Exterior wall insulation board anchor bolt and anchor nail abnormal state identification method

Abstract

The invention discloses an external wall insulation board anchor bolt and anchor nail abnormal state identification method, which belongs to the technical field of external wall insulation construction, and adopts the technical scheme that a target image is acquired from a construction video image, and a sample set containing anchor bolt abnormity and anchor nail abnormity is constructed; carrying out labeling and data enhancement processing on anchor bolt abnormity and anchor nail abnormity in the sample set, and generating a standardized training set and a test set; performing model training by using the training sample set to obtain an anchor bolt anomaly recognition model and an anchor nail anomaly recognition model; verifying a detection performance index of the model through the test set until the performance index meets a preset requirement; real-time video images of the construction site are transmitted to the model for state recognition; anchor bolt abnormity and anchor nail abnormity events are marked, and early warning reminding is sent. The method has the beneficial effect that the method for identifying the abnormal state of the anchor bolt and the anchor nail of the external wall insulation board is provided.

Description

Technical Field

[0001] The invention belongs to the technical field of exterior wall insulation construction, and in particular relates to a method for identifying abnormal states of anchor bolts and nails of exterior wall insulation panels. Background Art

[0002] Exterior wall insulation construction faces the dual challenges of quality monitoring and safety hazards associated with working at height. Every year, substandard insulation construction leads to frequent problems such as insulation peeling and cracking. These problems are costly to repair, requiring not only new investment but also potential property damage and compensation. According to the "Technical Standard for Exterior Wall Insulation Engineering," JGJ 144-2019, Specification for the Construction of Exterior Wall Insulation Panels, anchor bolt installation depth and anchor placement sequence are key indicators affecting construction quality and safety. Currently, traditional inspection methods primarily rely on manual visual inspections. Manual visual inspections are time-consuming and labor-intensive, pose challenges in achieving full coverage, lack real-time monitoring, are subject to significant errors, and are prone to missed inspections. In particular, they cannot accurately identify insufficient anchor depth or improperly placed anchors.

[0003] In order to solve the above problems, a method for identifying abnormal states of anchor bolts of exterior wall insulation panels is provided. Summary of the Invention

[0004] The purpose of the present invention is to provide a method for identifying abnormal states of anchor bolts of exterior wall insulation panels.

[0005] A method for identifying abnormal states of anchor bolts of exterior wall insulation boards, characterized by comprising: S1. Collect target images from construction video images and construct a sample set containing anchor bolt anomalies and anchor nail anomalies; S2. Annotate and perform data augmentation on the anchor bolt anomalies and anchor nail anomalies in the sample set to generate standardized training and test sets. Extract static image frames from the surveillance video stream when constructing the training and test sets. Classify and annotate the anchor bolt anomalies and anchor nail abnormalities in the images through manual labeling, generally marking them as "anchor bolt protrusion" and "anchor nail and anchor bolt sequence abnormality." Generate a standardized image and label dataset for training.

[0006] Data augmentation generally includes at least three of the following: rotation, translation, scaling, image stretching, brightness perturbation, up / down flipping, image blending, mosaic stitching, and region replication. Data augmentation uses existing technologies, and the specific steps are not detailed here.

[0007] S3. Perform model training using the training sample set to obtain an anchor bolt anomaly recognition model and an anchor nail anomaly recognition model; S4. Verify the detection performance indicators of the model through the test set until the performance indicators meet the preset requirements; the performance indicators include accuracy and recall rate; S5, transmitting the real-time video image of the construction site to the model for status recognition; S6. Mark abnormal anchor bolts and abnormal anchor nail events, and send early warning reminders.

[0008] Furthermore, the model training is performed using the training sample set, including: Use public datasets to pre-train the model, extract low-level boundary information and high-level semantic information of the image, and fuse the low-level boundary information and high-level semantic information; The above process is iterated multiple times to complete feature learning of all data; based on the training sample sets of abnormal anchor bolts and abnormal anchor nails, the model parameters are fine-tuned in a supervised manner to enable the model to learn abnormal morphological characteristics and background characteristics; the model output layer is adjusted to distinguish between "normal anchor bolts" and "abnormal anchor bolts" and "normal anchor nails" and "abnormal anchor nails".

[0009] Furthermore, real-time video images of the construction site are transmitted to the model for status recognition, including: Identify and determine the center point of the target object from the input video image; Divide the image into regular grids and locate the grid where the center point of the target object is located; Calculate the target boundary coordinates based on the target object center point and the predefined anchor reference frame; The activation function is applied to determine the category confidence of the target boundary coordinates, and the recognition result is output based on the determination result.

[0010] Furthermore, the early warning reminder includes the following forms: real-time display on the system platform, and push notifications to management personnel through SMS notifications, mobile APPs, or reminders to construction personnel through voice; Marked anchor bolt anomalies and anchor nail anomaly events are displayed in real time on the construction monitoring screen through the bounding box.

[0011] Furthermore, the anchor bolt anomaly is a protruding anchor bolt, and the anchor nail anomaly is a simultaneous protrusion of the anchor nail and the anchor bolt after the anchor nail is inserted into the anchor bolt. During normal use, the anchor bolt should be fully driven into the wall before installing the anchor nail. Therefore, if the anchor nail is inserted into the anchor bolt and the anchor bolt remains protruding, it indicates that there is an issue with the installation sequence of the anchor nail and the anchor bolt.

[0012] The video image is obtained by shooting the exterior wall insulation board at a 45° angle. The camera is usually on a lifting construction platform, such as a lift or hanging basket.

[0013] The technical solution provided by the embodiments of the present invention has the following beneficial effects: This application utilizes an existing database pre-trained model for multi-level feature fusion and supervised fine-tuning of the model using the training set. Specifically, a two-stage training strategy is employed with the existing database, improving model training efficiency. Insufficient anchor depth can be identified by recognizing the protruding state of the anchor bolt; incorrect anchor sequence can be identified by recognizing the insertion of the anchor bolt and the simultaneous protrusion of the anchor bolt and the anchor bolt. This effectively addresses and reduces the problems of insufficient anchor installation depth and incorrect anchor placement sequence in existing construction. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solution of the present invention, the following is a brief introduction to the drawings used in the embodiments. Obviously, the drawings listed 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 any creative work.

[0015] Figure 1 This is a flow chart of a method for identifying abnormal states of anchor bolts for exterior wall insulation panels according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of the anchor bolt protruding and the anchor bolt in a normal state (the circle shows the anchor bolt protruding, and the box shows the anchor bolt in a normal state); Figure 3 This is a schematic diagram of the structure with abnormal and normal anchors (the left side shows a normal anchor and the right side shows an abnormal anchor); Figure 4 This is the identification status diagram of insufficient anchor depth; Figure 5 This is the identification status diagram of the incorrect anchor bolt sequence (the blue bounding box represents the incorrect anchor bolt sequence, and the yellow bounding box represents insufficient anchor bolt depth). DETAILED DESCRIPTION

[0016] 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 in conjunction with the embodiments. Of course, the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0017] See also Figure 1-Figure 5 A method for identifying abnormal state of anchor bolts of external wall insulation boards, characterized by comprising S1. Collect target images from construction video images and construct a sample set containing anchor bolt anomalies and anchor nail anomalies; S2. Annotate and perform data augmentation on the anchor bolt anomalies and anchor nail anomalies in the sample set to generate standardized training and test sets. Extract static image frames from the surveillance video stream when constructing the training and test sets. Classify and annotate the anchor bolt anomalies and anchor nail abnormalities in the images through manual labeling, generally marking them as "anchor bolt protrusion" and "anchor nail and anchor bolt sequence abnormality." Generate a standardized image and label dataset for training.

[0018] Data enhancement generally includes at least three of the following: rotation, displacement, scaling, image stretching, brightness perturbation, up / down flipping / left / right flipping, image mixing, mosaic stitching, and area replication.

[0019] S3. Perform model training using the training sample set to obtain an anchor bolt abnormal state recognition model and an anchor nail abnormal state recognition model; S4. Verify the detection performance indicators of the model through the test set until the performance indicators meet the preset requirements; performance indicators include accuracy and recall rate; accuracy and recall rate are generally greater than 90% to meet application requirements. Subsequently, the model will be iterated by expanding the sample set and enriching the diversity of construction scene data to improve the recognition accuracy of the model.

[0020] S5, transmitting the real-time video image of the construction site to the model for status recognition; S6. Mark abnormal anchor bolts and abnormal anchor nail events, and send early warning reminders.

[0021] Use the training sample set to train the model, including: Use public datasets to pre-train the model, extract low-level boundary information and high-level semantic information of the image, and fuse the low-level boundary information and high-level semantic information; The above process is iterated multiple times to complete feature learning of all data; based on the training sample sets of abnormal anchor bolts and abnormal anchor nails, the model parameters are fine-tuned in a supervised manner to enable the model to learn abnormal morphological characteristics and background characteristics; the model output layer is adjusted to distinguish between "normal anchor bolts" and "abnormal anchor bolts" and "normal anchor nails" and "abnormal anchor nails".

[0022] Transmit real-time video images of the construction site to the model for status recognition, including: Identify and determine the center point of the target object from the input video image; Divide the image into regular grids and locate the grid where the center point of the target object is located; Calculate the target boundary coordinates based on the target object center point and the predefined anchor reference frame; The activation function is applied to determine the category confidence of the target boundary coordinates, and the recognition result is output based on the determination result.

[0023] Early warning reminders include the following forms: real-time display on the system platform, and push notifications to management personnel through SMS or mobile APP or reminders to construction personnel through voice; Figure 4 and Figure 5 Insufficient anchor depth and incorrect anchor sequence are displayed in real time on the construction monitoring screen through frame selection.

[0024] An anchor bolt abnormality occurs when the anchor bolt protrudes, while an anchor nail abnormality occurs when the anchor bolt is inserted into the wall and both the anchor bolt and the anchor bolt protrude simultaneously. During normal use, the anchor bolt should be fully driven into the wall before installing the anchor bolt. Therefore, if the anchor bolt remains protruding after being inserted into the wall, it indicates an issue with the installation sequence of the anchor bolt and the anchor bolt. This can result in insufficient pullout force for the anchor bolt, potentially causing the insulation board to fall off.

[0025] The video image is obtained by shooting the exterior wall insulation board at a 45° angle. The camera is usually on a lifting construction platform, such as a lift or hanging basket.

[0026] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. 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 method for identifying abnormal states of anchor bolts of exterior wall insulation boards, characterized in that: include S1. Collect target images from construction video images and construct a sample set containing anchor bolt anomalies and anchor nail anomalies; S2. Annotate and perform data augmentation on the anchor bolt anomalies and anchor nail anomalies in the sample set to generate standardized training and test sets; S3. Perform model training using the training sample set to obtain an anchor bolt abnormal state recognition model and an anchor nail abnormal state recognition model; S4. Verify the detection performance indicators of the model through the test set until the performance indicators meet the preset requirements; S5, transmitting the real-time video image of the construction site to the model for status recognition; S6. Mark abnormal anchor bolts and abnormal anchor nail events, and send early warning reminders.

2. The method for identifying abnormal status of anchor bolts of exterior wall insulation boards according to claim 1, characterized in that: Use the training sample set to train the model, including: Use public datasets to pre-train the model, extract low-level boundary information and high-level semantic information of the image, and fuse the low-level boundary information and high-level semantic information; The above process was iterated multiple times to complete feature learning of all data in the public dataset. Based on the training sample sets of abnormal anchor bolts and abnormal anchor nails, the model parameters were fine-tuned in a supervised manner to enable the model to learn abnormal morphological characteristics and background features. The model output layer was adjusted to distinguish normal anchor bolts from abnormal anchor bolts, and normal anchor nails from abnormal anchor nails.

3. The method for identifying abnormal status of anchor bolts of exterior wall insulation boards according to claim 1, characterized in that: Transmit real-time video images of the construction site to the model for status recognition, including: Identify and determine the center point of the target object from the input video image; Divide the image into regular grids and locate the grid where the center point of the target object is located; Calculate the target boundary coordinates based on the target object center point and the predefined anchor reference frame; The activation function is applied to determine the category confidence of the target boundary coordinates, and the recognition result is output based on the determination result.

4. The method for identifying abnormal status of anchor bolts of exterior wall insulation boards according to claim 1, characterized in that: The warning reminder includes the following forms: real-time display on the system platform and push notification to management personnel via SMS notification or mobile APP; Marked anchor bolt anomalies and anchor nail anomaly events are displayed in real time on the construction monitoring screen through boundary boxes.

5. The method for identifying abnormal status of anchor bolts of exterior wall insulation boards according to claim 1, characterized in that: The anchor bolt abnormality is that the anchor bolt protrudes, and the anchor nail abnormality is that the anchor nail is inserted into the anchor bolt and the anchor nail and the anchor bolt protrude at the same time.

6. The method for identifying abnormal status of anchor bolts of exterior wall insulation boards according to claim 1, characterized in that: The construction video image is obtained by shooting the exterior wall insulation board construction at an oblique angle of 45 degrees.

Images