Refined building construction reinforcing steel bar binding recognition method and system based on image recognition

Abstract

The invention discloses a refined building construction reinforcing steel bar binding recognition method and system based on image recognition. The method comprises the steps of obtaining a building construction image and a video resource file; obtaining a target frame image based on the building construction video image file; performing feature extraction on the target frame image to obtain a target frame feature; position information between the target frame features is judged, and the steel bar binding condition is obtained based on the interaction condition of the personnel and the steel bars; according to the reinforcing steel bar binding coordinate information, calculating a convex hull of each target object through a ConvexHull algorithm; according to the enhancement features and the target query, determining a steel bar binding type and a region corresponding to steel bar binding in the target frame image; and based on the reinforcing steel bar binding type and the area corresponding to reinforcing steel bar binding, the current construction stage is judged. According to the invention, construction management personnel can be helped to master the progress condition of field steel bar binding in real time, and the working condition of workers is judged through refined identification, so that the construction efficiency is guaranteed.

Description

Technical Field

[0001] The present invention relates to the field of intelligent construction, and in particular to a refined construction steel bar binding identification method and system based on image recognition. Background Art

[0002] Rebar binding is one of the key links in construction, and its progress has a direct impact on the overall project progress. Traditional methods of rebar binding progress control usually rely on manual monitoring and experience judgment, and there are problems such as untimely information feedback and susceptibility to subjective factors, which leads to the inability of construction management personnel to grasp the actual progress of the project in a timely manner, and may miss the best window for implementing progress adjustment measures. With the development of science and technology, the application of computer vision technology in construction has gradually increased. By installing cameras or using drones to capture image data of the construction site in real time, the progress of rebar binding is dynamically tracked and analyzed using intelligent image recognition technology. This method can not only monitor the completion progress of rebar binding in real time, but also accurately record the actual progress of each stage, provide data support for the adjustment and optimization of the construction plan, thereby reducing human intervention and errors, and improving the accuracy and efficiency of construction progress management.

[0003] Patent document CN115272691A discloses a training method, identification method and equipment for a steel bar binding state detection model. By collecting sample images of steel bar binding points, enhancing and processing to form a training set and annotating them, YOLOv4 is used for model training to obtain a steel bar binding state detection model, which is used to automatically determine whether the binding point has been tied. This patent improves the efficiency of steel bar binding state recognition and is a quality control method. It cannot provide the progress of the steel bar project during the construction phase.

[0004] Patent document CN116092012A discloses a method and device for monitoring the steel bar binding process based on video stream. By setting the construction personnel and the steel bars in the construction personnel ROI area as the first recognition object and the second recognition object respectively, the corresponding first recognition model and the second recognition model are trained. The first recognition model is used to roughly inspect the image to obtain the ROI area centered on the first object, and the second recognition model is used to finely inspect the number and position of the truncated steel bars in the expanded ROI area. Whether the number of truncated steel bars is greater than the predetermined number is used as the judgment standard to judge whether the steel bar binding of the component is approaching completion. The patent uses the bottom steel bar and the truncated steel bar as the signs of the start and end of the steel bar project respectively, and can detect the two nodes of the start and end of the steel bar project at the component level. However, for components with huge steel bar projects, the binding duration is often long. The patent cannot provide real-time feedback on the current construction activities and whether the workers are carrying out the work, and lacks progress feedback on the construction process. In addition, the second recognition object of the patent only judges the construction progress based on the number of truncated steel bars in the construction personnel ROI area, which is highly dependent on the accuracy of recognition and has a large uncertainty in recognition. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention provides a refined construction steel bar binding identification method and system based on image recognition.

[0006] In order to achieve the above technical objectives, the technical solution adopted by the present invention is: a refined construction steel bar binding identification method based on image recognition, comprising:

[0007] S1: Acquire building construction images and video resource files;

[0008] S2: Based on the building construction video image file, obtain the target frame image;

[0009] S3: extracting features from the target frame image to obtain target frame features;

[0010] S4: judging the position information between the features of each target frame, and obtaining the specific steel bar binding situation based on the interaction between the personnel and the steel bars;

[0011] S5: Calculate the convex hull of each target object using the ConvexHull algorithm according to the reinforcement binding coordinate information;

[0012] S6: Determine the steel bar binding type and the area corresponding to the steel bar binding in the target frame image according to the enhanced features and the target query.

[0013] As a possible implementation manner, further, in step S2, the target frame image is an image frame in which personnel and steel bars exist in the construction video image.

[0014] As a possible implementation, further, in step S3, edge feature information of construction workers, construction materials, steel bars, etc. in each image frame of the construction video image is extracted by an edge detection algorithm to obtain target frame features.

[0015] As a possible implementation, further, step S3 also includes: enhancing the precise regional contour features of the steel bar binding through an instance segmentation algorithm on the acquired edge feature information; and setting a template library of steel bar types to facilitate target query, provide a reference for the identification of steel bar types, and increase the identification accuracy.

[0016] As a possible implementation, further, step S4 specifically includes the following steps:

[0017] S41: Image coordinate point information extraction

[0018] By reading the file content containing the target object category and coordinate information line by line, and processing each line of data, splitting it into the target object category and the corresponding coordinate point, and then returning a list containing all target objects, each target object contains its type and coordinate information, based on which to create an identification information database;

[0019] S42: Coordinate information preprocessing

[0020] In order to ensure that the coordinate system meets the usage requirements, the identified coordinates are normalized to the [0,1] coordinate system, and the y-axis coordinates are mirrored around the line y=0.5 to unify the directions of all coordinates. All input coordinate points are traversed, and the flipped y-axis coordinates are calculated in turn to generate a new coordinate point list.

[0021] S43: Coordinate information preprocessing

[0022] Display the data of each target object type in different colors and line types according to its category to distinguish objects of different categories;

[0023] The ConvexHull algorithm is used to calculate the convex hull of each target object, that is, the minimum outer envelope of the area occupied by the target object, and draw it on the image;

[0024] S44: Determination of steel bar binding conditions and output of results

[0025] Determine whether the convex hull of the object to which the construction worker belongs intersects with the convex hull of other types of target objects. If so, it is determined that the construction worker is tying steel bars;

[0026] By comparing and analyzing the changes in the relative positions of personnel and steel bars between adjacent frames, the specific operations of personnel during the binding process can be judged.

[0027] As a possible implementation, further, in step S5, a pole (usually the leftmost or bottommost point of the point set) is selected from the object to which the identified steel bar belongs, and then the remaining points are sorted by pole angle, and then these points are scanned to maintain a convex hull stack to determine whether the point belongs to the convex hull;

[0028] As a possible implementation, further, in step S6, the area where the convex hull of the construction worker's object intersects with the convex hull of other types of target objects is used as the area corresponding to the steel bar binding;

[0029] The steel bar binding type corresponding to the steel bar binding area is determined based on the template library of steel bar types and the enhanced precise area contour features of the steel bar binding obtained through the instance segmentation algorithm.

[0030] As a possible implementation mode, further, the refined construction steel bar binding identification method based on image recognition also includes S7: judging the current construction stage based on the steel bar binding type and the area corresponding to the steel bar binding.

[0031] The present invention also provides a refined construction steel bar binding identification system based on image recognition, which uses the refined construction steel bar binding identification method based on image recognition, and the system includes:

[0032] Image resource acquisition module, used to acquire building construction images and video resource files;

[0033] The target frame recognition module is used to extract the image frames containing personnel and steel bars in the construction video images;

[0034] The image recognition module is used to extract edge feature information of construction personnel, building materials, and steel bars in each image frame of the construction video image to obtain target frame features;

[0035] The type discrimination module is used to judge the position information between the features of each target frame, obtain the specific steel bar binding situation based on the interaction between personnel and steel bars, and determine the steel bar binding type and the corresponding area of ​​steel bar binding in the target frame image according to the enhanced features and target query.

[0036] As a possible implementation, further, the refined construction steel bar binding identification system based on image recognition also includes: a result saving module, which is used to save the relevant results of the construction personnel's steel bar binding identification.

[0037] By adopting the above technical solution, the present invention has the following beneficial effects compared with the prior art:

[0038] 1) The present invention draws corresponding convex hull graphics according to the shape of the steel bars and the on-site actions of the construction workers, and determines the actual steel bar binding situation according to the interaction of different convex hull position information, which can greatly improve the accuracy of recognition. In addition, the present invention can separately determine the binding of beams, slabs, columns and other steel bars, thereby achieving accurate judgment of the current construction stage.

[0039] 2) The present invention provides real-time feedback on the progress of steel bar binding, so that management personnel can understand the construction status in a timely manner, reduce construction delays, and help to reasonably allocate resources.

[0040] 3) The present invention can help construction management personnel to control the on-site steel bar binding status in real time, and ensure construction efficiency by judging the workers' working conditions through refined identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0042] Figure 1 It is a technical flow chart of the present invention;

[0043] Figure 2 It is a schematic diagram for identifying the construction workers tying the beam and slab reinforcement.

[0044] Figure 3 It is a system interface diagram of the present invention. DETAILED DESCRIPTION

[0045] The present invention will be further described in detail below in conjunction with the accompanying drawings and examples. It is particularly noted that the following examples are only used to illustrate the present invention, but are not intended to limit the scope of the present invention. Similarly, the following examples are only partial embodiments of the present invention rather than all embodiments, and all other embodiments obtained by those of ordinary skill in the art without creative work are within the scope of protection of the present invention.

[0046] See attached Figure 1 As shown, this embodiment provides a refined construction steel bar binding identification method based on image recognition, including:

[0047] S1: Acquire building construction images and video resource files.

[0048] There are many methods for acquiring video images of building construction. First, fixed cameras set up on the construction site can be used for acquisition. These cameras can capture images of the construction process around the clock to ensure that the main construction areas of the entire construction site are covered. Secondly, designated workers can shoot the construction buildings at specific times of the day. Workers manually operate the camera equipment according to a predetermined schedule to ensure that the latest construction progress is captured. In addition, robots or drones equipped with video acquisition equipment can be used to uniformly collect images of various construction buildings on the construction site at preset time intervals (such as every 12 hours or every day) and predetermined routes. Finally, high-altitude camera equipment such as tower crane cameras can be used to shoot construction buildings from above. High-altitude shooting can provide a wider perspective and capture more construction details and overall progress. Through the above methods, video images of building construction can be efficiently acquired, the workload of workers can be reduced, and the comprehensiveness and clarity of the images can be ensured.

[0049] S2: Based on the construction video image file, a target frame image is obtained. The target frame image is an image frame in which people and steel bars exist in the construction video image. In this embodiment, an image recognition algorithm is used to detect and identify steel bars and people in the video image. Specifically, the system analyzes each pixel in the video frame and identifies the characteristics of steel bars and people through a pre-trained model. For example, steel bars usually have specific shape, color and texture characteristics, while people are identified by body contours, movements and clothing features, and then the target frame image is extracted.

[0050] S3: Extract features from the target frame image to obtain target frame features; in this embodiment, edge feature information of construction personnel, building materials, and steel bars in each image frame of the construction video image is extracted through an edge detection algorithm to obtain target frame features. Among them, edge information is a significant change area in the image, which usually represents the contour and structural features of the object. These edge information are the core part of the target frame features, reflecting the shape and position of the main objects in the image. This feature extraction method can not only improve the accuracy and efficiency of construction image analysis, but also provide rich basic data for subsequent image processing and decision-making. In addition, the acquired edge feature information is enhanced through an instance segmentation algorithm to enhance the precise regional contour features of the steel bar binding to improve the recognition efficiency; and a template library of steel bar types is set to facilitate target query, provide a reference for the identification of steel bar types, and increase recognition accuracy.

[0051] S4: judging the position information between the features of each target frame, and obtaining the specific steel bar binding situation based on the interaction between the personnel and the steel bars; specifically comprising the following steps:

[0052] S41: Image coordinate point information extraction

[0053] By reading the file content containing the target object category (including construction personnel, beams, walls, etc.) and coordinate information line by line, and processing each line of data, splitting it into the target object category and the corresponding coordinate point, and then returning a list containing all target objects, each target object contains its type and coordinate information, based on which to create an identification information database;

[0054] S42: Coordinate information preprocessing

[0055] In order to ensure that the coordinate system meets the usage requirements, the recognized coordinates are normalized to the [0,1] coordinate system, and the y-axis coordinates are mirrored around the line y=0.5 to unify the directions of all coordinates. All input coordinate points are traversed, and the flipped y-axis coordinates are calculated in turn to generate a new coordinate point list; the processed data can accurately reflect the position of the actual object in the image.

[0056] S43: Coordinate information preprocessing

[0057] The data of each target object type is displayed in different colors and line types according to its category to distinguish objects of different categories. At the same time, the convex hull of each target object, that is, the minimum outer envelope of the area occupied by the target object, is calculated using the ConvexHull algorithm and drawn on the image; the result of this step is a coordinate map showing different objects, which clearly shows the location and shape of the objects.

[0058] S44: Determination of steel bar binding conditions and output of results

[0059] Determine whether the convex hull of the object to which the construction worker belongs intersects with the convex hull of other types of target objects. If so, it is determined that the construction worker is tying steel bars; this method can be used to determine whether the worker is working. By comparing and analyzing the relative position changes between the personnel and the steel bars between adjacent frames, the specific operation of the personnel during the tying process can be determined.

[0060] S5: Calculate the convex hull of each target object using the ConvexHull algorithm based on the reinforcement binding coordinate information. Specifically, when constructing the convex hull, determine whether the three points (p1, p2, p3) form a counterclockwise turn. A common method is to calculate the cross product. The cross product formula is as follows:

[0061] cross(p1,p2,p3)=(x2-x1)(y3-y1)-(y2-y1)(x3-x1)

[0062] If cross>0, the three points form a counterclockwise rotation, which means that the current point is outside the convex hull and should be added.

[0063] If cross<0, the three points form a clockwise turn, which means that the current point is inside the convex hull, and the point on the top of the stack should be popped out until a counterclockwise turn is formed.

[0064] If cross=0, the three points are collinear, which means that the current point is on the already constructed boundary and can be directly added.

[0065] Through the above steps, we can finally get the boundary point set p1, p2, ..., p of the convex hull m , where m is the number of vertices of the convex hull, and these points are arranged in a counterclockwise order. Based on this, a specific steel bar edge contour image can be obtained.

[0066] S6: Determine the type of steel bar binding and the area corresponding to the steel bar binding in the target frame image according to the enhanced features and target query. Specifically, the area where the convex hull of the construction worker object intersects with the convex hull of other types of target objects is used as the area corresponding to the steel bar binding; determine the steel bar binding type corresponding to the steel bar binding area using the template library of steel bar types and the enhanced precise area contour features of the steel bar binding obtained through the instance segmentation algorithm.

[0067] According to the enhanced features and target query, the steel bar binding type and the corresponding binding area are determined in the target frame image, which can achieve accurate identification and analysis of the steel bar binding situation and help locate the actual position of the steel bar binding. This method not only improves the accuracy and efficiency of steel bar binding, but also provides strong technical support for quality control and supervision of construction sites.

[0068] For ease of understanding, an example is provided below to illustrate. After the image is drawn, first, the algorithm checks whether the convex hull of type 0 (construction personnel) intersects with the convex hulls of other types. Determine whether the line segments between different objects have intersections. Secondly, if the convex hull of type 1 (beam and slab) intersects with the convex hull of type 0 (construction personnel), the program will output a prompt "Judgment is tying beam and slab reinforcement"; if the convex hull of type 2 (wall column) intersects with type 0 (construction personnel), the prompt "Judgment is tying wall column reinforcement" is output. This judgment method based on geometric intersection helps users quickly determine the type of reinforcement tying in the current construction. Finally, the drawn image is presented to the user in a visual way. The image contains the coordinate points of each type of object and its convex hull boundary. Any intersecting convex hull line segments are highlighted with red dashed lines, allowing users to view the analysis results at a glance. It provides a clear basis for judging the current construction progress.

[0069] By judging the edge position information between each target person and each type of steel bar, the specific situation of steel bar binding can be accurately analyzed and understood. Figure 2As shown in the figure, the image processing technology is used to extract the features of the personnel and steel bars in the image, including the position and posture of the personnel and the shape, size and position of the steel bars. Then, by comparing and analyzing the relative position changes of the personnel and steel bars between adjacent frames, the specific operation of the personnel in the binding process is judged. The results show that there are two construction workers tying the beam and slab steel bars in this picture.

[0070] S7: Based on the type of rebar binding and the corresponding area of ​​rebar binding, determine the current construction stage. Based on the type of rebar binding and the corresponding binding area, the current construction stage can be effectively determined, thereby providing managers with accurate construction progress information. By analyzing the identified rebar types and the construction areas where they are located, the progress of rebar binding at each stage can be identified. Based on this progress information, managers can promptly identify potential delays or problems and more reasonably allocate resources, such as personnel, machinery, and materials, to ensure that the construction plan is executed on time and improve the overall construction management efficiency and project quality. In addition, this refined identification method helps to better control the progress and resource allocation of the project, optimize all aspects of engineering management, ensure the smooth progress of the project and reduce construction delays.

[0071] This embodiment also provides a refined construction steel bar binding identification system based on image recognition, which applies the refined construction steel bar binding identification method based on image recognition, as shown in the attached Figure 3 As shown, the system includes:

[0072] The image resource acquisition module is used to obtain construction image and video resource files. Users need to select the resource input method suitable for the current application scenario through the image resource acquisition module. This can include uploading images or videos from the local file system, capturing on-site construction images through a real-time camera, or obtaining image resources from a remote server. Users can choose the optimal input method according to actual needs to ensure that the system can successfully receive the required construction image data.

[0073] The target frame recognition module is used to extract the image frames containing personnel and steel bars in the construction video images;

[0074] The image recognition module is used to extract edge feature information of construction workers, building materials, and steel bars in each image frame of the construction video image to obtain the target frame features. In the image recognition module, the system automatically recognizes steel bars, construction workers, etc. based on the input image resources and set parameters. The module uses deep learning algorithms or specific target detection models to identify the edge shape and position of steel bars and the actions of construction workers from the image. The final output recognition result will reflect the key feature information of the construction site in the current image.

[0075] The type discrimination module is used to judge the position information between the features of each target frame, obtain the specific steel bar binding situation based on the interaction between the personnel and the steel bars, and determine the steel bar binding type and the corresponding area of ​​the steel bar binding in the target frame image based on the enhanced features and target query. In this module, the system will analyze and classify the identified features to specifically judge the construction personnel's operation of tying steel bars in the image. This includes judging whether the construction personnel are performing steel bar binding operations. The output results of this module will provide key supervision information for the construction site.

[0076] The result saving module is used to save the relevant results of the construction personnel's steel bar binding identification. After the type discrimination is completed, the final identification and analysis results will be saved in the result saving module. The system will record the construction personnel's steel bar binding situation as structured data for subsequent supervision and tracking. In this way, construction management personnel can call the identification results at any time to evaluate the key details of the construction process and ensure the standardization of the steel bar binding operation. The saved results can be further used to generate reports, data visualization or conduct historical record comparison analysis.

[0077] The above descriptions are only some embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Any equivalent device or equivalent process transformation made 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 refined construction steel bar binding identification method based on image recognition, characterized in that: include: S1: Acquire building construction images and video resource files; S2: Based on the building construction video image file, obtain the target frame image; S3: extracting features from the target frame image to obtain target frame features; S4: judging the position information between the features of each target frame, and obtaining the specific steel bar binding situation based on the interaction between the personnel and the steel bars; S5: Calculate the convex hull of each target object using the ConvexHull algorithm according to the reinforcement binding coordinate information; S6: Determine the steel bar binding type and the area corresponding to the steel bar binding in the target frame image according to the enhanced features and the target query.

2. The refined construction steel bar binding identification method based on image recognition according to claim 1 is characterized in that: In step S2, the target frame image is an image frame in which personnel and steel bars exist in the construction video image.

3. The refined construction steel bar binding identification method based on image recognition according to claim 1 is characterized in that: In step S3, edge feature information of construction workers, construction materials and steel bars in each image frame of the construction video image is extracted by an edge detection algorithm to obtain target frame features.

4. The refined construction steel bar binding identification method based on image recognition according to claim 3 is characterized in that: Step S3 also includes: enhancing the precise regional contour features of the steel bar binding through an instance segmentation algorithm for the acquired edge feature information; and setting a template library of steel bar types to facilitate target query, provide a reference for the identification of steel bar types, and increase the identification accuracy.

5. The refined construction steel bar binding identification method based on image recognition according to claim 4 is characterized in that: Step S4 specifically includes the following steps: S41: Image coordinate point information extraction By reading the file content containing the target object category and coordinate information line by line, and processing each line of data, splitting it into the target object category and the corresponding coordinate point, and then returning a list containing all target objects, each target object contains its type and coordinate information, based on which to create an identification information database; S42: Coordinate information preprocessing In order to ensure that the coordinate system meets the usage requirements, the identified coordinates are normalized to the [0,1] coordinate system, and the y-axis coordinates are mirrored around the line y=0.5 to unify the directions of all coordinates. All input coordinate points are traversed, and the flipped y-axis coordinates are calculated in turn to generate a new coordinate point list. S43: Coordinate information preprocessing Display the data of each target object type in different colors and line types according to its category to distinguish objects of different categories; The ConvexHull algorithm is used to calculate the convex hull of each target object, that is, the minimum outer envelope of the area occupied by the target object, and draw it on the image; S44: Determination of steel bar binding conditions and output of results Determine whether the convex hull of the object to which the construction worker belongs intersects with the convex hull of other types of target objects. If so, it is determined that the construction worker is tying steel bars; By comparing and analyzing the changes in the relative positions of personnel and steel bars between adjacent frames, the specific operations of personnel during the binding process can be judged.

6. The refined construction steel bar binding identification method based on image recognition according to claim 5 is characterized in that: Step S5 selects a pole, sorts the remaining points by pole angle, and then scans these points to maintain a convex hull stack to determine whether the point belongs to the convex hull.

7. The refined construction steel bar binding identification method based on image recognition according to claim 5 is characterized in that: In step S6, the area where the convex hull of the construction worker object intersects with the convex hull of other types of target objects is used as the area corresponding to the steel bar binding; The steel bar binding type corresponding to the steel bar binding area is determined based on the template library of steel bar types and the enhanced precise area contour features of the steel bar binding obtained through the instance segmentation algorithm.

8. The refined construction steel bar binding identification method based on image recognition according to claim 1 is characterized in that: The method further includes S7: judging the current construction stage based on the steel bar binding type and the area corresponding to the steel bar binding.

9. A refined construction steel bar binding identification system based on image recognition, characterized in that: The refined construction steel bar binding identification method based on image recognition as claimed in any one of claims 1 to 8 is applied, and the system comprises: Image resource acquisition module, used to acquire building construction images and video resource files; The target frame recognition module is used to extract the image frames containing personnel and steel bars in the construction video images; An image recognition module is used to extract edge feature information of construction personnel, construction materials and steel bars in each image frame of the construction video image to obtain target frame features; The type discrimination module is used to judge the position information between the features of each target frame, obtain the specific steel bar binding situation based on the interaction between personnel and steel bars, and determine the steel bar binding type and the corresponding area of ​​steel bar binding in the target frame image according to the enhanced features and target query.

10. The refined construction steel bar binding identification system based on image recognition according to claim 9 is characterized in that: Also includes: The result saving module is used to save the relevant results of the identification of the steel bars tied by the construction workers.

Images