Prefabricated wall thermal insulation layer and connecting piece detection method and device and readable medium
Through three-dimensional point cloud technology and neural network model, the prefabricated wall insulation layer and connectors are automatically detected, which solves the problems of low manual detection efficiency and difficulty in digital storage, and achieves fast and accurate detection results and real-time data call.
Patent Information
- Application Number
- CN202510352043.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-25
- Publication Date
- 2025-07-04
AI Technical Summary
In the prior art, the detection of prefabricated wall insulation layers and connectors is greatly affected by human subjective factors, and the detection efficiency is low. The detection cannot be completed before the concrete is initially set, and the detection results cannot be digitally stored and real-time callable.
Three-dimensional point cloud technology and neural network model are used for automated detection, and the overall photos are obtained through the photo equipment, and three-dimensional point cloud computing and analysis are performed. The pre-trained insulation board joint width, connector position and height detection model is used to generate detection results, and compare them with product requirements, and automatically warning and adjustment.
It realizes fast and accurate detection, and the detection time is controlled within three minutes, which improves the product pass rate, and realizes electronic storage and real-time call of detection data.
Smart Images
Figure CN120259248A_ABST
Abstract
Description
Technical Field
[0001] Embodiments of the present disclosure relate to the technical field of precast wall insulation layer and connector identification, and specifically to methods, devices, and readable media for detecting precast wall insulation layers and connectors. Background Art
[0002] Precast sandwich insulation wall panels integrate multiple functions such as thermal insulation, fire protection, decoration, and enclosure of building exterior walls, and have advantages such as good durability, excellent fire performance, and easy maintenance throughout the life cycle. The laying accuracy of the insulation layer directly affects the thermal insulation performance of the precast wall, and the installation position of the connector and the depth of the connector extending into the inner wall (i.e., the height of the connector protruding from the insulation board) directly affect the safety of the outer leaf wall. The laying accuracy of the insulation layer, the installation position of the connector, and the depth of the connector extending into the inner wall are key elements that need to be detected and controlled during the production process of precast walls. Currently, the laying accuracy of the insulation layer, the installation position of the connector, and the depth of the connector extending into the inner wall are all inspected by manual visual inspection and measurement with a ruler. Workers observe the width of the insulation board joints and make adjustments, and workers hold a steel tape measure to measure the position of the connector and the height of the connector protruding from the insulation board and compare it with the design requirements, and adjust the connectors that do not meet the requirements.
[0003] However, it has been found in practice that when detecting precast wall insulation layers and connectors in the above manner, the following technical problems exist:
[0004] First, manual inspection is greatly affected by human subjective factors and has low detection efficiency. The number of connectors on a precast wall is generally not less than 20, and it is impossible to effectively detect all connectors.
[0005] Second, both the insulation layer and the connectors are attached to the unhardened outer leaf wall concrete below. Manual inspection is slow, and it is impossible to detect and adjust all connectors before the concrete initial setting.
[0006] Third, the results of manual inspection are stored in a manual recording manner, and it is impossible to achieve digital storage and retrieval of inspection data, which is not convenient for production process supervision.
[0007] The above information disclosed in this background art section is only used to enhance the understanding of the background of the inventive concept, and therefore, it may include information that does not form the prior art known to those of ordinary skill in the art in this country. Summary of the Invention
[0008] This summary of the disclosure is used to briefly introduce concepts that will be described in detail in the subsequent detailed implementation section. This summary of the disclosure is not intended to identify the key features or essential features of the claimed technical solution, nor is it intended to limit the scope of the claimed technical solution.
[0009] Some embodiments of the present disclosure propose a method, device, and readable medium for detecting precast wall insulation layers and connectors to solve one or more of the technical problems mentioned in the above background art section.
[0010] In a first aspect, some embodiments of the present disclosure provide a method for detecting precast wall insulation layers and connectors, the method including: controlling a photographing device to take an overall photo of the precast wall insulation layer and the connectors; performing preprocessing calculations on the above overall photo, and performing 3D point cloud operations based on the processed overall photo to obtain the overall point cloud data of the precast wall insulation layer and the connectors; parsing the above overall point cloud data to obtain a point cloud data set of the precast wall insulation layer and the connectors; inputting the above point cloud data set of the precast wall insulation layer and the connectors into a pre-trained recognition model of the precast wall insulation layer and the connectors to obtain a recognition information set of the precast wall insulation layer and the connectors, wherein the pre-trained recognition model of the precast wall insulation layer and the connectors includes: a detection model for the width of the insulation board joint, a recognition model for the position of the connector, and a detection model for the height of the connector, and the image recognition information set of the precast wall insulation layer and the connectors includes: the width of the insulation board joint, the position of the connector, and the height of the connector; comparing the above recognition information set of the precast wall insulation layer and the connectors with the product requirements to obtain a detection result of the precast wall insulation layer and the connectors; in response to the detection result of the precast wall insulation layer and the connectors being qualified, proceeding to the next production process, and if the detection result of the precast wall insulation layer and the connectors is unqualified, giving an alarm and reminding the operator to make adjustments.
[0011] In some embodiments, the above parsing the overall point cloud data to obtain a point cloud data set of the precast wall insulation layer and the connectors includes: performing data preprocessing on the overall point cloud data to generate processed overall point cloud data; performing filtering processing on the processed overall point cloud data to generate filtered overall point cloud data; performing structured feature extraction on the filtered overall point cloud data to generate a structured point cloud coordinate sequence set; and performing point cloud data grouping processing on the overall point cloud data according to the connection lines of the respective structured point cloud coordinates in the structured point cloud coordinate sequence set to generate a point cloud data set of the precast wall insulation layer and the connectors.
[0012] In some embodiments, the above comparing the recognition information set of the precast wall insulation layer and the connectors with the product requirements includes: extracting the requirements for the width of the insulation board joint, the position of the connector, and the height of the connector through the precast wall design document to generate a product requirement data set; aligning the recognition information set of the precast wall insulation layer and the connectors with the product requirement data set, and comparing the width of the insulation board joint, the position of the connector, and the height of the connector.
[0013] Specifically, the above precast wall design document is a 2D drawing in dwg format or a 3D BIM model.
[0014] In some embodiments, the insulation board joint width detection model is trained through the following steps: Obtain a first training sample set, where the training samples in the first training sample set include insulation board sample images and sample joint width information; Based on the above first training sample set, perform the following processing steps: Input the insulation board sample images included in at least one training sample in the first training sample set into the initial insulation board joint width detection model respectively, and obtain the joint width information corresponding to each training sample in the at least one training sample; Compare the joint width corresponding to each training sample in the at least one training sample with the corresponding sample joint width; Determine whether the initial insulation board joint width detection model reaches a preset optimization target according to the comparison result; In response to determining that the initial insulation board joint width detection model reaches the above optimization target, use the initial insulation board joint width detection model as the trained insulation board joint width detection model; In response to determining that the initial insulation board joint width detection model does not reach the above optimization target, adjust the model parameters of the initial insulation board joint width detection model, and use the unused training samples to form a training sample set. Use the adjusted initial insulation board joint width detection model as the initial insulation board joint width detection model, and perform the above processing steps again.
[0015] In some embodiments, the connector position recognition model is trained through the following steps: Obtain a second training sample set, where the training samples in the second training sample set include connector sample images and sample connector position information; Based on the above second training sample set, perform the following processing steps: Input the connector sample images included in at least one training sample in the second training sample set into the initial connector position recognition model respectively, and obtain the connector position information corresponding to each training sample in the at least one training sample; Compare the connector position information corresponding to each training sample in the at least one training sample with the corresponding sample connector position information; Determine whether the initial connector position recognition model reaches a preset optimization target according to the comparison result; In response to determining that the initial connector position recognition model reaches the above optimization target, use the initial connector position recognition model as the trained connector position recognition model; In response to determining that the initial connector position recognition model does not reach the above optimization target, adjust the model parameters of the initial connector position recognition model, and use the unused training samples to form a training sample set. Use the adjusted initial connector position recognition model as the initial connector position recognition model, and perform the above processing steps again.
[0016] In some embodiments, the connecting piece height detection model is trained through the following steps: obtaining a third training sample set, wherein the training samples in the third training sample set include connecting piece sample images and sample protruding insulation board height information; based on the above-mentioned third training sample set, performing the following processing steps: respectively inputting the connecting piece sample images included in at least one training sample in the third training sample set into the initial connecting piece height detection model to obtain the protruding insulation board height information corresponding to each training sample in at least one training sample; comparing the protruding insulation board height corresponding to each training sample in the at least one training sample with the corresponding sample protruding insulation board height; determining whether the initial connecting piece height detection model reaches a preset optimization goal according to the comparison result; in response to determining that the initial connecting piece height detection model reaches the above optimization goal, using the initial connecting piece height detection model as the trained connecting piece height detection model; in response to determining that the initial connecting piece height detection model does not reach the above optimization goal, adjusting the model parameters of the initial connecting piece height detection model, and using the unused training samples to form a training sample set, using the adjusted initial connecting piece height detection model as the initial connecting piece height detection model, and performing the above processing steps again.
[0017] In a second aspect, some embodiments of the present disclosure provide a precast wall insulation layer and connecting piece detection device, including: an image capturing module, composed of a photographing device and a photographing controller, for capturing images of the precast wall insulation layer and connecting pieces; an image point cloud operation module, configured to perform point cloud budgeting and analysis on the overall photograph to generate a precast wall insulation layer and connecting piece point cloud data set; a point cloud recognition module, configured to recognize the precast wall insulation layer and connecting piece point cloud data set based on a pre-trained precast wall insulation layer and connecting piece recognition model to generate a precast wall insulation layer and connecting piece recognition information set, wherein the precast wall insulation layer and connecting piece image recognition information set includes: insulation board joint width, connecting piece position, and connecting piece height; a result warning module, configured to compare the precast wall insulation layer and connecting piece recognition information set with product requirements, and give a warning and remind the operator to adjust according to the comparison result to meet the product requirements.
[0018] Specifically, the above-mentioned photographing device is a four-eye three-dimensional matrix camera.
[0019] In a third aspect, some embodiments of the present disclosure provide a computer-readable medium, on which a computer program is stored, wherein the computer program, when executed by a processor, implements the precast wall insulation layer and connecting piece detection method described in any implementation manner of the first aspect above.
[0020] The above-mentioned various embodiments of the present disclosure have the following beneficial effects:
[0021] First, through the precast wall insulation layer and connector detection method of some embodiments of the present disclosure, the detection efficiency and detection accuracy can be improved. Specifically, the reasons for the low detection efficiency and low detection accuracy are as follows: Manual detection is greatly affected by human subjective factors and has low detection efficiency. There are a large number of connectors, and it is impossible to cover and measure the joints between the connectors and the insulation board within the production line cycle time. The precast wall insulation layer and connector detection method of some embodiments of the present disclosure can achieve rapid coverage automatic detection of the joints between the connectors and the insulation board through visual detection.
[0022] Second, the precast wall insulation layer and connector detection method of some embodiments of the present disclosure has a fast detection speed, and the detection time can be controlled within three minutes, which is convenient for operators to adjust the connectors and the joints of the insulation board that do not meet the product requirements in a timely manner according to the detection results, thereby improving the product qualification rate.
[0023] Third, the precast wall insulation layer and connector detection method and equipment of some embodiments of the present disclosure can directly store the detection data in an electronic data format without loss, avoiding the problems of difficult to call and easy to lose data in the traditional manual recording method, and can realize real-time call of the detection data in the supervision link. BRIEF DESCRIPTION OF THE DRAWINGS
[0024] In combination with the accompanying drawings and with reference to the following specific embodiments, the above and other features, advantages and aspects of the various embodiments of the present disclosure will become more obvious. Throughout the drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic, and the elements and elements are not necessarily drawn to scale.
[0025] Figure 1 is a flowchart of some embodiments of the precast wall insulation layer and connector detection method according to the present disclosure;
[0026] Figure 2 is the overall point cloud data of the precast wall insulation layer and connectors according to some embodiments of the precast wall insulation layer and connector detection method of the present disclosure.
[0027] Figure 3 is a schematic structural diagram of some embodiments of the precast wall insulation layer and connector detection equipment according to the present disclosure. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0028] The embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although some embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. On the contrary, these embodiments are provided to more thoroughly and completely understand the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not used to limit the protection scope of the present disclosure.
[0029] In addition, it should be noted that for the convenience of description, only the parts related to the relevant invention are shown in the drawings. Without conflict, the embodiments in the present disclosure and the features in the embodiments can be combined with each other.
[0030] It should be noted that the concepts such as "first" and "second" mentioned in the present disclosure are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.
[0031] The present disclosure will be described in detail below with reference to the drawings and in combination with embodiments.
[0032] Figure 1 Flow 100 of some embodiments of a method for detecting a precast wall insulation layer and connectors according to the present disclosure is shown. The method for detecting the precast wall insulation layer and connectors includes the following steps:
[0033] Step 101, control a photographing device to take an overall photograph of the precast wall insulation layer and connectors.
[0034] In some embodiments, the execution subject of the method for detecting the precast wall insulation layer and connectors can control the photographing device through wireless or wired signal transmission to take an overall photograph of the precast wall insulation layer and connectors. Among them, the photographing device is deployed above the precast wall production line or on a drone platform, and can cover the entire range of the insulation layer and connectors of the precast wall through top-down photographing.
[0035] Step 102, perform preprocessing calculations on the overall photograph, and perform three-dimensional point cloud operations based on the processed overall photograph to obtain the overall point cloud data of the precast wall insulation layer and connectors.
[0036] In some embodiments, the above-mentioned execution subject can perform preprocessing calculations on the overall photograph, and perform three-dimensional point cloud operations based on the processed overall photograph to obtain the overall point cloud data of the precast wall insulation layer and connectors. Among them, the above-mentioned execution subject can automatically process the above overall photograph through various processing methods in three links of grayscale conversion, image restoration and denoising in sequence, and the three-dimensional point cloud operation can be completed by a three-dimensional point cloud automatic calculation program installed in an FPGA three-dimensional operation box.
[0037] Step 103, analyze the overall point cloud data to obtain a point cloud data set of the precast wall insulation layer and connectors.
[0038] In some embodiments, the above-mentioned execution subject can analyze the overall point cloud data of the precast wall insulation layer and connectors to obtain a point cloud data set of the precast wall insulation layer and connectors.
[0039] In some alternative implementation manners of some embodiments, the above-mentioned execution subject analyzes the overall point cloud data of the precast wall insulation layer and the connecting piece to obtain the point cloud data set of the precast wall insulation layer and the connecting piece, which may include the following steps:
[0040] First step, perform data preprocessing on the above-mentioned overall point cloud data to generate processed overall point cloud data. Among them, during the data preprocessing, the point cloud data outside the precast wall insulation layer and the connecting piece is removed to obtain the processed overall point cloud data.
[0041] Second step, perform filtering processing on the above-mentioned processed overall point cloud data to generate filtered overall point cloud data. Among them, the filtering processing can be carried out by means of VoxelGrid filtering to remove the point cloud data of irrelevant items such as those affected by ambient light, steel bars, molds, personnel, tools, etc. in the point cloud data to obtain the filtered overall point cloud data.
[0042] Third step, perform structured feature extraction on the above-mentioned filtered overall point cloud data to generate a structured point cloud coordinate sequence set. Among them, through the point cloud structuring algorithm, perform structured feature extraction on the above-mentioned filtered overall point cloud data, and extract the point cloud coordinates representing the edges of the insulation board and the connecting piece as the structured point cloud coordinate sequence.
[0043] As an example, the above-mentioned point cloud structuring algorithm may include but is not limited to at least one of the following: ISS3D algorithm (keypoint extraction algorithm based on intrinsic shape features), NARF algorithm (radial feature extraction algorithm based on normal alignment), Harris3D algorithm (geometric feature extraction algorithm based on local surface), etc.
[0044] Fourth step, according to the connecting lines of each structured point cloud coordinate in the above-mentioned structured point cloud coordinate sequence set, perform point cloud data grouping processing on the above-mentioned overall point cloud data to generate the point cloud data set of the precast wall insulation layer and the connecting piece. Among them, the point cloud data can be divided according to the connecting lines of the point cloud coordinates of the edges of the insulation board and the connecting piece to obtain the point cloud data set of the precast wall insulation layer and the connecting piece.
[0045] As an example, the overall point cloud data of the precast wall insulation layer and the connecting piece can be as Figure 2 shown.
[0046] Step 104, input the point cloud data set of the precast wall insulation layer and the connecting piece into the pre-trained precast wall insulation layer and connecting piece recognition model to obtain the precast wall insulation layer and connecting piece recognition information set. Among them, the pre-trained precast wall insulation layer and connecting piece recognition model includes: insulation board joint width detection model, connecting piece position recognition model, and connecting piece height detection model, and the precast wall insulation layer and connecting piece image recognition information set includes: insulation board joint width, connecting piece position, and connecting piece height.
[0047] In some embodiments, the above-mentioned execution entity may input the point cloud data sets of the prefabricated wall insulation layer and the connectors into a pre-trained prefabricated wall insulation layer and connector recognition model to obtain a prefabricated wall insulation layer and connector recognition information set. Among them, the pre-trained prefabricated wall insulation layer and connector recognition model includes: an insulation board joint width detection model, a connector position recognition model, and a connector height detection model. The prefabricated wall insulation layer and connector image recognition information set includes: the insulation board joint width, the connector position, and the connector height. Among them, the insulation board joint width detection model may be a neural network model that takes an insulation board image as input and outputs information representing the insulation board joint width. For example, the insulation board joint width detection model may be a trained convolutional neural network model; the connector position recognition model may be a neural network model that takes a connector image as input and outputs information representing the connector position coordinates. For example, the connector position recognition model may be a trained convolutional neural network model; the connector height detection model may be a neural network model that takes a connector image as input and outputs information representing the height of the connector protruding from the insulation board. For example, the connector height detection model may be a trained convolutional neural network model.
[0048] In practice, the above-mentioned insulation board joint width detection model is trained through the following steps:
[0049] The first step is to obtain a first training sample set. Among them, the training samples in the first training sample set include: insulation board sample images and sample joint width information.
[0050] In some embodiments, the above-mentioned execution entity may obtain the training sample set from the terminal device through a wireless transmission method. Among them, the training samples in the training sample set include: insulation board sample images and sample joint width information. The sample joint width information may represent the joint width between the above-mentioned insulation boards. If the joint width is too large, it will affect the thermal performance of the insulation layer. For example, the sample joint width information may be 2, 4, or other numbers, indicating that the joint width of the insulation board sample is 2 mm, 4 mm, or other sizes.
[0051] The second step is to perform the following processing steps based on the above-mentioned training sample set:
[0052] The first sub-step is to input the insulation board sample images included in at least one training sample in the first training sample set into the initial insulation board joint width detection model respectively, and obtain the joint width information corresponding to each training sample in at least one training sample.
[0053] In some embodiments, the above-mentioned execution entity may respectively input the insulation board sample images included in at least one training sample in the above-mentioned first training sample set into the initial insulation board joint width detection model to obtain the joint width information corresponding to each training sample in the at least one training sample. Here, the initial insulation board joint width detection model may be YOLOv5, VGG19 or other image recognition models.
[0054] The second sub-step is to compare the joint width corresponding to each training sample in the above-mentioned at least one training sample with the corresponding sample joint width.
[0055] In some embodiments, the above-mentioned execution entity may compare the joint width corresponding to each training sample in the above-mentioned at least one training sample with the corresponding sample joint width. For example, subtract the joint width corresponding to each training sample from the corresponding sample joint width, and then take the absolute value of the difference.
[0056] The third sub-step is to determine whether the above-mentioned initial insulation board joint width detection model reaches the preset optimization target according to the comparison result.
[0057] In some embodiments, the above-mentioned execution entity may determine whether the above-mentioned initial insulation board joint width detection model reaches the preset optimization target according to the comparison result. The above-mentioned comparison result refers to the comprehensive comparison result obtained by comparing the joint width corresponding to each training sample in at least one training sample with the corresponding sample joint width. For example, subtract the joint width corresponding to each training sample from the corresponding sample joint width, and then take the absolute value of the difference as the comparison result of the joint width corresponding to each training sample and the corresponding sample joint width, and then take the average value of all comparison results as the comprehensive comparison result. The above-mentioned preset optimization target means that the comparison result is less than the preset threshold. The preset threshold is a preset fixed value. For example, the preset threshold may be 0.5, 0.8 or other fixed values. In practice, when the comparison result is less than the preset threshold, it indicates that the above-mentioned initial insulation board joint width detection model reaches the preset optimization target, and when the comparison result is greater than or equal to the preset threshold, it indicates that the above-mentioned initial insulation board joint width detection model does not reach the preset optimization target.
[0058] The fourth sub-step is to, in response to determining that the initial insulation board joint width detection model reaches the above-mentioned optimization target, use the initial insulation board joint width detection model as the insulation board joint width detection model that has been trained; in response to determining that the initial insulation board joint width detection model does not reach the above-mentioned optimization target, adjust the model parameters of the initial insulation board joint width detection model, and use the training samples that have not been used to form a training sample set, use the adjusted initial insulation board joint width detection model as the initial insulation board joint width detection model, and execute the above-mentioned processing steps again.
[0059] In some embodiments, the above-mentioned execution entity may, in response to determining that the initial insulation board joint width detection model reaches the above-mentioned optimization goal, use the initial insulation board joint width detection model as the trained insulation board joint width detection model. In practice, when the comparison result in the third sub-step is less than a preset threshold, it indicates that the initial insulation board joint width detection model reaches the preset optimization goal, and thus it can be considered that the initial insulation board joint width detection model is trained.
[0060] In some embodiments, the above-mentioned execution entity may, in response to determining that the initial insulation board joint width detection model does not reach the above-mentioned optimization goal, adjust the model parameters of the initial insulation board joint width detection model, and use unused training samples to form a training sample set. Then, use the adjusted initial insulation board joint width detection model as the initial insulation board joint width detection model and execute the above-mentioned processing steps again. In practice, when the comparison result in the third sub-step is greater than or equal to the preset threshold, it indicates that the initial insulation board joint width detection model does not reach the preset optimization goal, and it is considered that the initial insulation board joint width detection model is not trained. At this time, the backpropagation algorithm or the gradient descent method can be used to adjust the model parameters of the above-mentioned initial insulation board joint width detection model. Then, use the initial insulation board joint width detection model with adjusted model parameters as the initial insulation board joint width detection model, and use unused training samples to form a training sample set, and execute the above-mentioned processing steps again until the initial insulation board joint width detection model reaches the above-mentioned optimization goal.
[0061] In practice, the above-mentioned connector position recognition model is trained through the following steps:
[0062] First step, obtain a second training sample set, where the training samples in the second training sample set include: connector sample images and sample connector position information.
[0063] In some embodiments, the above-mentioned execution entity may obtain the training sample set from the terminal device through a wireless transmission method. Among them, the training samples in the training sample set include: connector sample images and sample connector position information. The sample connector position information can represent the position coordinates of the connector. For example, the sample connector position information can be (100, 100), (500, 500) or other numbers, indicating that the connector sample coordinates are (100, 100), (500, 500) or other coordinates, and the unit is millimeters.
[0064] Second step, based on the above-mentioned second training sample set, execute the following processing steps:
[0065] The first sub-step is to input the connector sample images included in at least one training sample in the above-mentioned second training sample set into the initial connector position recognition model respectively, so as to obtain the connector position information corresponding to each training sample in at least one training sample.
[0066] In some embodiments, the above-mentioned execution entity may input the connector sample images included in at least one training sample in the above-mentioned second training sample set into the initial connector position recognition model respectively, so as to obtain the connector position information corresponding to each training sample in at least one training sample. Here, the initial connector position recognition model may be YOLOv5, VGG19 or other image recognition models.
[0067] The second sub-step is to compare the connector position information corresponding to each training sample in the above-mentioned at least one training sample with the corresponding sample connector position information.
[0068] In some embodiments, the above-mentioned execution entity may compare the connector position information corresponding to each training sample in the above-mentioned at least one training sample with the corresponding sample connector position information. For example, compare the connector position information corresponding to each training sample with the corresponding sample connector position information, and then calculate the deviation distance.
[0069] The third sub-step is to determine whether the above-mentioned initial connector position recognition model reaches the preset optimization goal according to the comparison result.
[0070] In some embodiments, the above-mentioned execution entity may determine whether the above-mentioned initial connector position recognition model reaches the preset optimization goal according to the comparison result. The above-mentioned comparison result refers to the comprehensive comparison result obtained by comparing the connector position information corresponding to each training sample in at least one training sample with the corresponding sample connector position information. For example, compare the connector position information corresponding to each training sample with the corresponding sample connector position information, calculate the deviation distance as the comparison result of the connector position information corresponding to each training sample and the corresponding sample connector position information, and then take the average value of all comparison results as the comprehensive comparison result. The above-mentioned preset optimization goal means that the comparison result is less than the preset threshold. The preset threshold is a preset fixed value. For example, the preset threshold may be 10, 20 or other fixed values. In practice, when the comparison result is less than the preset threshold, it indicates that the above-mentioned initial connector position recognition model reaches the preset optimization goal, and when the comparison result is greater than or equal to the preset threshold, it indicates that the above-mentioned initial connector position recognition model does not reach the preset optimization goal.
[0071] Fourth sub-step: in response to determining that the initial connector position recognition model reaches the optimization goal, use the initial connector position recognition model as the trained connector position recognition model; in response to determining that the initial connector position recognition model does not reach the optimization goal, adjust the model parameters of the initial connector position recognition model, and use the unused training samples to form a training sample set, use the adjusted initial connector position recognition model as the initial connector position recognition model, and execute the above processing steps again.
[0072] In some embodiments, the above-mentioned execution entity may, in response to determining that the initial connector position recognition model reaches the above optimization goal, use the initial connector position recognition model as the trained connector position recognition model. In practice, when the comparison result in the third sub-step is less than a preset threshold, it indicates that the initial connector position recognition model reaches the preset optimization goal, so it can be considered that the initial connector position recognition model is trained.
[0073] In some embodiments, the above-mentioned execution entity may, in response to determining that the initial connector position recognition model does not reach the above optimization goal, adjust the model parameters of the initial connector position recognition model, and use the unused training samples to form a training sample set, use the adjusted initial connector position recognition model as the initial connector position recognition model, and execute the above processing steps again. In practice, when the comparison result in the third sub-step is greater than or equal to the preset threshold, it indicates that the initial connector position recognition model does not reach the preset optimization goal, and it is considered that the initial connector position recognition model is not trained. At this time, the backpropagation algorithm or the gradient descent method can be used to adjust the model parameters of the above initial connector position recognition model. Then, use the initial connector position recognition model with adjusted model parameters as the initial connector position recognition model, and use the unused training samples to form a training sample set, and execute the above processing steps again until the initial connector position recognition model reaches the above optimization goal.
[0074] In practice, the above-mentioned connector height detection model is trained through the following steps:
[0075] First step: obtain a third training sample set, where the training samples in the above third training sample set include: connector sample images and sample protruding insulation board height information.
[0076] In some embodiments, the above-mentioned execution entity can obtain a training sample set from a terminal device through a wireless transmission method. Among them, the training samples in the above-mentioned training sample set include: a connecting piece sample image and sample protruding insulation board height information. The sample protruding insulation board height information can represent the height information of the connecting piece protruding from the insulation board. For example, the sample protruding insulation board height information can be 40, 45 or other numbers, indicating that the connecting piece sample protrudes from the insulation board by 40 mm, 45 mm or other heights.
[0077] Second step, based on the above-mentioned third training sample set, perform the following processing steps:
[0078] The first sub-step is to input the connecting piece sample images included in at least one training sample in the above-mentioned third training sample set into the initial connecting piece height detection model respectively, and obtain the protruding insulation board height information corresponding to each training sample in the at least one training sample.
[0079] In some embodiments, the above-mentioned execution entity can input the connecting piece sample images included in at least one training sample in the above-mentioned third training sample set into the initial connecting piece height detection model respectively, and obtain the protruding insulation board height information corresponding to each training sample in the at least one training sample. Here, the initial connecting piece height detection model can be YOLOv5, VGG19 or other image recognition models.
[0080] The second sub-step is to compare the protruding insulation board height corresponding to each training sample in the above-mentioned at least one training sample with the corresponding sample protruding insulation board height.
[0081] In some embodiments, the above-mentioned execution entity can compare the protruding insulation board height corresponding to each training sample in the above-mentioned at least one training sample with the corresponding sample protruding insulation board height. For example, subtract the protruding insulation board height information corresponding to each training sample from the corresponding sample protruding insulation board height information.
[0082] The third sub-step is to determine whether the initial connecting piece height detection model meets the preset optimization goal according to the comparison result.
[0083] In some embodiments, the above-mentioned execution entity may determine whether the initial connecting piece height detection model reaches a preset optimization target according to the comparison result. The above-mentioned comparison result refers to the comprehensive comparison result obtained by comparing the protruding insulation board height information corresponding to each training sample in at least one training sample with the corresponding sample protruding insulation board height information. For example, the difference between the protruding insulation board height information corresponding to each training sample and the corresponding sample protruding insulation board height information is calculated, and then the average value of all comparison results is taken as the comprehensive comparison result. The above-mentioned preset optimization target means that the comparison result is less than a preset threshold value. The preset threshold value is a fixed value set in advance. For example, the preset threshold value may be 1, 2 or other fixed values. In practice, when the comparison result is less than the preset threshold value, it indicates that the above-mentioned initial connecting piece height detection model reaches the preset optimization target; when the comparison result is greater than or equal to the preset threshold value, it indicates that the above-mentioned initial connecting piece height detection model does not reach the preset optimization target.
[0084] The fourth sub-step: in response to determining that the initial connecting piece height detection model reaches the optimization target, use the initial connecting piece height detection model as the trained connecting piece height detection model; in response to determining that the initial connecting piece height detection model does not reach the optimization target, adjust the model parameters of the initial connecting piece height detection model, and use the unused training samples to form a training sample set, use the adjusted initial connecting piece height detection model as the initial connecting piece height detection model, and execute the above-mentioned processing steps again.
[0085] In some embodiments, the above-mentioned execution entity may, in response to determining that the initial connecting piece height detection model reaches the above-mentioned optimization target, use the initial connecting piece height detection model as the trained connecting piece height detection model. In practice, when the comparison result in the third sub-step is less than the preset threshold value, it indicates that the initial connecting piece height detection model reaches the preset optimization target, and thus it can be considered that the training of the initial connecting piece height detection model is completed.
[0086] In some embodiments, the above-mentioned execution entity may, in response to determining that the initial connector height detection model does not meet the above-mentioned optimization goal, adjust the model parameters of the initial connector height detection model, and use unused training samples to form a training sample set. Then, take the adjusted initial connector height detection model as the initial connector height detection model and execute the above-mentioned processing steps again. In practice, when the comparison result in the third sub-step is greater than or equal to a preset threshold, it indicates that the initial connector height detection model does not meet the preset optimization goal, and it is considered that the initial connector height detection model has not been trained successfully. At this time, the backpropagation algorithm or the gradient descent method can be used to adjust the model parameters of the above-mentioned initial connector height detection model. Then, take the initial connector height detection model with adjusted model parameters as the initial connector height detection model, and use unused training samples to form a training sample set, and execute the above-mentioned processing steps again until the initial connector height detection model meets the above-mentioned optimization goal.
[0087] Step 105: Compare the above-mentioned precast wall insulation layer and connector identification information set with the product requirements to obtain the precast wall insulation layer and connector detection result.
[0088] In some embodiments, the above-mentioned execution entity may compare the precast wall insulation layer and connector identification information set with the product requirements to obtain the precast wall insulation layer and connector detection result. Among them, the product requirements include the requirements for the joint width of the insulation board, the position deviation of the connector, and the height of the connector protruding from the insulation board.
[0089] In some optional implementation manners of some embodiments, the above-mentioned execution entity's comparison of the precast wall insulation layer and connector identification information set with the product requirements may include the following steps:
[0090] The first step: Extract the requirements for the joint width of the insulation board, the position of the connector, and the height of the connector from the precast wall design document to generate a product requirement data set. Among them, the precast wall design document is a two-dimensional drawing in dwg format or a three-dimensional BIM model, and a design file automatic reading program is used to read product requirements such as the joint width of the insulation board, the position of the connector, and the height of the connector.
[0091] The second step: Align the precast wall insulation layer and connector identification information set with the product requirement data set, and compare the joint width of the insulation board, the position of the connector, and the height of the connector. Among them, the center position of the precast wall can be used as the alignment reference, and then all the joint widths of the insulation boards, the positions of the connectors, and the heights of the connectors are compared.
[0092] Step 106: In response to the qualification of the precast wall insulation layer and connector detection result, continue to the next production process. If the precast wall insulation layer and connector detection result is unqualified, give an early warning and remind the operator to make adjustments.
[0093] In some embodiments, if the detection results of the precast wall insulation layer and the connectors are qualified, the above-mentioned execution entity proceeds to the next production process. If the detection results of the precast wall insulation layer and the connectors are unqualified, a warning is issued to remind the operator to make adjustments. Among them, if the detection results of the precast wall insulation layer and the connectors are unqualified, the above-mentioned execution entity can mark the positions that do not meet the product requirements on the overall photo of the precast wall insulation layer and the connectors, and give adjustment suggestions, which are automatically sent to the construction display terminal of the operator. The operator makes timely adjustments to the insulation layer and the connectors according to the image guidance to avoid being unable to make adjustments after the concrete initial setting.
[0094] The above relevant content, as an inventive point of the present disclosure, solves technical problems one to three mentioned in the background art. The reasons for the above technical problems are as follows: Manual inspection is greatly affected by human subjective factors and has low inspection efficiency. The number of connectors of precast walls is generally not less than 20. Manual inspection mainly relies on measurement with a ruler. The time required for effectively inspecting all insulation board joints and connectors is generally not less than 20 minutes, which does not match the production rhythm and exceeds the initial setting time of the concrete, making it impossible to adjust unqualified insulation boards or connectors. The results of manual inspection are archived in paper form, with high digital storage costs and difficult to be called in real time. Through visual inspection, the present disclosure can achieve rapid coverage automatic inspection of connectors and insulation board joints, with a fast inspection speed. The inspection time can be controlled within three minutes, facilitating the operator to make timely adjustments to the connectors and insulation board joints that do not meet the product requirements before the concrete initial setting, improving the product qualification rate. At the same time, the inspection data is stored losslessly in electronic data format, enabling real-time call of the inspection data in the supervision link.
[0095] Further referring to Figure 3 , as an implementation of the methods shown in the above figures, the present disclosure provides some embodiments of a precast wall insulation layer and connector detection device. These device embodiments correspond to those method embodiments shown in Figure 1 , and the device can be specifically applied to various electronic devices.
[0096] Such as Figure 3As shown in the figure, the precast wall insulation layer and connector detection device 300 of some embodiments includes: an image capture module 301, an image point cloud operation module 302, a point cloud recognition module 303, and a result warning module 304. Among them, the image capture module 301, which consists of a photographing device and a shooting controller, captures images of the precast wall insulation layer and connectors. As an example, the photographing device uses a four-eye three-dimensional matrix camera, which avoids the matching ambiguity defect of binocular vision technology; the image point cloud operation module 302 is configured to perform point cloud calculation and analysis on the overall photo to generate a point cloud data set of the precast wall insulation layer and connectors. As an example, the image point cloud operation module 302 can use an FPGA three-dimensional operation box to integrate the point cloud operation program into the FPGA three-dimensional operation box to achieve real-time and fast operation of the point cloud; the point cloud recognition module 303 is configured to identify the point cloud data set of the precast wall insulation layer and connectors based on a pre-trained precast wall insulation layer and connector recognition model, and generate a precast wall insulation layer and connector recognition information set. Among them, the precast wall insulation layer and connector image recognition information set includes: the width of the insulation board joint, the position of the connector, and the height of the connector; the result warning module 304 is configured to compare the precast wall insulation layer and connector recognition information set with the product requirements, and warn and remind the operator to adjust according to the comparison result to meet the product requirements. As an example, the result warning module 304 is configured with a warning information display terminal for the operator to view in real time.
[0097] It can be understood that the various units described in the device 300 correspond to the respective steps in the method described with reference to Figure 1 Therefore, the operations, features, and beneficial effects described above for the method also apply to the device 300 and the units included therein, and will not be repeated here.
[0098] In particular, according to some embodiments of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer-readable medium. For example, some embodiments of the present disclosure include a computer-readable medium that includes a computer program carried on the computer-readable medium, and the computer program includes program codes for executing the method shown in the flowchart and implementing the above functions defined in the method of some embodiments of the present disclosure.
[0099] In summary, in the present disclosure, through visual detection, rapid and comprehensive automatic detection of precast wall connectors and insulation board joints is achieved. The detection speed is fast, and the detection time can be controlled within three minutes, which is convenient for the operator to adjust the connectors and insulation board joints that do not meet the product requirements in time before the concrete initial setting according to the detection results, improving the product qualification rate. At the same time, the detection data is stored losslessly in electronic data format, and real-time call of the detection data in the supervision link can be realized.
[0100] The above description is only some preferred embodiments of the present disclosure and an explanation of the applied technical principles. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present disclosure is not limited to the technical solutions formed by the specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above inventive concept. For example, the technical solutions formed by mutually replacing the above features with the technical features (but not limited to) disclosed in the embodiments of the present disclosure that have similar functions.
Claims
1. Detection method for precast wall insulation layer and connectors, comprising: Controlling a photographing device to capture an overall photograph of the precast wall insulation layer and connectors; Performing preprocessing calculations on the overall photograph, and performing 3D point cloud operations based on the processed overall photograph to obtain the overall point cloud data of the precast wall insulation layer and connectors; Analyzing the overall point cloud data to obtain a point cloud data set of the precast wall insulation layer and connectors; Inputting the point cloud data set of the precast wall insulation layer and connectors into a pre-trained recognition model for the precast wall insulation layer and connectors to obtain a recognition information set for the precast wall insulation layer and connectors. Among them, the pre-trained recognition model for the precast wall insulation layer and connectors includes: an insulation board joint width detection model, a connector position recognition model, and a connector height detection model. The image recognition information set for the precast wall insulation layer and connectors includes: insulation board joint width, connector position, and connector height; Comparing the recognition information set for the precast wall insulation layer and connectors with the product requirements to obtain the detection result for the precast wall insulation layer and connectors; In response to the qualified detection result for the precast wall insulation layer and connectors, proceeding to the next production process. If the detection result for the precast wall insulation layer and connectors is unqualified, giving an alarm and reminding the operator to make adjustments.
2. The detection method of the precast wall insulation layer and the connector according to claim 1, wherein, The analyzing the overall point cloud data to obtain a point cloud data set of the precast wall insulation layer and connectors includes: Performing data preprocessing on the overall point cloud data to generate processed overall point cloud data; Performing filtering processing on the processed overall point cloud data to generate filtered overall point cloud data; Performing structured feature extraction on the filtered overall point cloud data to generate a structured point cloud coordinate sequence set; According to the connection lines of each structured point cloud coordinate in the structured point cloud coordinate sequence set, performing point cloud data grouping processing on the overall point cloud data to generate a point cloud data set of the precast wall insulation layer and connectors.
3. The precast wall insulation layer and connector detection method according to claim 1, wherein, The comparing the recognition information set for the precast wall insulation layer and connectors with the product requirements includes: Extracting the requirements for the insulation board joint width, connector position, and connector height through the precast wall design document to generate a product requirement data set; Aligning the recognition information set for the precast wall insulation layer and connectors with the product requirement data set, and comparing the insulation board joint width, connector position, and connector height.
4. The precast wall insulation layer and connector detection method according to claim 3, wherein, The precast wall design document is a 2D drawing in dwg format or a 3D BIM model.
5. The inspection method of the precast wall insulation layer and the connector according to claim 1, wherein, The insulation board joint width detection model is trained through the following steps: Obtaining a first training sample set, where the training samples in the first training sample set include: insulation board sample images and sample joint width information; Based on the first training sample set, perform the following processing steps: respectively input the insulation board sample images included in at least one training sample in the first training sample set into the initial insulation board joint width detection model to obtain the joint width information corresponding to each training sample in at least one training sample; compare the joint width corresponding to each training sample in the at least one training sample with the corresponding sample joint width; determine whether the initial insulation board joint width detection model reaches a preset optimization target according to the comparison result; in response to determining that the initial insulation board joint width detection model reaches the optimization target, use the initial insulation board joint width detection model as the trained insulation board joint width detection model; in response to determining that the initial insulation board joint width detection model does not reach the optimization target, adjust the model parameters of the initial insulation board joint width detection model, and use the unused training samples to form a training sample set, use the adjusted initial insulation board joint width detection model as the initial insulation board joint width detection model, and perform the above processing steps again.
6. The precast wall insulation layer and connector detection method according to claim 1, wherein, The connector position recognition model is trained through the following steps: Obtain a second training sample set, where the training samples in the second training sample set include: connector sample images and sample connector position information; Based on the second training sample set, perform the following processing steps: respectively input the connector sample images included in at least one training sample in the second training sample set into the initial connector position recognition model to obtain the connector position information corresponding to each training sample in at least one training sample; compare the connector position information corresponding to each training sample in the at least one training sample with the corresponding sample connector position information; determine whether the initial connector position recognition model reaches a preset optimization target according to the comparison result; in response to determining that the initial connector position recognition model reaches the optimization target, use the initial connector position recognition model as the trained connector position recognition model; in response to determining that the initial connector position recognition model does not reach the optimization target, adjust the model parameters of the initial connector position recognition model, and use the unused training samples to form a training sample set, use the adjusted initial connector position recognition model as the initial connector position recognition model, and perform the above processing steps again.
7. The detection method of the precast wall insulation layer and the connecting piece according to claim 1, wherein, The connector height detection model is trained through the following steps: Obtain a third training sample set, where the training samples in the third training sample set include: connector sample images and sample protruding insulation board height information; Based on the third training sample set, the following processing steps are performed: respectively input the connector sample images included in at least one training sample in the third training sample set into the initial connector height detection model to obtain the protruding insulation board height information corresponding to each training sample in the at least one training sample; compare the protruding insulation board height corresponding to each training sample in the at least one training sample with the corresponding sample protruding insulation board height; determine whether the initial connector height detection model reaches a preset optimization goal according to the comparison result; in response to determining that the initial connector height detection model reaches the optimization goal, use the initial connector height detection model as the trained connector height detection model; in response to determining that the initial connector height detection model does not reach the optimization goal, adjust the model parameters of the initial connector height detection model, and use the unused training samples to form a training sample set, use the adjusted initial connector height detection model as the initial connector height detection model, and perform the processing steps again.
8. A precast wall insulation layer and connector detection device, comprising: An image capturing module, composed of a photographing device and a photographing controller, for capturing images of the precast wall insulation layer and connectors; An image point cloud operation module, configured to perform point cloud budgeting and analysis on the overall photograph of the precast wall insulation layer and connectors to generate a precast wall insulation layer and connector point cloud data set; A point cloud recognition module, configured to recognize the precast wall insulation layer and connector point cloud data set based on a pre-trained precast wall insulation layer and connector recognition model to generate a precast wall insulation layer and connector recognition information set, wherein the precast wall insulation layer and connector image recognition information set includes: insulation board joint width, connector position, and connector height; A result warning module, configured to compare the precast wall insulation layer and connector recognition information set with product requirements, and give a warning and remind the operator to adjust according to the comparison result to meet the product requirements.
9. The precast wall insulation layer and connector detection device according to claim 8, wherein, The photographing device is a four-eye three-dimensional matrix camera.
10. A computer-readable medium having a computer program stored thereon, wherein, When the computer program is executed by a processor, it implements the precast wall insulation layer and connector detection method according to any one of claims 1-7.