A pre-work meeting intelligent monitoring method and system

Abstract

The present invention discloses a method and system for intelligent monitoring of pre-work meetings, which relate to the fields of production safety monitoring and deep learning. During the set meeting time, a video acquisition unit is used to acquire video of the meeting room, and a personnel detection unit is used to detect personnel on the acquired video images; timing is performed after the number of personnel in the meeting area exceeds the start threshold, and timing is ended after the number of personnel is lower than the end threshold; if the timing time exceeds the set time threshold, it indicates that the pre-work meeting is valid, and a pre-work meeting record is generated and archived; personnel detection is continuously performed during the set meeting time. The personnel detection unit includes a backbone feature extraction network, a multi-scale fusion network, and a feature decoder connected in sequence. The present invention can not only accurately detect personnel in a complex environment, but also monitor the start and end of the meeting in real time, and automatically record the duration of the meeting and the number of participants.

Description

Technical Field

[0001] The present invention relates to the field of production safety monitoring and deep learning technology, and in particular to a pre-work meeting intelligent monitoring method and system. Background Art

[0002] In the field of safety production monitoring, especially in construction and engineering projects, the effective execution of pre-work meetings is crucial to ensure construction safety and improve work efficiency. Pre-work meetings allow teams to discuss the risks, task allocation and safety measures of the upcoming work, and are a key step in preventing accidents and misunderstandings. However, in traditional meeting monitoring systems, manual inspections and reports are often relied on, which is not only time-consuming but also prone to omissions due to human factors.

[0003] In existing technologies, such as the "System and method for monitoring and reporting safety meetings in the workplace" disclosed in US Patent US20140249754A1, a solution for monitoring meetings in the workplace using video recording equipment is proposed. This technology can record the participation in meetings through video monitoring, but it mainly relies on video recording without in-depth image analysis or personnel recognition, so in practical applications it may not be able to accurately count the number of participants or evaluate the effectiveness of the meeting.

[0004] In addition, some current monitoring systems use basic image recognition technology to detect the presence of people, such as the "system and method for detecting human bodies" disclosed in European Patent EP3009118B1. Such systems usually use traditional image processing techniques, such as background subtraction or simple motion detection, to identify people in images. However, these methods are prone to misjudgment in complex construction environments (such as multiple people moving, light changes, etc.).

[0005] In summary, the existing technology mainly has the following disadvantages:

[0006] 1. High dependence on manual work: Most existing pre-work meeting monitoring technologies rely on manual execution and monitoring, such as confirming meeting participation through manual recording or inspection. This method is not only inefficient, but also easily affected by subjective judgment and human negligence, resulting in inaccurate records.

[0007] 2. Primitive technology: While some systems use video surveillance to track meetings, these systems often lack advanced image processing capabilities and are unable to accurately count the number of people in a meeting or identify specific participants, especially in environments with large crowds or limited field of view.

[0008] 3. Lack of adaptability and flexibility: Traditional monitoring systems usually do not have the ability to handle complex and dynamically changing environments, such as light changes, frequent entry and exit of people, etc., which will greatly reduce the accuracy and reliability of the system.

[0009] 4. Low degree of automation: In existing technologies, the start and end of meetings often require manual judgment or rely on a fixed schedule. It lacks the ability to adaptively adjust and cannot automatically adjust the monitoring status according to actual conditions, which greatly limits its effectiveness in practical applications. Summary of the invention

[0010] In order to overcome the defects in the above-mentioned prior art, the present invention provides a method for intelligent monitoring of pre-work meetings, which improves the accuracy and efficiency of meeting monitoring and provides an innovative technical solution for the field of safe production.

[0011] To achieve the above object, the present invention adopts the following technical solutions, including:

[0012] A pre-work meeting intelligent monitoring method comprises the following steps:

[0013] S1, according to the set meeting time, perform personnel detection in the meeting room within the set meeting time;

[0014] S2, counting the number of people in the meeting area, determining whether the number of people in the meeting area exceeds the starting threshold, if so, proceeding to step S3; if not, returning to step S1, continuing to perform personnel detection until the number of people in the meeting area exceeds the starting threshold, and then proceeding to step S3;

[0015] S3: When the number of people in the meeting area exceeds the start threshold, it means that the pre-work meeting officially begins. At this time, the timer starts, and during the timer period, the number of people in the meeting area is monitored in real time until the number of people in the meeting area is less than the end threshold, and the timer ends;

[0016] S4, judging whether the current timing time exceeds the set time threshold, if so, it means that the current pre-work meeting is valid, and the pre-work meeting record is generated and archived; if not, it means that the current pre-work meeting is invalid, and the pre-work meeting record is not generated;

[0017] S5, determine whether it is within the set meeting time. If it is within the set meeting time, return to step S1 and continue to detect personnel. After the number of people in the meeting area exceeds the starting threshold again, start timing again. If the timing time exceeds the set time threshold again, generate the pre-work meeting record again and archive it; if it exceeds the set meeting time, stop personnel detection.

[0018] Preferably, in step S1, a video acquisition unit is used to acquire video of the conference room, and a personnel detection unit is used to perform personnel detection on the acquired video images.

[0019] Preferably, in step S4, the video image data, meeting duration and number of people of the pre-job meeting are recorded, a pre-job meeting record is generated, and the pre-job meeting record is sent to the management personnel for review.

[0020] Preferably, the personnel detection unit comprises a backbone feature extraction network, a multi-scale fusion network, and a feature decoder connected in sequence; wherein,

[0021] The backbone feature extraction network includes four parts connected in sequence, each of which is composed of a downsampling layer and a multi-level CSP residual module; the output end of the downsampling layer in each part is connected to the input end of the multi-level CSP residual module; the output end of the multi-level CSP residual module in the previous part is connected to the input end of the downsampling layer in the next part; the input end of the downsampling layer of the first part inputs the image to be detected;

[0022] The features output by the multi-level CSP residual modules in the last three parts of the backbone feature extraction network are fused through a multi-scale feature fusion network to achieve the fusion of feature maps of different scales.

[0023] The feature decoder processes the fused feature map, extracts the information required for person detection, and generates person detection results.

[0024] Preferably, the main structure of the multi-level CSP residual module is composed of four residual modules connected in series, each residual module is equipped with a CSP branch structure, the output result of the main structure is fused with the output result of each branch structure through a 1×1 standard convolution module CBA, and then the input and the fused output are combined through the residual edge as the output of the multi-level CSP residual module;

[0025] The CSP branch structure is composed of a 1×1 standard convolution module CBA; the residual module includes a 1×1 standard convolution module CBA and an n×n deep convolution module DBA connected in series, the input of the residual module is processed by the 1×1 standard convolution module CBA and the n×n deep convolution module DBA in sequence, and the output of the deep convolution module DBA is added to the input of the residual module as the output of the residual module;

[0026] The 1×1 standard convolution module CBA is composed of a 1×1 standard convolution layer, a batch normalization layer and an activation function connected in sequence;

[0027] The n×n deep convolution module DBA is composed of an n×n deep convolution layer, a batch normalization layer, an activation function, a 1×1 standard convolution layer, a batch normalization layer and an activation function connected in sequence.

[0028] Preferably, the multi-scale feature fusion network adopts a PANet structure.

[0029] Preferably, the SiLU activation function is adopted.

[0030] The present invention also provides a pre-work meeting intelligent monitoring system, which is applicable to the above-mentioned pre-work meeting intelligent monitoring method, and the system includes: a video acquisition unit, a personnel detection unit, a central processing unit and a storage unit;

[0031] The video acquisition unit is used to acquire video of the conference room;

[0032] The personnel detection unit is used to perform personnel detection on the collected video images;

[0033] The central processing unit is used to determine whether the pre-work meeting is valid according to the number of people in the meeting area, and record the video image data, meeting duration and number of people of the valid pre-work meeting to form a pre-work meeting record;

[0034] The storage unit is used to store the pre-work meeting records.

[0035] The present invention also provides a computer program product, which includes a computer program / instruction, and when the computer program / instruction is executed by a processor, the above-mentioned pre-work meeting intelligent monitoring method is implemented.

[0036] The present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed, the above-mentioned pre-work meeting intelligent monitoring method is implemented.

[0037] The advantages of the present invention are:

[0038] (1) The present invention can automatically start and end the meeting timer according to the number of people in the meeting area, and record the duration of the meeting, thus achieving real-time monitoring and management. The present invention provides accurate data support for managers by automatically saving key meeting images and generating detailed meeting records, helping managers to better conduct quality control and security review.

[0039] (2) The present invention designs a dedicated deep neural network training model for personnel detection, which improves the generalization ability and stability of the model under different lighting and dynamic environments, automatically and accurately identifies and counts the number of meeting participants in complex environments, reduces manual intervention, and improves data accuracy.

[0040] (3) The present invention proposes a highly automated and intelligent solution, namely, an intelligent monitoring system for pre-work meetings based on deep neural networks. By designing a dedicated target detection algorithm, the present invention can not only accurately detect people in complex environments, but also monitor the start and end of meetings in real time, and automatically record the duration of meetings and the number of participants. The present invention significantly improves the accuracy and efficiency of meeting monitoring, and provides an innovative technical solution for the field of safe production.

[0041] (4) The present invention enhances recognition accuracy and efficiency. Specifically, the present invention specially designs a complex neural network including a CSP structure, a residual structure, and a PANet structure. This network architecture is particularly suitable for real-time personnel detection in a production environment and can efficiently process and parse data from complex scenes. By constructing a backbone network with a CSP structure and a residual structure, the present invention significantly improves the feature learning ability of the network. Through its unique branch design, it allows the reuse of gradient information and optimizes the sparse connections of the network, which not only enhances the network's ability to recognize subtle features, but also improves the processing speed, allowing the system to monitor in real time and accurately determine personnel changes in a meeting.

[0042] (5) The present invention realizes efficient meeting management. Specifically, the present invention integrates a set of automated meeting determination rules, which can automatically determine the start, progress and end of a meeting based on the results of real-time image analysis. The automated meeting determination rules greatly simplify the monitoring and management process of meetings. The present invention can automatically start and record the start and end time of a meeting, and automatically determine the effectiveness of the meeting based on real-time data on the number of people. This automation reduces the need for manual monitoring and reduces labor costs, while improving the accuracy and efficiency of meeting management.

[0043] (6) The present invention provides data-driven decision support. Specifically, the present invention not only monitors meetings, but also automatically saves key meeting data and images, providing data support for subsequent analysis and decision making. This function enables managers to evaluate the effectiveness of meetings and employee participation based on accurate data, thereby better optimizing meeting processes and improving work efficiency.

[0044] (7) The present invention is scalable and adaptable. Specifically, the present invention adopts a modular design, allowing the system to be flexibly configured and expanded according to different application environments and user needs. The design of the present invention takes into account conference environments of different sizes and types, ensuring wide applicability and high scalability. Whether it is a small team discussion or a large corporate meeting, the system can be flexibly configured to meet various monitoring needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1The present invention is a flow chart of a method for intelligently monitoring pre-work meetings.

[0046] Figure 2 It is a structural schematic diagram of the personnel detection unit of the present invention.

[0047] Figure 3 It is a schematic diagram of the structure of the multi-level CSP residual module of the present invention.

[0048] Figure 4 Schematic diagram of the structure of the residual module of the present invention. DETAILED DESCRIPTION

[0049] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0050] Example 1

[0051] The present invention relates to a meeting monitoring method based on deep learning technology, which is specially designed to automatically monitor and record the effectiveness and participation of pre-work meetings. The core of the system lies in the design of a dedicated personnel detection algorithm and finely defined meeting judgment rules to ensure the accuracy and automation of meeting management.

[0052] Depend on Figure 1 As shown, a pre-work meeting intelligent monitoring method proposed in this embodiment includes the following steps:

[0053] S1, according to the set meeting time, within the set meeting time, use the video acquisition unit to collect video of the meeting room, and use the personnel detection unit to perform personnel detection on the collected video images.

[0054] S2, determine whether the area where the personnel are located is a meeting area, and count the number of personnel in the meeting area, and determine whether the number of personnel in the meeting area exceeds the starting threshold. If so, proceed to step S3; if not, return to step S1 and continue personnel detection until the number of personnel in the meeting area exceeds the starting threshold, and then proceed to step S3.

[0055] S3, when the number of people in the meeting area exceeds the start threshold, it means that the pre-work meeting officially begins, and the timing starts at this time. During the timing period, the number of people in the meeting area is monitored in real time until the number of people in the meeting area is less than the end threshold, and the timing ends.

[0056] S4, judging whether the current timing time exceeds the set time threshold, if so, it means that the current pre-work meeting is valid, and the pre-work meeting record is generated and archived; if not, it means that the current pre-work meeting is invalid, and the pre-work meeting record is not generated;

[0057] S5, determine whether it is within the set meeting time. If it is within the set meeting time, return to step S1 and continue to detect personnel. After the number of people in the meeting area exceeds the starting threshold again, start timing again. If the timing time exceeds the set time threshold again, generate the pre-work meeting record again and archive it; if it exceeds the set meeting time, stop personnel detection.

[0058] In the present invention, by recording the video image data, meeting duration and number of people of the pre-work meeting, a pre-work meeting record is generated, and the pre-work meeting record is sent to relevant managers for review. Specifically, the number of effective pre-work meetings on the day can be counted, and the result can be sent to the project manager electronically.

[0059] In the present invention, the set meeting time is a time period (such as 7 to 8 in the morning), and multiple shorter pre-work meetings will be held during this time period. The purpose of the present invention is to detect how many pre-work meetings have been held and the situation of each pre-work meeting during this time period. Real-time personnel detection is always performed during this time period. If the number of people exceeds the threshold at a certain moment, it may be a meeting or just a flow of people passing by, so timing is required. If the number of people in the picture is always higher than a value within a certain time domain value (such as 30s), then it is judged that a meeting is indeed in progress. Until the number of people in the picture decreases to less than a value, it is considered that the meeting is over. At this time, the screenshot of the meeting, the meeting duration, and the number of participants can be recorded.

[0060] After the meeting, we had to continue the personnel check within the set time period, because there might be pre-work meetings later. When the number of people exceeded the threshold, we repeated the above steps.

[0061] Until the set meeting time is exceeded, the personnel detection ends and the number of meetings detected today is counted.

[0062] Depend on Figure 2 As shown, the personnel detection unit proposed in the present invention adopts an improved deep learning network, which is specially used for efficient and accurate personnel detection in production environments. The network consists of three parts: a backbone feature extraction network, a multi-scale fusion network, and a feature decoder.

[0063] The backbone feature extraction network includes four parts connected in sequence, each of which consists of a downsampling layer and a multi-level CSP residual module; the output end of the downsampling layer in each part is connected to the input end of the multi-level CSP residual module; the output end of the multi-level CSP residual module in the previous part is connected to the input end of the downsampling layer in the next part; the input end of the downsampling layer of the first part inputs the image to be detected.

[0064] The features output by the multi-level CSP residual modules in the last three parts of the backbone feature extraction network are fused through the multi-scale feature fusion network. The backbone feature extraction network adopts the PANet structure to achieve effective fusion of feature maps of different scales. The PANet structure enhances the expressiveness of features through bidirectional fusion, allowing the network to detect people more accurately.

[0065] Finally, the feature decoder is used to process the fused feature map, extract the information required for personnel detection, and generate the detection result. This design allows the gradient information to be reused during the transmission process, while optimizing the network structure through sparse connections, improving the recognition accuracy and processing speed of the network model for personnel targets in the detection area.

[0066] Depend on Figure 3 As shown in Figure 1, the main structure of the multi-level CSP residual module is composed of four residual modules connected in series, and each residual module is equipped with a CSP branch structure. By maximizing the number of branch paths, the width of the backbone network is expanded. The output of the main structure is fused with the output of each branch structure through a 1×1 standard convolution module CBA, and then the input and the fused output are combined through the residual edge as the output of the multi-level CSP residual module, which effectively avoids the loss of important features and provides a mapping channel for upper-level features.

[0067] The CSP branch structure consists of a 1×1 standard convolutional module CBA.

[0068] Depend on Figure 4 As shown, in order to ensure the effective transmission of image information, the residual module of the present invention includes a 1×1 standard convolution module CBA and an n×n deep convolution module DBA connected in series in sequence. The input of the residual module is processed by the 1×1 standard convolution module CBA and the n×n deep convolution module DBA in sequence, and the output of the deep convolution module DBA is added to the input of the residual module as the output of the residual module.

[0069] Among them, the 1×1 standard convolution module CBA is composed of a 1×1 standard convolution layer, a batch normalization layer and an activation function connected in sequence; the n×n deep convolution module DBA is composed of a n×n deep convolution layer, a batch normalization layer, an activation function, a 1×1 standard convolution layer, a batch normalization layer and an activation function connected in sequence.

[0070] The present invention adopts SiLU (Sigmoid Linear Unit) activation function, which has the characteristics of no upper bound but lower bound and non-monotonicity, and is used in the present invention to promote the effective propagation of personnel characteristics in the deep network.

[0071] Example 2

[0072] This embodiment provides a pre-work meeting intelligent monitoring system, which is applicable to a pre-work meeting intelligent monitoring method of the above-mentioned embodiment 1. The system includes: a video acquisition unit, a personnel detection unit, a central processing unit and a storage unit.

[0073] The video acquisition unit is used to acquire video of the conference room.

[0074] The personnel detection unit is used to perform personnel detection on the collected video images.

[0075] The central processing unit is used to determine whether the pre-job meeting is valid according to the number of people in the meeting area, and record the video image data, meeting duration and number of people of the valid pre-job meeting to form a pre-job meeting record.

[0076] The storage unit is used to store the pre-work meeting records.

[0077] Example 3

[0078] In addition to the above-mentioned methods and systems, an embodiment of the present invention may also be a computer program product, which includes computer program instructions, which, when executed by a processor, enable the processor to execute the steps of the decision-making behavior decision method according to various embodiments of the present application described in the above-mentioned embodiment 1 of this specification.

[0079] The computer program product may be written in any combination of one or more programming languages ​​to write program codes for performing the operations of the embodiments of the present application, including object-oriented programming languages, such as Java, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as an independent software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0080] Example 4

[0081] An embodiment of the present invention may also be a computer-readable storage medium having computer program instructions stored thereon, which, when executed by a processor, enables the processor to execute the steps of the decision-making behavior decision method according to various embodiments of the present invention described in the above embodiment 1 of this specification.

[0082] The computer readable storage medium can adopt any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can include, for example, but is not limited to, a system, device or device of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination of the above. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable 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.

[0083] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A pre-work meeting intelligent monitoring method, characterized in that: The following steps are involved: S1, according to the set meeting time, perform personnel detection in the meeting room within the set meeting time; S2, counting the number of people in the meeting area, determining whether the number of people in the meeting area exceeds the starting threshold, if so, proceeding to step S3; if not, returning to step S1, continuing to perform personnel detection until the number of people in the meeting area exceeds the starting threshold, and then proceeding to step S3; S3: When the number of people in the meeting area exceeds the start threshold, it means that the pre-work meeting officially begins. At this time, the timer starts, and during the timer period, the number of people in the meeting area is monitored in real time until the number of people in the meeting area is less than the end threshold, and the timer ends; S4, judging whether the current timing time exceeds the set time threshold, if so, it means that the current pre-work meeting is valid, and the pre-work meeting record is generated and archived; if not, it means that the current pre-work meeting is invalid, and the pre-work meeting record is not generated; S5, determine whether it is within the set meeting time. If it is within the set meeting time, return to step S1 and continue to detect personnel. After the number of people in the meeting area exceeds the starting threshold again, start timing again. If the timing time exceeds the set time threshold again, generate the pre-work meeting record again and archive it; if it exceeds the set meeting time, stop personnel detection.

2. The method for intelligent monitoring of pre-work meetings according to claim 1 is characterized in that: In step S1, a video acquisition unit is used to acquire video of the conference room, and a personnel detection unit is used to perform personnel detection on the acquired video images.

3. The method for intelligent monitoring of pre-work meetings according to claim 1 is characterized in that: In step S4, the video image data, meeting duration and number of people of the pre-work meeting are recorded, a pre-work meeting record is generated, and the pre-work meeting record is sent to the management personnel for review.

4. The method for intelligent monitoring of pre-work meetings according to claim 2 is characterized in that: The personnel detection unit includes a backbone feature extraction network, a multi-scale fusion network, and a feature decoder connected in sequence; wherein, The backbone feature extraction network includes four parts connected in sequence, each of which is composed of a downsampling layer and a multi-level CSP residual module; the output end of the downsampling layer in each part is connected to the input end of the multi-level CSP residual module; the output end of the multi-level CSP residual module in the previous part is connected to the input end of the downsampling layer in the next part; the input end of the downsampling layer of the first part inputs the image to be detected; The features output by the multi-level CSP residual modules in the last three parts of the backbone feature extraction network are fused through a multi-scale feature fusion network to achieve the fusion of feature maps of different scales. The feature decoder processes the fused feature map, extracts the information required for person detection, and generates person detection results.

5. The method for intelligent monitoring of pre-work meetings according to claim 4 is characterized in that: The main structure of the multi-level CSP residual module is composed of four residual modules connected in series, each residual module is equipped with a CSP branch structure, the output result of the main structure is fused with the output result of each branch structure through a 1×1 standard convolution module CBA, and then the input and the fused output are combined through the residual edge as the output of the multi-level CSP residual module; The CSP branch structure is composed of a 1×1 standard convolution module CBA; the residual module includes a 1×1 standard convolution module CBA and an n×n deep convolution module DBA connected in series, the input of the residual module is processed by the 1×1 standard convolution module CBA and the n×n deep convolution module DBA in sequence, and the output of the deep convolution module DBA is added to the input of the residual module as the output of the residual module; The 1×1 standard convolution module CBA is composed of a 1×1 standard convolution layer, a batch normalization layer and an activation function connected in sequence; The n×n deep convolution module DBA is composed of an n×n deep convolution layer, a batch normalization layer, an activation function, a 1×1 standard convolution layer, a batch normalization layer and an activation function connected in sequence.

6. The method for intelligent monitoring of pre-work meetings according to claim 4 is characterized in that: The multi-scale feature fusion network adopts the PANet structure.

7. The method for intelligent monitoring of pre-work meetings according to claim 4, characterized in that: The SiLU activation function is used.

8. An intelligent monitoring system for pre-work meetings, characterized in that: A pre-work meeting intelligent monitoring method applicable to any one of claims 1-7, the system comprising: a video acquisition unit, a personnel detection unit, a central processing unit and a storage unit; The video acquisition unit is used to acquire video of the conference room; The personnel detection unit is used to perform personnel detection on the collected video images; The central processing unit is used to determine whether the pre-work meeting is valid according to the number of people in the meeting area, and record the video image data, meeting duration and number of people of the valid pre-work meeting to form a pre-work meeting record; The storage unit is used to store the pre-work meeting records.

9. A computer program product, characterized in that It includes a computer program / instruction, which, when executed by a processor, implements a pre-work meeting intelligent monitoring method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed, the method for intelligently monitoring a pre-work meeting as described in any one of claims 1-7 is implemented.

Images