Anti-tailgating detection method and device based on dual detection model

Abstract

The present application provides an anti-tailing detection method and device based on a dual detection model. The method includes: obtaining a set of target images collected at at least two angles; using a first detection model to process the target image set, obtaining a first number of people output corresponding to each target image, and determining a first credible output from all first number of people output; if the number of first credible outputs is greater than 1, determining the target tailing result based on the first credible output; otherwise, using a second detection model to process the target image set, obtaining a second number of people output corresponding to each target image, and determining the target tailing result based on the second number of people output. By first using the first detection model with a faster processing speed for detection, the process ends when the recognition is successful. If the recognition fails, a second detection model with a higher accuracy is used for double detection, which effectively combines the advantages of the two types of detection models and can meet the requirements of processing speed and detection accuracy at the same time.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to an anti-tailgating detection method and device based on a dual detection model. Background Art

[0002] Existing inspection gates can usually identify tailgating situations. These gate equipment includes devices such as the fuselage, front and rear door wings, front and rear identification devices, video anti-tailgating devices, control systems and power systems. The front identification device mainly uses document identification (such as documents and tickets) or non-sensing identification technology. After the inspection is passed, the first door wing opens for secondary inspection. The rear identification device usually uses biometric technology (such as fingerprint, face and iris recognition). After the inspection is passed, the second door wing opens for travel. The video anti-tailgating device is installed between the front and rear door wings and is used to identify tailgating situations through video acquisition. The built-in processing algorithm can distinguish between passers-by and tailgating personnel, accompanying infants and young children, luggage and pets. Once tailgating occurs, the rear door wing will be locked and an alarm will be triggered.

[0003] Existing anti-tailgating detection algorithms based on deep learning include target monitoring algorithms, image layer analysis methods, and foreground-based target segmentation methods. Among them, the target monitoring algorithm uses a conventional target detection algorithm to identify pedestrians in the image, and determines whether there is a tailgating situation based on the number of pedestrians identified. The image layer analysis method determines whether there are two or more people at the same time by layering the depth map and performing connected domain analysis on each layer. The foreground-based target segmentation method (such as the instance segmentation algorithm Mask RCNN) predicts the category of the foreground target at the pixel level. It includes three parts: the first part is the backbone network used for feature extraction, the second part is the head structure used for bounding box recognition, and the third part is the mask prediction network used for pixel-level segmentation of each foreground target. This method can monitor the situation where one or more people enter the channel in various ways (for example, adults bring children into the channel), and can identify the luggage items that enter the channel with the passengers.

[0004] However, existing detection methods are difficult to meet the requirements of processing speed and detection accuracy at the same time, and are prone to false detection or missed detection. Summary of the invention

[0005] In view of the above-mentioned problems, the present application is proposed to provide an anti-tailgating detection method and device based on a dual detection model to overcome the problems or at least partially solve the problems, including:

[0006] An anti-tailgating detection method based on a dual detection model, comprising:

[0007] Acquire a target image set acquired from the target object at at least two preset angles; wherein the preset angles intersect each other;

[0008] Processing the target image set using a first detection model to obtain a first number of people output corresponding to each target image, and determining a first credible output from all the first number of people outputs; wherein the first detection model detects the target object in the target image through a target detection network;

[0009] If the number of the first credible outputs is greater than 1, determining a target trailing result according to the first credible outputs;

[0010] If the number of the first credible outputs is less than or equal to 1, the target image set is processed using a second detection model to obtain a second number of people output corresponding to each target image, and the target tracking result is determined based on the second number of people output; wherein the second detection model detects the target object in the target image through a region proposal network and a target detection network.

[0011] Preferably, the step of using the first detection model to process the target image set to obtain a first headcount output corresponding to each target image, and determining a first credible output from all the first headcount outputs comprises:

[0012] Processing the target image set using a first detection model to obtain a first number of people output and a first detection confidence corresponding to each target image;

[0013] A first credible output is determined from all of the first number of people output according to the first detection confidence.

[0014] Preferably, the step of determining a first credible output from all the first number of people outputs according to the first detection confidence level comprises:

[0015] The first number of people whose first detection confidence level is greater than a first preset value is output as a first credible output.

[0016] Preferably, the step of determining the target tracking result according to the first credible output includes:

[0017] Taking the first credible output with the largest value as the first maximum output;

[0018] If the value of the first maximum output is greater than 1, the target trailing result is set to have trailing;

[0019] If the value of the first maximum output is less than or equal to 1, the target trailing result is set to no trailing.

[0020] Preferably, the step of using the second detection model to process the target image set to obtain a second number of people output corresponding to each target image, and determining the target following result according to the second number of people output includes:

[0021] Processing the target image set using a second detection model to obtain a second number of people output and a second detection confidence corresponding to each target image;

[0022] Determining a second credible output from all of the second number of people outputs according to the second detection confidence level;

[0023] If the number of the second trusted outputs is greater than 1, determining a target trailing result according to the second trusted outputs;

[0024] If the number of the second credible outputs is less than or equal to 1, the manually identified target tracking result is obtained.

[0025] Preferably, the step of determining a second credible output from all the second number of people output according to the second detection confidence level comprises:

[0026] The second number of people whose second detection confidence level is greater than a second preset value is output as a second credible output.

[0027] Preferably, the step of determining the target tracking result according to the second trusted output includes:

[0028] Taking the second credible output with the largest value as the second maximum output;

[0029] If the value of the second maximum output is greater than 1, the target trailing result is set to have trailing;

[0030] If the value of the second maximum output is less than or equal to 1, the target trailing result is set to no trailing.

[0031] An anti-tailgating detection device based on a dual detection model comprises:

[0032] A target image acquisition module, used to acquire a target image set acquired from a target object at at least two preset angles; wherein the preset angles intersect each other;

[0033] A first number detection module, configured to process the target image set using a first detection model to obtain a first number output corresponding to each target image, and determine a first credible output from all the first number outputs; wherein the first detection model detects target objects in the target images through a target detection network;

[0034] a first tailing determination module, configured to determine a target tailing result according to the first credible output if the number of the first credible outputs is greater than 1;

[0035] A second tailing determination module is used to process the target image set using a second detection model if the number of the first credible outputs is less than or equal to 1, to obtain a second number of people output corresponding to each target image, and to determine a target tailing result based on the second number of people output; wherein the second detection model detects the target object in the target image through a region proposal network and a target detection network.

[0036] A computer device comprises a processor, a memory and a computer program stored in the memory and capable of running on the processor, wherein the computer program implements any of the above methods when executed by the processor.

[0037] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method described in any one of the above items is implemented.

[0038] This application has the following advantages:

[0039] In an embodiment of the present application, in view of the problem that the existing detection methods are difficult to simultaneously meet the requirements of processing speed and detection accuracy and are prone to false detection or missed detection, the present application provides a solution for anti-tailing detection based on multi-angle images and dual detection models, specifically: "obtain a set of target images acquired from the target object at at least two preset angles; wherein the preset angles intersect each other; use a first detection model to process the target image set to obtain a first number of people output corresponding to each target image, and determine a first credible output from all the first number of people output; wherein the first detection model detects the target object in the target image through a target detection network; if the number of the first credible outputs is greater than 1, determine the target tailing result based on the first credible output; if the number of the first credible outputs is less than or equal to 1, use a second detection model to process the target image set to obtain a second number of people output corresponding to each target image, and determine the target tailing result based on the second number of people output; wherein the second detection model detects the target object in the target image through a region proposal network and a target detection network". By detecting the target image at multiple angles, the detection error caused by viewing angle problems can be reduced, and a more comprehensive detection result can be obtained (even if two pedestrians overlap in the image from one viewing angle, they can be accurately distinguished through images from other viewing angles). By first using the first detection model with a faster processing speed for detection, the process ends when the recognition is successful. If the recognition fails, the second detection model with higher accuracy is used for double detection. This effectively combines the advantages of the two types of detection models, and can meet the requirements of processing speed and detection accuracy at the same time (in actual applications, the recognition success rate of the first detection model is extremely high, and the second detection model is rarely used for double detection. It can be seen that the processing method of the present application can effectively reduce computing consumption and improve processing speed while ensuring detection accuracy). BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solution of the present application, the drawings required for use in the description of the present application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0041] Figure 1 It is a flowchart of a step of an anti-tailgating detection method provided by an embodiment of the present application;

[0042] Figure 2 It is a flowchart of an anti-tailgating detection method provided in one embodiment of the present application;

[0043] Figure 3It is a structural schematic diagram of a first detection model in an anti-tailgating detection method provided in an embodiment of the present application;

[0044] Figure 4 It is a structural block diagram of an anti-tailgating detection device provided in one embodiment of the present application;

[0045] Figure 5 It is a structural schematic diagram of a computer device provided by an embodiment of the present application;

[0046] Figure 6 It is a structural schematic diagram of a self-service inspection channel provided in one embodiment of the present application.

[0047] The reference numerals in the drawings of the specification are as follows:

[0048] 12. Computer equipment; 14. External devices; 16. Processing unit; 18. Bus; 20. Network adapter; 22. I / O interface; 24. Display; 28. Memory; 30. Random access memory; 32. Cache memory; 34. Storage system; 40. Program / utility; 42. Program module. DETAILED DESCRIPTION

[0049] In order to make the objects, features and advantages of the present application more obvious and understandable, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods. Obviously, the described embodiments are part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present application.

[0050] The inventors have found through analysis of the prior art that the mainstream target detection framework based on deep learning mainly includes two categories, one is the "one-stage" method (such as YOLO and SSD), which completes the tasks of regression and classification through a single target detection network at the same time, and the other is the "two-stage" method (such as Faster R-CNN), which divides the detection task into two parts, regression and classification, first completing the task of regression through the region proposal network, and then completing the task of classification through the target detection network. However, it is difficult for the existing detection methods to meet the requirements of processing speed and detection accuracy at the same time. For example, the "one-stage" method has a faster processing speed but lower detection accuracy, while the "two-stage" method has a higher detection accuracy but a slower processing speed. In addition, the existing methods generally perform detection based on images from a single perspective. When two pedestrians overlap in the image, the algorithm may not be able to accurately distinguish them, which is prone to false detection or missed detection.

[0051] Reference Figure 1, 2 , shows an anti-tailgating detection method based on a dual detection model provided by an embodiment of the present application, including:

[0052] S110, acquiring a target image set acquired from the target object at at least two preset angles; wherein the preset angles intersect each other;

[0053] S120, using a first detection model to process the target image set to obtain a first number of people output corresponding to each target image, and determining a first credible output from all the first number of people output; wherein the first detection model detects the target object in the target image through a target detection network;

[0054] S130: If the number of the first credible outputs is greater than 1, determining a target trailing result according to the first credible outputs;

[0055] S140. If the number of the first credible outputs is less than or equal to 1, the target image set is processed using a second detection model to obtain a second number of people output corresponding to each target image, and a target tracking result is determined based on the second number of people output; wherein the second detection model detects the target object in the target image through a region proposal network and a target detection network.

[0056] In an embodiment of the present application, in view of the problem that the existing detection methods are difficult to simultaneously meet the requirements of processing speed and detection accuracy and are prone to false detection or missed detection, the present application provides a solution for anti-tailing detection based on multi-angle images and dual detection models, specifically: "obtain a set of target images acquired from the target object at at least two preset angles; wherein the preset angles intersect each other; use a first detection model to process the target image set to obtain a first number of people output corresponding to each target image, and determine a first credible output from all the first number of people output; wherein the first detection model detects the target object in the target image through a target detection network; if the number of the first credible outputs is greater than 1, determine the target tailing result based on the first credible output; if the number of the first credible outputs is less than or equal to 1, use a second detection model to process the target image set to obtain a second number of people output corresponding to each target image, and determine the target tailing result based on the second number of people output; wherein the second detection model detects the target object in the target image through a region proposal network and a target detection network". By detecting the target image at multiple angles, the detection error caused by viewing angle problems can be reduced, and a more comprehensive detection result can be obtained (even if two pedestrians overlap in the image from one viewing angle, they can be accurately distinguished through images from other viewing angles). By first using the first detection model with a faster processing speed for detection, the process ends when the recognition is successful. If the recognition fails, the second detection model with higher accuracy is used for double detection. This effectively combines the advantages of the two types of detection models, and can meet the requirements of processing speed and detection accuracy at the same time (in actual applications, the recognition success rate of the first detection model is extremely high, and the second detection model is rarely used for double detection. It can be seen that the processing method of the present application can effectively reduce computing consumption and improve processing speed while ensuring detection accuracy).

[0057] Next, an anti-tailgating detection method based on a dual detection model in this exemplary embodiment will be further described.

[0058] As described in step S110, a target image set acquired by capturing the target object at at least two preset angles is obtained; wherein the preset angles intersect with each other.

[0059] As an example, there are three preset angles, namely: front side (capturing the front outline of the target object), side side (capturing the side outline of the target object) and top side (capturing the top outline of the target object).

[0060] The preset angles intersect each other (ie, are not parallel to each other), which can ensure that the features of the target object are observed from different viewing angles, and is beneficial to improving the accuracy and robustness of recognition in the subsequent detection process.

[0061] As described in step S120, the target image set is processed using a first detection model to obtain a first number of people output corresponding to each target image, and a first credible output is determined from all the first number of people outputs; wherein the first detection model detects the target object in the target image through a target detection network.

[0062] The first detection model is a pre-trained "one-stage" model (including target detection networks such as YOLO and SSD) that can quickly identify people and quantities in images.

[0063] The target image is input into the first detection model to obtain a first number of people output and a first detection confidence corresponding to the target image, wherein the first number of people output represents the number of people in the target image, and the first detection confidence represents the degree of confidence of the model in the first number of people output.

[0064] Specifically, the model generates a single detection confidence for each person in the target image. The single detection confidence with the lowest value can be used as the first detection confidence, or the first detection confidence can be generated according to the average value of all single detection confidences.

[0065] A first credible output is determined from all of the first number of people outputs according to the first detection confidence.

[0066] Specifically, the first number of people whose first detection confidence level is greater than a first preset value is output as a first credible output.

[0067] Through the above method, high-confidence detection results can be effectively identified and screened from multi-angle images.

[0068] As described in step S130, if the number of the first trusted outputs is greater than 1, a target trailing result is determined according to the first trusted outputs.

[0069] If the number of the first credible outputs is greater than 1 (that is, the first detection model is successfully identified), the target following result is determined according to the first credible output.

[0070] Specifically, in one embodiment, the first credible output with the largest value is used as the first maximum output; if the value of the first maximum output is greater than 1, the target trailing result is set to have trailing; if the value of the first maximum output is less than or equal to 1, the target trailing result is set to no trailing. In another embodiment, the first credible output with the largest number of occurrences of the same value is used as the first maximum output (for example, the first credible output includes 3, and the values ​​are 2 people, 2 people, and 3 people, respectively, then the first credible output with the value of 2 people is used as the first maximum output); if the value of the first maximum output is greater than 1, the target trailing result is set to have trailing; if the value of the first maximum output is less than or equal to 1, the target trailing result is set to no trailing.

[0071] As described in step S140, if the number of the first credible outputs is less than or equal to 1, the target image set is processed using a second detection model to obtain a second number of people output corresponding to each target image, and the target tracking result is determined based on the second number of people output; wherein the second detection model detects the target object in the target image through a region proposal network and a target detection network.

[0072] The second detection model is a pre-trained "two-stage" model (including a region proposal network and an object detection network, such as Faster R-CNN), which can accurately identify the people and quantities in the image.

[0073] In one embodiment, if the number of the first credible outputs is less than or equal to 1 (that is, the first detection model fails to recognize), the target image is input into the second detection model to obtain a second number of people output corresponding to the target image, and the second number of people output represents the number of people in the target image.

[0074] The target following result is determined according to the second number of people output.

[0075] Specifically, in one embodiment, the second number of people with the largest value is output as the second maximum output; if the value of the second maximum output is greater than 1, the target trailing result is set to have trailing; if the value of the second maximum output is less than or equal to 1, the target trailing result is set to no trailing. In another embodiment, the second number of people with the largest number of occurrences of the same value is output as the second maximum output (for example, the second number of people output includes 3, and the values ​​are 2, 2, and 3, respectively, then the second number of people output with the value of 2 is output as the second maximum output); if the value of the second maximum output is greater than 1, the target trailing result is set to have trailing; if the value of the second maximum output is less than or equal to 1, the target trailing result is set to no trailing.

[0076] In another embodiment, if the number of the first credible outputs is less than or equal to 1, the target image is input into the second detection model to obtain a second number of people output and a second detection confidence corresponding to the target image, wherein the second number of people output represents the number of people in the target image, and the second detection confidence represents the model's degree of confidence in the second number of people output.

[0077] Specifically, the model generates a single detection confidence for each person in the target image. The single detection confidence with the lowest value can be used as the second detection confidence, or the second detection confidence can be generated according to the average value of all single detection confidences.

[0078] A second credible output is determined from all of the second number of people outputs according to the second detection confidence.

[0079] Specifically, the second number of people whose second detection confidence is greater than a second preset value is output as a second credible output. Preferably, the second preset value is greater than the first preset value.

[0080] Through the above method, detection results with higher confidence can be effectively identified and screened from multi-angle images.

[0081] If the number of the second credible outputs is greater than 1 (that is, the second detection model is successfully identified), the target following result is determined according to the second credible output.

[0082] Specifically, in one embodiment, the second credible output with the largest value is used as the second maximum output; if the value of the second maximum output is greater than 1, the target trailing result is set to have trailing; if the value of the second maximum output is less than or equal to 1, the target trailing result is set to no trailing. In another embodiment, the second credible output with the largest number of occurrences of the same value is used as the second maximum output (for example, the second credible output includes 3, and the values ​​are 2 people, 2 people, and 3 people, respectively, then the second credible output with the value of 2 people is used as the second maximum output); if the value of the second maximum output is greater than 1, the target trailing result is set to have trailing; if the value of the second maximum output is less than or equal to 1, the target trailing result is set to no trailing.

[0083] If the number of the second credible outputs is less than or equal to 1 (that is, the second detection model fails to identify), then the manually identified target tracking result is obtained.

[0084] In one embodiment of the present application, the first detection model is deployed on an edge computer; and the second detection model is deployed on a remote server.

[0085] By deploying the first detection model with a faster processing speed on a device with limited performance to implement edge detection, the detection efficiency can be effectively improved.

[0086] In one embodiment of the present application, the first detection model is e-YOLO, which includes an input layer, a convolutional layer, a prediction layer and an output layer.

[0087] The input layer is used to input images of a fixed size (for example, 448x448x3). This setting enables the model to process image data in a unified manner across the entire image collection, which helps the model to efficiently process image data in memory and speed up data copying and calculation.

[0088] The convolutional layer is provided with at least one, and each convolutional layer is responsible for extracting features of the image. By using convolution kernels of different sizes, the model can capture visual features of different scales, from detailed textures to larger shapes and contours. Preferably, the convolutional layer is provided with a size of 1x1, which can effectively reduce the amount of calculation and improve the model processing speed.

[0089] The prediction layer is used to output the detection result, which can predict the position (ie, bounding box) and category (ie, person) of each object in the image based on the extracted features.

[0090] The output layer is used to output the final result, such as the total number of statistical objects (ie, persons) and the detection confidence.

[0091] The overall architecture of e-YOLO is based on YOLOv5, using the "one-stage" mode to directly detect and classify targets through a single neural network, avoiding the candidate box generation and screening steps in the traditional two-stage detection method, and simplifying the algorithm process. The e-YOLO algorithm has a fast detection speed and can process a large amount of image data in real time or within an efficient time range. At the same time, because the e-YOLO algorithm introduces technologies such as multi-scale prediction of anchor boxes, it can more flexibly adapt to targets of different sizes and shapes, and improves its performance in small target detection and positioning.

[0092] Reference Figure 3 In one embodiment of the present application, the target detection network uses head and shoulder detection to complete target detection.

[0093] Head and shoulders are unique features of humans. Since the algorithm does not need to be classified, it only needs to determine whether there is an area containing the head and shoulders, which makes the model training results have a high degree of fit.

[0094] As for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0095] Reference Figure 4 , shows an anti-tailgating detection device based on a dual detection model provided by an embodiment of the present application, including:

[0096] The target image acquisition module 210 is used to acquire a target image set acquired from the target object at at least two preset angles; wherein the preset angles intersect each other;

[0097] A first person detection module 220 is used to process the target image set using a first detection model to obtain a first person output corresponding to each target image, and determine a first credible output from all the first person outputs; wherein the first detection model detects the target object in the target image through a target detection network;

[0098] A first tailing determination module 230, configured to determine a target tailing result according to the first credible output if the number of the first credible outputs is greater than 1;

[0099] The second tailing determination module 240 is used to process the target image set using a second detection model if the number of the first credible outputs is less than or equal to 1, to obtain a second number of people output corresponding to each target image, and to determine the target tailing result based on the second number of people output; wherein the second detection model detects the target object in the target image through a region proposal network and a target detection network.

[0100] In an embodiment of the present application, the first number detection module 220 includes:

[0101] A first headcount output submodule, configured to process the target image set using a first detection model to obtain a first headcount output and a first detection confidence corresponding to each target image;

[0102] The first credible output submodule is used to determine a first credible output from all the first number of people output according to the first detection confidence level.

[0103] In an embodiment of the present application, the first trusted output submodule includes:

[0104] The first credible output setting submodule is used to output the first number of people whose first detection confidence level is greater than a first preset value as a first credible output.

[0105] In one embodiment of the present application, the first tailing determination module 230 includes:

[0106] A first maximum output submodule, configured to use the first credible output with the largest value as a first maximum output;

[0107] A first trailing setting submodule, configured to set a target trailing result as having a trailing if the value of the first maximum output is greater than 1;

[0108] The first no-tailing setting submodule is configured to set the target tailing result to no tailing if the value of the first maximum output is less than or equal to 1.

[0109] In one embodiment of the present application, the second tailing determination module 240 includes:

[0110] A second headcount output submodule, configured to process the target image set using a second detection model to obtain a second headcount output and a second detection confidence corresponding to each target image;

[0111] A second credible output submodule, configured to determine a second credible output from all the second number of people output according to the second detection confidence level;

[0112] a second trusted output submodule, configured to determine a target tailing result according to the second trusted output if the number of the second trusted outputs is greater than 1;

[0113] The second tailing determination submodule is configured to obtain a manually identified target tailing result if the number of the second credible outputs is less than or equal to 1.

[0114] In an embodiment of the present application, the second trusted output submodule includes:

[0115] The second credible output setting submodule is used to output the second number of people whose second detection confidence level is greater than a second preset value as a second credible output.

[0116] In one embodiment of the present application, the second tailing determination submodule includes:

[0117] A second maximum output submodule, configured to use the second credible output with the largest value as the second maximum output;

[0118] A second trailing setting submodule, configured to set the target trailing result to have trailing if the value of the second maximum output is greater than 1;

[0119] The second no-tailing setting submodule is used to set the target tailing result to no tailing if the value of the second maximum output is less than or equal to 1.

[0120] Reference Figure 5, showing a computer device of the present application, the computer device 12 is in the form of a general-purpose computing device; the computer device 12 includes: one or more processors or processing units 16, a memory 28 and a bus 18 connecting different system components (including the memory 28 and the processing unit 16).

[0121] The bus 18 may be one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processor, or a local bus using any of a variety of bus architectures. For example, these architectures include, but are not limited to, an Industry Standard Architecture (ISA) bus, a Micro Channel Architecture (MAC) bus, an Enhanced ISA bus, an Audio Video Electronics Standards Association (VESA) local bus, and a Peripheral Component Interconnect (PCI) bus.

[0122] The computer device 12 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by the computer device 12, including volatile and non-volatile media, removable and non-removable media.

[0123] Memory 28 may include computer system readable media in the form of volatile memory, such as random access memory 30 and / or cache memory 32. Computer device 12 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 34 may be used to read and write to non-removable, non-volatile magnetic media (commonly referred to as a "hard drive"). Although Figure 5 Not shown, a disk drive for reading and writing to a removable non-volatile disk (such as a "floppy disk"), and an optical disk drive for reading and writing to a removable non-volatile optical disk (such as a CD-ROM, DVD-ROM or other optical medium) may be provided. In these cases, each drive may be connected to the bus 18 via one or more data medium interfaces. The memory may include at least one program product having a set (e.g., at least one) of program modules 42, which are configured to perform the functions of the various embodiments of the present application.

[0124] A program / utility 40 having a set (at least one) of program modules 42 may be stored in, for example, a memory, such program modules 42 including an operating system, one or more application programs, other program modules 42, and program data, each of which or some combination may include an implementation of a network environment. The program modules 42 generally perform the functions and / or methods of the embodiments described herein.

[0125] The computer device 12 may also communicate with one or more external devices 14 (e.g., keyboards, pointing devices, displays 24, cameras, etc.), one or more devices that enable an operator to interact with the computer device 12, and / or any device that enables the computer device 12 to communicate with one or more other computing devices (e.g., network cards, modems, etc.). Such communication may be performed through the I / O interface 22. Furthermore, the computer device 12 may also communicate with one or more networks (e.g., local area networks (LANs)), wide area networks (WANs), and / or public networks (e.g., the Internet) through a network adapter 20. Figure 5 As shown, the network adapter 20 communicates with other modules of the computer device 12 via the bus 18. It should be understood that although Figure 5 Not shown, other hardware and / or software modules may be used in conjunction with the computer device 12, including but not limited to: microcode, device drivers, redundant processing units 16, external disk drive arrays, RAID systems, tape drives, and data backup storage systems 34, etc.

[0126] The processing unit 16 executes various functional applications and data processing by running the programs stored in the memory 28, such as implementing the anti-tailgating detection method provided in any embodiment of the present application.

[0127] That is, when the processing unit 16 executes the program, it can achieve: obtaining a set of target images acquired from the target object at at least two preset angles; wherein the preset angles intersect each other; using the first detection model to process the target image set to obtain a first number of people output corresponding to each target image, and determining a first credible output from all the first number of people output; wherein the first detection model detects the target object in the target image through a target detection network; if the number of the first credible outputs is greater than 1, determining the target tracking result based on the first credible output; if the number of the first credible outputs is less than or equal to 1, using the second detection model to process the target image set to obtain a second number of people output corresponding to each target image, and determining the target tracking result based on the second number of people output; wherein the second detection model detects the target object in the target image through a region proposal network and a target detection network.

[0128] In one embodiment of the present application, a computer-readable storage medium is further provided, on which a computer program is stored. When the computer program is executed by a processor, the anti-tailgating detection method provided in any embodiment of the present application is implemented.

[0129] That is, when the program is executed by the processor, it can achieve: obtaining a set of target images obtained by capturing the target object at at least two preset angles; wherein the preset angles intersect each other; using the first detection model to process the target image set to obtain a first number of people output corresponding to each target image, and determining a first credible output from all the first number of people output; wherein the first detection model detects the target object in the target image through a target detection network; if the number of the first credible outputs is greater than 1, determining the target tracking result based on the first credible output; if the number of the first credible outputs is less than or equal to 1, using the second detection model to process the target image set to obtain a second number of people output corresponding to each target image, and determining the target tracking result based on the second number of people output; wherein the second detection model detects the target object in the target image through a region proposal network and a target detection network.

[0130] Any combination of one or more computer-readable media may be used. A computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium may be an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or any combination of the above. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this document, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in conjunction with an instruction execution system, device, or device.

[0131] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, which carry computer-readable program code. Such propagated data signals may take a variety of forms, including electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0132] The computer program code for performing the operation of the present application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" language or similar programming languages. The program code can be executed entirely on the operator's computer, partially on the operator's computer, as an independent software package, partially on the operator's computer, partially on the remote computer, or entirely on the remote computer or server. In the case of a remote computer, the remote computer can be connected to the operator's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (for example, using an Internet service provider to connect through the Internet). The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments, and the same and similar parts between the various embodiments can be referred to each other.

[0133] In one embodiment of the present application, the computer-readable storage medium adopts NKhadas VIM3 Pro based on 6 Cortex cores, 4G memory, and 2.5TOPS NPU.

[0134] Khadas VIM3 Pro is a high-performance computing platform with advanced hardware composition and powerful performance characteristics, suitable for a variety of application scenarios. It uses 6 independent Cortex series CPU cores, which can handle multiple tasks in parallel and improve overall processing power and efficiency; equipped with 4GB of RAM, it can ensure smooth multi-tasking operation and fast response of applications; equipped with 2.5TOPS NPU, it provides 2.5 trillion operations per second of computing power, supports a maximum operating frequency of 800MHz, INT8 inference performance of up to 1536MAC, with 512KB of internal L2 cache and 1MB of system workspace buffer, it can efficiently handle deep learning tasks; compatible with all major deep learning frameworks, such as TensorFlow and Caffe, allowing users to easily implement various AI applications.

[0135] Reference Figure 6 This application can be applied to self-service inspection channels in various inspection places, including ports, airports, high-speed railways, railway stations, communities and high-end scenic spots.

[0136] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the embodiments of the present application.

[0137] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or terminal device including the elements.

[0138] The above is a detailed introduction to the anti-tailgating detection method and device based on a dual detection model provided by the present application. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea; at the same time, for general technical personnel in this field, according to the idea of ​​the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. An anti-tailgating detection method based on a dual detection model, characterized in that: include: Acquire a target image set acquired from capturing the target object at at least two preset angles; wherein the preset angles intersect each other; Processing the target image set using a first detection model to obtain a first number of people output and a first detection confidence corresponding to each target image, and determining a first credible output from all the first number of people outputs according to the first detection confidence; wherein the first detection model detects the target object in the target image through a target detection network; If the number of the first credible outputs is greater than 1, determining a target trailing result according to the first credible outputs; If the number of the first credible outputs is less than or equal to 1, the target image set is processed using a second detection model to obtain a second number of people output corresponding to each target image, and the target tracking result is determined based on the second number of people output; wherein the second detection model detects the target object in the target image through a region proposal network and a target detection network.

2. The method according to claim 1, characterized in that The step of determining a first credible output from all the first number of people outputs according to the first detection confidence level comprises: The first number of people whose first detection confidence level is greater than a first preset value is output as a first credible output.

3. The method according to claim 1, characterized in that The step of determining a target tracking result according to the first credible output includes: Taking the first credible output with the largest value as the first maximum output; If the value of the first maximum output is greater than 1, the target trailing result is set to have trailing; If the value of the first maximum output is less than or equal to 1, the target trailing result is set to no trailing.

4. The method according to claim 1, characterized in that: The step of using the second detection model to process the target image set to obtain a second number of people output corresponding to each target image, and determining the target following result according to the second number of people output includes: Processing the target image set using a second detection model to obtain a second number of people output and a second detection confidence corresponding to each target image; Determining a second credible output from all of the second number of people outputs according to the second detection confidence level; If the number of the second trusted outputs is greater than 1, determining a target trailing result according to the second trusted outputs; If the number of the second credible outputs is less than or equal to 1, the manually identified target tracking result is obtained.

5. The method according to claim 4, characterized in that The step of determining a second credible output from all the second number of people output according to the second detection confidence level comprises: The second number of people whose second detection confidence level is greater than a second preset value is output as a second credible output.

6. The method according to claim 4, characterized in that The step of determining the target tracking result according to the second trusted output includes: Taking the second credible output with the largest value as the second maximum output; If the value of the second maximum output is greater than 1, the target trailing result is set to have trailing; If the value of the second maximum output is less than or equal to 1, the target trailing result is set to no trailing.

7. An anti-tailgating detection device based on a dual detection model, characterized in that: include: A target image acquisition module, used to acquire a target image set acquired from a target object at at least two preset angles; wherein the preset angles intersect each other; a first number detection module, configured to process the target image set using a first detection model to obtain a first number output and a first detection confidence corresponding to each target image, and determine a first credible output from all the first number outputs according to the first detection confidence; wherein the first detection model detects the target object in the target image through a target detection network; a first tailing determination module, configured to determine a target tailing result according to the first credible output if the number of the first credible outputs is greater than 1; A second tailing determination module is used to process the target image set using a second detection model if the number of the first credible outputs is less than or equal to 1, to obtain a second number of people output corresponding to each target image, and to determine a target tailing result based on the second number of people output; wherein the second detection model detects the target object in the target image through a region proposal network and a target detection network.

8. A computer device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein when the computer program is executed by the processor, the method according to any one of claims 1 to 6 is implemented.

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 method according to any one of claims 1 to 6 is implemented.

Images