Intelligent identification method and device for personnel tumbling in logistics center

Abstract

The invention relates to the technical field of logistics personnel safety protection methods, and discloses an intelligent identification method for personnel tumbling in a logistics center. Comprising the following steps: acquiring a picture sample monitored by a logistics transfer center camera, pre-processing the picture sample to generate a pre-processed picture sample, and labeling the pre-processed picture sample through LabelImg; the method comprises the following steps: training a YOLOv11 model through an IoU-based losses loss function; training a human body posture estimation OpenPose model, and adjusting the architecture of the OpenPose model according to personnel characteristics; the trained YOLOv11 model and the trained OpenPose model are fused, and a personnel wrestling identification model is obtained; pictures monitored by a logistics transfer center camera are input into a personnel wrestling recognition model frame by frame, whether personnel wrestling exists in the pictures or not is judged through the personnel wrestling recognition model, if yes, an alarm is triggered, monitoring personnel are notified through an audible and visual alarm, and related images and video clip storage records are stored; the system has the advantages that the working state of the employee is monitored in real time, wrestling events are found in time, and therefore the safety management level and the response speed are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of logistics personnel safety protection methods, and in particular to a method and device for intelligently identifying falls by personnel in a logistics center. Background Art

[0002] Employee safety is paramount in logistics and transportation centers, especially preventing accidents like falls. Falls not only cause injuries but can also damage cargo, disrupt operations, and severely impact the efficiency and safety of logistics and transportation centers. Existing technologies lack automated, intelligent monitoring systems, nor do they offer real-time monitoring of employee work status and the ability to detect falls promptly.

[0003] Therefore, a method and device for intelligently identifying falls by personnel in a logistics center are provided to solve the above problems. Summary of the Invention

[0004] The main purpose of the present invention is to solve the technical problem in the prior art that there is a lack of an automated and intelligent monitoring system and a method for real-time monitoring of employees' working status and timely detection of fall incidents.

[0005] A first aspect of the present invention provides an intelligent method for identifying falls by personnel in a logistics center, the method comprising: Obtain image samples monitored by the logistics transfer center camera, preprocess the image samples to generate preprocessed image samples, and annotate them using LabelImg; The YOLOv11 model is trained using the IoU-based losses function and the image samples annotated by Labelmg. Training an OpenPose model for human pose estimation, and using it to adjust its architecture based on the characteristics of the person, including clothing and movement. The trained YOLOv11 model and OpenPose model are integrated to obtain a fall recognition model. The images monitored by the logistics transfer center camera are input into the personnel fall recognition model frame by frame. The personnel fall recognition model is used to determine whether there is a person falling in the image. If so, an alarm is triggered, the monitoring personnel are notified through the sound and light alarm, and the relevant images and video clips are stored for record.

[0006] Optionally, obtaining image samples monitored by cameras in the logistics transfer center, preprocessing the image samples to generate preprocessed image samples, and labeling them using LabelImg includes: Image samples are obtained from cameras set up based on the personnel activity frequency evaluation system. Noise interference is removed from the image samples. An image cropping tool is used to crop out irrelevant background areas in the sample images. Images of different sizes are unified to a set size through a scaling algorithm. The pixel values ​​of the image samples are mapped from the original range to a specific interval to generate preprocessed image samples. LabelImg is used to annotate the preprocessed image samples, select the target objects and assign category labels to generate a sample dataset.

[0007] Optionally, the personnel activity frequency evaluation system includes: Obtain location name information, the number of people who have been active at the location in a day, and the duration of each person's activity. Calculate the average activity duration of each person at the location. Assign a frequency score to the location based on the duration of each person's activity at the location. Set cameras from high to low based on the frequency score.

[0008] Optionally, the training of the YOLOv11 model using the IoU-based losses function includes: Initialize the YOLOv11 model and adjust its parameters using the IoU-based losses function based on the target characteristics of the logistics transfer center, including the range of human movement and the shape of the goods. Mosaic technology randomly selects 2 or 4 different images from the sample data set, crops each image into 4 segments of the same shape, and then combines them into a new image by random arrangement, splicing, or superposition; Import the new image into the YOLOv11 model and train the model.

[0009] Optionally, the training of an OpenPose model for estimating human poses and adjusting the architecture according to the characteristics of the person using the OpenPose model include: Divide the sample dataset into training set, validation set and test set according to a certain ratio to train the OpenPose model; The OpenPose model learns posture features based on labeled human joint information and continuously adjusts model weights through the backpropagation algorithm to gradually reduce the error between the predicted posture and the true posture. After multiple rounds of training, the model performance is evaluated using the validation set; Based on the verification results, adjust the learning rate, optimize parameters or model results, and iterate repeatedly until the OpenPose model accurately identifies the working posture of logistics personnel on the verification set.

[0010] Optionally, fusing the trained YOLOv11 model with the OpenPose model to obtain a person fall recognition model includes: The feature maps of the intermediate layers of the YOLOv11 model and the OpenPose model are extracted, superimposed according to certain weights, and input into the subsequent fully connected layer to achieve the fusion of the YOLOv11 model and the OpenPose model to obtain the person fall recognition model.

[0011] A second aspect of the present invention provides an intelligent device for identifying falls by personnel in a logistics center, the intelligent device for identifying falls by personnel in a logistics center comprising: The image sample processing and labeling module is used to obtain image samples monitored by the logistics transfer center camera, preprocess the image samples to generate preprocessed image samples, and label them through LabelImg; YOLOv11 model training module, used to train the YOLOv11 model using the IoU-based losses function and the image samples annotated by Labelmg; The OpenPose model training module is used to train the OpenPose model for human pose estimation and adjust the OpenPose model architecture based on the characteristics of the person. A fall recognition system building module is used to fuse the trained YOLOv11 model with the OpenPose model to obtain a fall recognition model. The personnel fall recognition module is used to input the images monitored by the logistics transfer center camera frame by frame into the personnel fall recognition model, and use the personnel fall recognition model to determine whether there is a person fall in the image. If so, an alarm is triggered, and the monitoring personnel are notified through the sound and light alarm, and the relevant images and video clips are stored and recorded.

[0012] Optional. The image sample processing and annotation module includes an image acquisition unit and an image annotation unit; The image acquisition unit is used to obtain image samples from a camera set up based on the personnel activity frequency evaluation system, remove noise interference from the image samples, use an image cropping tool to crop irrelevant background areas in the sample images, unify images of different sizes to a set size through a scaling algorithm, and map the pixel values ​​of the image samples from the original range to a specific interval to generate preprocessed image samples; The image labeling unit is used to label the preprocessed image samples using LabelImg, select the target object and assign a category label to generate a sample data set; The YOLOv11 model training module includes: a YOLOv11 model adjustment unit, a picture splicing and combination unit, and a YOLOv11 model training unit; The YOLOv11 model adjustment unit is used to initialize the YOLOv11 model and adjust the YOLOv11 model parameters through the IoU-based losses function according to the target characteristics of the logistics transfer center. The target characteristics of the logistics operation center include the amplitude of personnel movements and the shape of goods; The image stitching and combining unit is used to randomly select 2 or 4 different images from the sample data set using Mosaic technology, crop each image into 4 segments of the same shape, and then combine them into a new image by random arrangement, stitching or superposition; The YOLOv11 model training unit is used to import the new image into the YOLOv11 model and train the model; The OpenPose model training module includes: a sample data set division unit, an OpenPose model weight adjustment unit, a model performance evaluation unit, and a model iteration unit; The sample data set division unit is used to divide the sample data set into a training set, a validation set and a test set according to a certain ratio for training the OpenPose model; The OpenPose model weight adjustment unit is used for the OpenPose model to learn posture features based on the labeled human joint information, and continuously adjust the model weights through the backpropagation algorithm to gradually reduce the error between the predicted posture and the true posture; The model performance evaluation unit is used to evaluate the model performance through the validation set after multiple rounds of training; The model iteration unit is used to adjust the learning rate, optimize parameters or model results based on the verification results, and iterate repeatedly until the OpenPose model accurately identifies the working posture of logistics personnel on the verification set.

[0013] A third aspect of the present invention provides an electronic device, comprising a memory and at least one processor, wherein instructions are stored in the memory; The at least one processor calls the instructions in the memory to enable the electronic device to execute the various steps of intelligent identification of falls of logistics center personnel as described above.

[0014] A fourth aspect of the present invention provides a computer-readable storage medium having instructions stored thereon, which, when executed by a processor, implement the various steps of the method for intelligently identifying falls of logistics center personnel as described above.

[0015] In the technical solution of the present invention, by taking into account the complexity of actual scenes, Mosaic data enhancement technology is adopted when training the model. Mosaic data enhancement technology can improve the adaptability of the model to different scenes and targets, use Gaussian filtering algorithm to remove noise from the image, and use Gaussian function to perform weighted averaging on the image; Gaussian filtering can smooth the image while better retaining image details; The IoU-based losses loss function is used to further improve the accuracy of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A flow chart of a method for intelligently identifying falls by personnel in a logistics center provided by an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of the intelligent device for identifying falls of personnel in a logistics center provided by the present invention; Figure 3 A schematic structural diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0017] An embodiment of the present invention provides an intelligent identification method for personnel falls in a logistics center, comprising obtaining image samples monitored by cameras in a logistics transfer center, preprocessing the image samples to generate preprocessed image samples, and annotating them through LabelImg; training a YOLOv11 model through an IoU-based losses function and the image samples after LabelImg annotation; training an OpenPose model for human posture estimation, and adjusting the architecture of the OpenPose model according to personnel characteristics through the OpenPose model, wherein the personnel characteristics include personnel clothing characteristics and personnel movement characteristics; fusing the trained YOLOv11 model and the OpenPose model to obtain a personnel fall recognition model; inputting the images monitored by cameras in a logistics transfer center frame by frame into the personnel fall recognition model, and judging whether there is a personnel fall in the image through the personnel fall recognition model, and if so, triggering an alarm, notifying the monitoring personnel through an audible and visual alarm, and storing relevant images and video clips for storage records; the present invention solves the technical problems in the prior art of the lack of an automated and intelligent monitoring system, and the lack of a method for real-time monitoring of employees' work status and timely detection of fall incidents.

[0018] The terms "first," "second," "third," "fourth," and so on (if any) in the description and claims of the present invention and in the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a particular order or precedence. It should be understood that the terms used in this manner are interchangeable where appropriate, so that the embodiments described herein can be implemented in an order other than that shown or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus that includes a series of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, product, or apparatus.

[0019] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 The first embodiment of the intelligent identification method for personnel falls in a logistics center according to the present invention includes: Obtain image samples monitored by the logistics transfer center camera, preprocess the image samples to generate preprocessed image samples, and annotate them using LabelImg; Among them, specifically include: Build a personnel activity frequency evaluation system and set up cameras in the transfer center based on the personnel activity frequency evaluation system; Specifically, they include: Obtain location name information, the number of people who visited the location in a day, and the duration of each person's activity. Calculate the average duration of each person's activity at the location and assign a frequency score to each location based on the duration of each person's activity at the location. Set cameras based on the frequency score from high to low. For example, multiple cameras can be set up in locations with high frequency, while a single camera can be set up in locations with low frequency. Obtain image samples through video recording, remove noise interference from the image samples using the cv2.GaussianBlur() function, use the image cropping tool to crop out irrelevant background areas in the sample images, use the scaling algorithm to unify images of different sizes to the set size, and map the pixel values ​​of the image samples from the original range to a specific interval to generate preprocessed image samples; LabelImg is used to label the preprocessed image samples, select the target objects and assign category labels to generate a sample dataset; The YOLOv11 model is trained using the IoU-based losses function and the image samples annotated by Labelmg; specifically: Initialize the YOLOv11 model and adjust its parameters using the IoU-based losses function based on the target characteristics of the logistics transfer center, including the range of human movement and the shape of the goods. Mosaic technology randomly selects 2 or 4 different images from the sample data set, crops each image into 4 segments of the same shape, and then combines them into a new image by random arrangement, splicing, or superposition; Import the new image into the YOLOv11 model and train the model; Training an OpenPose model for human pose estimation, and using it to adjust its architecture based on the characteristics of the person, including clothing and movement. The trained YOLOv11 model and OpenPose model are fused together to build a complete system for identifying people falling. This involves extracting feature maps from the intermediate layers of the YOLOv11 and OpenPose models, weighting them, and then feeding them into the subsequent fully connected layer to achieve fusion of the YOLOv11 and OpenPose models. Finally, a complete system for identifying people falling is built based on the fused model. The images monitored by the logistics transfer center camera are input into the personnel fall recognition model frame by frame. The personnel fall recognition model is used to determine whether there is a person falling in the image. If so, an alarm is triggered, the monitoring personnel are notified through the sound and light alarm, and the relevant images and video clips are stored for record.

[0020] See also Figure 1 The second embodiment of the intelligent identification method for logistics center personnel falls in the embodiment of the present invention includes: Obtain image samples monitored by the logistics transfer center camera, preprocess the image samples to generate preprocessed image samples, and annotate them using LabelImg; Among them, specifically include: Build a personnel activity frequency evaluation system and set up cameras in the transfer center based on the personnel activity frequency evaluation system; Specifically, they include: Obtain location name information, the number of people who visited the location in a day, and the duration of each person's activity. Calculate the average duration of each person's activity at the location and assign a frequency score to each location based on the duration of each person's activity at the location. Set cameras based on the frequency score from high to low. For example, multiple cameras can be set up in locations with high frequency, while a single camera can be set up in locations with low frequency. Obtain image samples through video recording, remove noise interference from the image samples using the cv2.GaussianBlur() function, use the image cropping tool to crop out irrelevant background areas in the sample images, use the scaling algorithm to unify images of different sizes to the set size, and map the pixel values ​​of the image samples from the original range to a specific interval to generate preprocessed image samples; LabelImg is used to label the preprocessed image samples, select the target objects and assign category labels to generate a sample dataset; The YOLOv11 model is trained using the IoU-based losses function and the image samples annotated by Labelmg; specifically: Initialize the YOLOv11 model and adjust its parameters using the IoU-based losses function based on the target characteristics of the logistics transfer center, including the range of human movement and the shape of the goods. Mosaic technology randomly selects 2 or 4 different images from the sample data set, crops each image into 4 segments of the same shape, and then combines them into a new image by random arrangement, splicing, or superposition; Import the new image into the YOLOv11 model and train the model; Training an OpenPose model for human pose estimation, and using it to adjust its architecture based on the characteristics of the person, including clothing and movement. Specifically, the sample data set is divided into training set, validation set and test set according to a certain ratio for training the OpenPose model; The OpenPose model learns posture features based on labeled human joint information and continuously adjusts model weights through the backpropagation algorithm to gradually reduce the error between the predicted posture and the true posture. After multiple rounds of training, the model performance is evaluated using the validation set; Based on the validation results, adjust the learning rate, optimize parameters, or model results, and iterate repeatedly until the OpenPose model accurately identifies the working postures of logistics personnel on the validation set. The trained YOLOv11 model and OpenPose model are fused together to build a complete system for identifying people falling. This involves extracting feature maps from the intermediate layers of the YOLOv11 and OpenPose models, weighting them, and then feeding them into the subsequent fully connected layer to achieve fusion of the YOLOv11 and OpenPose models. Finally, a complete system for identifying people falling is built based on the fused model. The images monitored by the logistics transfer center camera are input into the personnel fall recognition model frame by frame. The personnel fall recognition model is used to determine whether there is a person falling in the image. If so, an alarm is triggered, the monitoring personnel are notified through the sound and light alarm, and the relevant images and video clips are stored for record.

[0021] See also Figure 1 The third embodiment of the intelligent identification method for logistics center personnel falls in the embodiment of the present invention includes: Obtain image samples monitored by the logistics transfer center camera, preprocess the image samples to generate preprocessed image samples, and annotate them using LabelImg; Among them, specifically include: Build a personnel activity frequency evaluation system and set up cameras in the transfer center based on the personnel activity frequency evaluation system; Specifically, they include: Obtain location name information, obtain the number of people who have been active at the location in a day, and the duration of each person's activity, calculate the average duration of each person's activity at the location, and assign a frequency score to the location based on the duration of each person's activity at the location; set up cameras based on the frequency score from high to low; in the logistics transfer center, it is necessary to determine the key locations for camera installation based on the site layout, cargo flow path, and areas with frequent personnel activities. For example, a camera that can take a panoramic view from above is installed at a high point in the cargo loading and unloading area to capture the postures of people when they are carrying cargo; multi-angle cameras are installed at the corners of the aisles to ensure that there are no blind spots, thereby covering different angles and scenes in all directions, ensuring that the collected images can reflect various actual operating conditions. Based on the frequency, multiple cameras are installed in places with high frequency, and a single camera is installed in places with low frequency; Obtain image samples through the camera and remove noise interference from the image samples using the cv2.GaussianBlur() function. The cv2.GaussianBlur() function can conveniently perform Gaussian filtering on the image. Parameters such as kernel size and standard deviation can be adjusted according to the degree of image noise, effectively improving image quality and preserving key details for subsequent processing; Use image cropping tools to remove irrelevant background areas in sample images to focus the image on key targets; use scaling algorithms such as bilinear interpolation to unify images of different sizes to a set size, and map the pixel values ​​of image samples from the original range to a specific interval to generate preprocessed image samples. For example, map image pixel values ​​from the original range (such as 0-255) to a specific interval (such as 0-1 or -1 to 1). This can accelerate model convergence, enable the model to find the optimal solution more quickly during training, and improve training efficiency.

[0022] Preprocessed image samples are annotated using LabelImg, with the target objects selected and assigned category labels to generate a sample dataset. For objects such as people, goods, and equipment in logistics scenarios, LabelImg is also used to annotate them, selecting the target objects and assigning category labels to provide accurate supervision information for subsequent model training. The accuracy and meticulousness of annotations are directly related to the effectiveness of model learning and require rigorous operation.

[0023] The YOLOv11 model is trained using the IoU-based losses function and the image samples annotated by Labelmg; specifically: Initialize the YOLOv11 model and adjust its parameters using the IoU-based losses function based on the characteristics of the logistics transfer center target, such as large movements of personnel and diverse shapes of goods. This allows for fine-tuning parameters such as anchor box size, confidence threshold, and number of categories. For example, if the focus is on identifying small handheld tools, the initial anchor box size can be appropriately reduced to ensure a more precise and focused initial detection.

[0024] Mosaic technology randomly selects two or four different images from a sample dataset, crops each image into four identically shaped segments, and then combines them into a new image through random permutation, splicing, or superposition. Using Mosaic data augmentation technology improves the model's adaptability to different scenarios and targets. This not only increases data diversity but also allows the model to learn target features from multiple scenarios within a single image, simulating complex and ever-changing logistics environments and improving model generalization. For example, by splicing an image segment of a forklift carrying goods with an image segment of a person walking, the model can simultaneously learn information about both types of targets, enhancing its ability to recognize scenarios where different targets co-occur.

[0025] Import the new image into the YOLOv11 model and train the model; Train the OpenPose model for human pose estimation and use it to adjust the architecture based on the characteristics of the person. Specifically, the sample data set is divided into training set, validation set and test set according to a certain ratio for training the OpenPose model; The OpenPose model learns posture features based on labeled human joint information and continuously adjusts model weights through the backpropagation algorithm to gradually reduce the error between the predicted posture and the true posture. After multiple rounds of training, the model performance is evaluated using the validation set; Based on the validation results, adjust the learning rate, optimize parameters, or model results, and iterate repeatedly until the OpenPose model accurately identifies the working postures of logistics personnel on the validation set. The trained YOLOv11 model and OpenPose model are fused together, and a complete person fall recognition system is built based on the fused model. Specifically, the system extracts the feature maps of the intermediate layers of the YOLOv11 model and the OpenPose model, superimposes them according to certain weights, and inputs them into the subsequent fully connected layer to achieve the fusion of the YOLOv11 model and the OpenPose model. Alternatively, a decision-level fusion is adopted, where the two models output detection results and posture judgments respectively. Then, based on preset rules (such as combining target position and posture confidence), a comprehensive judgment of person fall events is made to improve system reliability. A complete person fall recognition system is built based on the fused model.

[0026] It also includes: inputting the images monitored by the logistics transfer center camera frame by frame into the fusion model of the YOLOv11 model and the OpenPose model; The fusion model determines whether there is a person falling in the picture. If so, it triggers the alarm system, notifies the monitoring personnel through sound and light alarms, and stores the relevant images and video clips for storage records.

[0027] The above describes the intelligent identification method for personnel falling in a logistics center according to an embodiment of the present invention. The following describes the intelligent identification device for personnel falling in a logistics center according to an embodiment of the present invention. Figure 2 In the embodiment of the present invention, the intelligent device for identifying falls of personnel in a logistics center includes the following steps for the above embodiment: The image sample processing and labeling module 201 is used to obtain image samples monitored by the logistics transfer center camera, preprocess the image samples to generate preprocessed image samples, and label them through LabelImg; YOLOv11 model training module 202, used to train the YOLOv11 model through the IoU-based losses loss function and the image samples after Labelmg annotation; An OpenPose model training module 203 is used to train an OpenPose model for estimating human poses, and to adjust the architecture of the OpenPose model according to the characteristics of the person through the OpenPose model; A fall recognition system building module 204 is used to fuse the trained YOLOv11 model with the OpenPose model to obtain a human fall recognition model; The personnel fall recognition module 205 is used to input the images monitored by the logistics transfer center camera frame by frame into the personnel fall recognition model, and use the personnel fall recognition model to determine whether there is a situation of personnel falling in the image. If so, an alarm is triggered, and the monitoring personnel are notified through the sound and light alarm, and the relevant images and video clips are stored and recorded.

[0028] The image sample processing and marking module includes an image acquisition unit and an image marking unit; The image acquisition unit is used to obtain image samples from a camera set up based on the personnel activity frequency evaluation system, remove noise interference from the image samples, use an image cropping tool to crop irrelevant background areas in the sample images, unify images of different sizes to a set size through a scaling algorithm, and map the pixel values ​​of the image samples from the original range to a specific interval to generate preprocessed image samples; The image labeling unit is used to label the preprocessed image samples using LabelImg, select the target object and assign a category label to generate a sample data set; The YOLOv11 model training module includes: a YOLOv11 model adjustment unit, a picture splicing and combination unit, and a YOLOv11 model training unit; The YOLOv11 model adjustment unit is used to initialize the YOLOv11 model and adjust the YOLOv11 model parameters through the IoU-based losses function according to the target characteristics of the logistics transfer center. The target characteristics of the logistics operation center include the amplitude of personnel movements and the shape of goods; The image stitching and combining unit is used to randomly select 2 or 4 different images from the sample data set using Mosaic technology, crop each image into 4 segments of the same shape, and then combine them into a new image by random arrangement, stitching or superposition; The YOLOv11 model training unit is used to import the new image into the YOLOv11 model and train the model; The OpenPose model training module includes: a sample data set division unit, an OpenPose model weight adjustment unit, a model performance evaluation unit, and a model iteration unit; The sample data set division unit is used to divide the sample data set into a training set, a validation set and a test set according to a certain ratio for training the OpenPose model; The OpenPose model weight adjustment unit is used for the OpenPose model to learn posture features based on the labeled human joint information, and continuously adjust the model weights through the backpropagation algorithm to gradually reduce the error between the predicted posture and the true posture; The model performance evaluation unit is used to evaluate the model performance through the validation set after multiple rounds of training; The model iteration unit is used to adjust the learning rate, optimize parameters or model results based on the verification results, and iterate repeatedly until the OpenPose model accurately identifies the working posture of logistics personnel on the verification set.

[0029] above Figure 2The intelligent device for identifying falls of personnel in a logistics center in an embodiment of the present invention is described in detail from the perspective of modular functional entities, and the electronic device in an embodiment of the present invention is described in detail from the perspective of hardware processing.

[0030] Figure 3 Figure 7 is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. This electronic device 700 may vary significantly due to different configurations or performance characteristics. It may include one or more processors 710 (e.g., one or more processors), memory 720, and one or more storage media 730 (e.g., one or more storage devices, including RAM, FLASH, etc.) that store application programs 733 or data 732. The memory 720 and storage medium 730 may be either transient or persistent storage. The program stored in the storage medium 730 may include one or more modules (not shown), each of which may include a series of instruction operations on the electronic device 700. Furthermore, the processor 710 may be configured to communicate with the storage medium 730 to execute the series of instruction operations stored in the storage medium 730 on the electronic device 700.

[0031] The electronic device 700 may further include one or more power supplies 740, one or more input / output interfaces 750, and / or one or more operating systems 731, such as FreeRTOS, Android, etc. It will be understood by those skilled in the art that Figure 3 The illustrated electronic device structure does not constitute a limitation on the electronic device, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0032] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. The computer-readable storage medium stores instructions, which, when executed on a computer, cause the computer to execute the steps of an intelligent method for identifying falls by personnel in a logistics center.

[0033] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, devices, and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0034] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, mobile device, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0035] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those 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 thereof can be replaced by equivalents. However, 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. An intelligent method for identifying falls by personnel in a logistics center, characterized in that: The intelligent identification method for personnel falls in a logistics center includes: Obtain image samples monitored by the logistics transfer center camera, preprocess the image samples to generate preprocessed image samples, and annotate them using LabelImg; The YOLOv11 model is trained using the IoU-based losses function and the image samples annotated by Labelmg. Training an OpenPose model for human pose estimation, and using it to adjust its architecture based on the characteristics of the person, including clothing and movement. The trained YOLOv11 model and OpenPose model are integrated to obtain a fall recognition model. The images monitored by the logistics transfer center camera are input into the personnel fall recognition model frame by frame. The personnel fall recognition model is used to determine whether there is a person falling in the image. If so, an alarm is triggered, the monitoring personnel are notified through the sound and light alarm, and the relevant images and video clips are stored for record.

2. The intelligent identification method for personnel falls in a logistics center according to claim 1, characterized in that: The steps of obtaining image samples monitored by cameras in the logistics transfer center, preprocessing the image samples to generate preprocessed image samples, and labeling them using LabelImg include: Image samples are obtained from cameras set up based on the personnel activity frequency evaluation system. Noise interference is removed from the image samples. An image cropping tool is used to crop out irrelevant background areas in the sample images. Images of different sizes are unified to a set size through a scaling algorithm. The pixel values ​​of the image samples are mapped from the original range to a specific interval to generate preprocessed image samples. LabelImg is used to annotate the preprocessed image samples, select the target objects and assign category labels to generate a sample dataset.

3. The intelligent identification method for personnel falls in a logistics center according to claim 2, characterized in that: The personnel activity frequency evaluation system includes: Obtain location name information, the number of people who have been active at the location in a day, and the duration of each person's activity. Calculate the average activity duration of each person at the location. Assign a frequency score to the location based on the duration of each person's activity at the location. Set cameras from high to low based on the frequency score.

4. The intelligent identification method for personnel falls in a logistics center according to claim 2, characterized in that: The YOLOv11 model is trained by the IoU-based losses function and the image samples after Labelmg annotation, including: Initialize the YOLOv11 model and adjust its parameters using the IoU-based losses function based on the target characteristics of the logistics transfer center, including the range of human movement and the shape of the goods. Mosaic technology randomly selects 2 or 4 different images from the sample data set, crops each image into 4 segments of the same shape, and then combines them into a new image by random arrangement, splicing, or superposition; Import the new image into the YOLOv11 model and train the model.

5. The intelligent identification method for personnel falls in a logistics center according to claim 2, characterized in that: The training of the OpenPose model for estimating human poses and adjusting the architecture based on the characteristics of the person using the OpenPose model include: Divide the sample dataset into training set, validation set and test set according to a certain ratio to train the OpenPose model; The OpenPose model learns posture features based on labeled human joint information and continuously adjusts model weights through the backpropagation algorithm to gradually reduce the error between the predicted posture and the true posture. After multiple rounds of training, the model performance is evaluated using the validation set; Based on the verification results, adjust the learning rate, optimize parameters or model results, and iterate repeatedly until the OpenPose model accurately identifies the working posture of logistics personnel on the verification set.

6. The intelligent identification method for personnel falls in a logistics center according to claim 1, characterized in that: The trained YOLOv11 model and the OpenPose model are integrated to obtain a fall recognition model, which includes: The feature maps of the intermediate layers of the YOLOv11 model and the OpenPose model are extracted, superimposed according to certain weights, and input into the subsequent fully connected layer to achieve the fusion of the YOLOv11 model and the OpenPose model to obtain the person fall recognition model.

7. An intelligent device for identifying falls of personnel in a logistics center, characterized in that: include: The image sample processing and labeling module is used to obtain image samples monitored by the logistics transfer center camera, preprocess the image samples to generate preprocessed image samples, and label them through LabelImg; YOLOv11 model training module, used to train the YOLOv11 model using the IoU-based losses function and the image samples annotated by Labelmg; The OpenPose model training module is used to train the OpenPose model for human pose estimation and adjust the OpenPose model architecture based on the characteristics of the person. A fall recognition system building module is used to fuse the trained YOLOv11 model with the OpenPose model to obtain a fall recognition model. The personnel fall recognition module is used to input the images monitored by the logistics transfer center camera frame by frame into the personnel fall recognition model, and use the personnel fall recognition model to determine whether there is a person fall in the image. If so, an alarm is triggered, and the monitoring personnel are notified through the sound and light alarm, and the relevant images and video clips are stored and recorded.

8. The intelligent device for identifying falls of personnel in a logistics center according to claim 7, characterized in that: The image sample processing and marking module includes an image acquisition unit and an image marking unit; The image acquisition unit is used to obtain image samples from a camera set up based on the personnel activity frequency evaluation system, remove noise interference from the image samples, use an image cropping tool to crop irrelevant background areas in the sample images, unify images of different sizes to a set size through a scaling algorithm, and map the pixel values ​​of the image samples from the original range to a specific interval to generate preprocessed image samples; The image labeling unit is used to label the preprocessed image samples using LabelImg, select the target object and assign a category label to generate a sample data set; The YOLOv11 model training module includes: a YOLOv11 model adjustment unit, a picture splicing and combination unit, and a YOLOv11 model training unit; The YOLOv11 model adjustment unit is used to initialize the YOLOv11 model and adjust the YOLOv11 model parameters through the IoU-based losses function according to the target characteristics of the logistics transfer center. The target characteristics of the logistics operation center include the amplitude of personnel movements and the shape of goods; The image stitching and combining unit is used to randomly select 2 or 4 different images from the sample data set using Mosaic technology, crop each image into 4 segments of the same shape, and then combine them into a new image by random arrangement, stitching or superposition; The YOLOv11 model training unit is used to import the new image into the YOLOv11 model and train the model; The OpenPose model training module includes: a sample data set division unit, an OpenPose model weight adjustment unit, a model performance evaluation unit, and a model iteration unit; The sample data set division unit is used to divide the sample data set into a training set, a validation set and a test set according to a certain ratio for training the OpenPose model; The OpenPose model weight adjustment unit is used for the OpenPose model to learn posture features based on the labeled human joint information, and continuously adjust the model weights through the backpropagation algorithm to gradually reduce the error between the predicted posture and the true posture; The model performance evaluation unit is used to evaluate the model performance through the validation set after multiple rounds of training; The model iteration unit is used to adjust the learning rate, optimize parameters or model results based on the verification results, and iterate repeatedly until the OpenPose model accurately identifies the working posture of logistics personnel on the verification set.

9. An electronic device comprising a memory and at least one processor, wherein the memory stores instructions; The at least one processor calls the instructions in the memory to enable the electronic device to execute the various steps of the intelligent identification method for logistics center personnel falls as described in any one of claims 1-6.

10. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions are executed by the processor, the various steps of the intelligent identification method for logistics center personnel falls as described in any one of claims 1-6 are implemented.

Images