Steel grating welding seam automatic detection method and device based on point cloud

Abstract

The invention provides a steel grating welding seam automatic detection method and device based on point cloud, and relates to the field of data processing. The method comprises the following steps: acquiring three-dimensional point cloud data for a target steel grating; converting the three-dimensional point cloud data by adopting an implicit surface modeling mode to obtain a steel grating physical model; acquiring temperature data for the target steel grating; fusing the temperature data and the steel grating physical model to obtain a fusion result; inputting the fusion result into a neural network to obtain a target detection model; the target detection model is adopted to predict the temperature of the target steel grating, a normal area and / or an abnormal area are / is obtained, the normal area indicates that the weld joint of the target steel grating is qualified, and the abnormal area indicates that the weld joint of the target steel grating is unqualified. By implementing the technical scheme provided by the invention, the accuracy of steel grating welding seam detection can be conveniently improved.

Description

Technical Field

[0001] This application relates to the technical field of data processing, and particularly relates to an automatic detection method and device for steel grating welds based on point cloud. Background Art

[0002] The detection of steel grating welds is an important link in the quality control of steel structures, directly affecting the safety and reliability of the project.

[0003] At present, the detection of steel grating welds mainly relies on manual detection and visual assessment, and there are obvious deficiencies in this method. Since the weld positions are usually relatively hidden and affected by multiple factors such as environmental light and the experience of operators, it is difficult for manual detection to comprehensively and accurately capture the true shape and subtle changes of the welds, resulting in a relatively low detection accuracy in subsequent visual assessment.

[0004] Therefore, there is an urgent need for an automatic detection method and device for steel grating welds based on point cloud. Summary of the Invention

[0005] This application provides an automatic detection method and device for steel grating welds based on point cloud, which is convenient for improving the accuracy of detecting the steel grating welds.

[0006] In the first aspect of this application, an automatic detection method for steel grating welds based on point cloud is provided. The method includes: obtaining three-dimensional point cloud data for a target steel grating; converting the three-dimensional point cloud data by using an implicit surface modeling method to obtain a physical model of the steel grating; obtaining temperature data for the target steel grating; fusing the temperature data and the physical model of the steel grating to obtain a fusion result; inputting the fusion result into a neural network to obtain a target detection model; using the target detection model to predict the temperature of the target steel grating to obtain a normal area and / or an abnormal area, where the normal area indicates that the welds of the target steel grating are qualified, and the abnormal area indicates that the welds of the target steel grating are unqualified.

[0007] By adopting the above technical solution, an implicit physical model of the steel grating is constructed by obtaining three-dimensional point cloud data, and then the temperature data is fused with the physical model, making full use of geometric and heat conduction information. The fusion of such multi-source data can more comprehensively reflect the actual state of the steel plate weld. The temperature data can reveal the discontinuity of material properties and abnormal heat diffusion at the weld. After being combined with the physical model and predicted through a neural network, the normal area and the abnormal area can be more accurately distinguished, thus effectively identifying whether the weld is qualified. Inputting the fusion result into the neural network to train the object detection model to achieve automated detection reduces the subjectivity and inconsistency of manual detection, and improves the detection efficiency and reliability. This method uses a data-driven neural network model, which can achieve real-time temperature prediction and weld state determination, is applicable to large-scale industrial applications and online monitoring, and is convenient for improving the accuracy of the weld detection of the steel grating.

[0008] Optionally, the obtaining of the three-dimensional point cloud data for the target steel grating specifically includes: receiving the original scan data for the target steel grating sent by a scanning device, where the scanning device includes a laser scanner, a 3D lidar, and a structured light camera; denoising and filtering the original scan data to obtain target scan data; performing voxel downsampling and multi-view registration on the target scan data to generate the three-dimensional point cloud data.

[0009] By adopting the above technical solution, it supports various scanning devices such as laser scanners, 3D lidars, and structured light cameras, can adapt to different detection scenarios and accuracy requirements, and improves the flexibility and practicality of the system. Through denoising and filtering processing, environmental interference, measurement errors, and noise points can be effectively removed, making the obtained point cloud data more accurate and providing a reliable basis for subsequent modeling and analysis. The voxel downsampling technology is used to reduce redundant points, improve the calculation efficiency, and at the same time ensure that key geometric features are not lost, making the subsequent processing more efficient. Through multi-view registration, the point cloud data collected from different angles can be fused, the occlusion effect can be eliminated, and a more complete and accurate three-dimensional point cloud data can be generated, thus ensuring the comprehensiveness and accuracy of the subsequent steel grating weld detection.

[0010] Optionally, the converting of the three-dimensional point cloud data by using the implicit surface modeling method to obtain the physical model of the steel grating specifically includes: obtaining multiple sampling points included in the three-dimensional point cloud data; calculating the influence value corresponding to each sampling point by using a preset smoothing function; superimposing the multiple influence values to obtain a continuous function; establishing a grid on the target steel grating and counting the height values corresponding to each grid point on the grid; using the continuous function to construct a surface model for the multiple height values to obtain the physical model of the steel grating.

[0011] By adopting the above technical solution, through the implicit surface modeling method, the discrete point cloud data is converted into a continuous physical model of the steel grating, which can more accurately describe the geometric features of the weld and the surrounding structure, and improve the accuracy of weld detection. The influence value of each sampling point is calculated by using a preset smoothing function and globally superimposed, so that the constructed surface model is smoother, which can effectively eliminate point cloud noise and measurement errors. By establishing a continuous function and calculating the height value at the grid points, the problems of incomplete or discontinuous surfaces that may occur in the traditional discrete point cloud method are avoided, making the physical model of the steel grating more realistic and reliable. The implicit modeling method can adapt to the complex geometric shape of the steel grating weld, and even if there are minor deformations or weld defects, it can be well presented, providing more refined geometric information support for subsequent weld detection. Through grid statistics and continuous function calculation, compared with directly processing the original point cloud data, this method can reduce the calculation amount, improve the modeling efficiency, and at the same time ensure the smoothness and integrity of the model.

[0012] Optionally, the fusion of the temperature data and the physical model of the steel grating to obtain a fusion result specifically includes: extracting a plurality of coordinate points from the physical model of the steel grating; discretizing the temperature data and determining the temperature values corresponding to each of the coordinate points according to the corresponding relationship between the temperature and the coordinate points; and corresponding and fusing the plurality of coordinate points and the plurality of temperature values to obtain the fusion result.

[0013] By adopting the above technical solution, by corresponding and fusing the temperature data with the coordinate points of the physical model of the steel grating one by one, the temperature change situation can be accurately matched in the spatial dimension, making the abnormal weld area more clearly visible. Discretizing the temperature data and mapping it to the coordinate points of the physical model can analyze the temperature distribution on a finer spatial scale, avoid the influence of coarse-grained temperature measurement, and improve the resolution of weld defect detection. Weld abnormalities are often accompanied by temperature abnormalities. After being fused with the geometric model, minor weld defects such as cracks, poor welding, or overheating of the weld can be more comprehensively identified, enhancing the reliability of detection. By discretizing the temperature data and mapping it according to the coordinate points, the calculation redundancy can be reduced, ensuring data consistency and calculation efficiency, and providing high-quality input data for subsequent neural network training. The fused data can visually display the temperature field distribution of the steel grating through visualization means, facilitating engineers to quickly judge whether there are defects in the weld and improving the interpretability and operability of detection.

[0014] Optionally, inputting the fusion result into a neural network to obtain a target detection model specifically includes: determining input features according to the plurality of coordinate points; determining output features according to the plurality of temperature values; and inputting both the input features and the output features into the neural network for training and iteratively obtaining the target detection model.

[0015] By adopting the above technical solution, taking the coordinate points as input features and the temperature values as output features enables the neural network to learn the relationship between the geometric features of the steel grating welds and the temperature distribution, thereby more accurately identifying abnormal areas. Training with a neural network can automatically extract features from the data, reduce the complexity of manually setting rules, avoid the subjectivity of parameter adjustment in traditional methods, and improve the intelligent level of detection. By continuously training and optimizing the weights, the neural network enables the target detection model to adapt to the weld states under different working conditions, improves the ability to identify weld defects, and reduces false detections and missed detections. Combining geometric information and temperature information, the neural network can more comprehensively learn the multi-dimensional features of weld quality. Compared with the detection method using a single data source, it can more accurately distinguish normal welds and abnormal welds, and improve the reliability of detection.

[0016] Optionally, predicting the temperature of the target steel grating using the target detection model to obtain a normal area and / or an abnormal area specifically includes: calculating a PDE residual based on the temperature data; predicting a temperature gradient field according to the target detection model; calculating a target predicted temperature corresponding to a target area based on the temperature gradient field and the PDE residual, where the target area includes the welds of the target steel grating; comparing the size relationship between the target predicted temperature and a preset threshold to obtain a comparison result; and determining the target area as the normal area or the abnormal area according to the comparison result.

[0017] By adopting the above technical solution, by calculating the partial differential equation (PDE) residual and the temperature gradient field, it not only relies on data-driven neural network prediction but also combines physical information constraints, improves the physical consistency of weld abnormality detection, and reduces false detections and missed detections. Calculating the predicted temperature of the target area based on the temperature gradient field and the PDE residual can more accurately identify the thermal abnormalities in the weld area compared to simply relying on the original temperature data, and improves the discrimination accuracy of abnormal areas. The target detection model automatically calculates the temperature threshold and compares the predicted temperature with the preset threshold to achieve automatic judgment of the normal area and the abnormal area, reduces the intervention of human experience, and improves the intelligent degree of detection. This method can not only identify whether the weld is qualified but also accurately locate the position of the abnormal area through the change of the temperature gradient field, facilitating subsequent maintenance and quality tracking, and improving the practicality of detection. The combination method of using the PDE residual and the neural network is not limited to specific steel grating materials or weld morphologies, has strong adaptability, and can be extended to the detection of other types of welded structures, improving the generality and engineering value of the method.

[0018] Optionally, determining the target area as the normal area or the abnormal area according to the comparison result specifically includes: if it is determined that the target predicted temperature is greater than or equal to the preset threshold, determining the target area as the abnormal area; if it is determined that the target predicted temperature is less than the preset threshold, determining the target area as the normal area.

[0019] By adopting the above technical solution, whether the weld area is abnormal can be quickly determined through simple threshold comparison, ensuring the efficiency and interpretability of the detection process, and facilitating the understanding and application by engineering personnel. This method only needs to calculate the temperature prediction value and make a comparison, without complex post-processing steps, so it can quickly complete the judgment of weld quality and meet the requirements of on-line detection. Setting a reasonable temperature threshold can reduce the influence of environmental noise or errors on the detection result and improve the stability and reliability of abnormal area recognition. The preset threshold can be adjusted according to different steel grating materials, welding processes and environmental temperatures, making the detection method highly adaptable and applicable to a variety of industrial application scenarios. This method can be seamlessly integrated into an automatic detection system to achieve continuous detection and alarm, reduce manual intervention, and improve the intelligent level and production efficiency.

[0020] In a second aspect of the present application, an automatic steel grating weld detection device based on point cloud is provided. The automatic steel grating weld detection device includes an acquisition module and a processing module. Among them, the acquisition module is used to acquire three-dimensional point cloud data of a target steel grating; the processing module is used to convert the three-dimensional point cloud data by using an implicit surface modeling method to obtain a physical model of the steel grating; the acquisition module is also used to acquire temperature data of the target steel grating; the processing module is also used to fuse the temperature data and the physical model of the steel grating to obtain a fusion result; the processing module is also used to input the fusion result into a neural network to obtain a target detection model; the processing module is also used to use the target detection model to predict the temperature of the target steel grating to obtain a normal area and / or an abnormal area, where the normal area indicates that the weld of the target steel grating is qualified, and the abnormal area indicates that the weld of the target steel grating is unqualified.

[0021] In a third aspect of the present application, an electronic device is provided. The electronic device includes a processor, a memory, a user interface and a network interface. The memory is used to store instructions. The user interface and the network interface are both used to communicate with other devices. The processor is used to execute the instructions stored in the memory so that the electronic device executes the method described above.

[0022] In a fourth aspect of the present application, a computer-readable storage medium is provided. The computer-readable storage medium stores instructions, and when the instructions are executed, the method described above is executed.

[0023] In summary, one or more technical solutions provided in this application have at least the following technical effects or advantages: By obtaining 3D point cloud data to construct an implicit physical model of the steel grating, and then fusing the temperature data with the physical model, geometric and heat conduction information is fully utilized. The fusion of such multi-source data can more comprehensively reflect the actual state of the steel plate weld. Temperature data can reveal discontinuities in material properties and abnormal heat diffusion at the weld. After combining with the physical model and predicting through a neural network, the normal area and the abnormal area can be more accurately distinguished, thus effectively identifying whether the weld is qualified. Inputting the fusion result into a neural network to train an object detection model to achieve automated detection reduces the subjectivity and inconsistency of manual detection, improves the detection efficiency and reliability. This method uses a data-driven neural network model to enable real-time temperature prediction and weld state determination, which is applicable to large-scale industrial applications and online monitoring, facilitating the improvement of the accuracy of the steel grating weld detection. Description of the Drawings

[0024] Figure 1 It is a schematic flow chart of an automatic steel grating weld detection method based on point cloud provided by an embodiment of this application; Figure 2 It is another schematic flow chart of an automatic steel grating weld detection method based on point cloud provided by an embodiment of this application; Figure 3 It is a schematic module diagram of an automatic steel grating weld detection device based on point cloud provided by an embodiment of this application; Figure 4 It is a schematic structural diagram of an electronic device provided by an embodiment of this application.

[0025] Description of the reference numerals: 31, acquisition module; 32, processing module; 41, processor; 42, communication bus; 43, user interface; 44, network interface; 45, memory. Detailed Embodiments

[0026] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of this specification. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all the embodiments.

[0027] In the description of the embodiments of the present application, words such as "for example" or "for illustration" are used to give examples, illustrations or explanations. Any embodiment or design solution described as "for example" or "for illustration" in the embodiments of the present application should not be construed as being more preferred or more advantageous than other embodiments or design solutions. Rather, the use of words such as "for example" or "for illustration" is intended to present the relevant concepts in a specific manner.

[0028] In the description of the embodiments of the present application, the term "a plurality of" means two or more. For example, a plurality of systems means two or more systems, and a plurality of screen terminals means two or more screen terminals. In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly specifying the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of such features. The terms "include", "comprise", "have" and their variants all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0029] The weld inspection of steel grating is a key link in the quality control of steel structures, which is directly related to the safety and reliability of the project.

[0030] At present, weld inspection mainly relies on manual inspection and visual assessment, but this method has many limitations. Since the weld positions are often relatively hidden and affected by factors such as environmental light, observation angle, and the experience of operators, it is difficult for manual inspection to comprehensively and accurately identify the true shape and subtle defects of the welds. This limitation not only reduces the accuracy of inspection but also may lead to instability in quality assessment and increase project risks.

[0031] To solve the above technical problems, the present application provides an automatic steel grating weld detection method based on point cloud, referring to Figure 1 , Figure 1 is a schematic flow chart of an automatic steel grating weld detection method based on point cloud provided by an embodiment of the present application. This automatic steel grating weld detection method based on point cloud is applied to a server and includes steps S110 to S160. The above steps are as follows: S110. Obtain three-dimensional point cloud data for the target steel grating.

[0032] Specifically, the server is the core computing and data processing center of the entire weld detection system, responsible for receiving, storing, and analyzing the point cloud data from the scanning device, and performing subsequent processing such as denoising, modeling, and defect identification. The target steel grating refers to a specific steel structure component to be detected, usually welded by multiple flat steels arranged in parallel and twisted steel bars intersecting vertically. Its weld quality directly affects the load-bearing capacity and service life of the overall structure. The point cloud data is mainly obtained through 3D scanning devices (such as lidar, structured light cameras, industrial 3D scanners, etc.), which can accurately capture the spatial coordinate information on the surface of the steel grating. The point cloud data is a three-dimensional space data set composed of X, Y, and Z coordinate points, and each point represents a physical position on the surface of the steel grating, which helps to build its accurate three-dimensional model. After the scanning device collects the data, it is usually transmitted to the server through a wired / wireless network (such as industrial Ethernet, 5G, or Wi-Fi) for subsequent processing.

[0033] In a possible implementation manner, obtaining the three-dimensional point cloud data for the target steel grating specifically includes: receiving the original scanning data for the target steel grating sent by the scanning device, where the scanning device includes a laser scanner, 3D lidar, and a structured light camera; denoising and filtering the original scanning data to obtain the target scanning data; performing voxel downsampling and multi-view registration on the target scanning data to generate the three-dimensional point cloud data.

[0034] Specifically, the laser scanner measures the distance of the object surface by emitting laser light, with high precision and is suitable for high-precision scanning of complex steel grating structures. The 3D lidar, based on lidar technology, can quickly obtain a large range of point cloud data and is suitable for real-time monitoring of the steel structure production line. The structured light camera constructs the three-dimensional shape of the object by projecting a specific grating pattern and analyzing its deformation, and is suitable for fine scanning of smaller-sized steel gratings. The scanning device is installed on a robotic arm or a track detection system and moves along the surface of the steel grating to collect the spatial information of the weld and its surrounding area point by point. A handheld scanning device can also be used for on-site detection, and the inspector moves the device to perform a full-range scan of the steel grating.

[0035] The point cloud data collected by the device is usually the original scanning data, which contains a large amount of noise, redundant points, and irregular point distributions and needs further processing. Statistical filtering removes abnormal points in the point cloud data, such as isolated points caused by reflection or environmental interference. Radius filtering eliminates outliers that are too far from adjacent points to improve the continuity of the point cloud data. Mean filtering smooths the coordinates of each point to make the point cloud more uniform and reduce measurement errors. Gaussian filtering is used to smooth the surface of the point cloud and remove noise while maintaining edge features. Downsampling filtering can reduce the amount of point cloud data and improve the efficiency of subsequent processing. The target scanning data, which is the data after denoising and filtering, is cleaner and more continuous and can more accurately reflect the weld details of the steel grating.

[0036] Due to the huge amount of original point cloud data, voxel downsampling can reduce the data scale while retaining key morphological information, improving the computational efficiency of subsequent processing. The server divides the space into small three-dimensional grids and retains only one representative point in each grid to reduce the data volume. Since a single scanning device may not be able to cover the entire steel grating, it is necessary to align and fuse the scanning data from different angles to generate a complete three-dimensional model. The ICP algorithm aligns the point clouds by iteratively calculating the best matching relationship between the point clouds from different scanning perspectives. Global registration combines feature point matching to stitch the point cloud data from multiple perspectives into a complete model. The data scale is reduced through voxel downsampling to improve the computational efficiency; multi-perspective registration ensures seamless fusion of different scanning data to obtain a complete and accurate three-dimensional structure of the steel grating.

[0037] For example, in a certain steel structure production workshop, an automatic weld detection system is installed. This system uses a fixed 3D lidar to collect the point cloud data of the steel grating welds. The 3D lidar scans the steel grating and transmits the original point cloud data to the server. The server denoises and filters the data, removes the interference points, and extracts the effective weld information. Since the steel grating is large, the lidar scans from different angles, and the ICP algorithm is used to register the data from multiple perspectives. The point cloud scale is reduced through voxel downsampling to improve the calculation speed, and finally a complete three-dimensional point cloud model of the steel grating is generated. This method realizes efficient and automated weld detection, avoids the instability of manual detection, and improves the quality control level of the production line.

[0038] S120. Convert the three-dimensional point cloud data using an implicit surface modeling method to obtain the physical model of the steel grating.

[0039] Specifically, implicit surface modeling is a modeling method that describes the surface of an object through a mathematical function and is commonly used to process complex geometric bodies, especially being very effective in the processing of three-dimensional point cloud data. Different from explicit modeling, implicit modeling does not directly store the surface of the object but describes the object surface through an implicit function (usually a scalar field). The implicit surface is represented by defining a scalar function. The place where the function value is zero is the surface of the object. That is to say, the object surface is the set of all points that satisfy. Implicit modeling can naturally handle irregular and complex shapes, especially being very effective for those situations where the surface is not easily expressed directly (such as those with cracks or irregular surfaces). Three-dimensional point cloud data is obtained through scanning devices (such as lidar, structured light cameras, etc.) and contains a large number of discrete points on the surface of the target object. The goal of implicit surface modeling is to convert these discrete points into a continuous mathematical model, thereby generating a three-dimensional physical model of the steel grating. The server inputs the point cloud data into the implicit surface modeling algorithm, converts each point in the point cloud data into a spatial coordinate, and then constructs a smooth surface through methods such as interpolation and fitting. These surfaces can approximately describe the true surface morphology of the steel grating, including its shape, thickness, texture and other physical properties. Through implicit surface modeling, the server generates a mathematically continuous surface model, which can not only be used for three-dimensional visualization display but also support subsequent analyses, such as detecting weld quality and stress analysis.

[0040] In a possible implementation manner, an implicit surface modeling method is adopted to convert three-dimensional point cloud data to obtain a physical model of the steel grating, specifically including: obtaining a plurality of sampling points included in the three-dimensional point cloud data; calculating the influence value corresponding to each sampling point by using a preset smoothing function; superimposing the plurality of influence values to obtain a continuous function; establishing a grid on the target steel grating and statistically obtaining the height values corresponding to each grid point on the grid; constructing a surface model by using the continuous function for the plurality of height values to obtain a physical model of the steel grating.

[0041] Specifically, the three-dimensional point cloud data is obtained by scanning devices such as laser scanners, 3D lidars, or structured light cameras. Each point represents a small part of the steel grating surface. This process involves a large number of data points, usually dense, containing the shape, size, and surface features of the object. Suppose a steel grating is scanned with a 3D lidar. The lidar emits thousands of laser beams per second and receives the time difference of the reflected light, and then calculates the spatial coordinates of the reflection points. These coordinate points constitute the original three-dimensional point cloud data. Since the point cloud data often contains noise and irregular sampling intervals, smoothing processing is required to eliminate these inaccurate parts. This step uses preset smoothing functions (such as Gaussian functions, hyperbolic functions, etc.) to process each sampling point and calculate the influence value. The influence value reflects the degree of influence of each sampling point on the surrounding points in the entire surface modeling. If a sampling point is far from other points, its influence value may be small; while if it is close to other sampling points, its influence value is large. In this way, the smoothing function helps to interpolate between points, making the transition of the entire model smoother.

[0042] By superimposing the influence values of all sampling points, a continuous implicit function can be obtained. This function represents the shape of the steel grating surface. Usually, the form of the implicit function is a certain mathematical expression, and the zero points of the function (i.e., where the function value is zero) correspond to the surface of the object. If a Gaussian smoothing function is used, the influence value of each sampling point is weighted by the Gaussian distribution function, and finally a smooth surface is formed. This means that the surface of the model will be smooth without sharp corners. Next, a grid will be established on the steel grating surface, usually through meshing processing (such as square or rectangular grids). Each grid point represents a position on the object surface. For these grid points, the corresponding height value is statistically calculated, that is, the distance from the steel grating surface to a certain reference plane. Suppose a uniformly distributed grid is created on the steel grating surface, and the position of each grid point can be obtained by interpolation calculation. For each grid point, its height value is calculated according to the continuous function.

[0043] Finally, using the continuous function obtained above and combining the height values on the grid, a complete surface model is constructed. This model is a smooth and continuous surface that can approximate the actual shape of the steel grating and reflect its surface features. Use continuous functions (such as quadratic functions or B-spline functions) to fit the height values of all grid points, and finally generate a smooth steel grating surface model. This surface model can reflect the details of the steel grating surface, such as corrugations, welding marks, etc.

[0044] Through these steps, the implicit surface modeling method can transform discrete 3D point cloud data into a complete and continuous physical model of the steel grating. The benefits of this process include: Smoothing reduces noise and irregularities in the point cloud data, making the final physical model smoother and more accurate. Through the interpolation and superposition of smoothing functions, more complex geometric shapes can be processed to adapt to different surface features. Implicit modeling can generate a continuous surface model, which is not only convenient for subsequent analysis but also better combines with other computational models (such as finite element analysis).

[0045] For example, assume that in a large construction project, accurate modeling of the steel grating is required. To obtain complete information about the surface of the steel grating, engineers use a 3D lidar scanner to scan the steel grating and obtain a large amount of point cloud data. Then, through the steps of implicit surface modeling, this method helps engineers obtain an accurate and smooth model of the steel grating, which can be effectively used for subsequent quality inspection, welding analysis, etc. to ensure that the quality of the steel grating meets the engineering requirements.

[0046] S130. Obtain temperature data for the target steel grating.

[0047] Specifically, obtaining temperature data means measuring the temperature of the steel grating in real time or regularly through different sensors or monitoring devices and transmitting the measurement results to the server. These devices are usually temperature sensors or thermal imagers, which can be used to accurately capture and record the temperature changes of the target object. The obtained temperature data usually has temporality, that is, the temperature data will also change over time. Assume that the target steel grating is a device installed in a high-temperature environment, such as a conveyor belt in steel production or a steel structure in a building. To ensure the safety of the structure, it is necessary to monitor the temperature of the steel grating to prevent weld cracking or structural deformation caused by overheating. By installing multiple temperature sensors (such as thermocouples or infrared sensors) on the surface of the steel grating, the temperature data of the steel grating can be obtained regularly.

[0048] According to application requirements and the actual environment, temperature data can be obtained in various ways. Common temperature sensors include: Thermocouple: Suitable for high-temperature environments, it calculates the temperature by measuring the voltage difference between two different metals. Infrared temperature sensor: Used for non-contact measurement, it can sense temperature changes on the surface of the target at a long distance. The infrared sensor can quickly respond to temperature changes and is not affected by the surface material. RTD (Platinum Resistance Thermometer): A high-precision temperature sensor, commonly used in industrial environments, can provide relatively stable and linear temperature data.

[0049] In this process, the main role of the server is to receive data from various temperature sensors, and store, process, and further analyze this data. The server saves this real-time or batch temperature data into a database, and conducts analysis, prediction, and processing according to subsequent requirements. The server may communicate with other devices (such as sensor gateways or data acquisition devices), receive data, and store and process it. Assume that multiple temperature sensors are installed on the steel grating, and the sensors send data to the central server via a wireless network. The server receives, stores, and, as needed, conducts real-time processing and analysis of this data, such as determining whether the temperature in a specific area exceeds a preset safety range.

[0050] For example, Case 1: Temperature monitoring of steel gratings in steel production. In a steel production plant, steel gratings are used to carry high-temperature materials. To ensure the safety of the equipment, the factory needs to monitor the temperature of these steel gratings in real time. For this purpose, various temperature sensors are installed in the plant area, such as thermocouples and infrared sensors, to monitor the temperature at different positions of the steel gratings respectively. The thermocouples installed at different positions of the steel grating measure the surface temperature in real time, while the infrared temperature sensors measure the overall thermal radiation of the steel grating at a certain distance. Each sensor sends the collected temperature data to the central server via a wireless network. After receiving this data, the server analyzes and stores the temperature data. Based on this real-time data, the server can detect whether the temperature in a local area exceeds the safety threshold. If an anomaly is found, the system can automatically alarm to remind the factory operator to conduct an inspection or take measures.

[0051] Case 2: Temperature monitoring of steel gratings in a building. In the steel structure of a building, steel gratings are used to support floors. To ensure that the steel gratings do not deform or get damaged during temperature changes, building engineers install multiple infrared temperature sensors on the steel gratings. The engineers fix these infrared temperature sensors at key parts of the steel gratings, especially the welded joints and support points, which are areas prone to being affected by temperature changes. These infrared sensors send the temperature data to the server via a wireless network, and the server regularly receives and analyzes the data. By analyzing this temperature data, the server can evaluate whether the steel grating has overheated and judge whether the structure will deform according to the temperature change. For example, the server may monitor the temperature of a certain part of the steel grating, and if the temperature exceeds the set safety threshold, the system will issue an alarm to prompt the management to conduct repairs or replacements.

[0052] S140. Integrate the temperature data and the physical model of the steel grating to obtain an integration result.

[0053] Specifically, the three-dimensional physical model of the steel grating is converted from point cloud data and includes the geometric structure of the steel grating, including the position of welds, the thickness and shape of the steel grating, etc. To combine temperature data with the physical model, it is first necessary to extract multiple key coordinate points from the physical model. These coordinate points represent specific positions on the surface of the steel grating, such as at welds or areas with greater stress. Assume that the server has constructed a physical model of the steel grating, which contains all the points on the surface of the steel grating, and the coordinates (X, Y, Z) of each point are known. These points can be extracted for subsequent data fusion. The temperature data is measured by sensors or thermal imagers and is usually two-dimensional or three-dimensional data. To combine it with the physical model of the steel grating, the temperature data needs to be discretized and mapped to the surface coordinates of the steel grating.

[0054] If a point-type temperature sensor (such as a thermocouple) is used, then each sensor corresponds to a specific spatial position (X, Y, Z), and the temperature data can be directly associated with this position. If an infrared thermal imager or 3D temperature scanner is used, the acquired temperature data is a two-dimensional thermal map (temperature matrix), and it needs to be mapped to the three-dimensional surface model of the steel grating through coordinate transformation. Interpolation or fitting methods are used to fill in the unmeasured temperature data points to ensure that the temperature data covers the entire surface of the steel grating. For example, an infrared thermal imager takes a thermal distribution map of the steel grating and finds that the temperature in the weld area is relatively high. The server needs to map this thermal map to the three-dimensional model of the steel grating so that each point in the physical model can be associated with a temperature value. After fusing the coordinate points in the three-dimensional point cloud model with the temperature data, each coordinate point will correspond to a temperature value, thus forming a complete fusion model. This fusion model can be used for subsequent analysis, such as anomaly detection and weld defect identification. After server calculation, a complete temperature distribution model of the steel grating is obtained, which can visually show which areas have high temperatures (possibly with weld defects) and which areas have normal temperatures.

[0055] For example, Case 1: Weld temperature monitoring in manufacturing. Assume that a steel structure manufacturing factory is producing a batch of steel gratings and uses automated welding equipment for welding. To detect the weld quality, they use an infrared thermal imager to scan the steel gratings after welding to obtain the temperature distribution of the welds. A three-dimensional scanning device (such as a 3D lidar) is used to obtain the three-dimensional point cloud data of the steel gratings and construct a physical model of the steel gratings. An infrared thermal imager takes a thermal distribution map of the welds to obtain the temperature data of each pixel point. The server fuses the data from the thermal imager with the physical model of the steel grating and maps the temperature information to the three-dimensional coordinate points of the welds. By analyzing the temperature distribution, if the temperature in some areas of the weld is significantly higher than the normal range, it may indicate that there are welding defects (such as incomplete penetration or overheating affecting the material strength).

[0056] Case 2, long-term monitoring of bridge steel grating. The steel grating of a certain bridge may be affected by environmental temperature changes, rain erosion, etc. due to long-term exposure outdoors. The bridge management department installed multiple temperature sensors to monitor the temperature of the steel grating in order to detect whether the structure is damaged due to temperature changes. The steel grating of the bridge is scanned by lidar to obtain its three-dimensional point cloud data, and a geometric model of the steel grating is generated. Temperature sensors are installed at key positions (such as welds and connection points), and the temperature data is uploaded to the server at regular intervals. The server fuses the temperature data with the three-dimensional coordinate points of the steel grating to generate a thermal distribution model of the steel grating that changes over time. If it is found that the temperature fluctuation in a certain area is abnormal, for example, there is local high temperature in winter, it may indicate that there are cracks or material fatigue in that area.

[0057] Therefore, by fusing temperature information with the physical model, the thermal abnormal areas of the steel grating can be analyzed more intuitively, and whether the weld quality is qualified can be judged. The fused data can be used to generate a three-dimensional temperature distribution map to help engineers visually view the thermal state of the steel grating and conduct key analysis in the weld area. The server can further train an AI model or neural network based on the fusion result to achieve automated weld defect detection and improve the detection efficiency. After the server obtains the temperature data, it fuses it with the three-dimensional physical model of the steel grating, so that each point of the physical model has temperature information. This fusion enables the temperature anomalies of the welds to be intuitively mapped to specific spatial positions, thereby improving the accuracy and automation of weld detection. This method is widely used in scenarios such as manufacturing and bridge monitoring, and can effectively improve the safety and service life of steel gratings.

[0058] In a possible implementation manner, the temperature data and the physical model of the steel grating are fused to obtain a fusion result, which specifically includes: extracting multiple coordinate points from the physical model of the steel grating; discretizing the temperature data, and determining the temperature values corresponding to each coordinate point according to the correspondence between the temperature and the coordinate points; corresponding and fusing the multiple coordinate points with the multiple temperature values to obtain a fusion result.

[0059] Specifically, the physical model of the steel grating is converted from point cloud data, and it contains the geometric information of the steel grating (such as the position of the weld, the thickness of the plate, etc.). In order to fuse the temperature data with the model, multiple coordinate points need to be extracted from this physical model. These coordinate points include: weld key points (spatial coordinates of the weld area), grid nodes (uniform sampling points on the surface of the steel grating), and feature points (important structural points such as corners and boundaries). For example, assume that after the point cloud data of the steel grating is processed, a structured grid model is generated, which contains 10,000 coordinate points, and the coordinates of each point can be expressed as (X, Y, Z). These points cover the entire surface of the steel grating, and the point density in the weld area is relatively high to ensure the accuracy of weld detection.

[0060] Temperature data is usually measured by sensors (such as infrared thermal imagers or temperature sensor arrays), which may be a continuously varying temperature field or discrete measurement points distributed at different positions. For the convenience of fusion, it is necessary to discretize the temperature data, that is, to associate the temperature values with specific spatial coordinate points. If the temperature data is collected by sensors at fixed positions and the coordinates are known, the coordinate points and temperature values can be directly matched. If the temperature data is a continuously distributed thermal image (such as a two-dimensional image captured by an infrared thermal imager), but the physical model is three-dimensional, the two-dimensional temperature data needs to be mapped to the three-dimensional coordinate points of the steel grating through projection transformation or interpolation calculation. If the temperature data is sparsely distributed, spatial interpolation algorithms (such as radial basis function interpolation, Kriging interpolation, etc.) can be used to estimate the temperature values of unmeasured points to align them with the point cloud data on the surface of the steel grating.

[0061] For example, Scenario 1 (direct matching): 100 temperature sensors are installed on the steel grating, each sensor corresponding to a known coordinate (X, Y, Z), and the corresponding temperature data is collected. For example, the temperature at (10, 20, 0) is 55 °C, and the temperature at (15, 25, 0) is 60 °C.

[0062] Scenario 2 (interpolation calculation): Use an infrared thermal imager to photograph the steel grating to obtain a temperature thermal image of 640×480. This is a 2D image, which needs to be projected into 3D space through coordinate transformation (such as perspective transformation), and the temperature value of each pixel point is mapped to the three-dimensional surface coordinate points of the steel grating.

[0063] Scenario 3 (meshing): If only 10 temperature points are measured and there are 10,000 points on the surface of the steel grating, interpolation algorithms need to be used to calculate the temperature values of the remaining 9,990 points in order to generate complete temperature distribution data.

[0064] After completing the extraction of coordinate points and the discretization of temperature data, the next step is to correspond and fuse the temperature data with the coordinate points of the physical model of the steel grating to form a three-dimensional temperature field model for subsequent weld detection and analysis. One-to-one correspondence is made between the temperature data and the surface coordinate points of the steel grating. A structured data table or three-dimensional matrix is constructed to record the (X, Y, Z) and temperature value T(X, Y, Z) of each point. Color mapping (such as pseudocolor) is used to visualize the temperature distribution of the fused steel grating, making different temperature regions clearly visible in the image.

[0065] S150. Input the fusion result into a neural network to obtain an object detection model.

[0066] Specifically, compared with manual visual inspection, the neural network can quickly and accurately detect weld anomalies, improving the detection efficiency. Even in different environments (temperature fluctuations, light changes), the neural network can still maintain a high detection accuracy. The model trained based on long-term data can not only identify current defects but also predict possible future damage trends for early maintenance. The server can output the detection results as a three-dimensional heat map to visually display the abnormal areas of the steel grating. The core of this step is to use the neural network to train the three-dimensional temperature field data of the steel grating to establish an automated weld anomaly detection model. By inputting features such as coordinates, temperature, and gradient and training the neural network, the server can predict which areas may have welding defects, thus achieving intelligent quality control and fault warning.

[0067] In a possible implementation, the fusion result is input into the neural network to obtain the target detection model, which specifically includes: determining the input features according to multiple coordinate points; determining the output features according to multiple temperature values; inputting both the input features and the output features into the neural network for training, and iteratively obtaining the target detection model.

[0068] Specifically, the server obtains the three-dimensional temperature field data of the steel grating and labels the normal / abnormal conditions of some welds. The server extracts the input features (coordinates, temperature, gradient, etc.). The server trains the neural network to generate a weld anomaly detection model. The server uses this model to predict the new steel grating to determine whether there are welding defects. The prediction results can be used in the intelligent maintenance system. For example, if an anomaly is detected, further inspection can be arranged. Example: The server detects that the temperature of a certain weld is abnormal (80°C) and predicts that there may be a "lack of fusion" defect at this point, and recommends further ultrasonic inspection for confirmation.

[0069] For example, the server obtains the temperature distribution of the bridge steel grating every day and stores the historical data. The historical data is used to train the neural network so that the model can identify abnormal patterns that change over time. The server predicts the temperature anomaly trend of the steel grating in real time to determine whether early maintenance is required. The server finds that the temperature of a certain steel grating has increased by 10°C compared to the previous few days and predicts that a crack may occur after 3 months, improving the preventive maintenance ability.

[0070] S160. Use the target detection model to predict the temperature of the target steel grating to obtain the normal area and / or abnormal area. The normal area indicates that the weld of the target steel grating is qualified, and the abnormal area indicates that the weld of the target steel grating is unqualified.

[0071] Specifically, the server inputs these temperature data into the target detection model, which predicts the temperature abnormality area through deep learning methods, mainly including the following processing: calculating the temperature gradient, calculating the rate of change of temperature in space, and areas with sudden temperature changes may have weld defects. The server uses the heat conduction equation to calculate the theoretical temperature distribution and compares it with the actual temperature data to determine the abnormal area. Based on a large amount of training data, the neural network predicts whether the temperature of the weld area conforms to the normal mode. According to the output of the target detection model, the server will divide the weld area of ​​the steel grating into normal area and abnormal area: Normal area: The temperature distribution is as expected, indicating that the weld is defective. Abnormal area: The temperature distribution is abnormal, and there may be welding defects, such as cracks, uneven welding, and cold welding.

[0072] In a possible implementation, a target detection model is used to predict the temperature of a target steel grating to obtain a normal area and / or an abnormal area, specifically including: calculating a PDE residual based on temperature data; predicting a temperature gradient field based on the target detection model; calculating a target predicted temperature corresponding to the target area based on the temperature gradient field and the PDE residual, wherein the target area includes a weld of the target steel grating; comparing the target predicted temperature with a preset threshold to obtain a comparison result; and determining whether the target area is a normal area or an abnormal area based on the comparison result.

[0073] Specifically, PDE is used in heat conduction analysis to describe how heat diffuses inside metal materials. Ideally, the temperature in the weld area should follow a certain mathematical model. However, the actual measured data may have deviations, so it is necessary to calculate the PDE residual. First, establish a heat conduction equation to calculate how the temperature in the weld area should be distributed based on the thermal conductivity characteristics of steel. Obtain the actual temperature of the weld from an infrared thermal imager or temperature sensor. Compare the actual temperature with the theoretically calculated temperature. Areas with large deviations may have welding defects. The temperature gradient field represents the spatial rate of change of temperature on the steel grating. Under normal circumstances, the temperature gradient around the weld should transition smoothly, but where there are welding defects, the temperature change may suddenly increase or decrease. Input temperature data to the target detection model (deep neural network) to predict the temperature gradient field, that is, calculate the temperature change rate of each point, and judge the abnormal area. The location where the temperature changes dramatically may have defects. Based on the PDE residual and the temperature gradient field, the server calculates the predicted temperature of the target area, that is, the temperature value it should reach if the area is a normal weld. Based on the temperature gradient field and the surrounding temperature, predict the normal temperature value of the current point. Compare with the actual measured temperature to find the area with large deviations. The server compares the calculated target predicted temperature with a preset threshold: if the temperature is abnormal (exceeds the threshold), there may be a weld defect. If the temperature is normal (below the threshold), the weld may be acceptable.

[0074] In a possible implementation, referring to Figure 2 , Figure 2 is another schematic flowchart of an automatic detection method for steel grating welds based on point clouds provided by an embodiment of the present application. The method includes steps S210 to S220, and the above steps are as follows: S210. If it is determined that the target predicted temperature is greater than or equal to a preset threshold, then determine the target area as an abnormal area; S220. If it is determined that the target predicted temperature is less than the preset threshold, then determine the target area as a normal area.

[0075] Specifically, the main objective of this step is to determine whether there are defects in the welds of the target steel grating based on the comparison between the target predicted temperature and the preset threshold. That is, if the target predicted temperature is greater than or equal to the preset threshold, it is considered that there may be welding defects in this area, and it is determined as an abnormal area. If the target predicted temperature is less than the preset threshold, it is considered that the weld quality is qualified and it is determined as a normal area.

[0076] The target predicted temperature is calculated based on temperature data, PDE residuals, and temperature gradient fields, and is used to characterize the ideal temperature state of the weld area. Under normal circumstances, the temperature of the steel grating weld area should be within a reasonable range. However, if there are defects in the welds, such as cracks, pores, lack of fusion, etc., it may cause local temperature to rise or fall abnormally. The preset threshold is the boundary temperature value used to distinguish between normal areas and abnormal areas. The preset threshold is determined by experimental data, industry standards, or historical data. For example: the normal temperature range of the weld area is 40°C - 80°C. If the temperature of a certain area is ≥85°C, there may be welding defects.

[0077] The present application also provides an automatic detection device for steel grating welds based on point clouds. Referring to Figure 3 , Figure 3 is a module schematic diagram of an automatic detection device for steel grating welds based on point clouds provided by an embodiment of the present application. The device is a server, and the server includes an acquisition module 31 and a processing module 32. Among them, the acquisition module 31 acquires three-dimensional point cloud data for the target steel grating; the processing module 32 converts the three-dimensional point cloud data by using an implicit surface modeling method to obtain a physical model of the steel grating; the acquisition module 31 acquires temperature data for the target steel grating; the processing module 32 fuses the temperature data and the physical model of the steel grating to obtain a fusion result; the processing module 32 inputs the fusion result into a neural network to obtain a target detection model; the processing module 32 uses the target detection model to predict the temperature of the target steel grating to obtain a normal area and / or an abnormal area. The normal area indicates that the welds of the target steel grating are qualified, and the abnormal area indicates that the welds of the target steel grating are unqualified.

[0078] In a possible implementation, the acquisition module 31 acquires three-dimensional point cloud data for the target steel grating, which specifically includes: the acquisition module 31 receives the original scan data for the target steel grating sent by the scanning device, and the scanning device includes a laser scanner, a 3D lidar, and a structured light camera; the processing module 32 denoises and filters the original scan data to obtain the target scan data; the processing module 32 performs voxel downsampling and multi-view registration on the target scan data to generate three-dimensional point cloud data.

[0079] In a possible implementation, the processing module 32 converts the three-dimensional point cloud data by using an implicit surface modeling method to obtain a physical model of the steel grating, which specifically includes: the acquisition module 31 acquires a plurality of sampling points included in the three-dimensional point cloud data; the processing module 32 calculates the influence value corresponding to each sampling point by using a preset smoothing function; the processing module 32 superimposes the plurality of influence values to obtain a continuous function; the processing module 32 establishes a grid on the target steel grating and counts the height values corresponding to the grid points on the grid; the processing module 32 constructs a surface model by using the continuous function for the plurality of height values to obtain a physical model of the steel grating.

[0080] In a possible implementation, the processing module 32 fuses the temperature data and the physical model of the steel grating to obtain a fusion result, which specifically includes: the processing module 32 extracts a plurality of coordinate points from the physical model of the steel grating; the processing module 32 discretizes the temperature data and determines the temperature value corresponding to each coordinate point according to the correspondence between the temperature and the coordinate points; the processing module 32 performs corresponding fusion on the plurality of coordinate points and the plurality of temperature values to obtain a fusion result.

[0081] In a possible implementation, the processing module 32 inputs the fusion result into a neural network to obtain a target detection model, which specifically includes: the processing module 32 determines an input feature according to the plurality of coordinate points; the processing module 32 determines an output feature according to the plurality of temperature values; the processing module 32 inputs both the input feature and the output feature into the neural network for training and iteratively obtains a target detection model.

[0082] In a possible implementation, the processing module 32 uses the target detection model to predict the temperature of the target steel grating to obtain a normal area and / or an abnormal area, which specifically includes: the processing module 32 calculates a PDE residual according to the temperature data; the processing module 32 predicts a temperature gradient field according to the target detection model; the processing module 32 calculates a target predicted temperature corresponding to a target area based on the temperature gradient field and the PDE residual, and the target area includes the weld of the target steel grating; the processing module 32 compares the size relationship between the target predicted temperature and a preset threshold to obtain a comparison result; the processing module 32 determines whether the target area is a normal area or an abnormal area according to the comparison result.

[0083] In a possible implementation, the processing module 32 determines whether the target area is a normal area or an abnormal area according to the comparison result, which specifically includes: if the processing module 32 determines that the target predicted temperature is greater than or equal to the preset threshold, it determines that the target area is an abnormal area; if the processing module 32 determines that the target predicted temperature is less than the preset threshold, it determines that the target area is a normal area.

[0084] It should be noted that when the device provided in the above embodiment realizes its functions, only the division of the above functional modules is used for illustration. In practical applications, the above functions can be allocated to different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the device and method embodiments provided in the above embodiment belong to the same concept, and the specific implementation process can be seen in the method embodiment, which will not be elaborated here.

[0085] This application also provides an electronic device. Refer to Figure 4 , Figure 4 which is a schematic structural diagram of an electronic device provided in an embodiment of this application. The electronic device may include: at least one processor 41, at least one network interface 44, a user interface 43, a memory 45, and at least one communication bus 42.

[0086] Among them, the communication bus 42 is used to realize the connection and communication between these components.

[0087] Among them, the user interface 43 may include a display screen (Display) and a camera (Camera). Optionally, the user interface 43 may further include a standard wired interface and a wireless interface.

[0088] Among them, the network interface 44 may optionally include a standard wired interface and a wireless interface (such as a Wi-Fi interface).

[0089] Among them, the processor 41 may include one or more processing cores. The processor 41 connects various parts within the entire server through various interfaces and lines. By running or executing instructions, programs, code sets, or instruction sets stored in the memory 45, and by calling the data stored in the memory 45, it performs various functions of the server and processes data. Optionally, the processor 41 may be implemented in at least one hardware form of digital signal processing (DSP), field-programmable gate array (FPGA), or programmable logic array (PLA). The processor 41 may integrate a combination of one or several of a central processing unit (CPU), a graphics processing unit (GPU), and a modem, etc. Among them, the CPU mainly processes the operating system, user interface, application programs, etc.; the GPU is responsible for rendering and drawing the content to be displayed on the display screen; the modem is used to process wireless communication. It can be understood that the above-mentioned modem may not be integrated into the processor 41 and may be implemented separately by a single chip.

[0090] Among them, the memory 45 may include random access memory (RAM) and may also include read-only memory. Optionally, the memory 45 includes a non-transitory computer-readable storage medium. The memory 45 can be used to store instructions, programs, code, code sets, or instruction sets. The memory 45 may include a program storage area and a data storage area. Among them, the program storage area may store instructions for implementing the operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-mentioned various method embodiments, etc.; the data storage area may store the data involved in the above-mentioned various method embodiments. Optionally, the memory 45 may also be at least one storage device located far from the aforementioned processor 41. As Figure 4 shown, the memory 45, as a computer storage medium, may include an operating system, a network communication module, a user interface module, and an application program for an automatic detection method of steel grating welds based on point clouds.

[0091] In Figure 4In the electronic device shown, the user interface 43 is mainly used to provide an interface for the user to input and obtain the data input by the user; the processor 41 can be used to call an application program stored in the memory 45 for an automatic detection method of steel grating welds based on point clouds. When executed by one or more processors, the electronic device is caused to execute the method of one or more of the above embodiments.

[0092] It should be noted that, for the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that this application is not limited by the described action sequence, because according to this application, certain steps can be in other sequences or performed simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0093] This application also provides a computer-readable storage medium storing instructions. When executed by one or more processors, the electronic device is caused to execute the method of one or more of the above embodiments.

[0094] In the above embodiments, the descriptions of the respective embodiments each have their own emphases. For the parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0095] In the several embodiments provided by this application, it should be understood that the disclosed device can be implemented in other ways. For example, the device embodiments described above are only illustrative. For example, the division of units is only a logical function division. In actual implementation, there can be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection between each other can be through some service interfaces. The indirect coupling or communication connection of the device or unit can be in an electrical or other form.

[0096] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they can be located in one place, or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0097] In addition, in each embodiment of this application, the functional units can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above integrated units can be implemented in the form of hardware or in the form of software functional units.

[0098] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present application. The aforementioned memory includes various media that can store program codes, such as USB flash drives, mobile hard disks, magnetic disks, or optical discs.

[0099] The foregoing are only exemplary embodiments of the present disclosure, and the scope of the present disclosure cannot be limited thereby. That is, any equivalent changes and modifications made in accordance with the teachings of the present disclosure still fall within the scope covered by the present disclosure. After considering the specification and the practice of the present disclosure, those skilled in the art will readily think of other implementation manners of the present disclosure. The present application is intended to cover any variations, uses, or adaptive changes of the present disclosure, and these variations, uses, or adaptive changes follow the general principles of the present disclosure and include common general knowledge or conventional technical means in the technical field not recorded in the present disclosure. The specification and the embodiments are only regarded as exemplary, and the scope and spirit of the present disclosure are defined by the claims.

Claims

1. An automatic detection method for steel grating welds based on point cloud, characterized in that, The method includes: Obtaining three-dimensional point cloud data for a target steel grating; Converting the three-dimensional point cloud data by using an implicit surface modeling method to obtain a physical model of the steel grating; Obtaining temperature data for the target steel grating; Fusing the temperature data and the physical model of the steel grating to obtain a fusion result; Inputting the fusion result into a neural network to obtain a target detection model; Using the target detection model to predict the temperature of the target steel grating to obtain a normal region and / or an abnormal region, where the normal region indicates that the weld of the target steel grating is qualified, and the abnormal region indicates that the weld of the target steel grating is unqualified.

2. The automatic detection method for steel grating weld seams based on point cloud according to claim 1, characterized in that, The obtaining of the three-dimensional point cloud data for the target steel grating specifically includes: Receiving the original scan data for the target steel grating sent by a scanning device, where the scanning device includes a laser scanner, a 3D lidar, and a structured light camera; Denosing and filtering the original scan data to obtain target scan data; Performing voxel downsampling and multi-view registration on the target scan data to generate the three-dimensional point cloud data.

3. The automatic detection method for steel grating welds based on point cloud according to claim 1, characterized in that, The converting of the three-dimensional point cloud data by using the implicit surface modeling method to obtain the physical model of the steel grating specifically includes: Obtaining a plurality of sampling points included in the three-dimensional point cloud data; Calculating an influence value corresponding to each of the sampling points by using a preset smoothing function; Superimposing a plurality of the influence values to obtain a continuous function; Establishing a grid on the target steel grating and counting the height values corresponding to the grid points on the grid; Constructing a surface model for a plurality of the height values by using the continuous function to obtain the physical model of the steel grating.

4. The automatic detection method for steel grating welds based on point cloud according to claim 1, characterized in that, The fusing of the temperature data and the physical model of the steel grating to obtain the fusion result specifically includes: Extracting a plurality of coordinate points from the physical model of the steel grating; Discretizing the temperature data and determining the temperature values corresponding to the respective coordinate points according to the correspondence between the temperature and the coordinate points; Correspondingly fusing a plurality of the coordinate points and a plurality of the temperature values to obtain the fusion result.

5. The automatic detection method for steel grating welds based on point cloud according to claim 4, wherein The inputting of the fusion result into the neural network to obtain the target detection model specifically includes: Determining input features according to a plurality of the coordinate points; Determining output features according to a plurality of the temperature values; Inputting both the input features and the output features into the neural network for training and iteratively obtaining the target detection model.

6. The automatic detection method for steel grating welds based on point cloud according to claim 1, wherein The using of the target detection model to predict the temperature of the target steel grating to obtain the normal region and / or the abnormal region specifically includes: Calculating a PDE residual according to the temperature data; Predicting a temperature gradient field according to the target detection model; Calculating a target predicted temperature corresponding to a target region based on the temperature gradient field and the PDE residual, where the target region includes the weld of the target steel grating; Comparing the magnitude relationship between the target predicted temperature and a preset threshold to obtain a comparison result; Determining the target region as the normal region or the abnormal region according to the comparison result.

7. The automatic detection method for steel grating welds based on point cloud according to claim 6, characterized in that Determining whether the target area is the normal area or the abnormal area according to the comparison result specifically includes: If it is determined that the target predicted temperature is greater than or equal to the preset threshold, determining that the target area is the abnormal area; If it is determined that the target predicted temperature is less than the preset threshold, determining that the target area is the normal area.

8. An automatic detection device for steel grating welds based on point cloud, characterized in that, The automatic steel grating weld detection device includes an acquisition module (31) and a processing module (32), where the acquisition module (31) is configured to acquire three-dimensional point cloud data for a target steel grating; the processing module (32) is configured to convert the three-dimensional point cloud data by using an implicit surface modeling method to obtain a physical model of the steel grating; the acquisition module (31) is further configured to acquire temperature data for the target steel grating; the processing module (32) is further configured to fuse the temperature data and the physical model of the steel grating to obtain a fusion result; the processing module (32) is further configured to input the fusion result into a neural network to obtain a target detection model; the processing module (32) is further configured to use the target detection model to predict the temperature of the target steel grating to obtain a normal area and / or an abnormal area, where the normal area indicates that the weld of the target steel grating is qualified, and the abnormal area indicates that the weld of the target steel grating is unqualified.

9. An electronic device, characterized in that, The electronic device includes a processor (41), a memory (45), a user interface (43), and a network interface (44). The memory (45) is used to store instructions. The user interface (43) and the network interface (44) are both used to communicate with other devices. The processor (41) is used to execute the instructions stored in the memory (45) so that the electronic device executes the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions, and when the instructions are executed, the method according to any one of claims 1 to 7 is executed.

Images