Construction personnel positioning and tracking method

Abstract

The invention relates to a constructor positioning and tracking method, which comprises the following steps of performing continuous video acquisition on site constructors to obtain real-time images of the constructors, performing real-time distinguishing on the real-time images to obtain constructors and non-constructors in the images, and removing the non-constructors in the real-time images; drawing a rectangular frame image based on the target contour of the constructor to obtain a rectangular frame image corresponding to the constructor; comparing the position change of the rectangular frame image corresponding to the constructor to obtain a movement track corresponding to the constructor; performing feature extraction on the target contour of the constructor by using a preset target contour algorithm to obtain features of the corresponding constructor; acquiring coding information of each constructor, and obtaining real-time positioning coordinates of the corresponding constructor based on the coding information of each constructor, the motion trail and the characteristics of the constructor; accurate tracking and management of constructors are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of positioning and tracking, and in particular to a method, device, equipment and storage medium for positioning and tracking construction personnel. Background Art

[0002] Safety monitoring and management is one of the most important tasks at the construction site. Ensuring the safety of construction workers not only involves personal health, but also affects the progress of the project and cost control. Construction sites have always relied on traditional methods such as manual inspections and regular inspections for management and monitoring. However, these methods are time-consuming and labor-intensive, and cannot effectively track the specific location of construction workers in real time. Once an emergency occurs, they cannot respond in time. In recent years, with the rapid development of computer vision technology, especially the widespread application of deep learning technology in the field of image processing, new technical approaches have been provided for the positioning and tracking of construction site personnel. Summary of the invention

[0003] The main purpose of the present invention is to provide a construction personnel positioning and tracking method, device, equipment and storage medium, so as to realize accurate tracking and management of construction personnel.

[0004] To achieve the above object, the present invention provides a method for positioning and tracking construction personnel, comprising the following steps: Continuously collect video of construction workers on site to obtain real-time images of construction workers, distinguish the real-time images in real time through target detection algorithms, obtain construction workers and non-construction workers in the images, and remove non-construction workers in the real-time images; Performing image segmentation masking on the construction workers in the image to obtain a fine mask of the construction workers, and performing morphological expansion and / or corrosion processing on the fine mask of the construction workers to obtain a target contour corresponding to the construction workers; Draw a rectangular frame image based on the target outline of the construction personnel, and obtain a rectangular frame image corresponding to the construction personnel; By comparing the position changes of the rectangular frame images corresponding to the construction personnel, the corresponding motion trajectory of the construction personnel is obtained; Extracting features of the target contour of the construction personnel using a preset target contour algorithm to obtain features of the corresponding construction personnel; The coding information of each construction worker is obtained, and the real-time positioning coordinates of the corresponding construction worker are obtained based on the coding information, movement trajectory and characteristics of each construction worker.

[0005] Furthermore, continuous video collection is performed on the construction personnel on site to obtain real-time images of the construction personnel, and the real-time images are distinguished in real time by a target detection algorithm to obtain construction personnel and non-construction personnel in the images, including: The on-site construction workers are continuously captured through preset high-resolution cameras to obtain real-time images of the construction workers; Based on digital image processing technology, the real-time image of the construction personnel is preprocessed to enhance the real-time image of the construction personnel and obtain a preprocessed image; Performing distortion correction and scale adjustment on the preprocessed image to obtain a calibrated construction image; The calibrated construction image is distinguished in real time by a target detection algorithm to obtain construction personnel and non-construction personnel in the image.

[0006] Furthermore, the calibration construction image is distinguished in real time by a target detection algorithm to obtain construction personnel and non-construction personnel in the image, including: Input the calibration construction image into a preset target detection algorithm for target detection; wherein the target detection algorithm includes a GSConv module, a CBAM module, a C2f-Faster module and a classification layer module; Extract construction features from the calibrated construction image using a GSConv module to obtain extracted construction features; wherein the extracted construction features include construction personnel features and non-construction personnel features; Attention is paid to the extracted construction features through a CBAM module to obtain attention features; Performing feature fusion on the attention features through a C2f-Faster module to obtain fused features; The fusion features are distinguished in real time through a classification layer module to obtain construction workers and non-construction workers in the image.

[0007] Furthermore, performing image segmentation masking on the construction workers in the image to obtain a fine mask of the construction workers, and performing morphological expansion and / or corrosion processing on the fine mask of the construction workers to obtain a target contour corresponding to the construction workers, includes: Performing image segmentation masking on the construction workers in the image to segment the image of the construction workers from the background, and obtaining an initial fine mask of the construction workers; Performing morphological expansion and / or corrosion processing on the initial fine mask of the construction personnel to obtain a fine mask of the construction personnel; The fine mask of the construction worker is mapped back to the original image to locate the precise position of the construction worker in the image; the fine mask of the construction worker is mapped back to the original image to perform edge detection or contour tracking to obtain the target contour corresponding to the construction worker.

[0008] Furthermore, the step of drawing a rectangular frame image based on the target outline of the construction worker to obtain a rectangular frame image corresponding to the construction worker includes: Determine the proportion and size of the rectangular frame based on the target outline of the construction worker to ensure that the rectangular frame can fit the outline of the construction worker and cover it accurately; Based on the proportion and size of the rectangular frame, the target outline of the construction worker is drawn to obtain a rectangular frame image corresponding to the construction worker.

[0009] Furthermore, by comparing the position changes of the rectangular frame images corresponding to the construction personnel, the motion trajectory corresponding to the construction personnel is obtained, including: The rectangular frame images corresponding to the construction personnel are synthesized by image synthesis technology to generate a rectangular video of the construction personnel; Traverse each frame of the rectangular video of the construction worker to ensure that the activity status of the construction worker at each time point is captured, and obtain the rectangular frame image of the construction worker; The rectangular frame position of each construction worker in the continuous rectangular frame images is recorded, and the position changes of the rectangular frame images corresponding to the recorded construction workers are compared to obtain the initial movement trajectory of the construction workers; The construction workers identified in each frame are tracked through a multi-target tracking algorithm, so as to track the position changes of the construction workers in consecutive frames and generate the corresponding verification motion trajectory of the construction workers; The verification motion trajectory is used to verify the initial motion trajectory. If the verification motion trajectory coincides with the initial motion trajectory, the initial motion trajectory is used as the motion trajectory corresponding to the construction personnel.

[0010] Furthermore, the target contour of the construction personnel is subjected to feature extraction using a preset target contour algorithm to obtain features of the corresponding construction personnel, including: Inputting the target contour of the construction worker into a preset target contour algorithm for feature extraction; wherein the target contour algorithm includes a C2f_Faster module, a Slim-Neck structure, a classification module and an attention mechanism EMA module; wherein the Slim-Neck structure includes a GSConv convolution layer and a VoV-GSCSP module; Performing preliminary feature extraction on the target contour of the construction worker through the C2f_Faster module to obtain preliminary contour features; Performing feature enhancement and fusion on the preliminary contour features through the Slim-Neck structure to obtain enhanced fusion features; Positioning and classifying the enhanced fusion features through the classification module to obtain classification features; The classification features are paid attention to by the attention mechanism EMA module to obtain the features of the corresponding construction workers; wherein the features of the construction workers include contour features, clothing color features, dynamic features and spatiotemporal features of the construction workers, and the spatiotemporal features are the position change and time change information of the construction workers in the video sequence.

[0011] The present invention also provides a construction worker positioning and tracking device, comprising: The acquisition module is used to continuously acquire video of construction workers on site to obtain real-time images of construction workers, distinguish the real-time images in real time through target detection algorithms, obtain construction workers and non-construction workers in the images, and remove non-construction workers in the real-time images; A segmentation module is used to perform image segmentation masking on the construction workers in the image to obtain a fine mask of the construction workers, and perform morphological expansion and / or corrosion processing on the fine mask of the construction workers to obtain a target contour corresponding to the construction workers; A drawing module, used for drawing a rectangular frame image based on the target outline of the construction personnel, and obtaining a rectangular frame image corresponding to the construction personnel; A comparison module, used for obtaining the movement trajectory corresponding to the construction personnel by comparing the position changes of the rectangular frame images corresponding to the construction personnel; An extraction module, used to extract features of the target contour of the construction personnel using a preset target contour algorithm to obtain features of the corresponding construction personnel; The positioning module is used to obtain the coding information of each construction worker, and obtain the real-time positioning coordinates of the corresponding construction worker based on the coding information, movement trajectory and characteristics of each construction worker.

[0012] The present invention also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any one of the above methods when executing the computer program.

[0013] The present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of any of the above-mentioned methods are implemented.

[0014] The method, device, computer equipment and storage medium for positioning and tracking construction personnel provided by the present invention include the following steps: continuously collecting video of construction personnel on site to obtain real-time images of construction personnel, distinguishing the real-time images in real time to obtain construction personnel and non-construction personnel in the images, and removing non-construction personnel in the real-time images; drawing rectangular frame images based on the target contours of the construction personnel to obtain rectangular frame images corresponding to the construction personnel; obtaining the corresponding motion trajectories of the construction personnel by comparing the position changes of the rectangular frame images corresponding to the construction personnel; extracting features of the target contours of the construction personnel using a preset target contour algorithm to obtain the features of the corresponding construction personnel; obtaining the coding information of each construction personnel, and obtaining the real-time positioning coordinates of the corresponding construction personnel based on the coding information, motion trajectory and features of each construction personnel. Based on the coding information, motion trajectory and feature extraction results of the construction personnel, the technical problem of not being able to effectively track the specific positions of the construction personnel in real time is solved, and the real-time positioning coordinates of each construction personnel are obtained, thereby achieving accurate tracking and management. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 is a schematic diagram of the steps of a method for positioning and tracking construction personnel in one embodiment of the present invention; Figure 2 is a structural block diagram of a construction personnel positioning and tracking device in one embodiment of the present invention; Figure 3 It is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention.

[0016] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0018] like Figure 1 As shown, Figure 1 This is a schematic diagram of the steps of a method for positioning and tracking construction personnel in one embodiment of the present invention; In one embodiment of the present invention, a method for positioning and tracking construction personnel is provided, comprising the following steps: Step S1, continuously collect video of on-site construction workers to obtain real-time images of the construction workers, distinguish the real-time images in real time through a target detection algorithm to obtain construction workers and non-construction workers in the images, and remove non-construction workers in the real-time images.

[0019] Specifically, continuous video acquisition: use a camera to continuously capture video of the construction site, and obtain a real-time video stream, that is, a series of continuous image frames. Real-time image: extract a single image frame from the video stream as the object to be processed. These image frames represent the state of the construction site at a certain moment. Target detection algorithm: use a pre-trained target detection model to analyze each frame of real-time image. This algorithm can identify different objects in the image and locate them. Target detection algorithms usually include object classification and bounding box regression. The classifier is responsible for determining the object category (such as construction workers, non-construction workers, tools, etc.), and the bounding box regression is used to accurately locate the object. Real-time distinction: classify the objects in the real-time image through the target detection algorithm to distinguish which are construction workers and which are non-construction workers. Construction workers may have specific appearance features, such as wearing safety helmets and work clothes, etc. These features will be used by the model to assist in distinction. Non-construction worker elimination: After the classification is completed, the information of non-construction workers is removed from the real-time image, and only the data related to construction workers is retained. This can be achieved by filtering out the bounding boxes of non-construction workers in the post-processing stage, or by retaining only the area where the construction workers are located in the image. In summary, the entire process can be summarized as follows: first, continuous video acquisition of the construction site is performed to extract real-time images from it; then, these real-time images are processed using a target detection algorithm to distinguish between construction workers and non-construction workers; finally, non-construction workers are removed from the image, and only information related to construction workers is retained.

[0020] Step S2, performing image segmentation masking on the construction workers in the image to obtain a fine mask of the construction workers, and performing morphological expansion and / or corrosion processing on the fine mask of the construction workers to obtain a target contour corresponding to the construction workers.

[0021] Specifically, a deep learning model suitable for image segmentation, such as Mask R-CNN, U-Net, etc., is selected and trained using labeled images of construction workers. The above model can output the probability that each pixel in the image belongs to a construction worker to form a mask. The real-time captured construction site image is input into the trained segmentation model to obtain a fine mask of the construction workers in the image. The above steps separate each construction worker from the background and generate a mask for each construction worker. Dilate processing: Dilate the mask, which is usually used to connect the neighboring areas of the object and fill the small holes in the object. In the case of the construction worker mask, dilation can make the mask contour fuller and easier to identify and analyze. Erosion of the mask is usually used to remove small parts of the object boundary. In some cases, erosion can be used to remove noise points in the construction worker mask. Apply a contour detection algorithm (such as the findContours function in OpenCV) on the mask that has been dilated and / or eroded to obtain accurate contour information for each construction worker. A clear construction worker contour can be extracted based on the mask image after morphological processing.

[0022] The following technical effects can be obtained through the above steps: Through image segmentation mask and morphological processing, more accurate positioning of construction workers can be achieved, including the specific posture and shape information of the personnel. Morphological expansion and corrosion processing can effectively remove or reduce background noise, making the mask of construction workers cleaner and more accurate, which is helpful for subsequent image processing and analysis. In the processed image, the outline of the construction workers is clearer and more accurate, which is conducive to further identification, tracking and analysis. This technology is not only suitable for construction worker analysis, but can also be extended to other fields such as crowd counting, safety monitoring, etc., and has broad application prospects. Through this series of processing steps, we can obtain high-quality construction worker image data, which provides strong technical support for construction site safety monitoring, personnel management, and later data analysis.

[0023] Step S3, drawing a rectangular frame image based on the target outline of the construction worker to obtain a rectangular frame image corresponding to the construction worker.

[0024] Specifically, first, according to the above steps, the fine mask and the processed clear outline of the construction workers have been obtained. Use image processing libraries such as OpenCV to extract outlines. For each extracted outline, calculate the minimum rectangular bounding box that can completely surround the outline. In OpenCV, you can use the cv2.boundingRect() function to implement this function. Draw a rectangular box on the original image or the processed image based on the rectangular bounding box calculated for each outline. Use functions such as cv2.rectangle() to pass in the coordinates of the upper left and lower right corners of the rectangle, as well as parameters such as the color and thickness of the box to draw. Use the corresponding function to display the image containing the rectangular box on the screen or save it to the file system.

[0025] The following technical effects can be obtained through the above steps: By drawing a rectangular box around the target outline of the construction personnel, the target object can be made more prominent visually, which is convenient for human eye recognition and computer processing. Simplifying the complex personnel outline into a rectangular bounding box helps to reduce the amount of calculation for subsequent processing, such as when performing advanced processing such as personnel tracking and behavior analysis. By clearly marking the location of the construction personnel, it is convenient to perform location statistics, activity range analysis, etc., to improve the efficiency and safety level of construction management. The above technical means are not limited to construction personnel, but can also be widely used in target detection and tracking in other fields, such as athlete analysis, vehicle identification, etc. In short, by drawing a rectangular box on the target outline of the construction personnel, we can effectively identify and track the construction personnel, which has important practical application value for ensuring the safety management of the construction site and improving work efficiency.

[0026] Step S4, obtaining the movement trajectory corresponding to the construction personnel by comparing the position changes of the rectangular frame images corresponding to the construction personnel.

[0027] Specifically, in each frame of the video stream, a trained target detection model (such as YOLOv5) is used to identify construction workers, and a rectangular bounding box is drawn around each identified construction worker. The coordinate position of each rectangular box is recorded (generally the coordinates of the upper left corner and the lower right corner of the rectangular box). The position changes of the rectangular boxes of the same construction worker in two consecutive frames are compared. This can be achieved by detecting the change in the coordinates of the center point of the rectangular box, that is, calculating the position difference of the center point of the rectangular box of the same person in two adjacent frames. Using tracking algorithms, such as Kalman filtering, MeanShift or CamShift algorithms, the motion trajectory and position of the target object can be predicted more accurately. For each tracked construction worker, its position in consecutive frames is recorded, and the motion trajectory of the person is rendered on an image or layer. The motion trajectory can be achieved by drawing line segments that connect the center points of the rectangular boxes of the construction workers in consecutive frames. The captured motion trajectory data is analyzed to extract information such as motion speed and activity range. The motion trajectory and analysis results can be stored in a database for subsequent behavior analysis, safety assessment or statistical report generation.

[0028] The following technical effects can be obtained through the above steps: By real-time tracking of the movement of construction workers on the construction site, managers can have a better grasp of the on-site situation and promptly discover potential safety risks or violations. By analyzing the movement trajectory and activity range of construction workers, safety accidents in operations can be effectively prevented, such as preventing entry into dangerous areas. By analyzing the activity trajectory of construction workers, the work efficiency of construction workers and the reasonable layout of on-site operations can be evaluated, providing data support for optimizing construction management and improving efficiency. The stored movement trajectory and analysis results can be used as historical data for future safety accident investigations, workflow improvements, or to provide examples for safety training. Overall, by analyzing the position changes and movement trajectories of the rectangular frame images of construction workers, the safety management level of the construction site can be greatly improved, accidents can be reduced, and strong data support can be provided for construction planning and personnel management.

[0029] Step S5, using a preset target contour algorithm to extract features of the target contour of the construction worker to obtain features of the corresponding construction worker.

[0030] Specifically, the preset target contour algorithm: The "preset target contour algorithm" mentioned here refers to a pre-set algorithm or a set of algorithms that are specifically used to extract the contour information of an object (in this case, a construction worker) from an image. Contour information refers to the shape of the edge of an object, which can help us understand the appearance characteristics of the object. Target contour of a construction worker: The "target contour of a construction worker" refers to the outer edges of individuals identified as construction workers in the real-time image after being processed by the target detection algorithm. These contours are automatically identified by the algorithm and are used to describe the appearance characteristics of the construction worker. Feature extraction: Feature extraction is an important step in computer vision. Its purpose is to extract useful information from the raw data (in this case, the target contour of the construction worker). This information can be color, texture, shape, or other characteristics that are helpful for subsequent processing. For the target contour of the construction worker, feature extraction means extracting specific features from the contour that can represent the identity of the construction worker, such as the shape, size, location of certain key points of the contour, etc. Corresponding construction worker features: After feature extraction, we have a set of data that can characterize the characteristics of each individual construction worker. These features can be used for subsequent analysis, such as tracking specific construction workers, monitoring their behavior patterns, or for higher-level decision support.

[0031] Step S6, obtaining the coding information of each construction worker, and obtaining the real-time positioning coordinates of the corresponding construction worker based on the coding information, movement trajectory and characteristics of each construction worker.

[0032] Specifically, it is necessary to first identify and encode each construction worker at the construction site. This can be achieved by wearing different colored or marked helmets, uniforms, and carrying RFID and other technologies. Computer vision technology can also be used to train deep learning models, such as using CNN or Transformer models, to identify construction workers from video frames and assign each person a unique code. Using image processing technology, as mentioned above, the position changes of construction workers are compared through continuous image frames to record the movement trajectory of construction workers. Target tracking algorithms, such as Kalman filtering, SORT or DeepSORT, can be used to continuously track the position of each person based on the coding information of the construction workers.

[0033] Deep learning models, such as Vision Transformer, are used to analyze images of construction workers and extract their features, including high-dimensional features such as clothing type, color, and posture. These features are crucial to improving the accuracy and robustness of construction worker positioning. Based on the construction worker's coding, tracked motion trajectory, and features extracted by deep learning models, certain algorithm logic, such as nearest neighbor matching and trajectory analysis, can be combined to calculate the exact coordinate position of the construction worker at each moment.

[0034] The following technical effects can be obtained through the above steps: Based on the above technology, the construction site manager can monitor the position of each construction worker in real time and obtain the distribution of personnel in time. Through real-time positioning, the behavior of construction workers entering dangerous areas can be discovered in time, and reminders or control measures can be issued in time to effectively avoid accidents. Through real-time monitoring and analysis of the movement of construction workers, managers can more effectively dispatch resources and improve work efficiency, such as optimizing the layout of the construction site by analyzing the flow path of personnel. After accumulating enough positioning and motion data, data analysis technology can also be used for pattern recognition to better understand the work dynamics of the construction site and further optimize construction management and work arrangements. Overall, combining the coding information, motion trajectory and characteristics of construction workers to obtain real-time positioning coordinates can not only effectively improve the safety management level of the construction site, but also provide data support for construction management to improve work efficiency and construction quality.

[0035] In a specific embodiment, continuous video collection is performed on the construction personnel on site to obtain real-time images of the construction personnel, and the real-time images are distinguished in real time by a target detection algorithm to obtain construction personnel and non-construction personnel in the images, including: The on-site construction workers are continuously captured through preset high-resolution cameras to obtain real-time images of the construction workers; Based on digital image processing technology, the real-time image of the construction personnel is preprocessed to enhance the real-time image of the construction personnel and obtain a preprocessed image; Performing distortion correction and scale adjustment on the preprocessed image to obtain a calibrated construction image; The calibrated construction image is distinguished in real time by a target detection algorithm to obtain construction personnel and non-construction personnel in the image.

[0036] Specifically, the on-site construction personnel are continuously video-captured by a preset high-resolution camera to obtain a real-time image of the construction personnel: Preset high-resolution camera: Here refers to the high-definition camera installed at the construction site, which is pre-set to ensure that high-quality video streams are captured. Continuous video acquisition: The camera continuously records the situation at the construction site, thereby generating a continuous video stream. Real-time image: A single frame image extracted from the video stream represents the state of the construction site at a certain moment. Based on digital image processing technology, the real-time image of the construction personnel is image preprocessed to enhance the real-time image of the construction personnel and obtain a preprocessed image: Digital image processing technology: refers to technical means for improving image quality, including but not limited to noise removal, contrast enhancement, etc. Image preprocessing: At this stage, the real-time image is preliminarily processed to improve the quality and clarity of the image. Enhance the real-time image of the construction personnel: Enhance the visibility of the construction personnel in the image through various image processing technologies. Preprocessed image: The image processed through the above steps has improved quality and is more suitable for subsequent analysis. The pre-processed image is subjected to distortion correction and scale adjustment to obtain a calibrated construction image: Distortion correction: Due to the physical limitations of the camera (such as lens distortion) or the installation position, the original image may have certain deformations. Distortion correction is to correct these deformations to make the image more accurate. Scale adjustment: Ensure that the size of the image is suitable for subsequent processing, which includes operations such as scaling and cropping to better meet the requirements of the target detection algorithm. Calibrated construction image: The image after distortion correction and scale adjustment has better accuracy and consistency. The calibrated construction image is distinguished in real time by the target detection algorithm to obtain construction workers and non-construction workers in the image: Target detection algorithm: A machine learning algorithm that can identify and locate multiple objects in an image. Real-time distinction: The algorithm can quickly distinguish between construction workers and non-construction workers when processing the calibrated construction image. Construction workers and non-construction workers in the image: The final result is to identify which objects are construction workers and which are not.

[0037] In a specific embodiment, the calibration construction image is distinguished in real time by a target detection algorithm to obtain construction personnel and non-construction personnel in the image, including: Input the calibration construction image into a preset target detection algorithm for target detection; wherein the target detection algorithm includes a GSConv module, a CBAM module, a C2f-Faster module and a classification layer module; Extract construction features from the calibrated construction image using a GSConv module to obtain extracted construction features; wherein the extracted construction features include construction personnel features and non-construction personnel features; Attention is paid to the extracted construction features through a CBAM module to obtain attention features; Performing feature fusion on the attention features through a C2f-Faster module to obtain fused features; The fusion features are distinguished in real time through a classification layer module to obtain construction workers and non-construction workers in the image.

[0038] Specifically, the Bottleneck module in C2f in the target detection algorithm is replaced with the FasterNetBlock module, a lightweight module C2f-Faster is designed, the CBAM lightweight feature extraction module is introduced into the forward network of the C2f-Faster module, and an additional classification layer module is added; the calibrated construction image is input into the preset target detection algorithm. This algorithm is designed to distinguish between construction workers and non-construction workers in the image in real time.

[0039] In this scheme, in a specific embodiment, the scheme involves real-time differentiation of the calibrated construction image through a target detection algorithm to identify construction workers and non-construction workers in the image. This process includes multiple steps and modules, each of which plays an important role in the overall process. First, the calibrated construction image is input into a preset target detection algorithm, which includes a GSConv module, a CBAM module, a C2f-Faster module and a classification layer module, aiming to achieve accurate target classification through a series of processing steps. The first step of target detection is to extract construction features from the calibrated construction image through the GSConv module. The GSConv module is an efficient feature extraction module for convolutional neural networks (CNNs) that can effectively extract construction-related features from input images. Specifically, the GSConv module processes the image using its specially designed convolution operation to obtain a detailed feature map including construction worker features and non-construction worker features. The extraction of construction worker features and non-construction worker features is the basis for distinguishing construction workers from non-construction workers, and provides important information for subsequent processing. Next, the extracted construction features are input into the CBAM module for attention. The CBAM module is used in this process to increase the model's attention to important features. Through the channel attention module, CBAM can weight the feature map of each channel to highlight the channel features that are most important for target recognition, while suppressing irrelevant or redundant channel features. Next, the spatial attention module weights the spatial area in the feature map, so that the model pays more attention to the important areas in the image, thereby enhancing the ability to distinguish between construction workers and non-construction workers. Finally, the attention features processed by the CBAM module can more accurately reflect the subtle differences between construction workers and non-construction workers. After obtaining the attention features, these features will be passed to the C2f-Faster module for feature fusion. The C2f-Faster module is an improved module that combines the C2f and FasterNet Block modules, in which the C2f part is responsible for feature fusion, and the FasterNet Block module replaces the Bottleneck module in C2f to achieve more efficient feature extraction and fusion. Through these improvements, the C2f-Faster module can effectively fuse the attention features from the CBAM module, integrate feature information from different levels and scales, and generate more comprehensive and accurate fusion features. This fusion process is crucial for accurately identifying construction workers and non-construction workers because it can integrate the information of each feature map to improve the overall classification effect. Finally, the fused features will be input into the classification layer module for real-time distinction. The main function of the classification layer module is to classify the target based on the fused features and finally output the category information of construction workers and non-construction workers in the image.This module processes the fused features through a fully connected layer or other classification algorithms to obtain the final classification results. In the entire target detection process, the classification layer module plays a key role in converting the extracted, focused and fused features into actual classification results, ensuring the accuracy and real-time performance of the detection. In order to further optimize the target detection algorithm, in a specific embodiment, the Bottleneck module in C2f is replaced by the FasterNet Block module, and a lightweight module C2f-Faster is designed. This improvement aims to improve the computational efficiency and real-time processing capability of the model. The C2f-Faster module introduces the CBAM lightweight feature extraction module in the forward network and adds an additional classification layer module. These improvements not only improve the processing speed of the model, but also improve the detection accuracy, making the algorithm more suitable for real-time application scenarios. In this optimized target detection algorithm, the calibrated construction image achieves accurate and real-time distinction between construction personnel and non-construction personnel in the image through a series of efficient feature extraction, attention mechanism, feature fusion and classification steps, thereby improving the efficiency and safety of construction site management.

[0040] In a specific embodiment, performing image segmentation masking on the construction workers in the image to obtain a fine mask of the construction workers, and performing morphological expansion and / or corrosion processing on the fine mask of the construction workers to obtain a target contour corresponding to the construction workers includes: Performing image segmentation masking on the construction workers in the image to segment the image of the construction workers from the background, and obtaining an initial fine mask of the construction workers; Performing morphological expansion and / or corrosion processing on the initial fine mask of the construction personnel to obtain a fine mask of the construction personnel; The fine mask of the construction worker is mapped back to the original image to locate the precise position of the construction worker in the image; the fine mask of the construction worker is mapped back to the original image to perform edge detection or contour tracking to obtain the target contour corresponding to the construction worker.

[0041] Specifically, first, it is necessary to perform image segmentation on the captured construction site image. The above process involves deep learning models, such as U-Net or Mask R-CNN, which can effectively segment the image of the construction worker from the background and obtain the initial fine mask of the construction worker. This initial mask is a two-dimensional array, in which the construction worker area is marked as 1 (or other specific values) and the background area is marked as 0. Next, the obtained initial fine mask is subjected to morphological processing, including dilation and erosion operations, which help improve the quality of the mask. This step increases the pixels of the construction worker mask area so that the mask edge expands outward, which is usually used to close small holes inside the mask and enhance the connectivity of the mask. The above steps shrink the mask edge by reducing the pixels of the construction worker mask area, which is usually used to remove small protrusions on the mask edge and smooth the contour of the mask. The above two operations can be used separately or alternately according to actual needs to obtain a more accurate construction worker fine mask. Mapping the refined mask after morphological processing back to the original image involves aligning the geometric position of the segmentation mask with the geometric position of the original image (real-time image) to locate the exact position of the construction worker in the image. After the mapping is completed, edge detection or contour tracking can be performed, such as using the Canny edge detection algorithm or the findContours function in the OpenCV library to obtain the target contour corresponding to the construction worker.

[0042] The following technical effects can be obtained through the above steps: Through the segmentation and morphological processing of the deep learning model, the construction workers can be accurately segmented from the complex background and their positions in the image can be accurately located. Morphological processing improves the quality of the mask, improves the clarity of the contour and the accuracy of recognition, and contour tracking further provides the accurate contour of the construction workers, which is conducive to further identification and analysis. The automated personnel positioning and contour tracking system improves the monitoring efficiency and safety management level of the construction site, and can monitor the location and activities of construction workers in real time and detect potential safety hazards in a timely manner. Combining technical means such as image segmentation, morphological processing and contour tracking can effectively improve the performance of the construction site monitoring system and provide strong technical support for construction safety and management.

[0043] In a specific embodiment, the step of drawing a rectangular frame image based on the target outline of the construction worker to obtain a rectangular frame image corresponding to the construction worker includes: Determine the proportion and size of the rectangular frame based on the target outline of the construction worker to ensure that the rectangular frame can fit the outline of the construction worker and cover it accurately; Based on the proportion and size of the rectangular frame, the target outline of the construction worker is drawn to obtain a rectangular frame image corresponding to the construction worker.

[0044] Specifically, first, use image processing or computer vision algorithms (such as OpenCV) to analyze the target outline of the construction worker and extract the geometric parameters of the outline, such as the minimum enclosing rectangle of the outline. Determine the size and proportion of the rectangular box based on the minimum enclosing rectangle of the target outline. Ensure that this rectangle can completely cover the target outline while being as compact as possible to reduce additional blank areas. Select rectangle parameters: According to the size and proportion calculated in the previous step, select appropriate parameters (for example, the coordinates of the upper left corner of the rectangle, width, height). Use image processing tools or libraries (such as the OpenCV function cv2.rectangle()) to draw a rectangular box on the original image based on the determined parameters. The rectangular box can use different colors or line types for easy identification. Mark additional information near or inside the rectangular box, such as the construction worker's ID, detection time, etc., to provide assistance for subsequent tracking and identification.

[0045] The following technical effects can be obtained through the above steps: By drawing a rectangular frame around the construction workers, the observer can quickly identify and locate the construction workers in the image, improving the monitoring effect and response speed. The accurately calculated rectangular frame size and proportion ensure that the target outline of the construction workers is accurately covered, reducing the chance of misidentification and missed identification. The additional information annotated helps to quickly obtain detailed information about the target object and improves the efficiency of subsequent processing, such as tracking analysis and behavior recognition. The visual rectangular frame image provides intuitive data support for construction site safety management and work efficiency analysis, which is conducive to managers' decision-making and intervention. Overall, it provides an effective visual aid for the safety monitoring system of the construction site, which helps to improve the safety and efficiency of site management by enhancing the ability to locate and track construction workers.

[0046] In a specific embodiment, the movement trajectory corresponding to the construction personnel is obtained by comparing the position changes of the rectangular frame images corresponding to the construction personnel, including: The rectangular frame images corresponding to the construction personnel are synthesized by image synthesis technology to generate a rectangular video of the construction personnel; Traverse each frame of the rectangular video of the construction worker to ensure that the activity status of the construction worker at each time point is captured, and obtain the rectangular frame image of the construction worker; The rectangular frame position of each construction worker in the continuous rectangular frame images is recorded, and the position changes of the rectangular frame images corresponding to the recorded construction workers are compared to obtain the initial movement trajectory of the construction workers; The construction workers identified in each frame are tracked through a multi-target tracking algorithm, so as to track the position changes of the construction workers in consecutive frames and generate the corresponding verification motion trajectory of the construction workers; The verification motion trajectory is used to verify the initial motion trajectory. If the verification motion trajectory coincides with the initial motion trajectory, the initial motion trajectory is used as the motion trajectory corresponding to the construction personnel.

[0047] Specifically, an image synthesis technique, such as video editing software or programming library (such as OpenCV), is used to synthesize a sequence of images marked with rectangular frames of construction workers into a video (rectangular video of construction workers). In this way, the position changes of construction workers at different time points can be dynamically observed. The rectangular video of construction workers is traversed frame by frame to ensure that the position and activity status of construction workers at consecutive time points are captured. Each frame of the image records the position information of the construction workers at that moment. For each frame in the video, the position of the rectangular frame of the construction workers is recorded, and the movement trajectory of the construction workers is preliminarily constructed by comparing the position changes of the rectangular frame of the construction workers between consecutive frames. The construction workers identified in each frame of the video are continuously tracked using a multi-target tracking algorithm (such as SORT or DeepSORT). The above steps can effectively determine the position changes of construction workers in consecutive frames, thereby generating a more accurate motion trajectory. The verification motion trajectory obtained by the multi-target tracking algorithm is compared and verified with the initial movement trajectory. If the trajectories of the two are roughly consistent or completely matched, it can be considered that the obtained trajectory accurately reflects the actual movement of the construction workers.

[0048] The following technical effects can be obtained through the above steps: The above process can accurately monitor the movement trajectory of construction workers at the construction site, which is helpful for analyzing work processes and behavior patterns, and optimizing construction management and safety monitoring. By real-time tracking and analyzing the position changes of construction workers, managers can obtain important information in a timely manner and respond quickly to potential safety risks or efficiency issues. The precise movement trajectory provides data support, which helps management to make more scientific and reasonable decisions based on actual work conditions. By tracking the movement trajectory of construction workers for a long time, specific work habits and behavior patterns can be analyzed, which is of great significance for improving construction processes, improving safety and work efficiency. In summary, this series of steps can effectively capture, record and analyze the location and movement trajectory of construction workers at the construction site, providing strong technical support for safety monitoring and management of the construction site.

[0049] In a specific embodiment, the rectangular frame position of each construction worker in the continuous rectangular frame images is recorded, and the position change of the rectangular frame image corresponding to the recorded construction worker is compared to obtain the initial movement trajectory of the construction worker, including: Dynamically locate and record each construction worker in the rectangular frame position in the continuous rectangular frame image to obtain the recording result; Based on the recording results, a space-time trajectory matrix of the construction workers is constructed; wherein the space-time trajectory matrix is ​​used to characterize the position changes of the construction workers in the time and space dimensions, each element in the space-time trajectory matrix represents the spatial position information of the position changes of the rectangular frame image of each construction worker at each time point, and the columns of the space-time trajectory matrix represent different time points, and the columns of the space-time trajectory matrix represent different construction workers; The space-time trajectory matrix is ​​analyzed to obtain a motion vector field corresponding to the construction worker; wherein the motion vector field is used to describe the movement direction and speed of the construction worker between each frame; Performing multi-scale analysis on the motion vector field to obtain a multi-scale trajectory curve; The multi-scale trajectory curve is nonlinearly calculated using a preset nonlinear trajectory estimation model to obtain an initial trajectory curve of the construction personnel.

[0050] Specifically, in this scheme, it is first necessary to dynamically locate and record the rectangular frame position of each construction worker in the continuous rectangular frame image. This process involves processing the real-time monitoring image of the construction site, identifying the position of each construction worker in the image, and marking these position information in the corresponding rectangular frame. Through such positioning and recording, the spatial position of the construction workers at different times can be accurately obtained, laying the foundation for subsequent trajectory analysis. After completing the dynamic area positioning and recording, based on the obtained recording results, we constructed the spatiotemporal trajectory matrix of the construction workers. The role of this spatiotemporal trajectory matrix is ​​to characterize the position changes of the construction workers in the time and space dimensions. Specifically, each element of the matrix represents the position change of the rectangular frame image corresponding to the construction worker at a specific time point. The columns of the matrix represent different time points, and the rows correspond to different construction workers. The construction of this matrix helps to present the location information of the construction workers in a structured form, which is convenient for subsequent parsing and analysis. Next, we parse the constructed spatiotemporal trajectory matrix, and the result of the parsing is a motion vector field corresponding to the construction worker. The motion vector field is used to describe the movement direction and speed of the construction workers between each frame, which means that we can not only see the static position of the construction workers, but also observe their movement trends and behavior patterns at different time points through the motion vector field. Through the analysis of the motion vector field, we can upgrade the movement of the construction workers from simple displacement information to a more dynamic and directional vector expression, so as to better capture the movement law of the construction workers. After obtaining the motion vector field, we further perform multi-scale analysis on it. The purpose of multi-scale analysis is to extract the characteristics of different levels of the construction workers' movement trajectory through data analysis at different scales. This process can reveal the movement trends at different time scales and synthesize these trends into a multi-scale trajectory curve. The acquisition of multi-scale trajectory curves allows us to have a more comprehensive understanding of the movement trajectory of the construction workers and analyze them at different time granularities in order to discover important movement features that may be obscured by a single scale. Finally, through the preset nonlinear trajectory estimation model, we perform nonlinear calculations on the multi-scale trajectory curve to obtain the initial trajectory curve of the construction workers. The nonlinear trajectory estimation model can more accurately fit the actual movement trajectory of the construction workers by performing complex nonlinear operations on the multi-scale trajectory curve. This nonlinear calculation can overcome the nonlinear motion patterns that may be ignored by linear methods, especially when the motion trajectory of construction workers has complex changes such as acceleration and steering. The nonlinear model can provide higher accuracy. In summary, this scheme achieves accurate extraction of the initial movement trajectory of construction workers through a series of steps such as dynamic area positioning and recording, construction of spatiotemporal trajectory matrix, analysis of motion vector field, multi-scale analysis and nonlinear trajectory estimation.Each step is closely linked to the results of the previous step, forming a coherent processing flow, thus ensuring the accuracy and reliability of the final initial trajectory curve of the construction personnel. This process can not only capture the static position of the construction personnel at different time points, but also deeply explore the dynamic behavior characteristics of the construction personnel through multi-scale and nonlinear analysis, providing strong technical support for the safety management and efficiency improvement of the construction process.

[0051] In a specific embodiment, a preset target contour algorithm is used to extract features of the target contour of the construction personnel to obtain features of the corresponding construction personnel, including: Based on the preset adaptive shape spectrum analysis model, the target contour of the construction personnel is analyzed by spectrum analysis to obtain multi-band contour features; Performing multi-scale convolution extraction and analysis on the multi-band profile features to obtain a multi-scale convolution feature map; Performing feature aggregation on the multi-scale convolutional feature map to obtain aggregated features; Performing morphological transformation on the aggregated features to obtain key morphological features; Noise filtering, enhancement and vector conversion are performed on the key morphological features to obtain a highly robust feature embedding vector; Reconstructing the highly robust feature embedding vector to obtain a reconstructed feature vector; Encoding the reconstructed feature vector to obtain an encoding vector; The encoding vector is normalized and mapped using a preset target contour algorithm to obtain the characteristics of the corresponding construction personnel.

[0052] Specifically, in this scheme, the target contour of the construction worker is firstly extracted using the preset target contour algorithm. The core of this process is to extract the unique contour features of the construction worker from the complex image data, thereby providing a reliable basis for subsequent analysis and application. First, the target contour of the construction worker is subjected to spectrum analysis based on the preset adaptive shape spectrum analysis model. The main purpose of spectrum analysis is to decompose the target contour into features of different frequency bands, thereby revealing the morphological features of the contour at different levels. Through this analysis, we can obtain multi-band contour features, which are not only a simple description of the contour shape, but also an accurate capture of its morphological changes at different scales. Then, the multi-band contour features are subjected to multi-scale convolution extraction and analysis. The process of multi-scale convolution extraction and analysis is to process the contour features of different frequency bands using a multi-scale convolutional neural network (CNN) to obtain a multi-scale convolution feature map. This feature map is a description of the contour features of the construction worker at different scales, which can capture the morphological changes from local to global, which is crucial for accurately describing the complex contour morphology. After obtaining the multi-scale convolution feature map, we perform feature aggregation on it. The purpose of feature aggregation is to integrate features of different scales in the multi-scale convolution feature map to form a unified aggregated feature. This process can eliminate redundant information between features of different scales and enhance meaningful features, thereby obtaining a more concise and information-rich feature representation. Subsequently, the aggregated features are subjected to morphological transformation. Morphological transformation is to process aggregated features through mathematical morphological methods to obtain key morphological features. These key morphological features are further refinements of the contour morphology of construction personnel and can better reflect the geometric structure and morphological characteristics of the contour. After obtaining the key morphological features, they are subjected to noise filtering and enhancement and converted into vector form. The process of noise filtering and enhancement is mainly to improve the robustness of the features by filtering out interference information and strengthening important features. Then, the enhanced features are converted into highly robust feature embedding vectors, which can better adapt to subsequent feature reconstruction and analysis. Next, we reconstruct the highly robust feature embedding vectors to obtain reconstructed feature vectors. The reconstruction process is based on a preset reconstruction model to reorganize and arrange the highly robust feature embedding vectors to generate reconstructed feature vectors that can better adapt to subsequent analysis. This process ensures the integrity and consistency of the feature vector, providing a stable foundation for subsequent encoding and mapping. After the reconstruction is completed, we encode the reconstructed feature vector to obtain the encoding vector. The purpose of encoding is to convert the reconstructed feature vector into a more compact and efficient representation, which is convenient for storage and processing. This process is usually implemented through a specific encoding algorithm, which ensures that the encoding vector can both maintain the original feature information and improve the efficiency of data processing.Finally, the encoding vector is normalized and mapped by the preset target contour algorithm to obtain the features of the corresponding construction personnel. The purpose of normalization is to eliminate the scale differences between the eigenvalues ​​so that all features can be compared and analyzed on a unified scale. The mapping process is to project the normalized features into a specific feature space for further analysis and application. Through this step, the features of the corresponding construction personnel are finally obtained, which can be used to identify, track or classify construction personnel, providing accurate and stable data support. In summary, this scheme forms a complete processing flow from the initial target contour extraction to the final feature acquisition through a series of closely connected steps. Each step is based on the previous step. Through layer-by-layer analysis, refinement and optimization, accurate, robust and efficient construction personnel features are finally obtained. These features can not only reflect the contour morphology of the construction personnel, but also provide strong support for subsequent intelligent analysis and application, and have a wide range of practical application value.

[0053] In a specific embodiment, a preset target contour algorithm is used to extract features of the target contour of the construction personnel to obtain features of the corresponding construction personnel, including: Inputting the target contour of the construction worker into a preset target contour algorithm for feature extraction; wherein the target contour algorithm includes a C2f_Faster module, a Slim-Neck structure, a classification module and an attention mechanism EMA module; wherein the Slim-Neck structure includes a GSConv convolution layer and a VoV-GSCSP module; Performing preliminary feature extraction on the target contour of the construction worker through the C2f_Faster module to obtain preliminary contour features; Performing feature enhancement and fusion on the preliminary contour features through the Slim-Neck structure to obtain enhanced fusion features; Positioning and classifying the enhanced fusion features through the classification module to obtain classification features; The classification features are paid attention to by the attention mechanism EMA module to obtain the features of the corresponding construction workers; wherein the features of the construction workers include contour features, clothing color features, dynamic features and spatiotemporal features of the construction workers, and the spatiotemporal features are the position change and time change information of the construction workers in the video sequence.

[0054] Specifically, the partial convolution (PConv) in FasterNet Block is used to improve the C2f module in the target contour algorithm to form a C2f_Faster structure. Improvement of the neck: The Neck part in the target contour algorithm is replaced by a Slim-Neck structure, which includes a GSConv convolution layer and a VoV-GSCSP module. The VoV-GSCSP module is a lightweight and high-precision module used for feature fusion, reducing the amount of calculation and improving detection accuracy. The classification module is a lightweight decoupling head Detect_G: The lightweight decoupling head Detect_G is a new lightweight and high-precision positioning and classification detection head. The lightweight decoupling head Detect_G merges the four standard convolution layers in the decoupling head of the target contour algorithm into two to reduce repeated feature extraction processes. Then, the standard convolution in the decoupling head is replaced by a group convolution, and the classification and detection heads are separated by a 1×1 convolution layer to reduce the computational complexity. It is understandable (the Detect-G module uses group convolution to replace standard convolution, thereby reducing the amount of calculation while maintaining the effectiveness of feature extraction, and further merging the original multiple standard convolution layers into a smaller number of layers, which not only reduces the number of model parameters, but also reduces the complexity of the model. After feature extraction and merging, a 1×1 convolution layer is used to further classify and locate the features, ensuring that the model can accurately identify and locate the target. Through this design, the Detect-G module achieves lightweight while maintaining high-precision detection capabilities, effectively improving the deployment efficiency and detection speed of the model on edge devices.) Add attention mechanism EMA: Introduce the attention mechanism EMA (Efficient Multi-scale Attention) based on cross-spatial multi-scale to increase attention to small target detection defects.

[0055] In a specific embodiment, the process of extracting features from the target contour of the construction personnel using a preset target contour algorithm can achieve efficient and accurate feature extraction and recognition through the synergy of multiple modules. First, the target contour of the construction personnel is input into the preset target contour algorithm, the core of which is composed of the C2f_Faster module, the Slim-Neck structure, the classification module and the attention mechanism EMA module. These modules play their respective roles at different stages of the entire algorithm, and ultimately achieve comprehensive extraction and analysis of the construction personnel's features through mutual cooperation.

[0056] Specifically, first, the target contour of the construction worker in the image or video is used as input data and provided to the preset target contour algorithm. The target contour of the construction worker first enters the C2f_Faster module for preliminary feature extraction. This module has been optimized and the partial convolution (PConv) technology in the FasterNet Block has been used to improve the original C2f module, thus forming the C2f_Faster structure. The application of partial convolution technology is mainly to improve the recognition ability of incomplete target contours by effectively filling in the incomplete areas in the input feature map, especially in complex environments or unclear contours. This improvement significantly improves the integrity and accuracy of feature extraction. Therefore, through the C2f_Faster module, the contour features of the construction workers can be preliminarily extracted, and these features will be further processed and optimized in subsequent steps. Next, these preliminarily extracted contour features will be passed to the Slim-Neck structure for feature enhancement and fusion. The design of the Slim-Neck structure includes the GSConv convolution layer and the VoV-GSCSP module. This structure is an improved alternative to the Neck part in the target contour algorithm. The GSConv convolutional layer is an efficient convolution method that can improve computational efficiency by reducing redundant computations, while the VoV-GSCSP module is a lightweight, high-precision module specifically used for feature fusion. The feature fusion process is particularly important at this stage because it can integrate feature information from different scales to enhance the final feature expression capability. Through this process, the initially extracted contour features are significantly enhanced, and the redundant information is further streamlined through the fusion process, improving the compactness and discriminability of the features. This enables the subsequent classification and recognition processes to obtain more accurate and robust results with fewer computing resources. The enhanced and fused features are then fed into the classification module for positioning and classification. The classification module is designed with a lightweight decoupling head Detect_G, which is a new lightweight, high-precision positioning and classification detection head. By merging the four standard convolutional layers in the traditional decoupling head into two convolutional layers, the Detect_G module successfully reduces the repeated feature extraction process. This merging not only reduces the complexity and computational complexity of the model, but also replaces the standard convolution with group convolution, further improving computational efficiency. In addition, through the 1×1 convolution layer, the classification and detection heads are effectively separated, thereby reducing the computational complexity while maintaining high accuracy of classification and positioning. This design ensures that in the edge computing environment, the model can still perform target detection tasks with high efficiency, especially in scenarios that require real-time response, the application of the Detect_G module is particularly suitable. Finally, the classification features are further processed through the attention mechanism EMA module.EMA (Efficient Multi-scale Attention) is an attention mechanism based on cross-spatial multi-scales, which is particularly suitable for improving the sensitivity of small target detection. In the process of target contour extraction, the contour, clothing color, dynamic features and spatiotemporal features of construction workers may vary in different spatial and temporal scales, and the EMA mechanism can pay attention to these multi-scale spaces, so that the algorithm can more accurately capture these subtle but important feature information. Through the processing of the EMA module, a comprehensive and detailed feature expression is finally obtained, which includes the contour features, clothing color features, dynamic features and spatiotemporal features of construction workers. Among them, spatiotemporal features refer to the position change and time change information of construction workers in video sequences, which are of great significance for real-time monitoring and behavior analysis. In summary, through the collaborative work of a series of modules, the target contour algorithm can efficiently extract features of the target contour of construction workers. The specific role of each module in the whole process plays a vital role in the accuracy and completeness of the final features. The C2f_Faster module lays the foundation through preliminary feature extraction, the Slim-Neck structure improves feature quality through feature enhancement and fusion, the classification module ensures accurate recognition through efficient positioning and classification, and the attention mechanism EMA module makes the final extracted features more discriminative through multi-scale attention. Through this series of processing steps, the comprehensive extraction of multi-dimensional features of construction personnel is finally achieved, providing a solid foundation for subsequent applications.

[0057] In a specific embodiment, the calculation process of the real-time positioning coordinates of the construction personnel includes: Indicates The real-time positioning coordinates of each construction worker, The coefficient matrix representing the transformation from the encoded information to the coordinates, Indicates The coding information of each construction worker, Indicates The current movement trajectory of each construction worker, Indicates The change in the trajectory of each construction worker, Indicates Characteristics of construction workers. The coefficient matrix representing the transformation from personnel characteristics to coordinates, The weight coefficient representing the trajectory change, represents the weight coefficient of the current position of the trajectory, The coefficient matrix representing the transformation from motion trajectory to coordinates.

[0058] Specifically, the calculation process of the real-time positioning coordinates of the construction personnel includes: Indicates The real-time positioning coordinates of each construction worker, The coefficient matrix representing the transformation from the encoded information to the coordinates, Indicates The coding information of each construction worker, Indicates The current movement trajectory of each construction worker, Indicates The change in the movement trajectory of each construction worker, Indicates Characteristics of construction workers. The coefficient matrix representing the transformation from personnel characteristics to coordinates, The weight coefficient representing the trajectory change, represents the weight coefficient of the current position of the trajectory, The coefficient matrix representing the transformation from motion trajectory to coordinates.

[0059] The following technical effects can be achieved through the above steps: By integrating information from different sources (coding, features, trajectories) and using the corresponding conversion coefficient matrices, the accuracy of positioning can be greatly improved. In particular, the introduction of construction personnel characteristics can provide deeper information about target posture, dynamic features, etc., which helps to efficiently and accurately complete positioning in complex scenarios. By considering the current position and change of the motion trajectory and introducing the corresponding weight coefficient, the above calculation process effectively realizes the tracking of the dynamic movement of construction personnel. Even in the case of frequent movement of personnel or changes in movement mode, the positioning information can be accurately captured and updated in real time. By considering the current position and change of the motion trajectory and introducing the corresponding weight coefficient, the above calculation process effectively realizes the tracking of the dynamic movement of construction personnel. Even in the case of frequent movement of personnel or changes in movement mode, the positioning information can be accurately captured and updated in real time. By integrating motion trajectory information and personnel characteristics, accurate positioning and tracking of construction personnel can be achieved, providing solid data support for construction site management. Based on these real-time data, management can make more scientific decisions, such as issuing safety warnings, responding to emergencies, and optimizing the allocation of human resources. In summary, the application of this calculation process can effectively improve the ability of construction sites to locate and track personnel, help improve construction safety, optimize site management, and has good adaptability and flexibility.

[0060] The above describes the method for positioning and tracking construction personnel in the embodiment of the present invention. The following describes the device for positioning and tracking construction personnel in the embodiment of the present invention. Figure 2 , an embodiment of the construction personnel positioning and tracking device in the embodiment of the present invention includes: The acquisition module 21 is used to continuously acquire video of construction workers on site to obtain real-time images of construction workers, distinguish the real-time images in real time through a target detection algorithm, obtain construction workers and non-construction workers in the images, and remove non-construction workers in the real-time images; The segmentation module 22 is used to perform image segmentation masking on the construction workers in the image to obtain a fine mask of the construction workers, and perform morphological expansion and / or corrosion processing on the fine mask of the construction workers to obtain a target contour corresponding to the construction workers; A drawing module 23, used for drawing a rectangular frame image based on the target outline of the construction personnel, and obtaining a rectangular frame image corresponding to the construction personnel; A comparison module 24, used to obtain the movement trajectory corresponding to the construction personnel by comparing the position changes of the rectangular frame images corresponding to the construction personnel; An extraction module 25 is used to extract features of the target contour of the construction personnel using a preset target contour algorithm to obtain features of the corresponding construction personnel; The positioning module 26 is used to obtain the coding information of each construction worker, and obtain the real-time positioning coordinates of the corresponding construction worker based on the coding information, movement trajectory and characteristics of each construction worker.

[0061] In this embodiment, for the specific implementation of each unit in the above device embodiment, please refer to the above method embodiment, which will not be repeated here.

[0062] Reference Figure 3 The present invention also provides a computer device in an embodiment, wherein the internal structure of the computer device can be as follows: Figure 3 As shown. The computer device includes a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. Among them, the processor designed by the computer is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the above method is implemented.

[0063] Those skilled in the art will understand that Figure 3 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0064] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the above method is implemented. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0065] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided by the present invention and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM.

[0066] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0067] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for positioning and tracking construction personnel, characterized in that: The following steps are involved: Continuously collect video of construction workers on site to obtain real-time images of construction workers, distinguish the real-time images in real time through target detection algorithms, obtain construction workers and non-construction workers in the images, and remove non-construction workers in the real-time images; Performing image segmentation masking on the construction workers in the image to obtain a fine mask of the construction workers, and performing morphological expansion and / or corrosion processing on the fine mask of the construction workers to obtain a target contour corresponding to the construction workers; Draw a rectangular frame image based on the target outline of the construction personnel, and obtain a rectangular frame image corresponding to the construction personnel; By comparing the position changes of the rectangular frame images corresponding to the construction personnel, the corresponding motion trajectory of the construction personnel is obtained; Extracting features of the target contour of the construction personnel using a preset target contour algorithm to obtain features of the corresponding construction personnel; The coding information of each construction worker is obtained, and the real-time positioning coordinates of the corresponding construction worker are obtained based on the coding information, movement trajectory and characteristics of each construction worker.

2. The method for positioning and tracking construction personnel according to claim 1, characterized in that: Continuously collect video of on-site construction personnel to obtain real-time images of construction personnel, and use target detection algorithms to distinguish the real-time images in real time to obtain construction personnel and non-construction personnel in the images, including: The on-site construction workers are continuously captured through preset high-resolution cameras to obtain real-time images of the construction workers; Based on digital image processing technology, the real-time image of the construction personnel is preprocessed to enhance the real-time image of the construction personnel and obtain a preprocessed image; Performing distortion correction and scale adjustment on the preprocessed image to obtain a calibrated construction image; The calibrated construction image is distinguished in real time by a target detection algorithm to obtain construction personnel and non-construction personnel in the image.

3. The method for positioning and tracking construction personnel according to claim 2, characterized in that: The calibrated construction image is distinguished in real time by a target detection algorithm to obtain construction personnel and non-construction personnel in the image, including: Input the calibration construction image into a preset target detection algorithm for target detection; wherein the target detection algorithm includes a GSConv module, a CBAM module, a C2f-Faster module and a classification layer module; Extract construction features from the calibrated construction image using a GSConv module to obtain extracted construction features; wherein the extracted construction features include construction personnel features and non-construction personnel features; Attention is paid to the extracted construction features through a CBAM module to obtain attention features; Performing feature fusion on the attention features through a C2f-Faster module to obtain fused features; The fusion features are distinguished in real time through a classification layer module to obtain construction workers and non-construction workers in the image.

4. The method for positioning and tracking construction personnel according to claim 1, characterized in that: The performing image segmentation masking on the construction workers in the image to obtain a fine mask of the construction workers, and performing morphological expansion and / or corrosion processing on the fine mask of the construction workers to obtain a target contour corresponding to the construction workers, includes: Performing image segmentation masking on the construction workers in the image to segment the image of the construction workers from the background, and obtaining an initial fine mask of the construction workers; Performing morphological expansion and / or corrosion processing on the initial fine mask of the construction personnel to obtain a fine mask of the construction personnel; Map the detailed mask of the construction worker back to the original image to locate the exact position of the construction worker in the image; The fine mask of the construction worker is mapped back to the original image to perform edge detection or contour tracking to obtain the target contour corresponding to the construction worker.

5. The method for positioning and tracking construction personnel according to claim 1, characterized in that: The step of drawing a rectangular frame image based on the target outline of the construction personnel to obtain a rectangular frame image corresponding to the construction personnel includes: Determine the proportion and size of the rectangular frame based on the target outline of the construction worker to ensure that the rectangular frame can fit the outline of the construction worker and cover it accurately; Based on the proportion and size of the rectangular frame, the target outline of the construction worker is drawn to obtain a rectangular frame image corresponding to the construction worker.

6. The method for positioning and tracking construction personnel according to claim 1, characterized in that: By comparing the position changes of the rectangular frame images corresponding to the construction personnel, the motion trajectory corresponding to the construction personnel is obtained, including: The rectangular frame images corresponding to the construction personnel are synthesized by image synthesis technology to generate a rectangular video of the construction personnel; Traverse each frame of the rectangular video of the construction worker to ensure that the activity status of the construction worker at each time point is captured, and obtain the rectangular frame image of the construction worker; The rectangular frame position of each construction worker in the continuous rectangular frame images is recorded, and the position changes of the rectangular frame images corresponding to the recorded construction workers are compared to obtain the initial movement trajectory of the construction workers; The construction workers identified in each frame are tracked through a multi-target tracking algorithm, so as to track the position changes of the construction workers in consecutive frames and generate the corresponding verification motion trajectory of the construction workers; The verification motion trajectory is used to verify the initial motion trajectory. If the verification motion trajectory coincides with the initial motion trajectory, the initial motion trajectory is used as the motion trajectory corresponding to the construction personnel.

7. The method for positioning and tracking construction personnel according to claim 1, characterized in that: The target contour of the construction personnel is subjected to feature extraction using a preset target contour algorithm to obtain features of the corresponding construction personnel, including: Inputting the target contour of the construction worker into a preset target contour algorithm for feature extraction; wherein the target contour algorithm includes a C2f_Faster module, a Slim-Neck structure, a classification module and an attention mechanism EMA module; wherein the Slim-Neck structure includes a GSConv convolution layer and a VoV-GSCSP module; Performing preliminary feature extraction on the target contour of the construction worker through the C2f_Faster module to obtain preliminary contour features; Performing feature enhancement and fusion on the preliminary contour features through the Slim-Neck structure to obtain enhanced fusion features; Positioning and classifying the enhanced fusion features through the classification module to obtain classification features; The classification features are paid attention to by the attention mechanism EMA module to obtain the features of the corresponding construction workers; wherein the features of the construction workers include contour features, clothing color features, dynamic features and spatiotemporal features of the construction workers, and the spatiotemporal features are the position change and time change information of the construction workers in the video sequence.

8. A construction worker positioning and tracking device, characterized in that: include: The acquisition module is used to continuously acquire video of construction workers on site to obtain real-time images of construction workers, distinguish the real-time images in real time through target detection algorithms, obtain construction workers and non-construction workers in the images, and remove non-construction workers in the real-time images; A segmentation module is used to perform image segmentation masking on the construction workers in the image to obtain a fine mask of the construction workers, and perform morphological expansion and / or corrosion processing on the fine mask of the construction workers to obtain a target contour corresponding to the construction workers; A drawing module, used for drawing a rectangular frame image based on the target outline of the construction personnel, and obtaining a rectangular frame image corresponding to the construction personnel; A comparison module, used for obtaining the movement trajectory corresponding to the construction personnel by comparing the position changes of the rectangular frame images corresponding to the construction personnel; An extraction module, used to extract features of the target contour of the construction personnel using a preset target contour algorithm to obtain features of the corresponding construction personnel; The positioning module is used to obtain the coding information of each construction worker, and obtain the real-time positioning coordinates of the corresponding construction worker based on the coding information, movement trajectory and characteristics of each construction worker.

9. A computer device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images