Real-time hoisting behavior detection system and method
By designing a real-time lifting and lifting behavior detection system, using lightweight YOLOv8 network and embedded system deployment, accurate and timely detection of lifting and lifting behavior is achieved, solving the problem of untimely detection in the existing technology, and improving operational safety guarantees.
Patent Information
- Application Number
- CN202411467807.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-21
- Publication Date
- 2025-03-28
AI Technical Summary
It is difficult for the prior art to achieve accurate and timely detection of lifting and lifting behavior, resulting in difficult accident prevention and operational safety guarantees.
A real-time lifting and hoisting behavior detection system is designed, and the lightweight YOLOv8 network and embedded system deployment is used to realize real-time detection and alarm of the status of the lifting and hoisting equipment through image capture module, object detection module, alarm reporting module and alarm platform.
While ensuring detection accuracy, it reduces the demand for hardware performance, improves inference efficiency, reduces model complexity, and ensures the quality of model training data through data cleaning methods.
Smart Images
Figure CN119723434A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of hoisting and lifting monitoring, and in particular to a real-time hoisting and lifting behavior detection system and method. Background Art
[0002] This section is intended to provide a background or context to the embodiments of the invention recited in the claims. No admission is made that the description herein is prior art by inclusion in this section.
[0003] Lifting and hoisting operations are usually classified as high-risk activities because they involve the handling and operation of a large number of heavy objects. Therefore, extremely high real-time requirements are placed on target detection technology in such operations. Accurate and timely detection of lifting and hoisting behaviors is crucial to preventing accidents and ensuring operational safety. Summary of the invention
[0004] In order to achieve accurate and timely detection of lifting and hoisting behaviors, the present invention proposes a real-time lifting and hoisting behavior detection system and method, which effectively supports the safety management and risk control of lifting and hoisting operations.
[0005] In order to achieve the above object, the present invention adopts the following technical solution.
[0006] First, on the one hand, the present invention provides a real-time lifting and hoisting behavior detection system, including: an image capture module, a target detection module, an alarm reporting module and an alarm platform, the image capture module is used to obtain a lifting and hoisting equipment status image, the lifting and hoisting equipment status image includes the opening angle of the boom, the weight and the ground, and the horizontal rotation of the boom, the target detection module detects and marks the lifting and hoisting equipment behavior, inputs the detection result to the alarm reporting module, and the alarm platform outputs the alarm information, the target detection module is provided with a lightweight target detection model, the BN layer and the convolution layer of the lightweight target detection model are pruned network layers that have undergone channel pruning.
[0007] In some embodiments, the training data of the target detection model is obtained in the following manner: under different lighting conditions and different weather conditions, images and / or video data of different types of lifting equipment at different angles, different scenes, different viewing distances, and different lifting and hoisting states are collected; the images and / or video data of different lifting and hoisting states include images or videos of the lifting and hoisting equipment at different times or stages during the lifting process, and the images or videos include the opening angle of the boom, the horizontal rotation of the weight and the ground, and the boom; feature extraction is performed on the collected images and / or video data; similar images are filtered through cosine similarity calculation; fuzzy image filtering is performed through Laplace operator and variance calculation to obtain training data for the target detection model.
[0008] In some embodiments of the present invention, the different lifting and hoisting states include not lifted, about to be lifted, lifted but not off the ground, just off the ground, low-altitude suspension, hollow suspension and / or high-altitude suspension.
[0009] In some embodiments of the present invention, the system further comprises a camera information storage module, and a user can add, delete, modify camera information and store it in the camera information storage module, which then updates the information to the camera's image capture module.
[0010] In some implementations, the image capture module may access the camera information storage module to query the camera information.
[0011] In some embodiments, the system also includes a calculation module, which receives several consecutive frames of images output by the target detection module, calculates the opening angle of the boom, the distance between the weight and the ground, and the change in the horizontal rotation angle of the boom through the images of the opening angle of the boom, the weight and the ground, and the horizontal rotation angle of the boom, determines the lifting stage of the lifting and hoisting equipment, and calculates the remaining time for the weight to move downward, and transmits the calculation results to the alarm reporting module or the camera information storage module, and the camera information storage module further adjusts the camera's next shooting angle, shooting range and other parameters, and / or makes timely adjustments based on the shooting parameters or plans input by the user.
[0012] The present invention also provides a real-time lifting and hoisting behavior detection method, which is implemented through the above-mentioned system, and the method includes: the image capture module obtains the lifting and hoisting equipment status image, and the lifting and hoisting equipment status image includes the opening angle of the boom, the weight and the ground, and the horizontal rotation of the boom; the target detection module detects and marks the lifting and hoisting equipment behavior, and a lightweight target detection model is provided in the target detection module, and the BN layer and convolution layer of the lightweight target detection model are pruned network layers that have undergone channel pruning; the target detection module inputs the detection result to the alarm reporting module, and the alarm platform outputs the alarm information.
[0013] Furthermore, it also includes the target detection module inputting the detection result into the operation module, the operation module receives several consecutive frames of images output by the target detection module, calculates the opening angle of the boom, the distance between the weight and the ground, and the horizontal rotation angle of the boom through the images of the opening angle of the boom, the weight and the ground, and the horizontal rotation angle of the boom, determines the lifting stage of the lifting equipment, and calculates the remaining time for the weight to move downward, and transmits the operation result to the alarm reporting module or the camera information storage module, the camera information storage module further adjusts the camera's next shooting angle, shooting range and other parameters, and / or makes timely adjustments based on the shooting parameters or plans input by the user.
[0014] Furthermore, the method also includes that the user adds, deletes, or modifies the camera information in the camera information storage module, and the camera information storage module updates the camera information to the image capture module of the camera.
[0015] Furthermore, the training method of the target detection model includes: collecting images and / or video data of different types of lifting equipment at different angles, different scenes, different viewing distances, and different lifting and hoisting states under different lighting conditions and different weather conditions; the images and / or video data of different lifting and hoisting states include images or videos of the lifting and hoisting equipment at different times or stages during the lifting process, and the images or videos contain the opening angle of the boom, the horizontal rotation of the weight and the ground, and the boom; performing feature extraction on the collected images and / or video data; filtering similar images through cosine similarity calculation; filtering fuzzy images through Laplace operator and variance calculation to obtain training data for the target detection model; and inputting the training data into the target detection network for training to obtain the optimal target detection model.
[0016] Beneficial effects of this application:
[0017] This application uses a real-time lifting and hoisting behavior detection system based on a lightweight YOLOv8 network and embedded system deployment to reduce the complexity of the model and reduce dependence on high-performance devices while maintaining model performance. The improved model is easy to deploy to edge embedded devices, while ensuring accuracy, greatly reducing hardware performance requirements, improving reasoning efficiency, and reducing model complexity. In addition, in order to ensure the quality of lifting and hoisting data used for model training, a data cleaning method for lifting and hoisting data is provided. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:
[0019] Figure 1 Schematic diagram of a real-time lifting and hoisting behavior detection system according to Embodiment 1 of the present invention;
[0020] Figure 2 It is a flowchart of a method for implementing convolutional pruning of a target detection model in Example 1 of the present invention;
[0021] Figure 3 This is the real-time lifting and hoisting behavior detection system process of Example 1 of the present invention. DETAILED DESCRIPTION
[0022] Lifting and hoisting operations are usually classified as high-risk activities because they involve the handling and operation of a large number of heavy objects. Therefore, extremely high real-time requirements are placed on target detection technology in such operations. Accurate and timely detection of lifting and hoisting behaviors is crucial to preventing accidents and ensuring operational safety. Real-time target detection technology has become an important research direction in the field of computer vision in recent years. Specific contents include the design of lightweight target detection networks, the construction of target data sets, and the selection of model deployment platforms. The design of real-time target detection systems needs to strike a balance between accuracy, response speed, and system stability to ensure that they can effectively support the safety management and risk control of lifting and hoisting operations.
[0023] Existing technologies use mainstream target detection models to detect lifting and hoisting behaviors, including the RCNN series of models and the YOLO series of models. Although mainstream models have high image recognition accuracy, the convolutional neural network (CNN) model often has millions or even billions of parameters due to its hierarchical structure and large number of nodes. This high computational intensity and storage intensity make traditional convolutional networks face huge challenges in terms of computing and memory consumption, and it is difficult to meet the strict requirements of real-time applications for speed and response time, especially some two-stage target detection methods, which have extremely high requirements for deployed equipment. In addition, considering the complex environment and changing conditions at the lifting site, such as challenges brought by changes in lighting, interference from obstructions, and equipment movement, the image quality obtained by existing data collection and processing methods often cannot meet the needs of model training.
[0024] In order to solve the problems of high model complexity and high equipment performance requirements in mainstream methods, the present invention provides a real-time lifting and hoisting behavior detection system based on lightweight YOLOv8 network and embedded system deployment, which reduces the complexity of the model and reduces the dependence on high-performance equipment while maintaining model performance. In order to ensure the quality of lifting and hoisting data used for model training, a data cleaning method for lifting and hoisting data is provided.
[0025] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with embodiments and drawings. The exemplary embodiments of the present invention and their description are only used to explain the present invention and are not intended to limit the present invention.
[0026] It should be noted that, under the premise of no conflict, the various embodiments or technical features described below can be arbitrarily combined to form new embodiments, which also fall within the scope of protection of the present invention.
[0027] In the description of the present invention, the following terms need to be explained:
[0028] As for directional words, the terms "center", "lateral", "longitudinal", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise" and the like indicating directions and positional relationships in the specification and claims of this application are based on the directions or positional relationships shown in the drawings and are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and should not be understood as limiting the specific scope of protection of the present invention.
[0029] The terms "first", "second", etc. in the description and claims of the present application are used to distinguish similar objects and are not necessarily used to describe a particular sequence or order.
[0030] The terms "including" and "having" and any variations thereof in the specification and claims of this application are intended to cover non-exclusive inclusions. For example, a process, method, system, product or apparatus comprising a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products or apparatuses.
[0031] When an element is referred to as being "fixed to" or "disposed on" another element, it may be directly on the other element or there may be an intermediate element. When an element is referred to as being "connected to" another element, it may be directly connected to the other element or there may be an intermediate element. When an element is referred to as being "provided with" another element, it may be provided on the surface or inside of the element.
[0032] Unless explicitly stated otherwise, throughout the specification and claims, the term “comprise” or variations such as “include” or “comprising”, etc., will be understood to include the stated elements or components but not to exclude other elements or components.
[0033] The term "electrical connection" in the specification and claims of this application may refer to a circuit connection with physical contact or a communication connection, which may be a wired communication connection or a wireless communication connection.
[0034] In order to overcome the limitations of the prior art, this application provides a real-time lifting and hoisting behavior detection system. Figure 1As shown, it includes: an image capture module, a target detection module, an alarm reporting module and an alarm platform, the image capture module is used to obtain the state image of the lifting and hoisting equipment, the state image of the lifting and hoisting equipment includes the opening angle of the boom, the weight and the ground, and the horizontal rotation of the boom, the target detection module detects and marks the behavior of the lifting and hoisting equipment, and inputs the detection result to the alarm reporting module, and the alarm platform outputs the alarm information, the target detection module is provided with a lightweight target detection model, the BN layer and convolution layer of the lightweight target detection model are pruned network layers after channel pruning, the lightweight target detection model.
[0035] The model structure of the target detection model in the target detection module can be replaced by other mainstream model network frameworks such as the RCNN series model. As long as the convolution module contains convolution and BN in series, it can be designed to be lightweight. This embodiment selects the YOLOv8 network, and performs channel pruning and model lightweighting on the YOLOv8 network based on the gamma value γ in batch normalization. With the help of high-quality training data, efficient model deployment under embedded devices is achieved, and real-time detection and reporting of lifting and hoisting behaviors are completed while ensuring detection accuracy. The above high-quality training data is obtained in the following ways:
[0036] (1) Systematically collect lifting and hoisting data, and collect images and / or video data of different types of lifting equipment at different angles, different scenes, different viewing distances, and different lifting and hoisting states under different lighting conditions and different weather conditions; the different lifting and hoisting states include images or videos of the lifting and hoisting equipment at different times or stages during the lifting process, and the different lifting and hoisting states include not lifting, about to lift, lifted but not off the ground, just off the ground, low-altitude suspension, hollow suspension, and / or high-altitude suspension, etc. The images or videos contain the opening angle of the boom, the distance between the weight and the ground, and the horizontal rotation of the boom. This is to ensure that the subsequent target detection model can identify the key parts such as the opening angle of the boom, the distance between the weight and the ground, and the horizontal rotation of the boom, so as to more comprehensively reflect the state of the lifting and hoisting equipment, and also provide a basis for the calculation module to calculate the changes in the opening angle of the boom, the distance between the weight and the ground, and / or the horizontal rotation angle of the boom from several consecutive frames of images output by the target detection module;
[0037] (2) Data cleaning: First, the DI NOv2 model is used to extract features from the collected data to obtain high-dimensional feature representation;
[0038] F=DINOv2(x),x∈R C×H×W
[0039] Where x is the input image, C, H, W are the number of channels, height and width of the image respectively, and R is the real number range;
[0040] Then, the data is deduplicated by calculating the cosine similarity between the features. If the similarity is greater than the set threshold, similar images are filtered out. The calculation method of cosine similarity is:
[0041]
[0042] Among them, are the features of the i-th image and the j-th image, F∈R D is the extracted feature, D is the feature dimension;
[0043] Finally, the Laplace operator is used to calculate the second-order derivative of the image, and the dataset is blurred and filtered based on the variance of each image. If the variance is greater than the set threshold, the image is filtered. The calculation method is:
[0044]
[0045] Among them, I(x,y) represents the pixel value of the image at position (x,y), L i is the pixel value of the image after Laplace transformation, u is the mean of all pixels, and N is the total number of pixels.
[0046] In addition to high-quality training data, this embodiment also improves the target detection model with a lightweight model. Based on the YOLOv8 detection model, the BN layer and the convolution layer are pruned. The pruning is based on the following: the importance of each channel is sorted based on the gamma value γ of the BN layer, and the channel pruning is performed from bottom to top according to the sorting results, and the corresponding channels that are less than the set threshold are pruned. The convolution pruning implementation method is as follows: Figure 2 As shown in the figure. After pruning, the model is fine-tuned. Based on the lightweight YOLOv8 network, the weights are fine-tuned using the hoisting dataset. Finally, the fine-tuned lightweight YOLOv8 model is deployed to the Ascend Atlas embedded development board to form an object detection module. In addition to Ascend Atlas, the deployment device here can also be deployed to Nvidia series embedded devices, such as Jetson Nano, Jetson Tx2, etc.
[0047] After the detection model is deployed, the target detection module detects the video stream input or image input to obtain the target frame of the ongoing lifting and hoisting behavior. After the detection is completed, the detection result is sent to the alarm reporting module. The alarm reporting module encapsulates the output results of the target detection module. The alarm message includes the lifting and hoisting coordinate information, operation time, number of lifting and hoisting behaviors, etc. The image is uploaded to the database and the alarm message is uploaded to the MQTT service through the http protocol.
[0048] Optionally, the system also includes a camera information storage module, and the user can add, delete, modify the camera information and store it in the camera information storage module, which updates the information to the camera's image capture module. For example, the user can adjust the camera's shooting parameters (such as shooting focal length, exposure, etc.) and shooting angle, shooting position, etc. according to the on-site environmental conditions and detection requirements, making the lifting and hoisting process management more flexible, intelligent and humane. Optionally, the image capture module can also actively access the camera information storage module to query the camera information.
[0049] Optionally, in some embodiments, the system further includes a computing module, which receives several consecutive frames of images output by the target detection module, and calculates the opening angle of the boom, the distance between the weight and the ground, and the change in the horizontal rotation angle of the boom through the images of the opening angle of the boom, the weight and the ground, and the horizontal rotation of the boom, to determine the stage of lifting of the lifting equipment, such as not lifted, about to lift, lifted but not off the ground, just off the ground, low-altitude suspension, hollow suspension and / or high-altitude suspension, and calculates the remaining time for the weight to move downward. For example, the operation module records the change in the distance between the heavy object and the ground in several consecutive frames of images, and determines whether the heavy object has been lifted from the ground to the highest point. This process can be sent to the alarm platform to remind nearby or other staff members to pay attention to the safety above, or it can also be provided to the user or camera information storage module at the same time. The user or camera storage module adjusts the image capture module of the camera according to the state technology of the lifting and hoisting equipment. After the heavy object completes the operation at high altitude, the heavy object may need to be put back on the ground, so there will be a descending process. At this time, when the operation module detects that the distance between the heavy object and the ground is decreasing, it is considered that the heavy object is in the descending process. At this time, it is necessary to remind nearby and other personnel to transfer safely, facilitate their actions and work arrangements, and reduce accidents. The above structure only takes the distance between the heavy object and the ground as an example to explain the judgment process of the operation module. Optionally, in the above process, the operation module can also combine the angle change analysis of the boom, the change analysis of the horizontal rotation angle of the boom, etc., to provide more comprehensive and accurate information for the monitoring and analysis of the state of the lifting and hoisting equipment, so that the risk judgment can be more objective, reliable and accurate.
[0050] The calculation module transmits the calculation result to the alarm reporting module or the camera information storage module. The camera information storage module further adjusts the camera's next shooting angle, shooting range and other parameters. Alternatively, the user can actively input new shooting parameters or plans to the camera information storage module according to the equipment status or detection requirements, so as to adjust the camera's shooting behavior in time and improve the management capability of lifting and hoisting work.
[0051] Based on the above system, this embodiment also proposes a real-time lifting and hoisting behavior detection method, which is efficient and highly applicable. The complete real-time lifting and hoisting behavior detection system process is as follows: Figure 3 As shown, the user adds, deletes, modifies the camera information and stores it in the camera information storage module, which updates the information to the camera's image capture module. The camera capture module obtains the video stream data of the device and intercepts the required part to save it, and inputs it to the target detection module. The target detection module detects the behavior of the lifting and hoisting equipment, and the detection result is input to the alarm reporting module, and the alarm platform outputs the alarm information.
[0052] Specifically, the above method comprises the following steps:
[0053] S1: The image capture module obtains the state image of the lifting equipment, which includes the opening angle of the boom, the position of the weight and the ground, and the horizontal rotation of the boom;
[0054] S2: The target detection module detects and marks the behavior of the lifting equipment. A lightweight target detection model is provided in the target detection module. The BN layer and convolution layer of the lightweight target detection model are pruned network layers that have undergone channel pruning.
[0055] S3: The target detection module inputs the detection result to the alarm reporting module, and the alarm platform outputs the alarm information.
[0056] Optionally, the above step S3 also includes: the target detection module inputs the detection result to the operation module, the operation module receives the images of several consecutive frames output by the target detection module, and calculates the opening angle of the boom, the distance between the weight and the ground, and the horizontal rotation angle of the boom through the images of the opening angle of the boom, the weight and the ground, and the horizontal rotation of the boom, and / or the change of the horizontal rotation angle of the boom, judges the stage of the lifting of the lifting and hoisting equipment, and calculates the remaining time for the weight to run downward, and transmits the operation result to the alarm reporting module or the camera information storage module, and the camera information storage module further adjusts the parameters such as the shooting angle and shooting range of the camera in the next step. Optionally, the method also includes that the user adds, deletes, and modifies the camera information in the camera information storage module, and the camera information storage module updates the camera information to the image capture module of the camera. The camera information storage module is adjusted in time in combination with the shooting parameters or schemes input by the user.
[0057] Finally, it should be pointed out that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them. Although the present invention has been described in detail with reference to the above embodiments, a person skilled in the art should understand that the technical solutions described in the above embodiments can still be modified, or some of the technical features can be replaced by equivalents, and these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A real-time lifting and hoisting behavior detection system, characterized in that: It includes an image capture module, a target detection module, an alarm reporting module and an alarm platform. The image capture module is used to obtain the state image of the lifting and hoisting equipment. The state image of the lifting and hoisting equipment includes the opening angle of the boom, the balance between the weight and the ground, and the horizontal rotation of the boom. The target detection module detects and marks the behavior of the lifting and hoisting equipment, and inputs the detection result to the alarm reporting module. The alarm platform outputs the alarm information. A lightweight target detection model is provided in the target detection module. The BN layer and convolution layer of the lightweight target detection model are pruned network layers that have undergone channel pruning.
2. The real-time lifting and hoisting behavior detection system according to claim 1 is characterized in that: The training data of the target detection model is obtained in the following way: Under different lighting conditions and weather conditions, collect images and / or video data of different types of lifting equipment at different angles, different scenes, different viewing distances, and different lifting and hoisting states; the images and / or video data of different lifting and hoisting states include images or videos of the lifting and hoisting equipment at different times or stages during the hoisting process, and the images and / or videos include the opening angle of the boom, the position of the weight and the ground, and the horizontal rotation of the boom; Perform feature extraction on the collected image and / or video data; Filter similar images by cosine similarity calculation; Blurred images are filtered using the Laplace operator and variance calculation to obtain training data for the target detection model.
3. The real-time lifting and hoisting behavior detection system according to claim 2 is characterized in that: The different lifting and hoisting states include not lifted, about to be lifted, lifted but not off the ground, just off the ground, low-altitude suspension, hollow suspension and / or high-altitude suspension.
4. The real-time lifting and hoisting behavior detection system according to claim 3 is characterized in that: It also includes a camera information storage module, where users can add, delete, modify camera information and store it in the camera information storage module, which then updates the information to the camera's image capture module.
5. The real-time lifting and hoisting behavior detection system according to claim 4 is characterized in that: The image capture module can access the camera information storage module to query the camera information.
6. The real-time lifting and hoisting behavior detection system according to claim 5, characterized in that: The system also includes a calculation module, which receives several consecutive frames of images output by the target detection module, and calculates the opening angle of the boom, the distance between the weight and the ground, and the change in the horizontal rotation angle of the boom through the images of the opening angle of the boom, the distance between the weight and the ground, and the horizontal rotation angle of the boom, so as to determine the lifting stage of the lifting and hoisting equipment, calculate the remaining time for the weight to move downward, and transmit the calculation result to the alarm reporting module or the camera information storage module. The camera information storage module further adjusts the parameters such as the next shooting angle and shooting range of the camera, and / or makes timely adjustments based on the shooting parameters or plans input by the user.
7. A real-time lifting and hoisting behavior detection method, implemented by the system according to any one of claims 1 to 6, characterized in that: include The image capture module obtains the state image of the lifting and hoisting equipment, which includes the opening angle of the boom, the horizontal rotation position of the weight and the ground, and the boom; The target detection module detects and labels the behavior of the lifting and hoisting equipment. A lightweight target detection model is provided in the target detection module. The BN layer and convolution layer of the lightweight target detection model are pruned network layers that have undergone channel pruning. The target detection module inputs the detection result to the alarm reporting module, and the alarm platform outputs the alarm information.
8. The real-time lifting and hoisting behavior detection method according to claim 7, characterized in that: It also includes the target detection module inputting the detection result into the operation module, the operation module receives several consecutive frames of images output by the target detection module, and calculates the opening angle of the boom, the distance between the weight and the ground, and the horizontal rotation angle of the boom through the images of the opening angle of the boom, the weight and the ground, and the horizontal rotation angle of the boom, to determine the lifting stage of the lifting and hoisting equipment, and calculate the remaining time for the weight to move downward, and transmit the operation result to the alarm reporting module or the camera information storage module, the camera information storage module further adjusts the camera's next shooting angle, shooting range and other parameters, and / or makes timely adjustments based on the shooting parameters or plans input by the user.
9. The real-time lifting and hoisting behavior detection method according to claim 8, characterized in that: The method also includes that the user adds, deletes, and modifies the camera information in the camera information storage module, and the camera information storage module updates the camera information to the image capture module of the camera.
10. The real-time lifting and hoisting behavior detection method according to claim 9, characterized in that: The training method of the target detection model includes: Under different lighting conditions and weather conditions, collect images and / or video data of different types of lifting equipment at different angles, different scenes, different viewing distances, and different lifting and hoisting states; the images and / or video data of different lifting and hoisting states include images or videos of the lifting and hoisting equipment at different times or stages during the hoisting process, and the images and / or videos include the opening angle of the boom, the position of the weight and the ground, and the horizontal rotation of the boom; Perform feature extraction on the collected image and / or video data; Filter similar images by cosine similarity calculation; The blurred image is filtered through the Laplacian operator and variance calculation to obtain the training data of the target detection model; The training data is input into the target detection network for training to obtain the optimal target detection model.