked as an abnormal representative frame; if no, the representative frame is marked as a normal representative frame.

[0021] Preferably, feature extraction is performed on the abnormal representative frame to obtain a binary image, including:

[0022] The abnormal representative frame is converted into a grayscale space. In the grayscale space, any pixel point is selected as the object, the pixel point is taken as the central pixel point, and at least one neighborhood of the central pixel point is constructed, wherein the at least one neighborhood has a plurality of adjacent pixels adjacent to the central pixel point;

[0023] In at least one neighborhood, a pixel value of a central pixel is compared with pixel values ​​of each adjacent pixel to obtain comparison results of each adjacent pixel, wherein the comparison results include: the pixel value of the central pixel is equal to the pixel values ​​of each adjacent pixel or the pixel value of the central pixel is not equal to the pixel values ​​of each adjacent pixel;

[0024] Calculate the coordinates of each adjacent pixel in at least one neighborhood according to the coordinates of the central pixel;

[0025] Sort the comparison results of each adjacent pixel point according to the coordinates of each adjacent pixel point in at least one neighborhood to obtain a feature vector corresponding to the pixel point;

[0026] Update the selection object of the central pixel point, repeatedly execute the pixel point as the central pixel point, construct at least one neighborhood of the central pixel point, and there are several adjacent pixels adjacent to the central pixel point in the at least one neighborhood; in at least one neighborhood, compare the pixel value of the central pixel point with the pixel value of each adjacent pixel point to obtain the comparison result of each adjacent pixel point, and the comparison result includes: the pixel value of the central pixel point is equal to the pixel value of the adjacent pixel point or the pixel value of the central pixel point is not equal to the pixel value of the adjacent pixel point; calculate the coordinates of each adjacent pixel point in at least one neighborhood according to the coordinates of the central pixel point; sort the comparison results of each adjacent pixel point according to the coordinates of each adjacent pixel point in at least one neighborhood; obtain the feature vector corresponding to each pixel point of the abnormal representative frame;

[0027] The feature vectors corresponding to all pixels of the abnormal representative frame are combined to obtain a binary image.

[0028] Preferably, calculating the secondary verification probability of the abnormal representative frame based on the binary image includes:

[0029] Counting the total number of binary images whose comparison results show that the pixel value of the central pixel is not equal to the pixel value of the adjacent pixel points and the total number of binary images whose comparison results show that the pixel value of the central pixel is equal to the pixel value of the adjacent pixel points;

[0030] The secondary verification probability of the abnormal representative frame is calculated based on the total number of pixel values ​​of the central pixel point that are not equal to the pixel values ​​of the adjacent pixels and the total number of pixel values ​​of the central pixel point that are equal to the pixel values ​​of the adjacent pixels.

[0031] Preferably, when the pixel value of the central pixel is equal to the pixel value of the adjacent pixel, the comparison result of the adjacent pixel is assigned to "1"; when the pixel value of the central pixel is not equal to the pixel value of the adjacent pixel, the comparison result of the adjacent pixel is assigned to "0", the feature vector is a binary string composed of "0" and / or "1", and the binary image includes multiple binary strings.

[0032] Preferably, the calculation expression of the probability of the string tagging behavior of the abnormal representative frame is:

[0033] ;

[0034] In the formula, is the probability of string labeling behavior of the abnormal representative frame, is the acceptable probability of the abnormal representative frame, z represents the currently selected video sample, z* represents the sampled sample in the video sample z, is the secondary verification probability of the abnormal representative frame, where is the length of "0" in the binary string, is the total length of the binary string.

[0035] In a second aspect, the present invention provides an anti-trapped label identification device based on video analysis, the device is used to implement the above-mentioned anti-trapped label identification method based on video analysis, and the device includes:

[0036] A video acquisition module is used to acquire a video stream of the bidding site, pre-process the video stream of the bidding site, and obtain a video image frame set;

[0037] An image sampling module is used to perform motion sampling on a video image frame set to obtain a representative frame set, wherein the representative frame set has a plurality of representative frames, and the representative frames are used to represent the motion changes of the actors at the bidding site;

[0038] A probability calculation module, used for calculating the acceptability probability of each representative frame based on the video image frame set and a plurality of representative frames;

[0039] An abnormal screening module is used to screen out representative frames whose acceptable probability exceeds a preset probability from the representative frame set and mark them as abnormal representative frames;

[0040] A feature extraction module is used to extract features from abnormal representative frames to obtain a binary image;

[0041] An anomaly calculation module, used to calculate the secondary verification probability of the anomaly representative frame based on the binary image;

[0042] The secondary verification module is used to determine the probability of the colluded labeling behavior of the abnormal representative frame based on the secondary verification probability of the abnormal representative frame and the acceptable probability corresponding to the abnormal representative frame.

[0043] In a third aspect, the present invention provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned video analysis-based anti-fraud identification method when executing the computer program.

[0044] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned video analysis-based anti-fraud identification method.

[0045] Beneficial effects:

[0046] 1. The present invention collects video streams of the bidding site, and then performs action sampling from the video image frames of the video stream to obtain representative frames. The representative frames are used to characterize the action changes of the actors at the bidding site, and then the acceptable probability of each representative frame is calculated. The acceptable probability can reflect the probability of whether the action of the actor is an abnormal behavior. The larger the value of the acceptable probability is, the greater the probability of abnormal behavior is. When the value of the acceptable probability is large, the representative frame is marked as an abnormal representative frame, and a secondary verification is performed again. The probability of colluded bidding behavior of the abnormal representative frame is calculated based on the secondary verification probability of the abnormal representative frame and the acceptable probability corresponding to the abnormal representative frame. The probability of colluded bidding behavior is used as the final abnormal behavior judgment standard, which can improve the accuracy of abnormal behavior judgment.

[0047] 2. The present invention can determine whether there are abnormal behaviors between bid evaluation experts and bidders, between bid evaluation experts, between bidders, individual bidders, and individual bid evaluation experts through the probability of collusion in bidding, which greatly improves the fairness of bidding. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the following specific implementations, they are used to explain the embodiments of the present invention, but do not constitute a limitation on the embodiments of the present invention. In the accompanying drawings:

[0049] Figure 1 It is a flow chart of an anti-fraud identification method based on video analysis provided by an embodiment of the present invention;

[0050] Figure 2 It is a block diagram of an anti-fraud identification device based on video analysis provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0051] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the present invention will be briefly introduced below in combination with the drawings and the description of the embodiments or the prior art. Obviously, the following description of the structure of the drawings is only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. It should be noted that the description of these embodiments is used to help understand the present invention, but does not constitute a limitation of the present invention.

[0052] Embodiment 1

[0053] Figure 1 It is a flow chart of an anti-bid-rigging identification method based on video analysis provided by an embodiment of the present invention. The application scenarios of the anti-bid-rigging identification method based on video analysis of this embodiment include: multiple cameras and at least one monitoring terminal, wherein the monitoring terminal can be a PC, and the multiple cameras are communicatively connected to the PC; the multiple cameras are deployed at the bidding site to collect video data of the behavior of the actors; the actors include: bidders and bid evaluation experts, and the behavior includes: behavior between bid evaluation experts and bidders, behavior between bid evaluation experts, behavior between bidders, personal behavior of bidders, and personal behavior of bid evaluation experts.

[0054] Based on the above application scenarios, this embodiment provides an anti-fraud identification method based on video analysis, such as Figure 1 As shown, the method includes:

[0055] Step S10: Acquire the video stream of the bidding site, pre-process the video stream of the bidding site, and obtain a video image frame set.

[0056] In this embodiment, each camera can collect video of the bidding site in real time, and the monitoring terminal can obtain the video stream of the bidding site from the camera. After obtaining the video stream and other data, the video stream needs to be decoded and converted into video frames. Common encoding formats include H.264, H.265, VP8, VP9, ​​etc.

[0057] After obtaining the video frame, it is necessary to preprocess the video frame, for example:

[0058] Scaling: resizing a video frame to accommodate processing needs or the target display device;

[0059] Cropping: removing unnecessary parts of the video frame, such as black edges;

[0060] Frame rate conversion: adjust the playback speed of video frames;

[0061] Color space conversion: such as converting from YUV to RGB for subsequent processing;

[0062] Noise reduction: reduce the noise of video frames and improve video quality;

[0063] Sharpening: Enhance the edges of the video frame to make the image clearer;

[0064] Lighting Correction: Adjust the lighting conditions of video frames to improve visual effects.

[0065] After completing the preprocessing of the video frames, a number of video image frames can be obtained, and all the video image frames are taken as a set to obtain a video image frame set.

[0066] Step S20: performing motion sampling on the video image frame set to obtain a representative frame set, wherein the representative frame set has a plurality of representative frames, and the representative frames are used to represent the motion changes of the actors at the bidding site.

[0067] In this embodiment, sampling is performed at specified time intervals in a video image frame set to collect video image frames in which the actors in the video image frame set have motion changes. The video image frames in which the actors have motion changes are taken as representative frames, and all representative frames are taken as a set to obtain a representative frame set.

[0068] Step S30: Based on the video image frame set and a number of representative frames, the acceptability probability of each representative frame is calculated.

[0069] In this embodiment, the calculation expression of the acceptable probability of each representative frame is:

[0070] = ;

[0071] Wherein, z represents the currently selected video sample, which can be the entire video image frame set or a portion of the video image frames selected from the video image frame set, and z* represents the sampled sample in the video sample z. represents the acceptable probability of each representative frame, represents the kth sample extracted from the video sample z, represents the representative frame of the sample extracted from the video sample z for the kth time, represents the representative frame of the sample z*, represents the sample extracted from the video for the kth time from the sampled sample z*, Indicated in the sample The representative frame appears below probability; represents the probability of the representative frame of the sample z*; Indicated in the sample The representative frame appears below probability; Indicated in the sample The representative frame appears below probability; represents the probability of the representative frame of the sample extracted from the video sample z for the kth time; Indicates that the representative frame appears under the sample z* probability.

[0072] Step S40: Select representative frames whose acceptable probability exceeds a preset probability from the representative frame set and mark them as abnormal representative frames.

[0073] Specifically, representative frames whose acceptable probability exceeds a preset probability are selected from the representative frame set and marked as abnormal representative frames, including:

[0074] Step S401: determining whether the acceptable probability of each representative frame in the representative frame set reaches a preset probability;

[0075] Step S401: If yes, mark the representative frame as an abnormal representative frame; if no, mark the representative frame as a normal representative frame.

[0076] In this embodiment, the acceptable probability of each representative frame is used to characterize the probability of abnormal behavior in each representative frame. The higher the acceptable probability, the higher the probability of abnormal behavior in the representative frame. The preset probability value of this embodiment is 1. When the acceptable probability of any representative frame is greater than or equal to 1, it means that the actor is very likely to have abnormal behavior, and the representative frame is marked as an abnormal representative frame. The abnormal representative frame is used for subsequent abnormality analysis. When the acceptable probability of any representative frame is less than 1, it means that the actor's behavior is normal, and the possibility of the actor having abnormal behavior is very low. At this time, this part of the representative frames is marked as normal representative frames.

[0077] This embodiment screens out representative frames that may have abnormal behaviors first, and then performs secondary verification and analysis only on the abnormal representative frames, thereby improving the efficiency and accuracy of abnormal behavior judgment.

[0078] Step S50: extracting features from the abnormal representative frame to obtain a binary image.

[0079] Specifically, feature extraction is performed on the abnormal representative frame to obtain a binary image, including:

[0080] Step S501: convert the abnormal representative frame into a grayscale space. In the grayscale space, take any pixel point as the selected object, take the pixel point as the central pixel point, and construct at least one neighborhood of the central pixel point. There are several adjacent pixel points adjacent to the central pixel point in the at least one neighborhood.

[0081] Step S502: In at least one neighborhood, the pixel value of the central pixel is compared with the pixel values ​​of each adjacent pixel to obtain comparison results of each adjacent pixel, wherein the comparison results include: the pixel value of the central pixel is equal to the pixel values ​​of each adjacent pixel, or the pixel value of the central pixel is not equal to the pixel values ​​of each adjacent pixel.

[0082] Step S503: Calculate the coordinates of each adjacent pixel in at least one neighborhood according to the coordinates of the central pixel;

[0083] The function expression of the coordinates of each adjacent pixel point in at least one neighborhood is:

[0084] ;

[0085] In the formula, gc represents the pixel value of the center pixel, ranging from 0 to 255, where 0 represents black and 255 represents white. x is the pixel value of the center pixel in the x direction of the grayscale space, gc y is the pixel value of the central pixel in the y direction of the grayscale space, p is the pth neighborhood, and n is the number of sampling points (the number of adjacent pixels) on the circle with a radius of R (circular neighborhood).

[0086] Step S504: sorting the comparison results of each adjacent pixel point according to the coordinates of each adjacent pixel point in at least one neighborhood to obtain a feature vector corresponding to the pixel point.

[0087] Step S505: Update the selection object of the central pixel point, and repeat steps S501 to S504 to obtain the feature vector corresponding to each pixel point of the abnormal representative frame.

[0088] Step S506: Combine the feature vectors corresponding to all the pixels of the abnormal representative frame to obtain a binary image.

[0089] In this embodiment, when the pixel value of the central pixel is equal to the pixel value of the adjacent pixel, the comparison result of the adjacent pixel is assigned to "1", and when the pixel value of the central pixel is not equal to the pixel value of the adjacent pixel, the comparison result of the adjacent pixel is assigned to "0". At this time, the feature vector is a one-dimensional vector (i.e., a binary string) composed of "0" and / or "1". In this embodiment, the grayscale value of a pixel is compared with the grayscale value of the adjacent pixel in its neighborhood. If the adjacent pixel in the neighborhood has a similar grayscale value to the pixel, it can be marked as "1", otherwise it is marked as "0".

[0090] In this embodiment, since the neighborhood is circular, the comparison results of each adjacent pixel point can be sorted in a counterclockwise manner to obtain a binary string. A binary string is used as a feature vector, and a feature vector is the local texture information in the area. By combining multiple feature vectors (multiple binary strings), a binary image can be obtained. The binary image has richer image features, which helps to improve recognition accuracy and robustness.

[0091] Step S60: Calculate the secondary verification probability of the abnormal representative frame based on the binary image.

[0092] Specifically, the secondary verification probability of the abnormal representative frame is calculated based on the binary image, including:

[0093] Step S601: Counting the total number of pixel values ​​of the central pixel and the adjacent pixel values ​​that are not equal to each other in the binary image and the total number of pixel values ​​of the central pixel and the adjacent pixel values ​​that are equal to each other in the binary image;

[0094] Step S602: Calculate the secondary verification probability of the abnormal representative frame based on the total number of pixel values ​​of the central pixel point that are unequal to the pixel values ​​of the adjacent pixels and the total number of pixel values ​​of the central pixel point that are equal to the pixel values ​​of the adjacent pixels.

[0095] In this embodiment, when the pixel value of the central pixel is equal to the pixel value of the adjacent pixel, the comparison result of the adjacent pixel is assigned to "1", and when the pixel value of the central pixel is not equal to the pixel value of the adjacent pixel, the comparison result of the adjacent pixel is assigned to "0"; the number of "0"s is screened out in the same feature vector (that is, the same binary string), and the secondary verification probability of the abnormal representative frame can be calculated according to the ratio of the number of "0"s in a binary string to the total number of "0"s and "1"s in the binary string.

[0096] Step S70: Based on the secondary verification probability of the abnormal representative frame and the acceptable probability corresponding to the abnormal representative frame, determine the probability of the colluded bidding behavior of the abnormal representative frame.

[0097] In this embodiment, ;

[0098] In the formula, is the probability of string labeling behavior of the abnormal representative frame, is the acceptable probability of the abnormal representative frame, is the secondary verification probability of the abnormal representative frame, where is the length of "0" in the binary string (the total number of "0"s), is the total length of the binary string (the total number of "0"s and "1"s).

[0099] In this embodiment, the probability of the tampering behavior is 0.8 as the reference probability threshold. When the probability of the tampering behavior is greater than 0.8, in this embodiment, when a certain abnormal representative frame has at least one binary string, the secondary verification probability Make When the value of is greater than 0.8, it means that the anomaly represents the behavior abnormality of the frame with string labeling. When the probability of string labeling behavior is less than or equal to 0.8, it means that the anomaly represents the behavior abnormality of the frame without string labeling.

[0100] In this embodiment, the abnormal behaviors of bid-rigging include the following behaviors:

[0101] 1. Frequent or unnatural blinking by bidders during online bidding may be an attempt to attract attention or send signals;

[0102] 2. Bidding experts may blink frequently or unnaturally during online bidding to attract attention or send signals;

[0103] 3. The bidder opens his mouth unnaturally during online bidding, probably to simulate the action of speaking, but no sound is actually produced;

[0104] 4. The bidding experts opened their mouths unnaturally during online bidding, probably to simulate the action of speaking, but actually no sound was made;

[0105] 5. Bidders frequently nod their heads up and down during online bidding, which may be to imitate affirmative actions, but does not actually indicate agreement;

[0106] 6. Bidding experts frequently nod their heads up and down during online bidding, which may be to imitate affirmative actions, but it does not actually mean agreement;

[0107] 7. Bidders frequently shake their heads left and right during online bidding, which may be to imitate a negative action, but it does not actually express opposition;

[0108] 8. Bidding experts frequently shake their heads left and right during online bidding, which may be to imitate negative movements, but it does not actually mean opposition;

[0109] 9. When a bidder raises his arm during online bidding without having to raise his hand, it may be to convey instructions or information;

[0110] 10. When a bidding expert raises his arm during online bidding without having to raise his hand, it may be to convey instructions or information;

[0111] 11. Bidders pretend to eat or drink during online bidding, but do not actually consume any food or liquid;

[0112] 12. Bidders pretend to eat or drink during online bidding, but do not actually consume any food or liquid;

[0113] 13. Bidders pretend to smoke in online bidding but do not actually light up the cigarette;

[0114] 14. Bidding experts pretend to smoke during online bidding, but do not actually light up the cigarette;

[0115] 15. The bidder holds the mobile phone close to his ear during online bidding but does not make a call;

[0116] 16. The bidding expert put the mobile phone close to his ear during online bidding, but did not make a call;

[0117] 17. The bidder pretends to operate the mobile phone in the online bidding, but does not actually interact with the mobile phone;

[0118] 18. Bidding experts pretend to operate mobile phones during online bidding, but do not actually interact with the mobile phones.

[0119] The present invention collects the video stream of the bidding site, and then performs action sampling from the video image frame set of the video stream to obtain representative frames. The representative frames are used to characterize the action changes of the actors at the bidding site, and then the acceptable probability of each representative frame is calculated. The acceptable probability can reflect the probability of whether the action generated by the actor is an abnormal behavior. When the value of the acceptable probability is large, the representative frame is marked as an abnormal representative frame, and a secondary verification is performed again. The probability of collusion in bidding of the abnormal representative frame is calculated according to the secondary verification probability of the abnormal representative frame and the acceptable probability corresponding to the abnormal representative frame. The probability of collusion in bidding is used as the final abnormal behavior judgment standard, which can improve the accuracy of abnormal behavior judgment. Secondly, through the probability of collusion in bidding, it can be judged whether there are abnormal behaviors between evaluation experts and bidders, between evaluation experts, between bidders, individual bidders, individual evaluation experts, etc., which greatly improves the fairness of bidding.

[0120] Embodiment 2

[0121] Figure 2 is a block diagram of an anti-fraud identification device based on video analysis provided by an embodiment of the present invention, such as Figure 2 As shown, this embodiment provides an anti-staggered label recognition device based on video analysis, and the device is used to implement the anti-staggered label recognition method based on video analysis of embodiment 1, and the device includes:

[0122] A video acquisition module is used to acquire a video stream of the bidding site, pre-process the video stream of the bidding site, and obtain a video image frame set;

[0123] An image sampling module is used to perform motion sampling on a video image frame set to obtain a representative frame set, wherein the representative frame set has a plurality of representative frames, and the representative frames are used to represent the motion changes of the actors at the bidding site;

[0124] A probability calculation module, used for calculating the acceptability probability of each representative frame based on the video image frame set and a plurality of representative frames;

[0125] An abnormal screening module is used to screen out representative frames whose acceptable probability exceeds a preset probability from the representative frame set and mark them as abnormal representative frames;

[0126] A feature extraction module is used to extract features from abnormal representative frames to obtain a binary image;

[0127] An anomaly calculation module, used to calculate the secondary verification probability of the anomaly representative frame based on the binary image;

[0128] The secondary verification module is used to determine the probability of the colluded labeling behavior of the abnormal representative frame based on the secondary verification probability of the abnormal representative frame and the acceptable probability corresponding to the abnormal representative frame.

[0129] This embodiment also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the anti-fraud identification method based on video analysis of the first embodiment when executing the computer program.

[0130] This embodiment further provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the anti-fraud identification method based on video analysis of the first embodiment is implemented.

[0131] The present invention collects the video stream of the bidding site, and then performs action sampling from the video image frame set of the video stream to obtain representative frames. The representative frames are used to characterize the action changes of the actors at the bidding site, and then the acceptable probability of each representative frame is calculated. The acceptable probability can reflect the probability of whether the action generated by the actor is an abnormal behavior. When the value of the acceptable probability is larger, the probability of abnormal behavior is larger. When the value of the acceptable probability is larger, the representative frame is marked as an abnormal representative frame, and a secondary verification is performed again. The probability of collusion in bidding of the abnormal representative frame is calculated according to the secondary verification probability of the abnormal representative frame and the acceptable probability corresponding to the abnormal representative frame. The probability of collusion in bidding is used as the final abnormal behavior judgment standard, which can improve the accuracy of abnormal behavior judgment. Secondly, through the probability of collusion in bidding, it can be judged whether there are abnormal behaviors between evaluation experts and bidders, between evaluation experts, between bidders, individual bidders, and individual evaluation experts, thereby greatly improving the fairness of bidding.

[0132] It should be understood by those skilled in the art that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0133] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0134] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A method for anti-fraud identification based on video analysis, characterized in that: The method comprises: Acquire a video stream of the bidding site, pre-process the video stream of the bidding site, and obtain a video image frame set; Performing motion sampling on a video image frame set to obtain a representative frame set, wherein the representative frame set includes a plurality of representative frames, and the representative frames are used to represent the motion changes of the actors at the bidding site; Based on the video image frame set and several representative frames, the acceptable probability of each representative frame is calculated; the calculation expression of the acceptable probability of each representative frame is: = ; In the formula, z represents the currently selected video sample, z* represents the sampled sample in the video sample z, represents the acceptable probability of each representative frame, represents the kth sample extracted from the video sample z, represents the representative frame of the sample extracted from the video sample z for the kth time, represents the representative frame of the sample z*, represents the sample extracted from the video for the kth time from the sampled sample z*, Indicated in the sample The representative frame appears below probability; represents the probability of the representative frame of the sample z*; Indicated in the sample The representative frame appears below probability; Indicated in the sample The representative frame appears below probability; represents the probability of the representative frame of the sample extracted from the video sample z for the kth time; Indicates that the representative frame appears under the sample z* probability; Select representative frames whose acceptable probability exceeds a preset probability from the representative frame set and mark them as abnormal representative frames; Feature extraction is performed on the abnormal representative frame to obtain a binary image, including: The abnormal representative frame is converted into a grayscale space. In the grayscale space, any pixel point is selected as the object, the pixel point is taken as the central pixel point, and at least one neighborhood of the central pixel point is constructed, wherein the at least one neighborhood has a plurality of adjacent pixels adjacent to the central pixel point; In at least one neighborhood, a pixel value of a central pixel is compared with pixel values ​​of each adjacent pixel to obtain comparison results of each adjacent pixel, wherein the comparison results include: the pixel value of the central pixel is equal to the pixel values ​​of each adjacent pixel or the pixel value of the central pixel is not equal to the pixel values ​​of each adjacent pixel; Calculate the coordinates of each adjacent pixel in at least one neighborhood according to the coordinates of the central pixel; Sort the comparison results of each adjacent pixel point according to the coordinates of each adjacent pixel point in at least one neighborhood to obtain a feature vector corresponding to the pixel point; Update the selection object of the central pixel point, repeatedly execute the pixel point as the central pixel point, construct at least one neighborhood of the central pixel point, and there are several adjacent pixels adjacent to the central pixel point in the at least one neighborhood; in at least one neighborhood, compare the pixel value of the central pixel point with the pixel value of each adjacent pixel point to obtain the comparison result of each adjacent pixel point, and the comparison result includes: the pixel value of the central pixel point is equal to the pixel value of the adjacent pixel point or the pixel value of the central pixel point is not equal to the pixel value of the adjacent pixel point; calculate the coordinates of each adjacent pixel point in at least one neighborhood according to the coordinates of the central pixel point; sort the comparison results of each adjacent pixel point according to the coordinates of each adjacent pixel point in at least one neighborhood; obtain the feature vector corresponding to each pixel point of the abnormal representative frame; Combine the feature vectors corresponding to all pixels of the abnormal representative frame to obtain a binary image; The secondary verification probability of the abnormal representative frame is calculated based on the binary image, including: Counting the total number of binary images whose comparison results show that the pixel value of the central pixel is not equal to the pixel value of the adjacent pixel points and the total number of binary images whose comparison results show that the pixel value of the central pixel is equal to the pixel value of the adjacent pixel points; Calculate the secondary verification probability of the abnormal representative frame according to the total number of pixel values ​​of the central pixel point that are unequal to the pixel values ​​of the adjacent pixels and the total number of pixel values ​​of the central pixel point that are equal to the pixel values ​​of the adjacent pixels; Based on the secondary verification probability of the abnormal representative frame and the acceptable probability corresponding to the abnormal representative frame, the probability of the collusion behavior of the abnormal representative frame is determined.

2. The method for anti-fraud identification based on video analysis according to claim 1 is characterized in that: Filter out representative frames whose acceptable probability exceeds the preset probability from the representative frame set and mark them as abnormal representative frames, including: Determine whether the acceptable probability of each representative frame in the representative frame set reaches a preset probability; If yes, the representative frame is marked as an abnormal representative frame; if no, the representative frame is marked as a normal representative frame.

3. The method for anti-fraud identification based on video analysis according to claim 1 is characterized in that: When the pixel value of the central pixel is equal to the pixel value of the adjacent pixel, the comparison result of the adjacent pixel is assigned to "1", and when the pixel value of the central pixel is not equal to the pixel value of the adjacent pixel, the comparison result of the adjacent pixel is assigned to "0". The feature vector is a binary string composed of "0" and / or "1", and the binary image includes multiple binary strings.

4. The method for anti-fraud identification based on video analysis according to claim 3 is characterized in that: The calculation expression of the probability of the string tagging behavior of the abnormal representative frame is: ; In the formula, is the probability of string labeling behavior of the abnormal representative frame, is the acceptable probability of the abnormal representative frame, z represents the currently selected video sample, z* represents the sampled sample in the video sample z, is the secondary verification probability of the abnormal representative frame, where is the length of "0" in the binary string, is the total length of the binary string.

5. A device for preventing label fraud based on video analysis, the device being used to implement the method for preventing label fraud based on video analysis as claimed in any one of claims 1 to 4, characterized in that: The device comprises: A video acquisition module is used to acquire a video stream of the bidding site, pre-process the video stream of the bidding site, and obtain a video image frame set; An image sampling module is used to perform motion sampling on a video image frame set to obtain a representative frame set, wherein the representative frame set has a plurality of representative frames, and the representative frames are used to represent the motion changes of the actors at the bidding site; A probability calculation module, used for calculating the acceptability probability of each representative frame based on the video image frame set and a plurality of representative frames; An abnormal screening module is used to screen out representative frames whose acceptable probability exceeds a preset probability from the representative frame set and mark them as abnormal representative frames; A feature extraction module is used to extract features from abnormal representative frames to obtain a binary image; An anomaly calculation module, used to calculate the secondary verification probability of the anomaly representative frame based on the binary image; The secondary verification module is used to determine the probability of the colluded labeling behavior of the abnormal representative frame based on the secondary verification probability of the abnormal representative frame and the acceptable probability corresponding to the abnormal representative frame.

6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the anti-fraud identification method based on video analysis described in any one of claims 1 to 4 is implemented.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the anti-fraud identification method based on video analysis described in any one of claims 1 to 4 is implemented.