Worker management method and system for labor dispatch

Abstract

The invention is suitable for the technical field of worker management, and particularly relates to a worker management method and system for labor dispatch, and the method comprises the steps: obtaining a building construction drawing, and constructing a building three-dimensional model based on the building construction drawing; acquiring a real-time binocular image, and preprocessing the real-time binocular image to obtain two groups of preprocessed edge images; performing depth information extraction according to the two groups of preprocessed edge images, and performing three-dimensional reconstruction based on the extracted depth information to obtain a local three-dimensional model; matching is carried out, the position of the worker is determined, a preprocessed edge image is recorded, and the state of the worker is monitored according to the preprocessed edge image. By wearing the safety helmet equipped with the binocular image collector, real-time image collection can be carried out in the construction process of a worker, the worker can be positioned, real-time image transmission can be carried out in a thumbnail image mode, the data transmission quantity is greatly reduced, and the safety of the construction worker is ensured.

Description

Technical Field

[0001] The present invention belongs to the technical field of worker management, and in particular relates to a worker management method and system for labor dispatch. Background Art

[0002] Construction worker management refers to the process of effectively organizing, coordinating, controlling, and providing services to construction workers throughout the lifecycle of a construction project. This includes aspects such as worker recruitment and training, task assignment, safety education and management, attendance and performance evaluation, salary payments, and the provision of necessary living and labor protections. Good construction worker management not only improves project efficiency and quality, but also ensures site safety, safeguards workers' legitimate rights and interests, and promotes harmonious labor-capital relations. Furthermore, with technological advancements, modern construction worker management is increasingly incorporating information technology, such as software for personnel scheduling, skills training, and safety management, to enhance management effectiveness and decision-making support capabilities.

[0003] In construction scenarios, labor dispatch is one of the common employment methods. Since dispatched workers are unfamiliar with the construction site, construction accidents are more likely to occur during the construction process. Therefore, there is an urgent need for a method to track and manage workers in order to monitor the safety of construction workers in real time. Summary of the Invention

[0004] The purpose of the present invention is to provide a worker management method for labor dispatch, aiming to solve the problem that in construction scenarios, labor dispatch is one of the common employment methods. Since labor dispatch workers are unfamiliar with the construction site, construction accidents are more likely to occur during the construction process. Therefore, there is an urgent need for a method that can track and manage workers to monitor the safety of construction personnel in real time.

[0005] The present invention is implemented as follows: a worker management method for labor dispatch, the method comprising:

[0006] Obtain building construction drawings, construct a 3D model of the building based on the drawings, assign workspaces to each worker, and have them wear safety helmets equipped with binocular image collectors.

[0007] Acquire a real-time binocular image, pre-process the real-time binocular image, and obtain two sets of pre-processed edge images;

[0008] Depth information is extracted from the two sets of pre-processed edge images, and stereo reconstruction is performed based on the extracted depth information to obtain a local stereo model;

[0009] The local stereo model is matched with the three-dimensional model of the building to determine the location of the workers, and the pre-processed edge image is recorded. The status of the workers is monitored based on the pre-processed edge image.

[0010] Preferably, the step of acquiring a real-time binocular image and preprocessing the real-time binocular image to obtain two sets of preprocessed edge images specifically includes:

[0011] Sending an image acquisition instruction to the binocular image collector at a preset time interval, and receiving a real-time binocular image acquired by the binocular image collector;

[0012] Crop the real-time binocular image and convert it into a grayscale image;

[0013] The edge of the grayscale image is identified based on the edge extraction algorithm, and the grayscale image is converted into a preprocessed edge image.

[0014] Preferably, the step of extracting depth information based on the two sets of pre-processed edge images, and performing stereo reconstruction based on the extracted depth information to obtain a local stereo model specifically includes:

[0015] Perform epipolar correction on the preprocessed edge image, and refine and clean the edges in the preprocessed edge image to obtain an edge image to be matched;

[0016] Use multiple matching methods to match the edge images to be matched, fuse the matching results, and generate edge point matching data. The edge point matching data is used to record the coordinates of the two points that are successfully matched;

[0017] Disparity calculation and depth calculation are performed based on edge point matching data, three-dimensional coordinate points are generated based on the calculation results, and a local stereo model is generated based on the three-dimensional coordinate points.

[0018] Preferably, the steps of matching the local stereo model with the three-dimensional model of the building, determining the location of the worker, recording the pre-processed edge image, and monitoring the worker's status based on the pre-processed edge image include:

[0019] Cutting the 3D model of the building based on the work space allocated to the workers to obtain a local 3D model;

[0020] Compare the local stereo model with the local three-dimensional model to determine the location of the local stereo model, and determine the location of the worker based on the real-time binocular image;

[0021] The pre-processed edge image is transmitted in the form of coordinate points, and the pre-processed edge image is spliced ​​to generate a thumbnail image. When an abnormality occurs in the thumbnail image, the real-time binocular image is transmitted in real time.

[0022] Preferably, the step of transmitting the preprocessed edge image in the form of coordinate points includes sampling the edges in the preprocessed edge image to extract a plurality of sampling points, generating sampling point coordinates according to the positions of the sampling points in the image, packaging the sampling point coordinates to obtain a sampling point coordinate data set, and transmitting the sampling point coordinate data set.

[0023] Another object of the present invention is to provide a worker management system for labor dispatch, the system comprising:

[0024] The model building module is used to obtain building construction drawings, build a three-dimensional building model based on the building construction drawings, and assign a workspace to each worker. The workers all wear safety helmets equipped with binocular image collectors;

[0025] An image preprocessing module is used to obtain real-time binocular images and preprocess the real-time binocular images to obtain two sets of preprocessed edge images;

[0026] A model reconstruction module is used to extract depth information from two sets of preprocessed edge images, and perform stereo reconstruction based on the extracted depth information to obtain a local stereo model;

[0027] The personnel monitoring module is used to match the local stereo model with the three-dimensional model of the building, determine the location of the workers, record the pre-processed edge image, and monitor the status of the workers based on the pre-processed edge image.

[0028] Preferably, the image preprocessing module includes:

[0029] The image acquisition unit is used to send image acquisition instructions to the binocular image collector according to a preset time interval and receive real-time binocular images collected by the binocular image collector;

[0030] An image processing unit, used to crop the real-time binocular image and process it into a grayscale image;

[0031] The edge recognition unit is used to perform edge recognition on the grayscale image based on the edge extraction algorithm and convert the grayscale image into a preprocessed edge image.

[0032] Preferably, the image preprocessing module includes:

[0033] The image acquisition unit is used to send image acquisition instructions to the binocular image collector according to a preset time interval and receive real-time binocular images collected by the binocular image collector;

[0034] An image processing unit, used to crop the real-time binocular image and process it into a grayscale image;

[0035] The edge recognition unit is used to perform edge recognition on the grayscale image based on the edge extraction algorithm and convert the grayscale image into a preprocessed edge image.

[0036] Preferably, the personnel monitoring module includes:

[0037] A model cutting unit, configured to cut the three-dimensional model of the building based on the work space allocated to the worker to obtain a local three-dimensional model;

[0038] The personnel positioning unit is used to compare the local stereo model with the local three-dimensional model to determine the location of the local stereo model and determine the location of the worker based on the real-time binocular image;

[0039] The image transmission unit is used to transmit the pre-processed edge image in the form of coordinate points, splice the pre-processed edge image, generate a thumbnail image, and transmit the real-time binocular image in real time when an abnormality occurs in the thumbnail image.

[0040] Preferably, the step of transmitting the preprocessed edge image in the form of coordinate points includes sampling the edges in the preprocessed edge image to extract a plurality of sampling points, generating sampling point coordinates according to the positions of the sampling points in the image, packaging the sampling point coordinates to obtain a sampling point coordinate data set, and transmitting the sampling point coordinate data set.

[0041] The present invention provides a worker management method for labor dispatch. By wearing a safety helmet equipped with a binocular image collector, real-time image capture can be performed during the construction process of the workers, and the workers can be positioned based on the images captured in real time. The real-time images are transmitted in the form of thumbnail images, which greatly reduces the data transmission volume and ensures the safety of the construction workers. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 A flowchart of a worker management method for labor dispatch provided by an embodiment of the present invention;

[0043] Figure 2 A flowchart of the steps of acquiring a real-time binocular image, preprocessing the real-time binocular image, and obtaining two sets of preprocessed edge images provided by an embodiment of the present invention;

[0044] Figure 3 A flowchart of the steps of extracting depth information from two sets of preprocessed edge images, performing stereo reconstruction based on the extracted depth information, and obtaining a local stereo model, provided by an embodiment of the present invention;

[0045] Figure 4A flowchart of the steps of matching a local stereo model with a three-dimensional building model, determining the location of a worker, recording a pre-processed edge image, and monitoring the worker's status based on the pre-processed edge image, provided in an embodiment of the present invention;

[0046] Figure 5 An architectural diagram of a worker management system for labor dispatch provided by an embodiment of the present invention;

[0047] Figure 6 An architectural diagram of an image preprocessing module provided by an embodiment of the present invention;

[0048] Figure 7 An architectural diagram of a model reconstruction module provided by an embodiment of the present invention;

[0049] Figure 8 An architectural diagram of a personnel monitoring module provided in an embodiment of the present invention;

[0050] Figure 9 A schematic diagram of a real-time binocular image provided by an embodiment of the present invention;

[0051] Figure 10 A schematic diagram of a preprocessed edge image provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to 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.

[0053] like Figure 1 FIG. 1 is a flowchart of a method for managing workers in labor dispatch according to an embodiment of the present invention, the method comprising:

[0054] S100, obtain the building construction drawings, construct a three-dimensional model of the building based on the building construction drawings, and allocate a workspace for each worker. The workers all wear safety helmets equipped with binocular image collectors.

[0055] In this step, construction drawings of the building are obtained. The construction drawings of the building are three-dimensional drawings, and a three-dimensional model of the building is obtained based on them. When workers enter the site, their work areas are determined according to their construction tasks, and corresponding areas are divided for them in the three-dimensional model of the building as their work spaces. A binocular image collector is equipped on the safety helmet worn by the worker. The binocular image collector is used to capture two images of the same position from different angles at the same time. The binocular image collector is connected to the worker's mobile device, and the data collected by the binocular image collector is transmitted to the mobile device for processing.

[0056] S200 , acquiring a real-time binocular image, and preprocessing the real-time binocular image to obtain two sets of preprocessed edge images.

[0057] In this step, real-time binocular images are acquired. After the workers enter the construction site, the binocular image collector is started. The binocular image collector continuously acquires images to obtain real-time binocular images. Two groups of real-time binocular images are generated each time. The real-time binocular images are preprocessed. The preprocessing steps include cropping and grayscale processing of the images. Finally, the edges in the preprocessed real-time binocular images are extracted through the edge recognition algorithm and converted into a preprocessed edge image. In the preprocessed edge image, only the edges of objects in the building interior and the edges of buildings are recorded.

[0058] S300 , extracting depth information from the two sets of pre-processed edge images, and performing stereo reconstruction based on the extracted depth information to obtain a local stereo model.

[0059] In this step, depth information is extracted based on the two sets of preprocessed edge images. During this process, epipolar line correction is performed to ensure that the corresponding points in the two sets of preprocessed edge images are located on the same horizontal scan line, and the edge map is refined and cleaned. Finally, edge line matching is performed. Based on the results of edge line matching, depth information is calculated, including parallax calculation, depth calculation and 3D point generation. The corresponding local stereo model is constructed based on the generated 3D points.

[0060] S400 , matching the local stereo model with the three-dimensional model of the building, determining the location of the worker, recording the pre-processed edge image, and monitoring the worker's status based on the pre-processed edge image.

[0061] In this step, the local stereo model is matched with the three-dimensional model of the building. During the matching process, in order to improve the matching efficiency, according to the work area assigned to each worker, the local three-dimensional model corresponding to the work area is extracted from the three-dimensional model of the building, and the local three-dimensional model is compared with the local stereo model to determine the position of the current local stereo model in the local three-dimensional model. The position of the worker is inferred based on the real-time binocular image, and whether the worker has any abnormal situation is determined based on the change in the worker's position.

[0062] like Figure 2 As shown, as a preferred embodiment of the present invention, the steps of acquiring a real-time binocular image, preprocessing the real-time binocular image, and obtaining two sets of preprocessed edge images specifically include:

[0063] S201 : Sending an image acquisition instruction to a binocular image collector at a preset time interval, and receiving a real-time binocular image acquired by the binocular image collector.

[0064] In this step, an image acquisition instruction is sent to the binocular image collector at a preset time interval. When performing image acquisition, the time interval can be set according to the speed of picture change. When the binocular image collector detects that the picture changes greatly, the time interval is shortened. Conversely, when the picture changes slightly, the time interval is increased. When evaluating the amount of picture change, the grayscale value of the statistical image can be used to determine it. For example, two adjacent images are obtained, grayscale processing is performed on them, and the grayscale values ​​of two pixels at the same position are compared. If the difference between the two is less than a preset value, it is determined to be an unchanged pixel, otherwise it is determined to be a changed pixel. When the proportion of changed pixels reaches a preset value, the time interval is shortened. The greater the proportion of changed pixels, the shorter the time interval, thereby continuously acquiring images to obtain real-time binocular images.

[0065] S202: cropping the real-time binocular image and converting it into a grayscale image.

[0066] S203 , performing edge recognition on the grayscale image based on an edge extraction algorithm, and converting the grayscale image into a preprocessed edge image.

[0067] In this step, the real-time binocular image is cropped, and the two sets of real-time binocular images are cropped into two sets of images with the same size. Grayscale processing is performed to remove the color information of the pixels in the image to obtain a grayscale image. Edge recognition is performed through the edge extraction algorithm, and the Canny edge detection algorithm is used for edge extraction. The specific steps include: gradient calculation: using the Sobel operator to calculate the gradient amplitude and direction of each pixel point, and the direction classification includes horizontal, vertical and diagonal; non-maximum suppression: retaining the local maximum value in the gradient direction, refining the edge, and ensuring that the edge width is a single pixel; dual threshold detection: high threshold, detecting strong edges, low threshold, set to 50% of the high threshold, the results are divided into three categories: strong edges, weak edges and non-edges; edge connection: connecting weak edges to strong edges; post-processing optimization stage: using the flood fill algorithm (FloodFill) to identify the edge area, and performing morphological operations to obtain a preprocessed edge image, such as Figure 9 and Figure 10 shown.

[0068] like Figure 3 As shown, as a preferred embodiment of the present invention, the steps of extracting depth information based on two sets of preprocessed edge images, performing stereo reconstruction based on the extracted depth information, and obtaining a local stereo model specifically include:

[0069] S301 , performing epipolar correction on the pre-processed edge image, and thinning and cleaning the edges in the pre-processed edge image to obtain an edge image to be matched.

[0070] In this step, epipolar correction is performed on the preprocessed edge images to obtain the camera parameters of the binocular image collector. The camera parameters include at least intrinsic parameters, distortion coefficients, and extrinsic parameters between the left and right cameras. The two sets of preprocessed edge images are reprojected, and then a thinning algorithm is used to convert the pixels in the preprocessed edge images to a single pixel width to represent the center of the line. Morphological operations (such as opening operations) or length-based threshold filtering are used to remove small, isolated noisy edge fragments, and attempts are made to reconnect the same edge fragments that are broken due to occlusion or noise but have the same direction.

[0071] S302 , using multiple matching methods to match the edge image to be matched, fusing the matching results, and generating edge point matching data, which is used to record the coordinates of two points that are successfully matched.

[0072] In this step, multiple matching methods are used to match the edge images to be matched. Two matching methods are used: point matching and line segment matching. In the process of point matching, two sets of preprocessed edge images are defined as the left image and the right image respectively. A series of points are uniformly sampled on an edge line of the left image. For each sampling point: a matching point is searched on the corresponding epipolar line (or its nearby neighborhood) of the right image. During matching, the NCC (normalized cross correlation) and SSD (sum of squared differences) between the matching points are calculated. According to the preset reference range, when the NCC (normalized cross correlation) and SSD (sum of squared differences) both fall within the corresponding reference range, the two sampling points are determined to have grayscale similarity. Secondly, the angle between the direction of the line connecting the sampling point pair and the baseline direction is checked. When the angle between the two is less than a preset value, it is determined that the geometric constraints between the sampling point pair are met. The edge direction consistency of the sampling point pair is further checked. When the grayscale similarity, geometric constraints and edge direction consistency are simultaneously met, the sampling point pair is considered to match each other to generate a point matching result.

[0073] Then, line segment matching is performed to extract the descriptive features of each line segment, including length, direction, endpoint coordinates, midpoint coordinates, average grayscale / color of the area, and relationship with adjacent line segments (such as connectivity, angle difference), etc. A distance function is defined, specifically using a weighted combination of Euclidean distance and angle difference to measure the similarity of the line segment features of the left and right images. When matching, local search and geometric constraints are used. For the left image, Figure 1Line segments are searched for candidate line segments near the possible extreme line positions in the right figure, and line segments are matched according to feature similarity and geometric constraints to obtain line segment matching results. After the matching is completed, the line segments in the line segment matching results are sampled and converted into multiple points, that is, according to the preset pixel interval, verification point pairs are extracted from each line segment (including two verification points, located in the left and right figures respectively), and the verification point pairs are compared with the matching sampling point pairs in the point matching results. If the two coincide, they are retained, otherwise they are discarded to obtain edge point matching data.

[0074] S303 , performing disparity calculation and depth calculation based on the edge point matching data, generating three-dimensional coordinate points based on the calculation results, and generating a local stereo model based on the three-dimensional coordinate points.

[0075] In this step, disparity calculation and depth calculation are performed based on the edge point matching data. When calculating disparity, for each pair of successfully matched points (x left ,y),(x right ,y), the disparity is expressed as d=x left -x right ;

[0076] Depth calculation, based on the binocular vision geometry principle, calculate the three-dimensional coordinates (x, y, z) of the point,

[0077] Where Z = (f*B) / d; X = (x left *B) / d; Y=(y*B) / d;

[0078] Where f is the focal length of the camera after correction, B is the baseline length of the binocular camera (the distance between the optical centers of the left and right cameras), and d is the parallax, based on which the coordinates (x, y, z) of the three-dimensional coordinate point are obtained;

[0079] When constructing a local stereo model, the three-dimensional coordinate points are marked in the three-dimensional model, and the three-dimensional coordinate points are connected according to the connection relationship between the three-dimensional coordinate points to obtain a framework diagram composed of points and lines in the three-dimensional space, that is, the local stereo model.

[0080] like Figure 4 As shown, as a preferred embodiment of the present invention, the steps of matching the local stereo model with the three-dimensional building model, determining the location of the worker, recording the pre-processed edge image, and monitoring the worker's status based on the pre-processed edge image include:

[0081] S401 , cutting the building 3D model based on the work space allocated to the worker to obtain a local 3D model.

[0082] S402 : Compare the local stereo model with the local three-dimensional model to determine the location of the local stereo model, and determine the location of the worker based on the real-time binocular image.

[0083] In this step, the three-dimensional model of the building is cropped. By cropping the model, the amount of data processing can be greatly reduced. When matching, the edge of the cropped local three-dimensional model is extracted, and the corresponding model edge is marked in the three-dimensional coordinate system. The model edge is the edge of the wall. The model edge is compared with the local stereo model, and the area of ​​the building that the worker is currently facing is determined based on the overlapping position. The position of the binocular image collector can be inferred based on the real-time binocular image, which is the position of the worker.

[0084] S403, transmitting the pre-processed edge image in the form of coordinate points, splicing the pre-processed edge image to generate a thumbnail image, and transmitting the real-time binocular image in real time when an abnormality occurs in the thumbnail image.

[0085] In this step, when transmitting the picture, a group of pre-processed edge images is selected for transmission, as shown in the left figure. In order to reduce the amount of data transmission, the pre-processed edge image is used as the processing object, a two-dimensional coordinate system is constructed, the coordinates of each pixel are determined, the edges in the image are sampled, the coordinates of the pixels corresponding to the sampling points are extracted, and a sampling point coordinate data set is obtained. The sampling point coordinate data set is transmitted. At the data receiving end, data is restored according to the sampling points and marked in the two-dimensional coordinate system to generate a corresponding grayscale image. The angle of the building in the grayscale image is used to determine whether the worker is abnormal. If the building in the grayscale image is tilted, it means that the worker is in a state of falling. At this time, the real-time binocular image is directly transmitted and an alarm is issued. Through the present invention, the position of the worker can be determined in real time, and the worker's position can be determined in a thumbnail image manner. If an abnormality occurs, an alarm is directly issued and a real-time image is transmitted to ensure the safety of the worker.

[0086] like Figure 5 As shown, a worker management system for labor dispatch provided by an embodiment of the present invention includes:

[0087] The model building module 100 is used to obtain building construction drawings, build a three-dimensional building model based on the building construction drawings, and allocate a workspace to each worker. The workers all wear safety helmets equipped with binocular image collectors.

[0088] In this system, the model building module 100 obtains the building construction drawings, which are three-dimensional drawings, and obtains the three-dimensional model of the building based on them. When the workers enter the site, their work areas are determined according to the workers' construction tasks, and the corresponding areas are divided for them in the three-dimensional model of the building as the workers' work spaces. The safety helmets worn by the workers are equipped with binocular image collectors, which are used to capture two images of the same position from different angles at the same time. The binocular image collectors are connected to the workers' mobile devices, and the data collected by the binocular image collectors are transmitted to the mobile devices for processing.

[0089] The image preprocessing module 200 is used to acquire real-time binocular images and preprocess the real-time binocular images to obtain two sets of preprocessed edge images.

[0090] In this system, the image preprocessing module 200 acquires real-time binocular images. After the workers enter the construction site, the binocular image collector is started. The binocular image collector continuously acquires images to obtain real-time binocular images. Two groups of real-time binocular images are generated each time. The real-time binocular images are preprocessed. The preprocessing steps include cropping and grayscale processing of the images. Finally, the edges in the preprocessed real-time binocular images are extracted through the edge recognition algorithm and converted into a preprocessed edge image. In the preprocessed edge image, only the edges of objects in the building interior and the edges of buildings are recorded.

[0091] The model reconstruction module 300 is used to extract depth information from the two sets of pre-processed edge images, and perform stereo reconstruction based on the extracted depth information to obtain a local stereo model.

[0092] In this system, the model reconstruction module 300 extracts depth information based on two sets of preprocessed edge images. During this process, epipolar line correction is performed to ensure that the corresponding points in the two sets of preprocessed edge images are located on the same horizontal scan line, and edge map refinement and cleaning are performed. Finally, edge line matching is performed. Based on the results of edge line matching, depth information is calculated, including parallax calculation, depth calculation and three-dimensional point generation. The corresponding local stereo model is constructed based on the generated three-dimensional points.

[0093] The personnel monitoring module 400 is used to match the local stereo model with the three-dimensional model of the building, determine the location of the workers, record the pre-processed edge images, and monitor the status of the workers based on the pre-processed edge images.

[0094] In this system, the personnel monitoring module 400 matches the local stereo model with the three-dimensional model of the building. During the matching process, in order to improve the matching efficiency, the local three-dimensional model corresponding to the work interval is extracted from the three-dimensional model of the building according to the work interval assigned to each worker. The local three-dimensional model is compared with the local stereo model to determine the position of the current local stereo model in the local three-dimensional model, and the position of the worker is inferred based on the real-time binocular image, and whether the worker has any abnormal situation is determined based on the change in the worker's position.

[0095] like Figure 6 As shown, as a preferred embodiment of the present invention, the image preprocessing module 200 includes:

[0096] The image acquisition unit 201 is configured to send an image acquisition instruction to the binocular image collector at a preset time interval and receive a real-time binocular image acquired from the binocular image collector.

[0097] In this system, the image acquisition unit 201 sends an image acquisition instruction to the binocular image collector at a preset time interval. When performing image acquisition, the time interval can be set according to the speed of picture change. When the binocular image collector detects that the picture changes greatly, the time interval is shortened. Conversely, when the picture changes slightly, the time interval is increased. When evaluating the amount of picture change, the grayscale value of the statistical image can be used to determine it. For example, two adjacent images are obtained, grayscale processing is performed on them, and the grayscale values ​​of two pixels at the same position are compared. If the difference between the two is less than a preset value, it is determined to be an unchanged pixel. Otherwise, it is determined to be a changed pixel. When the proportion of changed pixels reaches a preset value, the time interval is shortened. The greater the proportion of changed pixels, the shorter the time interval, thereby continuously acquiring images to obtain real-time binocular images.

[0098] The image processing unit 202 is used to crop the real-time binocular image and process it into a grayscale image.

[0099] The edge recognition unit 203 is configured to perform edge recognition on the grayscale image based on an edge extraction algorithm, and convert the grayscale image into a pre-processed edge image.

[0100] In this system, the image processing unit 202 performs cropping processing on the real-time binocular image, crops the two sets of real-time binocular images into two sets of images with the same size, performs grayscale processing, removes the color information of the pixels in the image to obtain a grayscale image, performs edge recognition through an edge extraction algorithm, and adopts the Canny edge detection algorithm to extract the edge. The specific steps include: performing gradient calculation: using the Sobel operator to calculate the gradient amplitude and direction of each pixel point, and the direction classification includes horizontal, vertical and diagonal; non-maximum suppression: retaining the local maximum value in the gradient direction, refining the edge, and ensuring that the edge width is a single pixel; dual threshold detection: high threshold, detecting strong edges, low threshold, set to 50% of the high threshold, the results are divided into three categories: strong edge, weak edge and non-edge; edge connection: connecting weak edges to strong edges; post-processing optimization stage: using the flood fill algorithm (Flood Fill) to identify the edge area, and performing morphological operations to obtain a pre-processed edge image, such as Figure 9 and Figure 10 shown.

[0101] like Figure 7 As shown, as a preferred embodiment of the present invention, the model reconstruction module 300 includes:

[0102] The edge refinement unit 301 is configured to perform epipolar correction on the pre-processed edge image, and to refine and clean the edges in the pre-processed edge image to obtain an edge image to be matched.

[0103] In this module, the edge refinement unit 301 performs epipolar correction on the preprocessed edge image, obtains the camera parameters of the binocular image collector, which include at least intrinsic parameters, distortion coefficients, and extrinsic parameters between the left and right cameras, performs reprojection transformation on the two sets of preprocessed edge images, and then uses a refinement algorithm to convert the pixels in the preprocessed edge image into a single pixel width to represent the center of the line. Morphological operations (such as opening operations) or length-based threshold filtering are used to remove small, isolated noisy edge fragments, and retry to connect the same edge fragments that are broken due to occlusion or noise but have the same direction.

[0104] The data matching unit 302 is used to match the edge images to be matched using multiple matching methods, fuse the matching results, and generate edge point matching data. The edge point matching data is used to record the coordinates of two points that are successfully matched.

[0105] In this module, the data matching unit 302 uses multiple matching methods to match the edge images to be matched. Two matching methods are used: a point matching method and a line segment matching method. In the point matching process, two sets of pre-processed edge images are defined as a left image and a right image respectively. A series of points are uniformly sampled on an edge line of the left image. For each sampling point: a matching point is searched on the corresponding epipolar line (or its nearby neighborhood) of the right image. During matching, the NCC (normalized cross correlation) and SSD (sum of squared differences) between the matching points are calculated. According to a preset reference range, when both the NCC (normalized cross correlation) and SSD (sum of squared differences) fall within the corresponding reference range, the two sampling points are determined to have grayscale similarity. Secondly, the angle between the direction of the line connecting the sampling point pair and the baseline direction is checked. When the angle between the two is less than a preset value, it is determined that the geometric constraints between the sampling point pair are satisfied. The edge direction consistency of the sampling point pair is further checked. When the grayscale similarity, geometric constraints, and edge direction consistency are simultaneously satisfied, the sampling point pair is considered to be matched to generate a point matching result.

[0106] Then, line segment matching is performed to extract the descriptive features of each line segment, including length, direction, endpoint coordinates, midpoint coordinates, average grayscale / color of the area, and relationship with adjacent line segments (such as connectivity, angle difference), etc. A distance function is defined, specifically using a weighted combination of Euclidean distance and angle difference to measure the similarity of the line segment features of the left and right images. When matching, local search and geometric constraints are used. For the left image, Figure 1 Line segments are searched for candidate line segments near the possible extreme line positions in the right figure, and line segments are matched according to feature similarity and geometric constraints to obtain line segment matching results. After the matching is completed, the line segments in the line segment matching results are sampled and converted into multiple points, that is, according to the preset pixel interval, verification point pairs are extracted from each line segment (including two verification points, located in the left and right figures respectively), and the verification point pairs are compared with the matching sampling point pairs in the point matching results. If the two coincide, they are retained, otherwise they are discarded to obtain edge point matching data.

[0107] The stereo model generating unit 303 is configured to perform disparity calculation and depth calculation based on the edge point matching data, generate three-dimensional coordinate points based on the calculation results, and generate a local stereo model based on the three-dimensional coordinate points.

[0108] In this module, the stereo model generation unit 303 performs disparity calculation and depth calculation based on the edge point matching data. When calculating disparity, for each pair of successfully matched points (x left ,y),(x right ,y), the disparity is expressed as d=x left -x right ;

[0109] Depth calculation, based on the binocular vision geometry principle, calculate the three-dimensional coordinates (x, y, z) of the point,

[0110] Where Z = (f*B) / d; X = (x left *B) / d; Y=(y*B) / d;

[0111] Where f is the focal length of the camera after correction, B is the baseline length of the binocular camera (the distance between the optical centers of the left and right cameras), and d is the parallax, based on which the coordinates (x, y, z) of the three-dimensional coordinate point are obtained;

[0112] When constructing a local stereo model, the three-dimensional coordinate points are marked in the three-dimensional model, and the three-dimensional coordinate points are connected according to the connection relationship between the three-dimensional coordinate points to obtain a framework diagram composed of points and lines in the three-dimensional space, that is, the local stereo model.

[0113] like Figure 8 As shown, as a preferred embodiment of the present invention, the personnel monitoring module 400 includes:

[0114] The model cutting unit 401 is used to cut the building 3D model based on the work space allocated to the worker to obtain a local 3D model.

[0115] The personnel positioning unit 402 is used to compare the local stereo model with the local three-dimensional model to determine the position of the local stereo model, and determine the position of the worker based on the real-time binocular image.

[0116] In this system, the three-dimensional model of the building is cropped. By cropping the model, the amount of data processing can be greatly reduced. When matching, the edge of the cropped local three-dimensional model is extracted, and the corresponding model edge is marked in the three-dimensional coordinate system. The model edge is the edge of the wall. The model edge is compared with the local stereo model, and the area of ​​the building that the worker is currently facing is determined based on the overlapping position. The position of the binocular image collector can be inferred based on the real-time binocular image, which is the position of the worker.

[0117] The image transmission unit 403 is used to transmit the pre-processed edge image in the form of coordinate points, splice the pre-processed edge image to generate a thumbnail image, and transmit the real-time binocular image in real time when an abnormality occurs in the thumbnail image.

[0118] In this system, when the image transmission unit 403 transmits the image, it selects a group of pre-processed edge images for transmission, as shown in the left figure. In order to reduce the amount of data transmission, the pre-processed edge image is used as the processing object, a two-dimensional coordinate system is constructed, the coordinates of each pixel are determined, the edges in the image are sampled, the coordinates of the pixels corresponding to the sampling points are extracted, and a sampling point coordinate data set is obtained. The sampling point coordinate data set is transmitted. At the data receiving end, data is restored according to the sampling points and marked in the two-dimensional coordinate system to generate a corresponding grayscale image. The angle of the building in the grayscale image is used to determine whether the worker is abnormal. For example, if the building is tilted in the grayscale image, it means that the worker is in a state of falling. At this time, the real-time binocular image is directly transmitted and an alarm is issued. Through the present invention, the position of the worker can be determined in real time, and the worker's position can be determined by thumbnail images. If an abnormality occurs, an alarm is directly issued and the real-time image is transmitted to ensure the safety of the worker.

[0119] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A worker management method for labor dispatch, characterized in that: The method comprises: Obtain building construction drawings, construct a 3D model of the building based on the drawings, assign workspaces to each worker, and have them wear safety helmets equipped with binocular image collectors. Acquire a real-time binocular image, pre-process the real-time binocular image, and obtain two sets of pre-processed edge images; Depth information is extracted from the two sets of pre-processed edge images, and stereo reconstruction is performed based on the extracted depth information to obtain a local stereo model; The local stereo model is matched with the three-dimensional model of the building to determine the location of the workers, and the pre-processed edge image is recorded. The status of the workers is monitored based on the pre-processed edge image.

2. The worker management method for labor dispatch according to claim 1, characterized in that: The step of acquiring a real-time binocular image and preprocessing the real-time binocular image to obtain two sets of preprocessed edge images specifically includes: Sending an image acquisition instruction to the binocular image collector at a preset time interval, and receiving a real-time binocular image acquired by the binocular image collector; Crop the real-time binocular image and convert it into a grayscale image; The edge of the grayscale image is identified based on the edge extraction algorithm, and the grayscale image is converted into a preprocessed edge image.

3. The worker management method for labor dispatch according to claim 1, characterized in that: The steps of extracting depth information based on the two sets of pre-processed edge images, and performing stereo reconstruction based on the extracted depth information to obtain a local stereo model specifically include: Perform epipolar correction on the preprocessed edge image, and refine and clean the edges in the preprocessed edge image to obtain an edge image to be matched; Use multiple matching methods to match the edge images to be matched, fuse the matching results, and generate edge point matching data. The edge point matching data is used to record the coordinates of the two points that are successfully matched; Disparity calculation and depth calculation are performed based on edge point matching data, three-dimensional coordinate points are generated based on the calculation results, and a local stereo model is generated based on the three-dimensional coordinate points.

4. The worker management method for labor dispatch according to claim 1, characterized in that: The steps of matching the local stereo model with the three-dimensional model of the building, determining the location of the worker, recording the pre-processed edge image, and monitoring the worker's status based on the pre-processed edge image include: Cutting the 3D model of the building based on the work space allocated to the workers to obtain a local 3D model; Compare the local stereo model with the local three-dimensional model to determine the location of the local stereo model, and determine the location of the worker based on the real-time binocular image; The pre-processed edge image is transmitted in the form of coordinate points, and the pre-processed edge image is spliced ​​to generate a thumbnail image. When an abnormality occurs in the thumbnail image, the real-time binocular image is transmitted in real time.

5. The worker management method for labor dispatch according to claim 4, characterized in that: The step of transmitting the preprocessed edge image in the form of coordinate points includes sampling the edges in the preprocessed edge image to extract a plurality of sampling points, generating sampling point coordinates according to the positions of the sampling points in the image, packaging the sampling point coordinates to obtain a sampling point coordinate data set, and transmitting the sampling point coordinate data set.

6. A worker management system for labor dispatch, characterized in that: The system comprises: The model building module is used to obtain building construction drawings, build a three-dimensional building model based on the building construction drawings, and assign a workspace to each worker. The workers all wear safety helmets equipped with binocular image collectors; An image preprocessing module is used to obtain real-time binocular images and preprocess the real-time binocular images to obtain two sets of preprocessed edge images; A model reconstruction module is used to extract depth information from two sets of preprocessed edge images, and perform stereo reconstruction based on the extracted depth information to obtain a local stereo model; The personnel monitoring module is used to match the local stereo model with the three-dimensional model of the building, determine the location of the workers, record the pre-processed edge image, and monitor the status of the workers based on the pre-processed edge image.

7. The worker management system for labor dispatch according to claim 6, characterized in that: The image preprocessing module includes: The image acquisition unit is used to send image acquisition instructions to the binocular image collector according to a preset time interval and receive real-time binocular images collected by the binocular image collector; An image processing unit, used to crop the real-time binocular image and process it into a grayscale image; The edge recognition unit is used to perform edge recognition on the grayscale image based on the edge extraction algorithm and convert the grayscale image into a preprocessed edge image.

8. The worker management system for labor dispatch according to claim 6, characterized in that: The model reconstruction module includes: An edge refinement unit is used to perform epipolar correction on the preprocessed edge image, and to refine and clean the edges in the preprocessed edge image to obtain an edge image to be matched; A data matching unit is used to match the edge image to be matched using multiple matching methods, fuse the matching results, and generate edge point matching data. The edge point matching data is used to record the coordinates of the two points that are successfully matched; The stereo model generating unit is used to perform parallax calculation and depth calculation based on edge point matching data, generate three-dimensional coordinate points based on the calculation results, and generate a local stereo model based on the three-dimensional coordinate points.

9. The worker management system for labor dispatch according to claim 6, characterized in that: The personnel monitoring module includes: A model cutting unit, configured to cut the three-dimensional model of the building based on the work space allocated to the worker to obtain a local three-dimensional model; The personnel positioning unit is used to compare the local stereo model with the local three-dimensional model to determine the location of the local stereo model and determine the location of the worker based on the real-time binocular image; The image transmission unit is used to transmit the pre-processed edge image in the form of coordinate points, splice the pre-processed edge image, generate a thumbnail image, and transmit the real-time binocular image in real time when an abnormality occurs in the thumbnail image.

10. The worker management system for labor dispatch according to claim 9, characterized in that: The step of transmitting the preprocessed edge image in the form of coordinate points includes sampling the edges in the preprocessed edge image to extract a plurality of sampling points, generating sampling point coordinates according to the positions of the sampling points in the image, packaging the sampling point coordinates to obtain a sampling point coordinate data set, and transmitting the sampling point coordinate data set.

Images