Gambling detection behavior detection method, device and system for chess and card room scene, and storage medium

Abstract

The invention is suitable for the technical field of intelligent monitoring, and provides a gambling detection behavior detection method, device, storage medium and system for a chess and card room scene, and the method comprises the steps: obtaining a monitoring image, and obtaining a personnel target and a desktop target; obtaining ROI regions of the personnel and the desktop; marking the currency target by using a currency marking frame with a rotation angle; and performing secondary verification on the currency target in the currency standard frame, and generating report information when a comparison result meets a report condition. According to the method, gambling behaviors can be accurately and efficiently detected, and the response capability to events is enhanced through a real-time reporting mechanism. Compared with existing gambling detection, the method has the advantages that the detection accuracy and the false alarm probability are remarkably improved, currency detection is only carried out on key areas, consumption of calculation power is reduced, efficient automatic processing is achieved, important information can be fed back in time, and the method is suitable for being applied to real-time monitoring scenes of chess and card rooms.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent monitoring, and particularly relates to a method, device, storage medium and system for detecting gambling behavior in a chess and card room scenario. Background Technique

[0002] In recent years, with the increasing development of artificial intelligence, intelligent video surveillance has been applied in more and more industrial fields, such as unmanned supermarkets, unmanned chess and card rooms, smart stores, security patrol surveillance, etc.

[0003] In particular, with the development of the unmanned chess and card room industry, how to carry out standardized management is particularly important. Using intelligent monitoring means to analyze the data of the monitoring video to determine whether there is illegal gambling behavior in the monitoring video has received more and more attention. In addition, how to accurately and quickly analyze the content of the monitoring video has become a recent research hotspot.

[0004] Current detection algorithms usually adopt a scheme based on deep learning object detection, that is, input the image information collected by video surveillance equipment into some object detection algorithms, such as SSD, retinaNet, Yolo series, etc., to detect whether there is a specified object in the image, such as RMB or other currencies, and judge whether there is gambling behavior according to whether there is a specified object in the image.

[0005] However, since the camera is generally installed at a high place, and RMB or other currencies are generally relatively small, the proportion of these RMB or currencies in the picture will be very small at this time. In computer vision, small object detection is a challenging task because small objects occupy fewer pixels in the image, so directly detecting currencies using the detection scheme will cause more missed detections. In addition, due to the lack of details, directly adopting the detection scheme will lead to many false detections. Therefore, there is a need to further improve the existing detection algorithms. Summary of the Invention

[0006] The purpose of the embodiments of the present application is to provide a method for detecting gambling behavior in a chess and card room scenario, aiming to solve the problem of high false detection and missed detection rates in the current video detection scheme for whether there is gambling behavior in an unmanned chess and card room.

[0007] The embodiments of the present application are implemented as follows. A method for detecting gambling behavior in a chess and card room scenario is provided, and the method includes:

[0008] Obtain a monitoring image, obtain a person target and a table target from the monitoring image, and generate a rectangular annotation box for each person target and table target;

[0009] Based on the coordinate and dimension information of the rectangular annotation boxes for the personnel target and the desktop target, obtain the ROI regions of the personnel and the desktop;

[0010] Obtain the ROI region image, perform currency target detection on the ROI region image to obtain currency targets; use currency annotation boxes with rotation angles to mark the currency targets;

[0011] Perform secondary verification on the currency targets within the currency standard boxes, compare the verification results of the secondary verification with the reporting threshold, and generate a reporting message when the comparison result meets the reporting conditions.

[0012] Another object of the embodiments of the present application is to provide a gambling behavior detection device for a chess and card room scene. The gambling behavior detection device for the chess and card room scene includes:

[0013] A rectangular annotation box generation module, configured to obtain a monitoring image, obtain personnel targets and desktop targets from the monitoring image, and generate rectangular annotation boxes for each of the personnel targets and desktop targets;

[0014] A ROI region acquisition module for personnel and desktop, configured to obtain the ROI regions of the personnel and the desktop based on the coordinate and dimension information of the rectangular annotation boxes for the personnel target and the desktop target;

[0015] A currency target marking module, configured to obtain the ROI region image, perform currency target detection on the ROI region image to obtain currency targets; use currency annotation boxes with rotation angles to mark the currency targets;

[0016] A secondary verification module, configured to perform secondary verification on the currency targets within the currency standard boxes, compare the verification results of the secondary verification with the reporting threshold, and generate a reporting message when the comparison result meets the reporting conditions.

[0017] Another object of the embodiments of the present application is to provide a computer-readable storage medium. A computer program is stored in the computer-readable storage medium. When the computer program is executed by a processor, the processor is caused to execute the steps of the gambling behavior detection method for the chess and card room scene as described above.

[0018] Another object of the embodiments of the present application is to provide a gambling behavior detection system for a chess and card room scene, including a memory and a processor. A computer program is stored in the memory. When the computer program is executed by the processor, the processor is caused to execute the steps of the gambling behavior detection method for the chess and card room scene as described above.

[0019] A method for detecting gambling behavior in a chess and card room scenario provided by an embodiment of the present application has prominent advantages. Through a multi-level scheme of target detection, ROI region extraction, currency target recognition, and secondary verification carried out successively, the present application can accurately and efficiently detect gambling behavior, and enhances the response ability to events through a real-time reporting mechanism. Compared with existing gambling detection, it significantly improves the detection accuracy and false alarm probability, and only performs currency detection on key areas, reducing the consumption of computing power, achieving efficient automated processing, and being able to timely feedback important information, which is suitable for application in the real-time monitoring scenario of a chess and card room. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 FIG. is an application environment diagram of a method for detecting gambling behavior in a chess and card room scenario provided by an embodiment of the present application;

[0021] Figure 2 FIG. is a flowchart of a method for detecting gambling behavior in a chess and card room scenario provided by an embodiment of the present application;

[0022] Figure 3 FIG. is a structural diagram of a convolutional neural network provided by an embodiment of the present application;

[0023] Figure 4 FIG. is a schematic diagram of the effect of generating a rectangular annotation box provided by an embodiment of the present application;

[0024] Figure 5 FIG. is a structural diagram of a currency detection convolutional network provided by an embodiment of the present application;

[0025] Figure 6 FIG. is a schematic diagram of the effect of generating a currency annotation box provided by an embodiment of the present application;

[0026] Figure 7 FIG. is a structural block diagram of a device for detecting gambling behavior in a chess and card room scenario provided by an embodiment of the present application;

[0027] Figure 8 FIG. is an internal structural block diagram of a computer device in an embodiment. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0028] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0029] It can be understood that the terms "first", "second", etc. used in this application may be used herein to describe various elements, but unless otherwise specified, these elements are not limited by these terms. These terms are only used to distinguish a first unit or module from another unit or module. For example, without departing from the scope of this application, the first script may be referred to as the second script, and similarly, the second script may be referred to as the first script.

[0030] Figure 1 The figure is an application environment diagram of the gambling behavior detection method for the chess and card room scenario provided by the embodiments of this application. As Figure 1 shown, in this application environment, it includes a monitoring device 110 and a computer device 120.

[0031] The computer device 120 can be an independent physical server or terminal, or a server cluster composed of multiple physical servers. It can be a cloud server, a tablet computer, a laptop computer, a desktop computer, etc. that provide basic cloud computing services such as cloud servers, cloud databases, cloud storage, and CDN.

[0032] The monitoring device 110 can be various types of monitoring cameras. The monitoring device 110 and the computer device 120 can be connected through a network, and this application does not limit this here.

[0033] As Figure 2 shown, in one embodiment, a gambling behavior detection method for the chess and card room scenario is proposed. This embodiment mainly takes the application of this method to the computer device 120 in the above Figure 1 as an example. A gambling behavior detection method for the chess and card room scenario may specifically include the following steps:

[0034] Step S10, obtain a monitoring image, obtain human targets and desktop targets from the monitoring image, and generate a rectangular annotation box for each human target and desktop target.

[0035] In this embodiment, first, a real-time or recorded monitoring image is obtained, aiming to capture the human and desktop targets in the gambling scenario. The human and desktop targets are automatically identified through the object detection algorithm in the image, and a rectangular annotation box is generated for each target.

[0036] Step S20, based on the coordinate and size information of the rectangular annotation boxes of the human targets and desktop targets, obtain the ROI regions of the humans and the desktop.

[0037] In this embodiment, based on the rectangular annotation boxes of the personnel and the tabletop, the ROI (Region of Interest) region is further determined, which represents the key region related to gambling. The ROI region is mainly concentrated in the interaction region between the personnel and the tabletop, reducing the calculation of irrelevant regions. By focusing on the target region, the waste of computing resources is reduced, and the detection efficiency and accuracy are improved.

[0038] Step S30: Obtain the ROI region image, perform currency target detection on the ROI region image to obtain currency targets; use currency annotation boxes with rotation angles to mark the currency targets.

[0039] In this embodiment, within the determined ROI region, currency target detection is continued. The target detection model is used to identify currency targets in the region, and a convolutional neural network can also be used for identification. This embodiment enables the detection to focus on currency targets, reducing background interference, especially improving the accuracy of currency detection in complex environments.

[0040] Step S40: Perform secondary verification on the currency targets within the currency standard box, compare the verification result of the secondary verification with the reporting threshold, and generate a reporting message when the comparison result meets the reporting condition.

[0041] In this embodiment, currency annotation boxes with rotation angles are used to mark the detected currency targets. This is because the currency may rotate in the image due to reasons such as angle and occlusion, so the annotation box needs to support rotation detection. Perform secondary verification on the currency targets within the annotation box, and ensure that the detected currency meets the standards of gambling behavior by comparing the verification result with the set reporting threshold. Moreover, the system can automatically generate reports, enhancing the real-time response ability of the system and ensuring that gambling behavior can be identified and processed in a timely manner.

[0042] The method provided by the embodiment of the present application is highly automated throughout the process. From the acquisition of monitoring images, the detection of personnel and tabletop targets, to the identification and verification of currency targets, no manual intervention is required, greatly improving the detection efficiency and real-time performance.

[0043] Through the multi-level scheme of target detection, ROI region extraction, currency target recognition, and secondary verification carried out successively in this application, gambling behavior can be detected accurately and efficiently, and the response ability to events is enhanced through a real-time reporting mechanism. Compared with existing gambling detections, the accuracy of detection and the false alarm probability are significantly improved. Moreover, currency detection is only performed on key regions, reducing the consumption of computing power, achieving efficient automated processing, and being able to provide important information in a timely manner, which is suitable for applications in scenarios such as chess and card rooms that require real-time monitoring.

[0044] In a preferred embodiment, the method for obtaining human targets and desktop targets from the monitoring images and generating rectangular annotation frames for each of the human targets and desktop targets is as follows:

[0045] Obtain a pre-trained convolutional neural network, input the monitoring image into the convolutional neural network, and obtain the output of the convolutional neural network;

[0046] The convolutional neural network is configured to: perform target classification on the input image to be detected, identify human targets and desktop targets in the image to be detected, obtain the target contours of the human targets and desktop targets, and use rectangular annotation frames to frame and mark the target contours;

[0047] The loss function of the convolutional neural network is:

[0048] ;

[0049] Wherein, represents the overall loss function of the convolutional neural network, represents the loss function for frame selection and marking, represents the loss function for target classification, represents the proportional hyperparameter.

[0050] In the embodiment of the present application, by using the convolutional neural network shown in Figure 3 for target classification and rectangular frame annotation, and by setting a multi-task loss function, an efficient, accurate and adaptable target detection scheme is provided, and the detection effect is as shown in Figure 4 .

[0051] In this embodiment, the design of the hyperparameter of the loss function enables the model to simultaneously optimize the localization and classification of targets, so as to better identify human and desktop targets in a complex monitoring environment. By adjusting the proportional hyperparameter in the loss function, the attention degree of the model to different tasks can be flexibly adjusted according to actual needs. To balance the above parameters, the preferred values of the proportional hyperparameters , are 5.5 and 1.2 respectively.

[0052] In a preferred embodiment, the loss function for frame selection and marking is:

[0053]

[0054] Wherein, represents the intersection over union of the predicted box A and the ground truth box B, that is, , , respectively represent the center points of the predicted bounding box A and the ground truth bounding box B, represents the Euclidean distance between these two center points, represents the diagonal distance of the smallest closed region that can simultaneously contain the predicted bounding box A and the ground truth bounding box B, represents a balance parameter, denoted as , and is a parameter used to measure the aspect ratio consistency, denoted as , respectively represent the widths of the predicted bounding box A and the ground truth bounding box B, , respectively represent the heights of the predicted bounding box A and the ground truth bounding box B.

[0055] In the embodiments of the present application, the IOU part is used to directly measure the overlap degree of the bounding boxes. The larger the value, the smaller the loss. The Euclidean distance part is used to focus on the position accuracy of the bounding boxes, especially the position of the center point. By normalizing with the diagonal length of the smallest closed region, this item has a certain stability when the size of the bounding box changes. The aspect ratio consistency part is used to optimize the shape of the bounding box through consistency. This part is particularly important when the shape and size of the target vary.

[0056] The loss function provided in this embodiment plays a role in comprehensive optimization in the object detection task. It not only requires the model to improve the overlap degree and positioning accuracy of the bounding boxes, but also requires the model to be as close as possible to the ground truth bounding box in terms of shape. By reasonably adjusting the parameters, the loss function can adapt to different types of object detection tasks and has strong robustness, and can effectively improve the detection accuracy under different backgrounds and changes in the target shape.

[0057] In a preferred embodiment, the loss function for object classification is:

[0058]

[0059] where, is the ground truth label of the th object, is the value predicted by the model for the th object, represents all objects.

[0060] The loss function provided in the embodiments of the present application measures the accuracy of the model prediction by calculating the cross entropy between the ground truth label and the predicted probability of each object. During the training process, by minimizing this loss function, the model can gradually optimize its parameters to make the prediction results as consistent with the ground truth labels as possible.

[0061] In a preferred embodiment, the method for obtaining the ROI regions of the person and the desktop based on the coordinate and size information of the rectangular annotation boxes of the person target and the desktop target is as follows:

[0062]

[0063] Among them, (cx, cy) represents the center point coordinates of the rectangular annotation box, represents the width and height of the rectangular annotation box, (xmin, ymin) represents the upper left corner coordinates of the ROI region, and (xmax, ymax) represents the lower right corner coordinates of the ROI region, respectively represent the maximum value and minimum value functions, 、 respectively represent the width and height of the original image, represents the scaling factor.

[0064] In the embodiment of the present application, from the coordinate and size information of the rectangular annotation box, the ROI regions of the person and the desktop target are calculated according to the given scaling factor. The core objective of this step is to extract a suitable region of interest by adjusting the range of the rectangular annotation box, so as to achieve accurate positioning and processing of the target. The preferred value of the scaling factor is 0.2.

[0065] In a preferred embodiment, the method for detecting currency targets in the ROI region image and marking the currency targets with currency annotation boxes with rotation angles is as follows:

[0066] Obtain a pre-trained currency detection convolutional network, input the ROI region image into the currency detection convolutional network, and obtain the output of the currency detection convolutional network;

[0067] The currency detection convolutional network is used to identify currency targets in the input image to be tested, and the output of the currency detection convolutional network is:

[0068]

[0069] Among them, represents the center point coordinates of the currency annotation box, represents the width and height of the currency annotation box, represents the rotation angle of the currency annotation box relative to the vertical direction of the screen, and the rotation angle is used to make the currency annotation box coincide with the currency pattern.

[0070] In the embodiment of the present application, the rotation angle of the currency annotation box is used to more accurately calibrate the currency target, ensure that the currency annotation box completely coincides with the currency pattern, and achieve high-precision detection and annotation of the currency target in the image. By using such as Figure 5The pre-trained convolutional neural network shown can detect currency targets from ROI region images and mark them with the accurate position, size, and rotation angle of the annotation box. The obtained detection effect diagram is as shown in Figure 6 shown.

[0071] In a preferred embodiment, the method for secondarily verifying the currency target within the currency standard box, comparing the verification result of the secondary verification with the reporting threshold, and generating a reporting message when the comparison result meets the reporting condition is as follows:

[0072] Obtain the currency image of the detected currency target, perform grayscale and normalization processing on the currency image to obtain a preprocessed image;

[0073] Obtain the gradients of the preprocessed image in the horizontal and vertical coordinate directions, and accordingly obtain the gradient direction values at each pixel position;

[0074] Divide the preprocessed image into multiple sub-cells, and within each sub-cell, classify the gradient magnitudes into different histogram channels according to the gradient direction values;

[0075] Merge adjacent histogram channels to form a new region block;

[0076] Stitch all the region blocks in the currency image to obtain the currency feature of the currency image;

[0077] Compare the currency feature with several preset features in the preset feature library, and generate a reporting message when any similarity comparison result exceeds the threshold.

[0078] In the embodiment of the present application, when a currency extractor detects RMB or other currencies, in order to prevent false alarms caused by misdetection of the detector, a feature verification method is added to this solution. The feature extraction steps are as follows: First, in order to reduce the influence of illumination, it is necessary to perform grayscale processing on the input image and perform normalization processing, that is:

[0079]

[0080] where gamma is a hyperparameter, and in this solution, the value of gamma is 0.5.

[0081] Then, calculate the gradients in the horizontal and vertical coordinate directions of the image, and accordingly calculate the gradient direction values at the positions of each pixel, that is:

[0082]

[0083]

[0084] Among them, gradient magnitude, represents the gradient direction, and respectively represent the horizontal gradient and vertical gradient at the pixel point (x, y) in the image, that is:

[0085]

[0086] Subsequently, the image is divided into multiple small cells. Inside each cell, according to the gradient direction gradient, the gradient magnitude is classified into different histogram channels. In this solution, preferably 9 histogram channels can be used. To improve the robustness of the features, adjacent gradient histograms are merged to form a new region block. In this solution, 3x3 cells are spliced into a new region block. All the regions in the target image are spliced to form the final feature. Finally, the target feature to be verified is compared with the features of each currency in the feature library. If the similarity is greater than a certain threshold τ, it is considered a correct target, otherwise it is considered a false detection, that is:

[0087]

[0088] Among them, sim represents the final similarity, cos() represents the cosine function, and respectively represent the true feature in the feature library of the nth currency and the feature extracted from the predicted target, The preferred value in this solution is 0.76. Based on the above method, the occurrence probability of false alarms can be significantly reduced.

[0089] As Figure 7 shown, in one embodiment, a gambling behavior detection device for a chess and card room scene is provided. The gambling behavior detection device for the chess and card room scene can be integrated into the above computer device 120. The gambling behavior detection device for the chess and card room scene includes:

[0090] A rectangular annotation box generation module 510, configured to obtain a monitoring image, obtain a personnel target and a desktop target from the monitoring image, and generate a rectangular annotation box for each of the personnel target and the desktop target;

[0091] A ROI region acquisition module 520 for personnel and desktop, configured to obtain the ROI regions for personnel and desktop based on the coordinate and size information of the rectangular annotation boxes of the personnel target and the desktop target;

[0092] A currency target marking module 530, configured to obtain the ROI region image, perform currency target detection on the ROI region image to obtain a currency target; and mark the currency target with a currency annotation box with a rotation angle.

[0093] The secondary verification module 540 is used to perform secondary verification on the currency target within the currency standard box, compare the verification result of the secondary verification with the reporting threshold, and generate a reporting message when the comparison result meets the reporting condition.

[0094] In the embodiments of the present application, the explanations and descriptions of the gambling behavior detection device for the above-mentioned chess and card room scenario can refer to the explanations and descriptions of the corresponding methods above. For the description of the gambling behavior detection method for the above-mentioned chess and card room scenario, please refer to the above text and will not be repeated here.

[0095] The outstanding advantage of the embodiments of the present application is that through a multi-level scheme of target detection, ROI region extraction, currency target recognition, and secondary verification carried out successively, gambling behaviors can be detected accurately and efficiently, and the response ability to events is enhanced through a real-time reporting mechanism. Compared with existing gambling detections, the accuracy of detection and the false alarm probability are significantly improved, and currency detection is only performed on key areas, reducing the consumption of computing power, achieving efficient automated processing, and being able to timely feedback important information, which is suitable for application in the real-time monitoring scenario of chess and card rooms.

[0096] Figure 8 The internal structure diagram of a computer device in an embodiment is shown. The computer device may specifically be Figure 1 the computer device 120 in Figure 8 As shown, the computer device includes a processor, a memory, a network interface, an input device, and a display screen connected through a system bus. Among them, the memory includes a non-volatile storage medium and an internal memory. The non-volatile storage medium of the computer device stores an operating system and may also store a computer program. When the computer program is executed by the processor, the processor can implement the gambling behavior detection method for the chess and card room scenario. The internal memory may also store a computer program. When the computer program is executed by the processor, the processor can execute the gambling behavior detection method for the chess and card room scenario. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device may be a touch layer covered on the display screen, or a button, a trackball, or a touchpad provided on the housing of the computer device, or an external keyboard, touchpad, or mouse, etc.

[0097] Those skilled in the art can understand that Figure 8 the structure shown in

[0098] In one embodiment, the gambling behavior detection device for the chess and card room scenario provided by the present application can be implemented in the form of a computer program, and the computer program can run on a computer device as shown in Figure 8 . Each program module constituting the gambling behavior detection device for the chess and card room scenario can be stored in the memory of the computer device. For example, Figure 7 the rectangular annotation box generation module 510, the ROI area acquisition module 520 for personnel and the desktop, etc. shown in. The computer program constituted by each program module enables the processor to execute the steps in the gambling behavior detection method for the chess and card room scenario of each embodiment of the present application described in this specification.

[0099] For example, Figure 8 the computer device shown in can execute step S10 through the rectangular annotation box generation module 510 in the gambling behavior detection device for the chess and card room scenario shown in Figure 7 . The computer device can execute step S20 through the ROI area acquisition module 520 for personnel and the desktop. And so on.

[0100] In one embodiment, a computer-readable storage medium is provided. A computer program is stored on the computer-readable storage medium. When the computer program is executed by a processor, the processor is enabled to execute the steps of the gambling behavior detection method for the chess and card room scenario as described above.

[0101] In the embodiments of the present application, for the description of the above-mentioned gambling behavior detection method for the chess and card room scenario, please refer to the above, and details are not described herein again.

[0102] In the embodiments of the present application, the program running based on the method stored in the storage medium of the embodiments of the present application has the advantage that through a multi-level scheme of target detection, ROI area extraction, currency target recognition, and secondary verification carried out successively, it can accurately and efficiently detect gambling behavior, and enhances the response ability to events through a real-time reporting mechanism. Compared with the existing gambling detection, it significantly improves the detection accuracy and false alarm probability, and only performs currency detection on key areas, reducing the consumption of computing power, realizing efficient automatic processing, and can timely feedback important information, which is suitable for application in the real-time monitoring scenario of the chess and card room.

[0103] In one embodiment, a gambling behavior detection system for the chess and card room scenario is provided, including a memory and a processor. A computer program is stored in the memory. When the computer program is executed by the processor, the processor is enabled to execute the steps of the gambling behavior detection method for the chess and card room scenario as described above.

[0104] In the embodiments of the present application, the system may be a computer software or program system that runs on hardware containing a processing device. When the above system is running, it executes its corresponding method. For the description of the gambling behavior detection method in the above-mentioned chess and card room scenario, please refer to the above text and will not be elaborated here.

[0105] In the embodiments of the present application, the advantages of the present system are as follows. Through a multi-level scheme of target detection, ROI region extraction, currency target recognition, and secondary verification carried out successively, it can accurately and efficiently detect gambling behavior, and enhances the response ability to events through a real-time reporting mechanism. Compared with existing gambling detections, it significantly improves the detection accuracy and false alarm probability, and only performs currency detection on key areas, reducing the consumption of computing power, achieving efficient automated processing, and being able to timely feedback important information, which is suitable for application in the real-time monitoring scenario of chess and card rooms.

[0106] It should be understood that although the steps in the flowcharts of the embodiments of the present application are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear indication in this article, the execution of these steps has no strict order limit, and these steps can be executed in other orders. Moreover, at least a part of the steps in each embodiment may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed alternately or alternately with at least a part of other steps or sub-steps or stages of other steps.

[0107] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the various embodiments provided in the present application can include non-volatile and / or volatile memories. Non-volatile memories can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memories can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM), etc.

[0108] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, 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, it should be considered as the scope described in this specification.

[0109] The above embodiments merely represent several implementation manners of the present application. The description is relatively specific and detailed, but it should not be construed as a limitation on the patent scope of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the appended claims.

Claims

1. A method for detecting gambling-related behaviors in a chess and card room scenario, characterized in that, The method includes: Obtain a surveillance image, obtain human targets and desktop targets from the surveillance image, and generate rectangular annotation frames for each of the human targets and desktop targets; Based on the coordinate and dimension information of the rectangular annotation frames of the human targets and desktop targets, obtain the ROI regions of the humans and the desktop; Obtain the ROI region image, perform currency target detection on the ROI region image to obtain currency targets; use currency annotation frames with rotation angles to mark the currency targets; Perform secondary verification on the currency targets within the currency standard frame, compare the verification result of the secondary verification with a reporting threshold, and generate a reporting message when the comparison result meets the reporting condition.

2. The gambling behavior detection method for a chess and card room scenario according to claim 1, wherein The method for obtaining human targets and desktop targets from the surveillance image and generating rectangular annotation frames for each of the human targets and desktop targets is: Obtain a pre-trained convolutional neural network, input the surveillance image into the convolutional neural network, and obtain the output of the convolutional neural network; The convolutional neural network is configured to: perform target classification on the input image to be detected, identify human targets and desktop targets in the image to be detected, obtain the target contours of the human targets and desktop targets, and use rectangular annotation frames to frame and mark the target contours; The loss function of the convolutional neural network is: ； Among them, represents the overall loss function of the convolutional neural network, represents the loss function for box selection and marking, represents the loss function for target classification, represents the proportional hyperparameter.

3. The gambling behavior detection method for a chess and card room scene according to claim 2, characterized in that Loss function for performing box selection marking It is as follows: Among them, represents the intersection over union of the predicted box A and the ground truth box B, that is , , respectively represent the center points of the predicted box A and the ground truth box B, represents the Euclidean distance between these two center points, represents the diagonal distance of the smallest closed region that can simultaneously contain the predicted box A and the ground truth box B, represents a balance parameter, denoted as , and is a parameter used to measure the aspect ratio consistency, denoted as , respectively represent the widths of the predicted box A and the ground truth box B, , respectively represent the heights of the predicted box A and the ground truth box B.

4. A method for detecting gambling inspection behavior in a chess and card room scene according to claim 2, characterized in that, Loss function for target classification is as follows: Among them, is the true label of the th target, is the value of the th target predicted by the model, indicating all targets.

5. The gambling behavior detection method for a chess and card room scene according to claim 1, wherein The method for obtaining the ROI regions of the humans and the desktop based on the coordinate and dimension information of the rectangular annotation frames of the human targets and desktop targets is: Among them, (cx, cy) represents the center point coordinates of the rectangular annotation box, represents the width and height of the rectangular annotation box, (xmin, ymin) represents the upper left corner coordinates of the ROI region, and (xmax, ymax) represents the lower right corner coordinates of the ROI region, respectively represent the functions of taking the maximum value and the minimum value, and respectively represent the width and height of the original image, represents the scaling factor.

6. The gambling behavior detection method for a chess and card room scene according to claim 1, wherein, The method for performing currency target detection on the ROI region image to obtain currency targets and using currency annotation frames with rotation angles to mark the currency targets is: Obtain a pre-trained currency detection convolutional network, input the ROI region image into the currency detection convolutional network, and obtain the output of the currency detection convolutional network; The currency detection convolutional network is used to identify currency targets in the input image to be detected, and the output of the currency detection convolutional network is: Among them, represents the center point coordinates of the currency annotation box, represents the width and height of the currency annotation box, represents the rotation angle of the currency annotation box relative to the vertical direction of the screen. The rotation angle is used to align the currency annotation box with the currency pattern.

7. A method for detecting gambling inspection behavior in a chess and card room scene according to claim 1, characterized in that The method for performing secondary verification on the currency targets within the currency standard frame, comparing the verification result of the secondary verification with a reporting threshold, and generating a reporting message when the comparison result meets the reporting condition is: Obtain the currency image of the detected currency target, perform grayscale and normalization processing on the currency image to obtain a preprocessed image; Obtain the gradients of the preprocessed image in the horizontal and vertical coordinate directions, and accordingly obtain the gradient direction values at each pixel position; Divide the preprocessed image into multiple sub-cells, and within each sub-cell, classify the gradient magnitudes into different histogram channels according to the gradient direction values; Merge adjacent histogram channels to form a new region block; Stitch all the region blocks in the currency image to obtain the currency feature of the currency image; Compare the currency feature with several preset features in a preset feature library, and generate a reporting message when any similarity comparison result exceeds the threshold.

8. A gambling behavior detection device for a chess and card room scene, characterized in that, The gambling behavior detection device for the chess and card room scenario includes: A rectangular annotation box generation module, configured to obtain a monitoring image, obtain person targets and desktop targets from the monitoring image, and generate rectangular annotation boxes for each of the person targets and desktop targets; A ROI region acquisition module for persons and desktops, configured to obtain the ROI regions for persons and desktops based on the coordinate and dimension information of the rectangular annotation boxes of the person targets and desktop targets; A currency target marking module, configured to obtain the ROI region image, perform currency target detection on the ROI region image to obtain currency targets; and mark the currency targets with currency annotation boxes with rotation angles; A secondary verification module, configured to perform secondary verification on the currency targets within the currency standard boxes, compare the verification results of the secondary verification with a reporting threshold, and generate a reporting message when the comparison result meets the reporting conditions.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor is caused to execute the steps of the gambling behavior detection method for a chess and card room scenario according to any one of claims 1 to 7.

10. A gambling behavior detection system for a chess and card room scene, characterized in that, Comprising a memory and a processor, the memory stores a computer program, and when the computer program is executed by the processor, the processor is caused to execute the steps of the gambling behavior detection method for a chess and card room scenario according to any one of claims 1 to 7.

Images