A method and system for detecting a reflective vest of a person in a large scene

Abstract

The application provides a personnel reflective vest detection method and system in a large scene. The method comprises the following steps: S1, using a main camera to capture a video stream in real time, identifying personnel appearing in the video stream, and detecting, tracking and predicting each personnel to obtain a target position of each personnel after prediction; S2, determining a plurality of associated ball machines according to the target position, and controlling at least one of each associated ball machine to perform a centering and magnifying process on the personnel in a ball machine picture; S3, using a segmentation model to segment the upper body and the reflective vest of the personnel in the ball machine picture after the centering and magnifying process, so as to determine whether the personnel wear the reflective vest; and S4, if the personnel do not wear the reflective vest, outputting an alarm signal. The application can realize accurate detection on whether the personnel wear the reflective vest.

Description

Technical Field

[0001] This invention relates to the field of image recognition technology, and more specifically, to a method and system for detecting reflective clothing worn by people in large scenes. Background Technology

[0002] In large-scale environments such as ports and airports, ensuring the safe operation of personnel is paramount, and ensuring that workers are correctly wearing reflective vests is a key aspect of safety management. However, existing personnel detection and recognition technologies have several limitations when applied to detecting reflective vests in large scenes. These conventional methods suffer from the following problems: First, because people occupy a relatively small proportion of the image, traditional target detection methods struggle to accurately determine whether a person is wearing reflective vests. Second, reflective vests are worn over regular clothing, and the variety of regular clothing types necessitates traditional target detection methods to continuously accumulate samples to handle various clothing types. This approach is not only inefficient and difficult to maintain but also fails to meet accuracy requirements. Finally, in complex scenes, there are often instances of occlusion, overlap, and small proportions of reflective vests on the sides of individuals. In such cases, traditional algorithms struggle to accurately identify the reflective vests of each individual, leading to both false negatives and false positives.

[0003] Therefore, improving the accuracy of detecting reflective vests on personnel is a technical problem that needs to be solved. Summary of the Invention

[0004] In order to at least solve the technical problems existing in the background art, the present invention provides a method, system, electronic device, computer storage medium and computer program product for detecting reflective clothing of people in large scenes.

[0005] The first aspect of the present invention provides a method for detecting reflective clothing worn by people in a large scene, comprising the following steps:

[0006] S1. Use the main camera to capture video stream in real time, identify the people appearing in the video stream, and detect, track and predict each person to obtain the predicted target location of each person.

[0007] S2. Determine a number of associated PTZ cameras based on the target location, and control at least one of the associated PTZ cameras to center and zoom in on the person in the PTZ camera's view.

[0008] S3. In the centered and magnified PTZ camera image, the upper body and reflective vest of the person are segmented using a segmentation model to determine whether the person is wearing a reflective vest.

[0009] S4. If any of the personnel are not wearing reflective clothing, an alarm signal will be output.

[0010] Further, step S1 specifically includes:

[0011] use a human detection model based on deep learning to identify each person in the video stream;

[0012] use a multi-target tracking algorithm to maintain continuous tracking of each detected person;

[0013] According to the pixel speed of each person and the ball machine rotation time t, the target position of each person after t time is predicted.

[0014] Further, step S2 specifically includes:

[0015] According to the installation position and coverage area of the main camera, determine a plurality of ball machines associated with the main camera;

[0016] Control each ball machine to rotate to the target position of the predicted person;

[0017] Secondary identification of the person in the ball machine screen of each ball machine, and determine the ball machine that identifies the person as the associated ball machine;

[0018] Adjust the shooting parameters of each associated ball machine so that the person is located at the center position of the ball machine screen, and perform zoom-in processing on the ball machine screen.

[0019] Further, further, step S3 specifically includes:

[0020] In the ball machine screen after the centering and zooming-in processing, use a human detection model based on deep learning to identify the body image of the person;

[0021] Use a segformer semantic segmentation model to segment the upper body image of the person from the body image, and use a segformer semantic segmentation model to segment the reflective vest region image in the upper body image, to determine whether the person wears a reflective vest.

[0022] Further, the use of a segformer semantic segmentation model to segment the upper body image of the person from the body image includes:

[0023] Determine whether the pixel aspect ratio of the person based on the body image is less than a first preset value;

[0024] If yes, determine that the person is in a non-side state, and use a segformer semantic segmentation model to segment the upper body image of the person from the body image;

[0025] If not, it is determined that the person is in a side state, and the result of the current frame is labeled, and the corresponding result weight of the frame is correspondingly reduced when subsequent multi-frame result fusion is performed.

[0026] Further, the using the segformer semantic segmentation model to segment a reflective vest region image in the upper body image, so as to determine whether the personnel wear a reflective vest comprises:

[0027] Using the segformer semantic segmentation model to segment the reflective vest region image in the upper body image, and calculating the proportional relationship of the total pixel number N of the segmented reflective vest region and the total pixel M of the upper body image of the personnel;

[0028] If the proportional relationship is greater than a second preset value, it is determined that the personnel in the frame wear a reflective vest, otherwise it is determined that the personnel in the frame do not wear a reflective vest;

[0029] The number of frames determined to wear a reflective vest by the personnel accounts for a percentage of the total number of all frames, and if the percentage is higher than a third preset value, it is determined that the personnel wear a reflective vest, otherwise it is determined that the personnel do not wear a reflective vest.

[0030] Optionally, in step S4, specifically comprising:

[0031] Multiple people in the ball machine result summary: when multiple people are found in the associated ball machine, one person is found not wearing a reflective vest, and an alarm signal is output;

[0032] Multiple ball machine result summary:

[0033] 1) When the number of associated ball machines is 1, the analysis result of the current associated ball machine on each personnel is directly output; wherein the analysis result includes whether the personnel wear a reflective vest, and the corresponding alarm signal when the personnel do not wear a reflective vest;

[0034] 2) When the number of associated ball machines is greater than 1:

[0035] a) When the number of personnel in all associated ball machines is 1, the analysis result of the associated ball machine with the most analysis result is output; if the number of analysis results of each associated ball machine is the same, the analysis result of the associated ball machine that can see the non-side of the personnel is output;

[0036] b) If the number of personnel in the associated ball machine is greater than 1, if any ball machine generates an alarm, it is considered that the personnel do not wear a reflective vest.

[0037] The second aspect of the present application provides a personnel reflective vest detection system in a large scene, comprising a main camera, a plurality of associated ball machines, a processing module, and a storage module; the processing module is connected with the main camera, each ball machine, and the storage module;

[0038] The storage module is configured to store executable computer program codes.

[0039] The main camera and the associated spherical camera are configured to capture pictures containing people and transmit the pictures to the processing module.

[0040] The processing module is configured to execute the method according to any one of the preceding embodiments by invoking the executable computer program codes in the storage module.

[0041] The third aspect of the present application provides an electronic device, comprising a memory storing executable program codes, a processor coupled to the memory, and the processor invokes the executable program codes stored in the memory to execute the method according to any one of the preceding embodiments.

[0042] The fourth aspect of the present application provides a computer storage medium, which stores a computer program, and the computer program is executed by a processor to execute the method according to any one of the preceding embodiments.

[0043] The fifth aspect of the present application further provides a computer program product, which comprises a computer program stored on a non-transitory computer readable medium, and the computer program is executed by a processor to execute the method according to any one of the preceding embodiments.

[0044] The present application has at least the following beneficial effects:

[0045] 1) The detection of the wearing of reflective clothing by small targets (people) in large scene images can be realized, avoiding the problem of inaccurate detection caused by the small image proportion of people;

[0046] 2) The limitations of conventional target detection algorithms based on deep learning on the dependence of clothing types are broken, reducing the problem of excessive sample demand and model updating difficulty caused by the large variety of clothing types;

[0047] 3) It can effectively handle the problems of people occlusion, overlap, and small proportion of reflective clothing on the side of people in large scene videos, reduce the false negative and false positive rates, and improve the accuracy and reliability of the overall detection;

[0048] 4) It has strong adaptability and scalability and can be used for the detection of reflective clothing of people in various large scenes. BRIEF DESCRIPTION OF DRAWINGS

[0049] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some embodiments of the present application, and therefore should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can also be obtained without creative labor on the basis of these drawings.

[0050] Figure 1 is a flow diagram of a method for detecting a reflective vest of a person in a large scene according to an embodiment of the present application.

[0051] Figure 2 is a structural diagram of a system for detecting a reflective vest of a person in a large scene according to an embodiment of the present application.

[0052] Figure 3 is a structural diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0053] In order to make the objectives, technical solutions and advantages of the present application clearer, the following will further describe the present application with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative effort belong to the scope of the present application.

[0054] The terms used in the embodiments of the present application are merely for the purpose of describing particular embodiments and are not intended to limit the present application. The singular forms "a", "said" and "the" used in the embodiments of the present application and the appended claims are also intended to include the plural forms, unless the context clearly indicates otherwise. "Plural" generally includes at least two.

[0055] It should be understood that the term "and / or" used herein merely describes an association relationship of associated objects, which means that there can be three relationships, for example, A and / or B can mean that A exists alone, A and B exist together, and B exists alone. In addition, the character " / " herein generally represents an "or" relationship between the front and rear associated objects.

[0056] It should be understood that although the terms first, second, third, etc. may be used in the embodiments of the present application to describe …, these … should not be limited to these terms. These terms are only used to distinguish … from each other. For example, without departing from the scope of the embodiments of the present application, the first … can also be called the second …, and similarly, the second … can also be called the first ….

[0057] Depending on the context, the word "if" as used herein can be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting". Similarly, depending on the context, the phrase "if determined" or "if detecting (a stated condition or event)" can be interpreted as "when determined" or "in response to determining" or "when detecting (a stated condition or event)" or "in response to detecting (a stated condition or event)".

[0058] It is also important to note that the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus. An element proceeded by "comprises... a" does not, without more constraints, exclude the presence of additional identical elements in the process, method, article, or apparatus that comprises the element.

[0059] The preferred embodiments of the present application will be described in detail below with reference to the drawings.

[0060] As shown in the Figure 1 The embodiment of the present application discloses a method for detecting reflective vest of personnel in a large scene, comprising the following steps:

[0061] S1, using a main camera to capture a video stream in real time, identifying personnel appearing in the video stream, and detecting, tracking and predicting each personnel to obtain a target position of each personnel after prediction;

[0062] S2, determining a plurality of associated ball machines according to the target position, and controlling at least one of the associated ball machines to perform centering and magnification processing on the personnel in the ball machine picture;

[0063] S3, in the ball machine picture after the centering and magnification processing, using a segmentation model to segment the upper body and the reflective vest of the personnel respectively, so as to judge whether the personnel wears a reflective vest or not;

[0064] S4, if the personnel does not wear a reflective vest, an alarm signal is output.

[0065] Compared with the existing method mentioned in the background art, the above-mentioned scheme of the present application respectively uses personnel tracking, multi-ball machine linkage, picture centering and magnification processing, target segmentation and other means, which can realize the detection of the reflective vest wearing of small targets (personnel) in a large scene image, avoid the problem of inaccurate detection caused by small image proportion of personnel, and effectively solve the problems of personnel occlusion, overlap, small proportion of reflective vest on the side of personnel in a large scene video, reduce the false negative and false positive rates, and improve the accuracy and reliability of the overall detection.

[0066] It should be noted that the main camera in the present application can be a pre-set camera for identifying personnel, or any camera in the monitoring area, i.e. a gun camera or a ball camera, which is not limited in the present application.

[0067] Further, step S1 specifically comprises:

[0068] using a human body detection model based on deep learning to identify each personnel in the video stream;

[0069] maintain continuous tracking on each of the detected persons by using a multi-target tracking algorithm;

[0070] According to the pixel speed of each of the persons and the ball machine rotation time t, the target position of each of the persons after t time is predicted.

[0071] In this embodiment, this step is to ensure that the persons recognized can be in the ball machine picture after the ball machine is zoomed in in the subsequent step, and sufficient analysis time is reserved for the analysis of the reflective vest after the ball machine is zoomed in. The human body detection model can be, for example, YOLO11, and the ball machine rotation time t can be prior knowledge, which is estimated according to different ball machine positions and other information in the early stage.

[0072] An improved scheme of the embodiment is provided: the continuous tracking on each of the detected persons by using a multi-target tracking algorithm comprises:

[0073] Each of the persons is continuously tracked within a preset time length, and the fluctuation amplitude of the number of the persons within the preset time length is counted, and the extension time length is determined according to the fluctuation amplitude; wherein the extension time length is positively correlated with the fluctuation amplitude;

[0074] The continuous tracking on each of the persons is continued within the extension time length.

[0075] When there are many persons in the video stream, the prediction of the target position of each of the persons after t time can be deviated due to overlapping, shielding and other situations, and the deviation can be larger as the number of persons in the video stream increases. To this end, the present application firstly continuously tracks each of the persons within a preset time length (for example, 5s), counts the fluctuation amplitude of the number of the persons within the preset time length (overlapping, shielding and other situations can cause the number of persons in each video image to fluctuate, decrease or increase), and determines the extension time length according to the fluctuation amplitude (absolute value) in a positive correlation, that is, the longer the extension time length is set, the larger the fluctuation amplitude is, that is, the longer the tracking time is to ensure the accuracy of the prediction of the target position of the persons.

[0076] Further, step S2 specifically comprises:

[0077] According to the installation position and coverage area of the main camera, a plurality of ball machines associated with the main camera are determined;

[0078] Each of the ball machines is controlled to rotate to the target position of the predicted person;

[0079] The person is identified again in the ball machine picture of each of the ball machines, and the ball machine in which the person is recognized is determined as the associated ball machine;

[0080] The shooting parameters of each of the associated ball cameras are regulated so that the person is located in the center of the ball camera image, and the ball camera image is enlarged.

[0081] In this embodiment, part of the cameras can be associated, i.e., offline bound, in advance by considering factors such as proximity of installation positions, proximity of coverage areas, and whether they are located in a personnel walkable area, so as to facilitate subsequent triggering of the linkage between the main camera and the associated ball cameras, and to ensure that the ball cameras can capture the person from multiple angles.

[0082] Since the rotation and zoom operations of the ball camera all require a certain time, which is related to the rotation mechanism of the ball camera itself and the degree of rotation and zoom required, in order to ensure that the target is as centered as possible in the ball camera image after rotation, the ball camera needs to be rotated to the target position of the person after t seconds. Since the rotation speed of the ball camera is usually fast, and the zoom is slow, t is set in relation to the zoom (Z) here, and a (ti, Zi) table is usually set according to the actual situation, and the corresponding t value is obtained by using the table lookup method in actual use. Thus, the position of the corresponding person after the rotation and zoom of the ball camera are completed is calculated according to the prediction method in step S1.

[0083] In addition, it needs to be further explained that:

[0084] 1) In order to ensure that the person can be clearly seen from multiple angles, usually at least two ball cameras are linked (from the perspective of the person at mutually perpendicular angles), and more ball cameras (such as four) can be called according to the actual situation on site to view the person from multiple angles in front, back, left and right;

[0085] 2) Multi-angle viewing of the ball camera can solve the problem that one or more persons are blocked at some angles, thereby solving the problem of false reporting caused by conventional methods;

[0086] 3) Multi-angle viewing of the ball camera can also solve the problem of false reporting caused by the person appearing sideways in the video at a single angle (when the person appears sideways, the proportion of the reflective vest in the person's pixels is extremely small, which may cause the models on the market to fail to accurately detect);

[0087] 4) Multi-angle viewing of the ball camera can avoid the problem of overlap of the person in the video at some angles (the probability of overlap of the person at multiple angles will be greatly reduced), which will cause the conventional reflective vest detection algorithm on the market to detect that the person region carries part of the body of other persons (including the reflective vest), thereby causing false reporting or missing reporting problems.

[0088] Further, step S3 specifically includes:

[0089] In the ball machine picture after the centering and magnifying processing, a human body detection model based on deep learning is used to identify the body image of the personnel;

[0090] A segformer semantic segmentation model is used to segment the upper body image of the personnel from the body image, and a segformer semantic segmentation model is used to segment the reflective vest region image in the upper body image, so as to determine whether the personnel wears a reflective vest.

[0091] In this embodiment, after the identified personnel is processed by centering and magnifying, the personnel in the ball machine picture after the centering and magnifying processing needs to be identified again (also YOLO11 can be used), so as to determine the body image of the personnel in the ball machine picture after the centering and magnifying processing. Then, a preset segformer semantic segmentation model is used to segment the upper body image of the personnel from the body image, and the reflective vest region image from the upper body image.

[0092] Further, the use of the segformer semantic segmentation model to segment the upper body image of the personnel from the body image comprises:

[0093] Based on the body image, it is determined whether the pixel aspect ratio of the personnel is less than a first preset value;

[0094] If yes, it is determined that the personnel is in a non-side state, and a segformer semantic segmentation model is used to segment the upper body image of the personnel from the body image;

[0095] If no, it is determined that the personnel is in a side state, and the result of the current frame is labeled, and the corresponding result weight of the frame is correspondingly reduced when the multi-frame result fusion is performed subsequently.

[0096] In this embodiment, in order to prevent the model from being inaccurate due to the personnel turning sideways, it is necessary to determine whether the personnel is in a side state when the segformer semantic segmentation model is used to segment the upper body of the personnel. For example, it is determined whether the pixel aspect ratio of the personnel is less than 6 (a first preset value), if yes, it indicates a non-side state, otherwise, it indicates a side state.

[0097] If the person is determined to be in a front-facing state, the segmentation model can be directly used to better segment the upper body of the person, thereby solving the problem of false positives that may occur in subsequent reflective clothing detection caused by the introduction of other personnel bodies and clothing in the upper body bounding box of the person detected by the conventional detection model when the personnel overlap. If the person is determined to be in a side-facing state, the result of the current frame (i.e., whether the reflective clothing is worn) needs to be labeled, and the results of multiple frames need to be fused in the future. The weight of the result corresponding to each frame is different, and the fusion weight corresponding to the frame in the side-facing state is correspondingly reduced based on the labeling.

[0098] Further, the using the segformer semantic segmentation model to segment the reflective clothing area image in the upper body image to determine whether the person wears the reflective clothing comprises:

[0099] using the segformer semantic segmentation model to segment the reflective clothing area image in the upper body image, and calculating the proportional relationship between the total number of pixels N of the segmented reflective clothing area and the total pixels M of the upper body image of the person;

[0100] If the proportional relationship is greater than a second preset value, it is determined that the person in the frame wears the reflective clothing, otherwise it is determined that the person in the frame does not wear the reflective clothing;

[0101] Calculating the percentage of the number of frames determined to wear the reflective clothing to the total number of all frames, if the percentage is higher than a third preset value, it is determined that the person wears the reflective clothing; otherwise, it is determined that the person does not wear the reflective clothing.

[0102] In this embodiment, by segmenting the total number of pixels N of the reflective clothing area and the total pixels M of the upper body image of the person, if the proportional relationship is greater than 0.3 (the second preset value), it is determined that the reflective clothing is worn, otherwise it is determined that the reflective clothing is not worn, and the determination result is continuously determined in multiple frames, for example, 50 frames, greater than 60% (the third preset value) to determine that the reflective clothing is worn, thereby comprehensively determining whether the person wears the reflective clothing under the current ball camera view.

[0103] The main reason for selecting the segformer semantic segmentation model here is:

[0104] Efficiency: Segformer is relatively efficient in terms of computing resources and model parameters. It reduces the number of model parameters and computational complexity through a simple decoder design and hierarchical Transformer encoder. Compared with some traditional CNN-based complex segmentation models, it has faster inference speed while maintaining high segmentation accuracy, and is suitable for application in scenarios with high real-time requirements.

[0105] Multi-scale feature utilization: thanks to its hierarchical encoder structure, Segformer can make good use of multi-scale features. In the semantic segmentation task, different scales of objects and scene details can be effectively captured and processed, and this way of multi-scale feature fusion helps to improve the segmentation accuracy of objects of different sizes.

[0106] Adaptability and generalization ability: due to the characteristics of the Transformer architecture itself, Segformer has good adaptability to different types of image data and segmentation tasks. It does not rely on fixed convolution kernels and local receptive fields like some CNN models based on handcrafted features or trained on specific datasets. The self-attention mechanism of Transformer can automatically learn the long-distance dependencies between features, making the model have better generalization ability when facing different image distributions and task requirements.

[0107] Optionally, in step S4, specifically comprising:

[0108] Multiple people in the ball machine result summary: when multiple people are found in the associated ball machine, an alarm signal is output if one person is not wearing a reflective vest;

[0109] Multi-ball machine result summary:

[0110] 1) When the number of associated ball machines is 1, the analysis result of the current associated ball machine for each personnel is directly output; wherein the analysis result includes the result of whether the personnel is wearing a reflective vest, and the corresponding alarm signal when the personnel is not wearing a reflective vest;

[0111] 2) When the number of associated ball machines is greater than 1:

[0112] a) When there is only one person in all associated ball machines, the analysis result of the associated ball machine with the most analysis results is output; if the number of analysis results of each associated ball machine is the same, the analysis result of the associated ball machine that can see the non-side of the personnel is output (when the number of linked ball machines is 2 or more, at least two ball machine angles are basically perpendicular to each other, that is, there must be at least one ball machine that can see the front or back of the personnel);

[0113] b) If the number of personnel in the associated ball machine is greater than 1 (in this case, it indicates that there may be problems such as occlusion and overlap), if any ball machine generates an alarm, it is considered that the personnel is not wearing a reflective vest.

[0114] For example, Figure 2As shown in the figure, this invention discloses a system for detecting reflective clothing on people in large scenes, including a main camera, several associated PTZ cameras, a processing module, and a storage module; the processing module is connected to the main camera, each of the PTZ cameras, and the storage module.

[0115] The storage module is used to store executable computer program code;

[0116] The main camera and the associated PTZ camera are used to capture images containing people and transmit them to the processing module.

[0117] The processing module is configured to execute the method described in the preceding one by invoking the executable computer program code in the storage module.

[0118] The specific functions of the large-scale personnel reflective clothing detection system in this embodiment are the same as those in the above embodiments. Since the system in this embodiment adopts all the technical solutions of the above embodiments, it has at least all the beneficial effects brought about by the technical solutions of the above embodiments, which will not be described in detail here.

[0119] like Figure 3 As shown, an embodiment of the present invention discloses an electronic device, including: a memory storing executable program code; a processor coupled to the memory; the processor calls the executable program code stored in the memory to execute the method as described in the foregoing embodiments.

[0120] This invention also discloses a computer storage medium storing a computer program, which is executed by a processor to perform the methods described in the foregoing embodiments.

[0121] This invention also discloses a computer program product, including a computer program stored on a non-transitory computer-readable medium, the computer program being executed by a processor as described in the foregoing embodiments.

[0122] The apparatus / system according to embodiments of the present disclosure can include a processor, a memory for storing program data and executing the program data, a permanent memory such as a disk drive, a communication port for processing communication with an external apparatus, and a user interface apparatus, etc. The method is implemented as a software module or can be stored as computer-readable code or program commands executable by a processor on a computer-readable recording medium. Examples of the computer-readable recording medium can include magnetic storage media (e.g., read only memory (ROM), random access memory (RAM), floppy disks, hard disks, etc.), optical recording media (e.g., CD-ROMs, digital versatile disks (DVDs), etc.), etc. The computer-readable recording medium can be distributed among computer systems connected to a network, and the computer-readable code can be stored and executed in a distributed manner. The medium can be computer-readable, stored in the memory and executed by the processor.

[0123] Embodiments of the present disclosure can be directed to functional block components and various processing operations. The functional blocks can be implemented as various numbers of hardware and / or software components that perform specific functions. For example, embodiments of the present disclosure can implement direct circuit components that can perform various functions under the control of one or more microprocessors or other control apparatus, such as memories, processing circuits, logic circuits, lookup tables, etc. The components of the present disclosure can be implemented by software programming or software components. Similarly, embodiments of the present disclosure can include various algorithms implemented by combinations of data structures, processes, routines or other programming components, and can be implemented by programming or scripting languages such as C, C++, Java, assembler, etc. Functional aspects can be implemented by algorithms executed by one or more processors. Furthermore, embodiments of the present disclosure can implement related technologies for electronic environment settings, signal processing and / or data processing. Terms such as "mechanism," "element," "unit," etc. can be used broadly and are not limited to mechanical and physical components. These terms can represent a series of software routines related to a processor or the like.

[0124] Specific embodiments are described in the present disclosure as examples, and the scope of the embodiments is not limited thereto.

[0125] Although embodiments of the present disclosure have been described, it will be understood by those of ordinary skill in the art that various changes in form and details can be made therein without departing from the spirit and scope of the present disclosure defined by the appended claims. Therefore, the above-described embodiments of the present disclosure should be construed as examples, and do not limit the embodiments in all aspects. For example, each component described as a single unit can be performed in a distributed manner, and likewise, components described as distributed can be performed in a combined manner.

[0126] All examples or example terms (e.g., and the like) used herein are intended to be descriptive of the embodiments of the disclosure, and not to limit the scope of the embodiments of the disclosure.

[0127] Also, unless explicitly described otherwise, expressions such as "essential", "important", or the like associated with certain components can not indicate an absolute need for the components.

[0128] Those of ordinary skill in the art will understand that the embodiments of the disclosure can be implemented in modified forms without departing from the spirit and scope of the disclosure.

[0129] Since the disclosure allows various changes to be made to the embodiments of the disclosure, the disclosure is not limited to specific embodiments, and it will be understood that all changes, equivalents, and alternatives that do not depart from the spirit and technical scope of the disclosure are included in the disclosure. Accordingly, the embodiments of the disclosure described herein should be understood in all aspects as examples, and should not be interpreted as limiting.

[0130] In addition, terms such as "unit", "module", and the like mean a unit that processes at least one function or operation that can be implemented as hardware or software or a combination of hardware and software. The "unit" and "module" can be stored in a storage medium to be addressed, and can be implemented as a program that can be executed by a processor. For example, the "unit" and "module" can refer to components such as software components, object-oriented software components, class components, and task components, and can include processes, functions, properties, procedures, sub-routines, program code segments, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays, or variables.

[0131] In the disclosure, the expression "A can include one of a1, a2, and a3" can broadly mean that examples that can be included in element A include a1, a2, or a3. This expression should not be interpreted as limited to the meaning that examples included in element A must be limited to a1, a2, and a3. Therefore, examples included in element A should not be interpreted as excluding elements other than a1, a2, and a3. In addition, the expression means that element A can include a1, a2, or a3. The expression does not mean that the elements included in element A must be selected from a specific set of elements. That is, the expression should not be interpreted restrictively as meaning that a1, a2, or a3 must be included in element A selected from a set including a1, a2, and a3.

[0132] Further, in the present disclosure, the expression "at least one of a1, a2, and / or a3" indicates one of "a1", "a2", "a3", "a1 and a2", "a1 and a3", "a2 and a3", and "a1, a2, and a3". Accordingly, it should be noted that the expression "at least one of a1, a2, and / or a3" should not be interpreted to mean "at least one of a1", "at least one of a2", and "at least one of a3", unless explicitly described as such.

Claims

1. A method for detecting a reflective vest of a person in a large scene, characterized in that, The method comprises the following steps: S1, capturing a video stream in real time using a main camera, identifying personnel appearing in the video stream, and detecting, tracking and predicting each of the personnel to obtain a target position of each of the personnel after prediction; S2, determining a plurality of associated ball cameras according to the target position, and controlling a plurality of ball cameras in each of the associated ball cameras to perform centering and magnification processing on the personnel in the ball camera image; S3, in the ball camera image after centering and magnification processing, using a segmentation model to segment the upper body and reflective vest of the personnel respectively, so as to determine whether the personnel wear a reflective vest, specifically comprising: In the ball camera image after centering and magnification processing, using a human body detection model based on deep learning to identify the body image of the personnel; Using a segformer semantic segmentation model to segment the upper body image of the personnel from the body image, and using a segformer semantic segmentation model to segment the reflective vest region image in the upper body image, so as to determine whether the personnel wear a reflective vest; S4, if the personnel do not wear a reflective vest, output an alarm signal; Step S1, specifically comprising: Using a human body detection model based on deep learning to identify each personnel in the video stream; Using a multi-target tracking algorithm to maintain continuous tracking of each of the detected personnel; According to the pixel speed of each of the personnel and the ball camera rotation time t, the target position of each of the personnel after t time is predicted; The multi-target tracking algorithm is used to maintain continuous tracking of each of the detected personnel, comprising: Continuously tracking each of the personnel within a preset time length, and counting the number fluctuation amplitude of the personnel within the preset time length, and determining the extension time length according to the number fluctuation amplitude; wherein the extension time length is positively correlated with the number fluctuation amplitude; Continuously tracking each of the personnel within the extension time length.

2. The method of claim 1, wherein the method comprises: Step S2, specifically comprising: Pre-determining a plurality of ball cameras associated with the main camera according to the installation position and coverage area of the main camera; Controlling each of the ball cameras to rotate to the target position of the personnel after prediction; Identifying the personnel again in the ball camera image of each of the ball cameras, and determining the ball camera that identifies the personnel as the associated ball camera; Controlling the shooting parameters of each of the associated ball cameras to make the personnel located at the center position of the ball camera image, and performing magnification processing on the ball camera image.

3. The method of claim 1, wherein the method comprises: The segformer semantic segmentation model is used to segment the upper body image of the personnel from the body image, comprising: Determining whether the pixel aspect ratio of the personnel is less than a first preset value based on the body image; If yes, it is determined that the personnel is in a non-side state, and the segformer semantic segmentation model is used to segment the upper body image of the personnel from the body image; If not, it is determined that the personnel is in a side state, and the result of the current frame is labeled, and the result weight corresponding to the frame is correspondingly reduced when subsequent multi-frame result fusion is performed.

4. The method of claim 1, wherein the method further comprises: The using the segformer semantic segmentation model to segment a reflective vest region image in the upper body image, so as to determine whether the personnel wear a reflective vest comprises: Using the segformer semantic segmentation model to segment a reflective vest region image in the upper body image, and calculating the proportional relationship between the total pixel number N of the segmented reflective vest region and the total pixel M of the upper body image of the personnel; If the proportional relationship is greater than a second preset value, it is determined that the personnel in the frame wear a reflective vest, otherwise it is determined that the personnel in the frame do not wear a reflective vest; The percentage of the number of frames determined to wear a reflective vest to the total number of all frames is calculated, and if the percentage is higher than a third preset value, it is determined that the personnel wear a reflective vest, otherwise it is determined that the personnel do not wear a reflective vest.

5. The method of claim 1, wherein: Step S4 specifically comprises: Multi-person result summary in the ball machine: when multiple people are found in the associated ball machine, one person is found not wearing a reflective vest, and an alarm signal is output; Multi-ball machine result summary: 1) When the number of associated ball machines is 1, the analysis result of the current associated ball machine for each personnel is directly output; wherein the analysis result includes whether the personnel wear a reflective vest, and the corresponding alarm signal when the personnel do not wear a reflective vest; 2) When the number of associated ball machines is greater than 1: a) When the number of personnel in all associated ball machines is 1, the analysis result of the associated ball machine with the most analysis result is output; if the number of analysis results of each associated ball machine is the same, the analysis result of the associated ball machine that can see the non-side of the personnel is output; b) If the number of personnel in the associated ball machine is greater than 1, if any ball machine generates an alarm, it is considered that the personnel do not wear a reflective vest.

6. A personnel reflective vest detection system in a large scene, comprising a main camera, a plurality of associated ball machines, a processing module, and a storage module; the processing module is connected with the main camera, each ball machine, and the storage module; The storage module is used to store executable computer program codes; The main camera and the associated ball machines are used to shoot pictures containing personnel and transmit them to the processing module; characterized in that The processing module is used to execute the method of any one of claims 1-5 by calling the executable computer program codes in the storage module, so as to realize the superimposed display of the cargo content in the video image.

7. An electronic device, comprising: A memory storing executable program codes; A processor coupled with the memory; characterized in that: the processor calls the executable program codes stored in the memory to execute the method of any one of claims 1-5.

8. A computer storage medium having stored thereon a computer program, characterized in that: The computer program is executed by the processor to execute the method of any one of claims 1-5.

Images